Form-focused instruction and the development of … than logical that teaching gramma ... either overtly or covertly, ... pressing in the light of the oft heard complaint in pedagogy

Form-focused instruction and the development of second

language proficiency

Sible Andringa

Copyright © 2005 by Sible Andringa Cover design: Ronald Koekoek Printed by Print Partners Ipskamp, Enschede

The work in this thesis has been carried out under the auspices of the Behavioral and Cognitive Neurosciences (BCN) research school, Groningen

Groningen Dissertations in Linguistics 56 ISSN 0928-0030

RIJKSUNIVERSITEIT GRONINGEN

Form-focused instruction and the development of second language proficiency

PROEFSCHRIFT

ter verkrijging van het doctoraat in de Letteren

aan de Rijksuniversiteit Groningen op gezag van de

Rector Magnificus, dr. F. Zwarts in het openbaar te verdedigen op

donderdag 10 november 2005 om 14.45 uur

door

Sible Johannes Andringa

geboren op 17 april 1975 te Roodhuis

Promotor: Prof. dr. C.M. de Glopper Copromotor: Dr. H.I. Hacquebord Beoordelingscommissie: Prof. dr. C.L.J. de Bot

Prof. dr. R. Ellis Prof. dr. J.H. Hulstijn

Dankwoord

Het schrijven van een proefschrift is geen gemakkelijke opgaaf. Dat het me is gelukt, is te danken aan de wijze raad, de goede adviezen, de belangstelling, de steun en het begrip van de mensen om me heen. Ik wil iedereen hier heel hartelijk voor bedanken.

Allereerst, Kees en Hilde, wil ik jullie bedanken voor jullie betrokkenheid bij dit onderzoek. Ik heb veel vrijheid van jullie gekregen om mijn onderzoek vorm te geven zoals het nu is geworden, en ik heb dat vertrouwen ontzettend gewaardeerd. Jullie inspiratie, wijze raad, praktische adviezen en aanmoediging waren onmisbaar, en hebben ontzettend veel richting gegeven aan dit onderzoek. Ik hoop dat ik hier ook in de toekomst af en toe nog eens gebruik van mag maken.

Nanette, ook jij bent ontzettend belangrijk geweest voor het welslagen van dit onderzoek. Onze gesprekken in de beginfase waren een belangrijke bron van inspiratie. Met jouw ervaring op het gebied van onderzoek naar vormgerichte instructie kon je me wijzen op de mogelijkheden en de gevaren van dit soort onderzoek. Bedankt hiervoor.

Ik ben ook veel dank verschuldigd aan de ISK-leerlingen die deel hebben genomen aan dit onderzoek, en hun docenten die dat mogelijk hebben gemaakt. Ik ben in de loop van dit onderzoek op veel ISK-scholen geweest. Steeds weer werd ik getroffen door het optimisme en de toewijding van de leerlingen - ondanks de vaak moeilijke persoonlijke omstandigheden waarin zij zich bevonden, en de bevlogenheid van hun docenten. Het komt mij voor dat het onderwijs zoals dat op ISK’s wordt vormgegeven in veel opzichten ideaal is, en navolging verdient in het reguliere voortgezet onderwijs. Het is niet mogelijk om iedereen hier met naam en toenaam te bedanken: ik noem alleen Henk Vruggink en zijn toemalige leerlingen, bij wie ik een poos lang elke week hartelijk werd ontvangen om les- en toetsmateriaal uit te proberen.

Joanneke en Anke, mijn paranimfen, jullie wil ik bedanken voor je vriendschap en collegialiteit. Ik kon bij jullie altijd mijn verhaal kwijt, en jullie zorgden voor de nodige luchtigheid. Ik noem bespiegelingen over het belang van feng shui en de ‘koop-gordijnencampagne’ (die zelfs de nationale pers haalde) als voorbeelden van de ontelbare vrolijke noten die er waren. Het geeft vertrouwen te weten dat de twee personen die zo belangrijk voor me zijn geweest op mijn werk ook straks naast me lopen.

Ook wil ik alle Harmoniecollega’s bedanken voor hun gezelligheid en collegialiteit. Ik noem mijn kamergenoten, Danielle, Nesli en Marjolein, alle collega-aio’s, taalkundigen, de medewerkers van het secretariaat, de Anglisten, en

de Etoc’ers. De vele lunches, de koffie en de printergesprekken met jullie (en met de komst van de service pantry natuurlijk ook de koffiemachinegesprekken), hebben ervoor gezorgd dat ik me altijd heel erg thuis heb gevoeld op de Harmonie, en dat ik nu terug kan zien op een mooie periode. Bedankt allemaal.

Verder, Wander en Marjolijn, wil ik jullie bedanken voor het lezen van hoofdstukken van dit proefschrift en je nuttige commentaar daarop. Ruurd, je technische adviezen bij het maken het computerprogramma waren onmisbaar. Je kwam zelfs speciaal naar Groningen om me te helpen en voorkwam zo dat dit project vroegtijdig strandde: dank je wel. Heit en Froukje, dank je wel voor het uitproberen van toetsmateriaal bij jullie op school. En Ronald, dank je voor alle tijd die je hebt gestoken in de vormgeving van dit boek. Het resultaat is geweldig!

Vrienden en familie, en heit en mem in het bijzonder, jullie hebben op indirecte wijze ook een belangrijke bijdrage geleverd aan dit proefschrift. Bedankt voor de ontspanning, jullie nimmer aflatende belangstelling en morele steun, maar ook voor jullie begrip als ik het weer eens liet afweten vanwege dit project. Oh ja... mem: ik heb gewonnen! Wat hadden we er tussen gezet?

Kris, aan jou heb ik meer te danken dan ik kan zeggen: zonder jou geen proefschrift! Je moedigde me aan en leidde me af, en altijd op het juiste moment. Bedankt voor je geduld. Nu jij!

Contents 1 Introduction __________________________________________________ 1 2 Form-focused instruction and the development of L2 proficiency________ 7

2.1 Introduction ______________________________________________ 7 2.2 The interface debate________________________________________ 8

2.2.1 Introduction __________________________________________ 8 2.2.2 Three interface positions ________________________________ 9 2.2.3 Developmental readiness and natural orders of acquisition____ 14 2.2.4 The type of target structure _____________________________ 16 2.2.5 Individual differences__________________________________ 18 2.2.6 Summarizing the interface debate________________________ 19

2.3 Second language knowledge, use, and learning __________________ 22 2.3.1 Introduction _________________________________________ 22 2.3.2 The nature of second language grammatical knowledge_______ 23 2.3.3 Second language use___________________________________ 25 2.3.4 Implicit learning processes______________________________ 28 2.3.5 Explicit knowledge and learning _________________________ 30 2.3.6 Reappraising the interface debate ________________________ 32

2.4 The interface debate and form-focused instruction research _______ 36 2.4.1 Introduction _________________________________________ 36 2.4.2 Measures of explicit and implicit knowledge________________ 37 2.4.3 FFI and its impact on implicit knowledge: Ellis (2002) ________ 38 2.4.4 The interface between explicit and implicit knowledge _______ 40 2.4.5 Developmental readiness, structure complexity and ID’s ______ 46 2.4.6 Conclusion __________________________________________ 49

2.5 Summary _______________________________________________ 50 3 Methodology ________________________________________________ 55

3.1 Research questions and research design _______________________ 55 3.2 The instruction___________________________________________ 59

3.2.1 The target structures __________________________________ 59 3.2.2 The instruction_______________________________________ 63 3.2.3 Input matching_______________________________________ 65

3.3 Assessing grammatical development__________________________ 66 3.3.1 Introduction _________________________________________ 66 3.3.2 Explicit grammatical knowledge _________________________ 68

3.3.3 Implicit grammatical knowledge _________________________ 70 3.4 The participants __________________________________________ 73

3.4.1 Introduction _________________________________________ 73 3.4.2 The subjects and their educational background _____________ 73 3.4.3 Mortality ___________________________________________ 74 3.4.4 General L2 proficiency _________________________________ 75 3.4.5 Exposure to the second language_________________________ 77

3.5 Assessment of individual characteristics_______________________ 78 3.5.1 Introduction _________________________________________ 78 3.5.2 Developmental readiness _______________________________ 78 3.5.3 First (and second) language background ___________________ 79 3.5.4 Age ________________________________________________ 80 3.5.5 Aptitude ____________________________________________ 80 3.5.6 Affective variables ____________________________________ 81

3.6 Procedures and data processing______________________________ 83 3.6.1 Introduction _________________________________________ 83 3.6.2 Data collection _______________________________________ 83 3.6.3 Organisation of the instruction __________________________ 85 3.6.4 Coding _____________________________________________ 86 3.6.5 Statistical procedures __________________________________ 90

4 Findings ____________________________________________________ 93

4.1 Introduction _____________________________________________ 93 4.2 Learning to use grammar ___________________________________ 93

4.2.1 Introduction _________________________________________ 93 4.2.2 Structure realizations and correct use _____________________ 94 4.2.3 Developing the ability to use the target structure___________ 101 4.2.4 Conclusion _________________________________________ 105

4.3 FFI, explicit knowledge, and grammatical development __________ 106 4.3.1 Introduction ________________________________________ 106 4.3.2 FFI and the development of the degrees of comparison ______ 107 4.3.3 FFI and the development of subordinate clauses ___________ 110 4.3.4 The interface between explicit and implicit knowledge ______ 113 4.3.5 Conclusion _________________________________________ 115

4.4 FFI and learner characteristics______________________________ 117 4.4.1 Introduction ________________________________________ 117 4.4.2 Developmental readiness ______________________________ 118 4.4.3 L1 similarity ________________________________________ 123 4.4.4 Aptitude, cognitive style, motivation and age ______________ 125 4.4.5 Conclusion _________________________________________ 129

4.5 Summary: differences in the use of DoC and SubC ______________ 130

4.5.1 Introduction ________________________________________ 130 4.5.2 Differences in characteristics of use _____________________ 130 4.5.3 FFI, grammatical development, and structure complexity ____ 131 4.5.4 Individual characteristics and structure complexity _________ 132 4.5.5 Conclusion _________________________________________ 134

5 Discussion _________________________________________________ 135

5.1 Introduction ____________________________________________ 135 5.2 Evaluation of the research design ___________________________ 135 5.3 Evaluation of theory______________________________________ 139

5.3.1 Introduction ________________________________________ 139 5.3.2 The nature of implicit L2 knowledge _____________________ 139 5.3.3 FFI and the development of explicit and implicit knowledge __ 142 5.3.4 Individual differences and the effectiveness of instruction ___ 146 5.3.5 Structure complexity and the effectiveness of FFI __________ 151 5.3.6 The interface issue ___________________________________ 153

5.4 Implications for second language pedagogy ___________________ 155 5.5 Conclusion _____________________________________________ 157

Appendices _____________________________________________________ 161

A1: Screenshots of the computer programme ______________________ 161 A2: A detailed outline of the instruction __________________________ 162 A3: Rule explanation of the degrees of comparison _________________ 166 A4: Rule explanation of subordination ___________________________ 168 B1: Grammaticality judgement test______________________________ 171 B2: Controlled production _____________________________________ 173 B3: The free written response task ______________________________ 175 C1: The C-test_______________________________________________ 182 C2: the rote memory test ______________________________________ 184 C3: the grammatical sensitivity test _____________________________ 185 C4: the teacher questionnaire __________________________________ 187

References _____________________________________________________ 189 Nederlandse samenvatting ________________________________________ 197 Groningen dissertations in linguistics (GRODIL) ________________________ 211

1 Introduction

The common way to talk about language is in terms of its grammar (Hulstijn, 2002). In fact, probably the most common conception of language is that it consists of a large set of lexical items and a set of rules, and if you are learning a second language, the task at hand is to learn to put the words and rules of a second language to proper use. Seeing language as a logically organised system of words and rules is not just a popular view; many linguists hold it too. Given the pervasiveness of the notion of language consisting of a finite set of rules, it is no more than logical that teaching grammar is an important component of second language instruction. Indeed, second language learning programmes, even those that are functionally and communicatively oriented, are invariably based on so-called pedagogical grammars: either overtly or covertly, they involve a structured approach to the presentation of the rules of grammar.

In fact, course books and resource books such as New ways in teaching grammar (Carswell Pennington, ed., 1995) testify to the countless different ways in which grammar can be taught, and a question that has concerned educators and researchers alike is how grammar should be taught to achieve second language proficiency. After all, the ability to use the second language (L2) fluently in speech and writing is the ultimate goal of any teaching effort. This question is especially pressing in the light of the oft heard complaint in pedagogy that grammar teaching does not lead to the changes in student behaviour sought for. Kwakernaak (1995), for example, expresses reservations about the value of practicing grammar rules in isolation in that there does not seem be any transfer to writing, let alone speaking. Similarly, Macrory and Stone (2000) demonstrated that their students, after four years of French, were able to score well on perfect tense gap-filling exercises. They were able to express this knowledge in terms of rule-like statements, too. However, performance in spontaneous discourse was poor, and did not reflect what they explicitly knew to be correct. This raises doubts as to the practical value of having conscious knowledge of the grammar of the L2 to becoming proficient.

This concern has been addressed by theories of instructed second language acquisition. These theories usually make a distinction between explicit and implicit linguistic knowledge, although different labels have sometimes been used (Bialystok, 1994b; DeKeyser, 1998; R. Ellis, 1990; 1994b; Krashen, 1981). Explicit

2 CHAPTER 1

knowledge, also referred to as declarative or learned knowledge, denotes factual and conscious knowledge about the rules of the second language. Implicit knowledge, or procedural or acquired knowledge, refers to the knowledge that enables someone to use the second language appropriately in spontaneous situations of language use. In terms of explicit and implicit knowledge, the question is whether there can be an interface between explicit and implicit knowledge. Teaching and practising the rules of grammar in isolation seems a valuable experience only when explicit knowledge directly affects the acquisition or use of implicit knowledge, i.e., when there is an interface between explicit and implicit knowledge.

Within the field of second language acquisition, so-called form-focused instruction (FFI) research has put theories of instructed second language acquisition to the test, dealing with issues concerning the effectiveness of different types of instruction in different circumstances of learning. In a discussion of FFI, R. Ellis defines it as “any planned or incidental instructional activity that is intended to induce language learners to pay attention to linguistic form” (2001: p. 1/2). Although the majority of FFI studies have focused on the domain of grammar, the term form in form-focused instruction actually refers to all formal aspects of language: to grammar, but also to pronunciation, spelling, intonation, etc. It should also be pointed out that incidental in this definition is not the opposite of planned. Incidental FFI is equally planned, in that it is intentionally provided when communication problems incidentally arise. The term FFI, then, covers a broad range of activities all focusing the learner’s attention on formal aspects of the L2.

FFI research dates back to the 1960’s, and FFI has been a popular research topic throughout the years (2001). The picture that emerges from a number of reviews of FFI studies (Ellis, 1994b; Larsen-Freeman & Long, 1991; Lightbown, 2001; Long, 1983; Norris & Ortega, 2000) is first that FFI should be integrated in curricula that are primarily meaning-oriented, focusing on functional use of the L2 and communication. Another conclusion is that there is – at least to some extent – an interface between explicit and implicit knowledge, and that grammar should play a considerable role in second language learning programmes. Consequently, FFI research has moved from the question of whether FFI is effective to the investigation of what kinds of FFI are effective (Ellis, 2001). Currently, an issue of primary interest is how FFI can be adapted to promote input processing mechanisms. ‘Focus on form’ research, for example, investigates the value of shortly shifting the learner’s attention to form during communication, based on the claim that the window of opportunity for effective

INTRODUCTION 3

FFI is small (Doughty, 2001a; see: Doughty & Williams, eds., 1998). VanPatten’s ideas of Processing Instruction have also generated considerable interest (VanPatten, 1996; VanPatten & Cadierno, 1993; VanPatten & Sanz, 1995). His argument is that FFI should optimize L2 processing by informing learners about mismatches between their L1 and the L2. Finally, another research interest relates to how FFI interacts with numerous other variables, such as the nature of grammar structures, aptitude, L1 background, etc. (Norris & Ortega, 2000).

There are, however, a number of reasons to reconsider the yieldings of FFI research so far. Both Ellis (2001) and Hulstijn (1997) point out that despite the amount of attention FFI has received, it is still difficult to draw conclusions with a fair degree of certainty. The reason lies in the complexity of the FFI research agenda: many different methods of investigation have been used; few replicating studies have been done; and potentially intervening variables have been recognized or controlled for insufficiently. In addition, Norris and Ortega (2000) have demonstrated that FFI research has suffered from considerable bias: L2 progress has been assessed with measures of explicit knowledge much more than with measures of implicit knowledge. In a meta-analysis of FFI research, they synthesized the results of 49 FFI studies and found that the largest effects were generally obtained with explicit types of instruction. However, no more than eight studies included in the analysis measured progress by means of free constructed response tasks. On this note, Norris & Ortega remark about FFI research that “the measurement of change induced by instruction is typically carried out on instruments that seem to favour more explicit types of treatments by calling on explicit memory-based performance.” (p. 483). The implication of this research bias is that claims of superiority of explicit types of instruction over implicit types of instruction need to be reconsidered. As Doughty puts it: “the case for explicit instruction has been overstated.” (2003, p. 274).

This study assesses the importance of explicit knowledge resulting from

explicit instruction to the development of second language proficiency. The experiment that this book reports on involved monitoring second language development of learners of Dutch as a second language receiving either explicit or implicit instruction about two grammar structures. Measures of both decontextualized and natural language performance were used to assess progress. The explicit instruction was designed to provide the language learners with conscious knowledge of the targeted grammar structures, while the implicit instruction was designed to expose the language learners to the target structures, so that the amount and nature of exposure to the target structures were equal

4 CHAPTER 1

between the compared groups. In doing so, both theoretical and practical purposes are served. Theoretically, the study addresses the interface issue and appraises claims regarding the presence or absence of an interface between explicit and implicit knowledge. In addition, this study serves to evaluate FFI research findings, and the extent to which conclusions drawn so far have been ‘misguided’ due to the already mentioned bias in measurement of progress. The practical value of this study is rather obvious: the question of how to integrate grammar teaching into second language learning programmes is still very much alive in everyday teaching. Knowledge about how L2 learners put explicit knowledge of the second language to use can provide important insights into curriculum design and the development of instructional material.

It should be pointed out that this study is not intended to evaluate the merits of a particular type of grammar instruction. In both pedagogy and research, a wide variety of different types of instruction have been used. However, throughout this book, only explicit and implicit instruction are contrasted. The terms explicit instruction and implicit instruction are used to refer to two types of instruction in which attention to form is either overt or covert. As soon as the instruction involves explanation of rules, or if learners are asked to discover rules, the instruction must be considered explicit. Conversely, when rules are not discussed and learners are not asked to attend to rules during L2 tasks, the instruction is implicit (Norris & Ortega, 2000). The most typical example of explicit instruction, and one which has often been used in form-focused instruction research, is traditional teacher-fronted rule explanation (e.g., DeKeyser, 1997). Another example of explicit instruction that has been investigated quite frequently is input processing and practise, which consists of tasks designed to promote or practise forms and their meanings (e.g., Salaberry, 1997; VanPatten & Cadierno, 1993). Flooding techniques (e.g., Williams & Evans, 1998) and input enhancement (e.g., Jourdenais, Ota, Stauffer, Boyson & Doughty, 1995; Trahey & White, 1993; White, 1998) in which forms are made salient through high frequent presentation or typographical enhancement (e.g., underlining) are the most notable examples of implicit FFI. It is possible to characterize instructional treatments further, for example according to their primary focus (e.g., Norris & Ortega, 2000). However, as the goal is to compare and contrast explicit and implicit instruction rather than specific types of instruction, no further distinctions will be made throughout this report.

The organisation of this book is straightforward. Chapter 2 discusses the

theoretical likelihood of and empirical evidence for the presence of an interface

INTRODUCTION 5

between explicit and implicit knowledge. The interface debate itself is outlined; it is evaluated in the light of recent developments in SLA theory; and FFI research findings are scrutinized in search of support for claims related to the interface issue. In Chapter 3, research questions are proposed based on the analysis of the interface issue. In addition, the design and organisation of the experiment conducted are expounded. In Chapter 4, the findings are presented, and these are discussed in Chapter 5.

2 Form-focused instruction and the development of L2 proficiency

2.1 Introduction

The goal of this chapter is to evaluate the importance of explicit knowledge in relation to second language proficiency. It should be pointed out that this is an issue of considerable controversy in two ways. One point of controversy is the role of explicit knowledge in communication. Although most researchers agree that spontaneous and undeliberate language use is based in implicit knowledge, some argue that highly planned instances of language use are in fact based on explicit knowledge rather than implicit knowledge. Sharwood Smith provides some examples of language use driven by explicit knowledge: preparing a question, a speech or a telephone conversation (1988). The second point of controversy concerns the development of L2 proficiency: does knowledge about the second language in some way provide scaffolds for L2 proficiency? Is it the starting point or the catalyst of second language acquisition? This study addresses this latter point: whether explicit knowledge promotes the development of implicit knowledge.

The three sections that this chapter consists of all deal with this issue from very different angles. Section 2.2 discusses the interface debate, as this debate focuses on the organisation of linguistic knowledge; and as such, on the role of explicit knowledge in L2 acquisition. The purpose of this section is to describe (rather than critically discuss) the potential relationships between explicit and implicit knowledge, and how this relationship affects FFI. Section 2.3 and 2.4 both provide critical evaluations of the interface debate. In 2.3, the interface debate is considered in the light of recent developments in SLA theory. In doing so, different strands of SLA theory are brought together with a heavy focus on precisely defining implicit and explicit linguistic knowledge. Second language acquisition is considered in the light of the nature and representation of linguistic knowledge; how linguistic knowledge is put to use in communication; and what drives the acquisition of linguistic knowledge. Section 2.4 takes the insights provided by 2.2 and 2.3 as a starting point, and evaluates to what extent FFI

8 CHAPTER 2

research has provided definite answers to the contribution that explicit knowledge makes to the development of second language proficiency. Section 2.5 summarizes this chapter, and identifies issues in need of further study.

2.2 The interface debate

2.2.1 Introduction

As pointed out in the introduction (Chapter 1), the potential power of form-focused instruction can be discussed in terms of possible interfaces between explicit and implicit knowledge. The interface debate discusses the role that explicit knowledge plays in the acquisition of implicit knowledge. Three positions can be distinguished: the no interface position, the strong interface position and the weak interface position. Each of these positions claims a different role for explicit knowledge in the course of acquiring implicit knowledge, and consequently, each provides an alternative answer as to how form should be taught. As this study hopes to contribute to the interface debate, this discussion will be outlined in the present section. It is important to stress, though, that a sketchy overview of the debate is provided, discussing only the main proponents of each position. The purpose of this section is simply to identify the positions: what role they attribute to explicit knowledge, and what predictions they make concerning how and when grammar teaching is most effective in improving second language proficiency (2.2.2). In addition, a number of issues are highlighted that are related to the interface debate and that are important to the development and instruction of L2 proficiency. First, developmental readiness, the notion that FFI can only be successful when the language learner is in a particular stage of development is addressed in 2.2.3. In addition, the interface debate is related to differences between grammar structures (2.2.4), and to personal factors (2.2.5). In doing so, this section tries to reveal the central questions and concerns of the interface debate, and it will provide the framework of discussion for the following sections.

FFI AND THE DEVELOPMENT OF L2 PROFICIENCY 9

2.2.2 Three interface positions

The no interface position

The no interface position posits that the explicit and implicit knowledge systems are completely separate from each other, and it is strongly associated with Krashen and his theory of second language acquisition (1981; 1985; 1994; Krashen & Terrell, 1983). Krashen proposed the distinction between acquisition and learning. Acquisition is a subconscious process that leads to acquired knowledge (implicit knowledge), whereas learning requires conscious effort on behalf of the learner, resulting in ‘learned’ knowledge (explicit knowledge). In other words, Krashen claims that aspects of language can be internalized in two fundamentally different ways, resulting in two fundamentally different knowledge bases. Most important to the discussion here is Krashen’s claim that explicit knowledge resulting from learning plays only a very limited role in the development of second language proficiency.

For Krashen, second language acquisition is an unstoppable and inescapable process that will occur as soon as L2 learners try to understand messages in the second language. He sees second language acquisition as very similar to the way children develop their L1: it too is driven by an innate Language Acquisition Device (LAD) as proposed by Chomsky (1965), and there is a natural order in which the rules of language are internalized. However, successful second language acquisition requires that certain conditions be met. First, the input has to be comprehensible, and it has to suit to the learner’s stage of development. Therefore, if a learner is at stage “i”, comprehensible input with “i+1” qualities will lead to acquisition. The second requirement has to do with the language learner. Krashen refers to an ‘affective filter’ which when up would prevent acquisition to occur, because the input does not reach the LAD. Whether the filter is up or down depends on affective factors such as motivation, anxiety, self-confidence, etc. Krashen summarizes his theory in one single claim which he refers to as the fundamental principle in second language acquisition: “people acquire second languages only if they obtain comprehensible input and if their affective filters are low enough to allow the input ‘in’.” (1985: p. 4). Comprehensible input, then, is the only way to second language proficiency. “All other factors thought to encourage or cause second language acquisition work only when they contribute to comprehensible input and/or a low affective filter.” (1985: p. 4).

Krashen does acknowledge that the kind of knowledge accrued via learning may be advantageous. First and foremost, he attributes a very limited role to

10 CHAPTER 2

explicit knowledge in students’ production. It may serve as a monitor to edit speech or writing online, but only if the second language user is “consciously concerned about correctness; and he or she must know the rule.” (1985: p. 2). Also, it remains to be seen whether students are able to use the explicit knowledge appropriately. Second, having explicit knowledge might be advantageous if it could somehow support or aid the acquisition process. Krashen does indeed acknowledge that there may be an indirect contribution of learned knowledge to second language acquisition (1985). One possibility is that teaching grammar may lower students’ affective filters; for instance, because it satisfies certain students’ desires to know about the structure of the language they are learning, or because it positively influences their self-confidence. A third way in which teaching grammar may be helpful is that it in fact helps to make the input comprehensible, and in this way stimulates the acquisition process. If a learned rule is encountered in text, it may be more comprehensible to a student and thus increase the possibility of acquisition of other rules. Finally, Krashen allows for the possibility that the students’ own output may serve as comprehensible input. Thus, if a student can be induced to use a learned structure (at i+1) productively, acquisition may occur on the basis of his or her own input. Although explicit knowledge may help, Krashen sticks to his claim that conscious learning contributes little to second language proficiency, and that acquisition can take place without error correction, skill-building or output, but not without comprehensible input (1994).

Krashen’s downplay of the role of explicit knowledge in second language acquisition has impacted significantly on pedagogy. It led many teachers to change traditional rule-oriented language education to communication-oriented education. The title of Krashen and Terrell’s book, The Natural Approach (1983), conveys Krashen’s recommendations for pedagogy clearly. Krashen and Terrell underscore the importance of receptive types of learning. However, although natural language is the most important ingredient in his view, they do not merely advocate for students to engage in conversation. Free conversation would not constitute optimal input, because the students may not understand what is being said. Krashen, then, acknowledges that second language acquisition in formal contexts (i.e. the classroom) is more efficient than in informal contexts (1981). He feels that it requires professionals to design classroom activities that provide students with comprehensible and catching input, and that allow for or stimulate real communication.


The strong interface position

The strong interface position has its roots in cognitive psychology. This position entails a strong relation between explicit and implicit knowledge: they are typically seen as the extremes of one continuum. This means that nature of linguistic knowledge changes in the course of acquisition in such a way that it becomes increasingly more available in communicative settings. One can distinguish two variants of the strong interface position, which seem to be in complete opposition. On the one hand, Bialystok (1989; 1994a; 1994b) argues that linguistic knowledge starts out as implicit and becomes more explicit as the language learner becomes more proficient. On the other hand, researchers such as, O’Malley, Chamot, and Walker (1987), Sharwood-Smith (1988),1 and DeKeyser (1998) argue that the development of second language proficiency is a process of automatizing explicit knowledge so that it becomes implicit. Both views will shortly be discussed.

Most striking about Bialystok’s theory of second language acquisition is that

she argues that L2 proficiency develops along two dimensions: analysis and control. Analysis refers to “awareness of structure”, and it is “represented as a proposition in which the formal structure and the relationship to meaning are apparent” (1989: p. 33). Based on the assumption that language is a structured knowledge system, Bialystok argues that one of the main goals of L2 learners is to develop awareness of the structure of language. As long as a particular linguistic structure is unanalysed, it functions as a pattern or routine. But as the learner develops awareness for that structure, it becomes available for use in new contexts. In other words, as a particular piece of linguistic knowledge moves along the analysis dimension, the knowledge itself remains inherently the same, but it gradually becomes available in functionally more creative contexts (1989). Bialystok’s control dimension refers to the ability to access linguistic information (1989), or the degree of automatization. Second language learners may differ in their ability to access a particular linguistic structure, irrespective of its degree of analysis. Control, then, accounts for differences in fluency between L2 learners.

With reference to the interface hypothesis, it is important to point out that for Bialystok, explicit analysed knowledge evolves from implicit unanalysed knowledge (1994; 1994b). Therefore, all linguistic knowledge necessarily starts out as being implicit and nonautomatic. This means that explicit knowledge does not exist independently: it exists by virtue of implicit knowledge. Bialystok herself uses the

1 Originally printed in 1981 in Applied Linguistics, 2, 2: 159–168.

12 CHAPTER 2

terms implicit and explicit knowledge to refer to unanalysed and analysed knowledge, respectively. However, knowledge being explicit does not mean that it is conscious knowledge; rather, it can be brought to consciousness if called upon. Bialystok’s idea of how a second language is learned, then, is perhaps best illustrated by the following quote: “Indeed, increasing explicitness can almost serve as a definition for what we mean by ‘learning’”(1994b: p. 567).

For most proponents of the strong interface hypothesis, learning is better characterized as increasing implicitness. DeKeyser’s (1998) account of second language acquisition, for example, which is based on Anderson’s ACT model of skill acquisition (1995), proposes that linguistic knowledge typically moves through three developmental stages. It starts out as declarative (factual) then becomes procedural (knowing how), and eventually will become fully automatized.2 And once the learner reaches the final stages, the declarative knowledge that formed the basis of learning may actually be lost. How does factual knowledge become procedural? The crucial step in this process, according to DeKeyser, is proceduralization: a learner starts to engage “in target behaviour – or procedure – while leaning on declarative crutches” (p. 49). This basically means that the language learner has to start using a particular structure while keeping the declarative knowledge in mind. Repeated practice of this kind will lead to knowledge that is more procedural and that encodes behaviour rather than factual knowledge. The final step from procedural to fully automatized knowledge requires “strengthening, fine-tuning, and automatization...” (p. 49) through practice. Or, in the words of Sharwood Smith (1988: p. 52):

“Armed with explicit information about particular tasks, the learner can use conscious applications of rules to practice in and out of class and to communicate in the target language at a higher level of proficiency, albeit without the speed and spontaneity associated with the notion of ‘fluency’. Fluency is assumed to come later and as a result of practicing TL structures in formal and informal, naturalistic ways.”

The pedagogical recommendations made by proponents of the strong interface position evolve around the establishment and automatization of L2 knowledge. DeKeyser (1998), for example, recommends that declarative knowledge is established by means of mechanical drills: exercises that focus exclusively on

2 In the spirit of skill acquisition theory, DeKeyser follows Anderson in using declarative and procedural knowledge to refer to explicit and implicit knowledge. In a footnote he points out that – despite slight differences in meaning – the terms are often used interchangeably (1998: p. 48).


linguistic form and do not require the L2 learner to pay attention to meaning. However, the goal of such mechanical drills should be “to develop, test, and refine declarative knowledge, which means that the student should have ample time to think, an should never be rushed or put through activities that are so repetitive as to pre-empt all conscious rule application” (1998: p. 55). Subsequently, DeKeyser proposes ‘communicative drills’ to bring about proceduralization. Ideal communicative drills in the eyes of DeKeyser are those that require the language learner to convey new information while keeping declarative knowledge in mind. In contrast to Krashen, then, DeKeyser heavily emphasizes careful planning of classroom activities.

The weak interface position

The weak interface position has been put forward by R. Ellis (1990; 1994a; 1997). He, too, argues that implicit and explicit knowledge are two separately coexisting knowledge systems. Ellis’s theory tries to allow for findings that suggest that – for some rules – formal instruction is effective only if properly timed: these rules seem developmentally constrained (for a review, see Ellis, 1990). The weak interface position states that explicit knowledge can become implicit. However, for rules that are developmentally constrained, this can only happen if the instruction is properly timed. Also, explicit knowledge can positively affect implicit learning processes, but in such cases, the effects of instruction will be delayed rather than immediate. Another important feature of Ellis’s theory is that knowledge does not necessarily start out as being explicit. In fact, L2 knowledge mostly starts out as implicit.

For Ellis, acquisition is not so much driven by learners’ needs to understand messages, as Krashen supposes. Rather, he argues that language learners pay attention to features of the input and compare them to their own output: mechanisms referred to by Ellis as noticing and comparing. What attracts the learner’s attention may depend on various things, but not understanding the message may be one of them. Frequency, salience, particular task demands, all these aspects of the input may be noticed and compared. This process may lead the language learner to reconsider particular hypotheses he or she has about the target language, and perhaps lead to new hypotheses. Ellis refers to this latter process as integration. The importance of explicit knowledge may be that it helps the L2 learner to notice a particular rule, especially when it is communicatively redundant. On top of that, it may also help the L2 learner to notice the gap: with

14 CHAPTER 2

explicit knowledge, it may be easier to see how the input and the L2 learner’s own output differ.

Like Bialystok, Ellis tries to disentangle implicit and explicit knowledge systems and controlled and automatic processing, which he feels have unjustly been conflated in the course of the interface debate. Ellis argues that both implicit and explicit knowledge can be processed in varying degrees of automaticity. Thus, he arrives at four different types of knowledge (summarized in Table 2.1) and he does not exclude the possibility that for a particular structure all four types coexist (Ellis, 1994b): TABLE 2.1: Types of L2 knowledge, based on Ellis (1997). Type of knowledge Controlled processing Automatic processing Explicit Conscious and intentional

use of newly taught rules in grammar exercises.

Conscious and intentional use of an older rule in a variety of tasks.

Implicit Deliberate use of rules that were noticed in the input

Intuitive use of rules as in everyday use.

In line with Ellis’s views on the organisation of linguistic knowledge, Ellis’s

recommendations for pedagogy involve two pedagogical practices aimed at developing implicit and explicit knowledge (1997). The acquisition of implicit knowledge, he argues, stands to benefit most from interpretation tasks. Such tasks facilitate noticing because understanding the message crucially depends on properly interpreting the meaning expressed by the targeted grammatical structure. Explicit knowledge should be promoted by means of consciousness-raising tasks, which typically aim for discovery of a particular rule, for example, identification tasks. In addition, one of Ellis’s main concerns seems to be timing the instruction properly: when to teach which rule of grammar. As it is virtually impossible to make well-informed decisions (not enough is known about developmental orders), Ellis does not believe in approaches that offer the rules of grammar step by step. Rather, he argues in favour of receptive rather than productive tasks, aiming at comprehension rather than production.

2.2.3 Developmental readiness and natural orders of acquisition

A notion that frequently occurs in the interface debate is the notion of developmental readiness. This notion results from the observation that second


languages develop according to natural orders of acquisition. Most notably associated with the finding of natural orders of acquisition are Pienemann and associates (e.g., Meisel, Clahsen, & Pienemann, 1981; Pienemann, 1988). They reported that fixed stages can be identified in the acquisition of the word order rules of German second language learners. They hypothesized that the acquisition of German word order is subject to psychological constraints which are grounded in UG. These findings of course raise the question of whether and how instruction can impact on this process. In fact, Pienemann himself has reported that instruction is effective only if it matches the stage of development the language learner is in (e.g., 1988). Consequently, timing the instruction of a particular structure may be of crucial importance, which in turn may severely limit the contribution of explicit knowledge to the acquisition process.

However, DeKeyser (1998) downplays the importance of developmental readiness. He takes issue with the methodology of studies that have reported natural orders (often no control group, and improper operationalization of the instruction). But most importantly, he contests their cause: psychological constraints. Goldschneider and DeKeyser (2001) performed a meta-analysis on studies reporting a developmental sequence, trying to identify alternative causes for these sequences. They found that – to a very large extent – developmental sequences can be predicted by five variables: frequency, phonological salience, semantic complexity, morphological regularity, and syntactic category. This finding is difficult to match with the notion of psychological constraints, and alternatively suggests that fixed orders of acquisition are caused by the extent to which structures stand out in the input.

Irrespective of their cause, it is fair to assume that particular features of a second language develop according to what may be called ‘natural orders’. Such orders may limit the potential effectiveness of FFI, in that instruction would only be effective if the L2 learner is sensitive to it, and it suggests that instruction has to be dynamically organised, continuously adapting to the L2 learners’ needs as they move through the orders. All this implies that the success of the FFI crucially depends on knowing when to teach which structure, and in which order. However, little is known yet about what parts of language develop in a fixed order (e.g., DeKeyser, 1998; Ellis, 1997; Lightbown, 1998). But perhaps the scope of FFI is not as limited as the above suggests. First, the effect of instruction may be delayed (Ellis, 1997). The knowledge that results from FFI may not immediately affect acquisition in cases where the L2 learner was not ready, but once the learner does reach the proper stage, the knowledge that was taught previously may facilitate the acquisition process. This is an important observation for

16 CHAPTER 2

assessing the effect of instructional practices, and most FFI research includes an assessment of delayed learning effects for this very reason. Second, if natural orders are caused by factors such as frequency and salience, as suggested by Goldschneider and DeKeyser’s (2001) study, then instruction may effective if it achieves to manipulate the input in such a way that L2 learners can benefit from it.

The pedagogical recommendations that have been put forward to deal with natural orders are quite diverse. Krashen, for example, does not recommend trying too hard to match the language learner’s exact stage of acquisition. The risk of missing the proper stage is too large, he feels; as is the individual variation within classrooms. Therefore, Krashen recommends that teachers try to make themselves understood in a similar way as caretakers do for young children; they need to adjust to the speaker’s level of proficiency (Krashen, 1985). However, most consider the presence natural orders of acquisition to have severe consequences for the effectiveness of instruction. Pienemann obviously argues that instruction must match the learner’s stage of development to be effective. Similarly, one of Ellis’s main concerns is how to design a structural syllabus, which is “a list of grammatical items, usually arranged in the order in which they are to be taught” (1997: p. 135). This is obviously difficult on the basis of the limited amount of information available about natural orders.

2.2.4 The type of target structure

Another potentially constraining factor for effective FFI is the type of grammar structure to be taught. In the light of the interface debate, this implies that the practical value of explicit knowledge varies between grammar structures. It has proven difficult, though, to pinpoint exactly which features of a rule of grammar would cause such differences. First of all, as already explained in the previous subsection, the notion of developmental readiness may be involved: Ellis (Ellis, 1994a) has claimed that only those structures that are not developmentally constrained can be successfully taught. FFI about developmentally constrained structures need to match the learner’s stage of acquisition. Similarly, the features identified by Goldschneider and DeKeyser (2001) may determine the effectiveness of instructional efforts. However, other aspects have been identified as possibly interfering with the ‘teachability’ of a particular structure as well, all pertaining to the nature of the grammar structures themselves.

Several researchers have hypothesized that the effectiveness of instruction depends on differences in structure complexity. In fact, Krashen addresses this


issue, claiming that complex rules can only be learned implicitly (1981). Complexity can be described in terms of formal and functional complexity. DeKeyser (1998) provides definitions: a structure may be formally complex when it requires complex processing operations, and it may be functionally complex when the relation between form and function is opaque. However, DeKeyser also points out that there is little agreement on how this works out in practice. He gives the example of the argument between Krashen and Ellis over the formal complexity of the third person –s, which Krashen deems formally simple, whereas Ellis thinks it complex because it has to agree with the subject (in number). Also concerning functional complexity, DeKeyser argues that the common assumption that third person –s is functionally simple may not hold, because “one morpheme expresses several semantic concepts at the same time (present tense, singular, third person)” (1998: p. 44).

Hulstijn & De Graaff (1994) identify scope and reliability as factors that influence the effectiveness of FFI instruction. These factors relate to whether learning is rule-based or exemplar-based. Scope refers to the absolute number of instances and reliability refers to the percentage of instances for which the rule holds (number of exceptions). They hypothesize that explicit instruction is valuable for rules of large scope and high reliability only. Rules that are unreliable need not be taught because a learner cannot safely apply the rule, and rules that are small in scope do not require FFI because their infrequent occurrence does not justify the effort.

Finally, effectiveness of FFI has been related to L1 – L2 contrasts. Harley (1993), for example, has argued that structures that may be taught effectively are those that differ from the students’ L1 and are not salient. VanPatten (1996) has similarly argued that instruction should inform learners about mismatches between their default processing strategies (often based on their L1) and the strategies needed to successfully process L2 input. It is rather difficult, though, to predict how the L1 may affect L2 acquisition, because of the intricate ways in which grammar structures may interact. Thus, it is dangerous to generalize on the basis of findings for one particular structure and one particular set of languages (Odlin, 2003). Besides, Doughty and Williams (1998) point out that there is little knowledge about what this implies for instruction. Cross-linguistic influence may both undermine and strengthen particular instructional efforts, which makes decisions-making very difficult.

Identifying which characteristics of grammar structures affect their teachability is just the beginning of an explanation. Ultimately, claims regarding differences between grammar structures concerning their teachability need to be

18 CHAPTER 2

motivated by theories of language acquisition. For now, it will suffice to point out that effects of FFI do not necessarily apply universally to all aspects of grammar. Thus, it is possible that there is an interface between explicit and implicit grammar only for some aspects of grammar.

2.2.5 Individual differences

The possibility that individuals may differ with respect to their ability to use explicit knowledge for acquiring implicit knowledge has not inspired the interface debate very much. Only Krashen addresses the issue explicitly (1981). He relates individual variation in second language performance to differences between learners in the ability to use consciously learned knowledge as a monitor. Krashen describes how some language learners are able to monitor their output effectively and at appropriate moments, whereas other learners tend to overuse or underuse their explicit knowledge, symptomatic of respectively ‘self-conscious’ or ‘outgoing’ types of learners. In addition, Krashen relates the concepts of attitude and aptitude to acquisition and learning, respectively. For Krashen, attitude predicts acquisition, while aptitude predicts learning. A positive attitude serves to lower learners’ affective filters, and thus allowing the process of acquisition to take place. Aptitude, on the other hand, taps the ability to learn explicitly. According to Krashen, then, especially differences in attitude will predict differences between individuals in L2 proficiency development.

Krashen’s dismissal of aptitude as not being relevant to acquisition has cleared it out of the centre of attention for quite some time. However, in the last decade or so, a number of researchers (e.g., Robinson, 1997; 2001; Skehan, 1998a) have claimed that the construct of aptitude as used in aptitude test batteries such as Carroll and Sapon’s Modern language aptitude test; the MLAT (1959) was too restricted in its operationalization, tapping only explicit learning abilities, and they have attempted to redefine the construct. Robinson (2002), for example, distinguishes at least four types of aptitude complexes that are called upon in different settings of second language learning. Apart from aptitude for the ability to learn explicitly, he also distinguishes aptitude for focus on form via recasts, aptitude for incidental learning via oral content and aptitude for incidental learning via written text. Each of these complexes is facilitated by particular cognitive resources, such as attention, memory and processing speed. Skehan (Skehan, 1998a) relates aptitude to learning style, and argues that aptitude should be characterized in terms of learners’ inclinations towards analysis and memory. He argues that some learners are more naturally inclined to analytic


processing, leading to rule-based representations of language, while others are predisposed to the use of memory, leading to a large store of exemplars. It would lead too far to discuss both Robinson’s and Skehan’s models in more detail here, but both models imply that the effectiveness of developing explicit and implicit knowledge is dependent on a particular mix of cognitive abilities.

It has also been suggested that age affects second language acquisition. The Fundamental Difference hypothesis poses that older learners may require explicit information to successfully learn a second language, while young learners can do without and can learn languages entirely implicitly (Bley-Vroman, 1988). Comparing ultimate attainment levels of child and adult immigrants by means of a grammaticality judgement test, DeKeyser (2000) found that those who learned their second language at a young age substantially outperformed older learners. The adult learners were able to match the levels of attainment of young learners only if they had high verbal aptitude abilities as measured with the MLAT (Carroll & Sapon, 1959). DeKeyser argued that maturation causes changes in cognitive functioning, leading either to a loss of the ability to learn implicitly, or to an increased reliance on explicit learning abilities, or both. The implication is that for adult second language learners, but not for child L2 learners, explicit and implicit knowledge interface.

In sum, some individuals may have advantages over others in learning a second language, and these advantages seem to be related to differences in aptitude. Recently, it has been claimed that underlying the construct of aptitude are a number of cognitive resources, and differences between individuals are due to differences in reliance on such resources. In addition, some of these resources may be subject to maturation. The potential explanatory scope of aptitude for language learning is wide, in that it may also explain age and attitude (negative attitudes may result from low aptitude) effects. It is important to point out, though, that the precise involvement of aptitude is as yet rather speculative, and that a lot of work needs to be done in this area.

2.2.6 Summarizing the interface debate

In this section, three very different views on the relationship between explicit and implicit knowledge have been outlined, each attributing a different role to explicit knowledge in the development of second language proficiency. Each of the three interface positions are the result of an analysis of the process of second language acquisition, and the underlying ideas about L2 acquisition are indeed quite different. Krashen’s no interface position is the result of the parallelism he

20 CHAPTER 2

TAB

LE 2

.2

A su

mm

ary

of th

e in

terf

ace p

ositi

ons

Posi

tion

M

ain

prop

onen

ts

Org

anis

atio

n of

kn

owle

dge

Syst

em(s

) of l

earn

ing

The

role

of e

xplic

it

know

ledg

e Pe

dago

gica

l im

plic

atio

ns

No

inte

rfac

e K

rash

en

(198

1, 1

982)

Ex

plic

it a

nd

impl

icit

kno

wle

dge

are

two

sepa

rate

ly

orga

nise

d kn

owle

dge

syst

ems

Two

syst

ems:

acq

uist

ion

and

lear

ning

for i

mpl

icit

an

d ex

plic

it k

now

ledg

e,

resp

ecti

vely

.

Expl

icit

kno

wle

dge

cont

ribu

tes

only

in

dire

ctly

to L

2 pr

ofic

ienc

y de

velo

pmen

t.

Expl

icit

inst

ruct

ion

affe

cts

lear

ning

onl

y;

it h

as li

ttle

val

ue fo

r ac

quis

itio

n, a

nd m

ay

only

indi

rect

ly

prom

ote

acqu

isit

ion

Stro

ng

inte

rfac

e Bi

alys

tok

(198

9, 1

994)

; D

eKey

ser

(199

8)

Expl

icit

and

im

plic

it k

now

ledg

e ar

e th

e ex

trem

es o

f on

e co

ntin

uous

kn

owle

dge

syst

em

One

sys

tem

; For

Bi

alys

tok:

Ana

lysi

s; F

or

DeK

eyse

r: a

sta

ged

lear

ning

pro

cess

dur

ing

whi

ch d

ecla

rati

ve

know

ledg

e is

pr

oced

ural

ized

and

au

tom

atiz

ed.

For B

ialy

stok

, exp

licit

in

stru

ctio

n m

akes

le

arne

rs a

war

e of

st

ruct

ural

reg

ular

itie

s.

For D

eKey

ser,

exp

licit

kn

owle

dge

conv

erts

in

to im

plic

it k

now

ledg

e

For B

ialy

stok

: un

clea

r, b

ut a

imed

at

eith

er a

naly

sis

or

cont

rol:

For

DeK

eyse

r:

esta

blis

hmen

t and

au

tom

atiz

atio

n of

ex

plic

it k

now

ledg

e

Wea

k in

terf

ace

R. E

llis

(199

4a)

Expl

icit

and

im

plic

it k

now

ledg

e ar

e tw

o se

para

tely

or

gani

sed

know

ledg

e sy

stem

s

A s

peci

aliz

ed s

yste

m fo

r ac

quis

itio

n of

impl

icit

lin

guis

tic

know

ledg

e;

gene

ral

lear

ning

m

echa

nism

s fo

r ex

plic

it

know

ledg

e

Dep

endi

ng o

n w

heth

er

a st

ruct

ure

is p

art o

f a

deve

lopm

enta

l ord

er,

expl

icit

kno

wle

dge

eith

er c

onve

rts

into

im

plic

it k

now

ledg

e, o

r in

dire

ctly

faci

litat

es

impl

icit

lear

ning

.

Two

peda

gogi

es, o

ne

for e

ach

type

of

know

ledg

e (S

ee E

llis,

19

97)


assumes between first and second language acquisition. DeKeyser based his ideas on Skill acquisition theory derived from cognitive psychology, and Ellis’s theory is based on information processing theories as proposed by Gass (1988), Schmidt (1990; 1994) and VanPatten (1987). In Table 2.2, the three positions are summarized. In its essence, the interface debate centres on the question of how linguistic knowledge is organised. All three positions make a distinction between explicit and implicit knowledge. Both the no interface position and the weak interface position see the two types of knowledge as completely separate, while the strong interface position views linguistic knowledge to be continuous, varying with respect to the level of awareness and the level of control involved. In addition, both the no and the weak interface position make a distinction between explicit and implicit learning, while the strong interface position recognizes just one learning system.

Obviously, the role of explicit instruction in second language acquisition is most prominent in the strong interface view, and considerable importance is attached to explicit instruction of the rules of grammar. As explicit knowledge is mostly considered to be the starting point of second language proficiency (e.g., DeKeyser, 1998; O'Malley, Chamot, & Walker, 1987; Sharwood Smith, 1988), there is a direct relationship between teaching grammar and second language proficiency. The same applies to Bialystok’s (1994b) views, as she argues that L2 learners benefit from explicit information for developing analysed linguistic knowledge. Proponents of the no interface position acknowledge that explicit knowledge may to some extent facilitate the development of implicit knowledge (Krashen, 1981), but see this as an indirect effect of instruction. Explicit instruction may lower students’ affective filters, or it may make the input more comprehensible. Also, in some instances, some learners may be able to use explicit knowledge as a monitor. The role of explicit knowledge is limited to a facilitating role, though, in that it may positively affect implicit learning processes, and effects of explicit knowledge on L2 proficiency development are severely limited. The weak interface position allows for both direct and indirect effects of instruction, depending on whether the structure is subject to developmental constraints. Explicit knowledge can become implicit if the language learner is developmentally ready. Explicit knowledge can also facilitate implicit learning processes, although the effects will then be delayed.

Although the organisation of knowledge and learning processes make up an important part of the practical value of explicit knowledge, other aspects may weaken or strengthen the effect of FFI as well. An important moderating aspect may be the notion that language development is somehow subject to a natural

22 CHAPTER 2

orders of acquisition, and that instruction has to match the L2 learners’ stage of development for them to be able to benefit from the instruction. Another potentially interacting variable is the nature of the target structure. And finally, individuals may differ in their ability to exploit explicit knowledge to the benefit of L2 proficiency development. The problem with all these factors, however, is that there is little clarity on how these factors interact.

The goal of this study is to assess the value of explicit knowledge to L2

proficiency development. As the practical value of explicit knowledge is largely determined by whether there is an interface between explicit and implicit knowledge, the concern of the following two sections is to reappraise the interface issue. Both an analysis of second language acquisition theories and FFI research can provide father insights into this debate.

2.3 Second language knowledge, use, and learning

2.3.1 Introduction

One way to further the interface debate is to relate the constructs of explicit and implicit knowledge to recent developments in SLA theory in order to come to more precise definitions of explicit and implicit knowledge. In the previous section, implicit knowledge has been characterized as the ability to use the L2 grammar fluently and accurately, and as the prime source of knowledge underpinning L2 proficiency. The aim of this section is to provide more insights into the construct of implicit knowledge by teasing apart various aspects of proficiency. First, the nature and representation of the grammatical system will be considered in 2.3.2, which should shed light onto what grammar is. In 2.3.3, the focus shifts to how L2 learners use their grammars when they produce language, followed by a discussion of the mechanisms that enable second language acquisition in 2.3.4. In addition, explicit knowledge will be more carefully defined in 2.3.5. With this information, the questions that are central to the interface debate can be reappraised (2.3.6). Given what is known about the nature of explicit and implicit knowledge, can there be conversion of explicit into implicit knowledge, as supposed by the strong interface position? And what is the scope for explicit knowledge for use in communication, and can it facilitate acquisition? All in all, the section brings together various strands of second language acquisition theory to evaluate instructed SLA theories.


2.3.2 The nature of second language grammatical knowledge

This subsection focuses on the nature of the grammatical system underpinning second language proficiency. Broadly speaking, there are two schools of thought on what grammatical knowledge is and how grammatical knowledge develops: formal linguistics favours instantaneous or innatist models of language development, while constructivist approaches, to which cognitive linguistics, emergentism, connectionism, and functional linguistics belong (Ellis, 2003), see language development as emergent in the course of acquisition. In this section, second language development is approached from a constructivist perspective.

Formal linguistics argues that linguistic knowledge is governed by a universal grammar: innately specified linguistic principles or categories that guide the acquisition process. Formal linguists have primarily been concerned with representation of linguistic knowledge, and they have come a long way in describing linguistic knowledge from one stage to the next (Gregg, 2001). Second language acquisition is seen as a process of setting and resetting parameters, which means that from the very start, second language use – or at least those aspects of language that are indeed UG constrained – is rule-governed (As Sharwoord-Smith (1993) points out, principles and parameters are strictly speaking not rules, but adherence to them would result in highly systematic, rule-like behaviour). Thus, the rules of language resemble those of – say – mathematics, in that they are hard and exceptionless algorithms. The theory does not include mechanisms of learning, and has generally had a blind spot for the experiences that cause language to change and develop (Bialystok, 2002). In addition, the theory is difficult to apply to language pedagogy, because the goals of UG theory and language pedagogy are too divergent, as Ellis points out in an appraisal of UG theory (1995). It makes predictions that run counter to what pedagogical practice knows to be effective. This may in fact be a serious flaw of the theory, as one could argue that the strength of any second language acquisition theory is also determined by how well it can address pedagogical concerns. This is at least what Ellis (1995: p. 88) suggests when citing Brumfit: “second language acquisition theory needs to be compatible with the practice of teaching, as much as teaching needs to be compatible with second language acquisition theory” (Brumfit, 1994: p. 270).

In recent years, gradualist or constructivist approaches have gained currency. They posit that grammar is emergent in the course of acquisition rather than innate, and the primary mechanisms that cause such a grammar to emerge are general associative learning mechanisms. Cognitive psychologists have pointed

24 CHAPTER 2

out that the human mind specializes in the ability to note ‘if A then B’ concurrences, and these kinds of associations are at the heart of learning. Learning is largely ‘attending to regularities’ (e.g., Goodson, 2003). Most constructivists would agree that from these regularities an abstract grammar can ultimately emerge. Associations between A and B become stronger when they appear together often, until at some point the occurrence of A actually cues B: they have become a chunk. In the course of interacting with language, literally millions of chunks are established, represented in a complex network of interrelated associations. A hierarchical organisation, a grammar, originates when chunks are associated with other chunks (Ellis, 2003). Therefore, rules of grammar are essentially regularities rather than rules: strong associations that enter into rule-like behaviour. Thus, language can to some extent be captured in what seem to be rules, but also has probabilistic properties. Indeed, the work of corpus linguists (e.g., Nattinger & DeCarrico, 1992; Pawley & Syder, 1983; Sinclair, 1991) has demonstrated that there are indeed many features of language that are not in keeping with rule-like behaviour.

N. Ellis (2003) suggests that the typical course of language development is from ‘formulae’ to ‘limited scope patterns’ to ‘constructions’. The argument is basically that language learning starts out with accumulating formulae or exemplars of the second language; for L2 learners, these will primarily be words and short phrases that are stored as exemplars and called upon as such. The store of exemplars that is gradually built up this way is the source out of which ‘limited scope patterns’ grow. Limited scope patterns can be defined as lexically specified ‘rules’: i.e.: slot-and-frame patterns whose operation depends on the presence of a particular lexical environment. N. Ellis (2003) illustrates how hierarchical structure slowly emerges with an example from first language acquisition. Suppose a child is able to piece together the exemplars (Lulu)(gone), (Teddy)(gone), (the ball)(gone). Once a number of such exemplars are available, the child may abstract an overarching chunk containing an open element (X gone), where X can be filled with any concrete object or person. The scope of this slot-and-frame pattern is still limited. However, at some point it may be slotted in with a chunk as well ((funny)(man))(gone). Thus, the lexically specific nature of slot-and-frame patterns is gradually lost, and they become available for use in more, and more productive settings (Ellis, 2003; Pine & Lieven, 1997). A hierarchical structure much like a phrase grammar is the result.

Ultimately, this abstraction process leads to a grammar much like construction grammar as proposed by among others Goldberg (1995). A construction is a conventionalized form-meaning pair, and it is the basic building block of language


(Ellis, 2003; Goldberg, 1995). Each utterance is an assembly of constructions within constructions. Examples of constructions are [Adj Noun], which is a schematic construction that may be slotted in in numerous ways. Constructions also operate below word level, in the case of [Verb stem-past tense marker], which leads to walked, signed, and sometimes goed. However, a construction can be specific as well. Roman Catholic, red herring, and side mirror are examples of constructions in their own right, because together these words designate a conventionalized meaning that cannot be computed from the words’ independent meanings, and/or because the two words have become a collocation due to frequency of occurrence. In order to fully capture this, red, herring, and red herring must be independently represented constructions. The most important feature of constructions, then – whether schematic or specific – is that they carry meaning. As a final example, consider Goldberg’s sentence, Pat sneezed the napkin off the table, (1995: p. 224) an example that shows how even abstract sentence level constructions carry meaning. The verb sneeze is used in a ditransitive sentence here which is normally impossible for this verb. In this case, the sentence is still grammatical, because the prototypical meaning of ditransitive constructions – agent-successfully-causes-recipient-to-move-patient – is not compromised.

In short, the development of rule-governed L2 proficiency, or implicit knowledge, is a slow and gradual process of figuring language out on the basis of frequency of co-occurrence. As Ellis puts it: “… the acquisition of grammar is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them.” (2003: p. 67). L2 learners face the task of internalizing enormous amounts of chunks and abstracting schematic constructions on the basis of these. And, of course, they have to learn to use this ‘grammar’.3

2.3.3 Second language use

The central concern in this subsection is how developing second language grammars as put forward in the previous subsection are put to use by L2 learners. In line with the notion that second language grammar has both rule-like and probabilistic properties, it will be argued that proficiency has computational properties, but is also to some extent exemplar-based. A number of researchers have indeed argued in favour of a dual system of language processing (e.g., Carr & Curran, 1994; Pawley & Syder, 1983; Sinclair, 1991; Skehan, 1998b). They see 3 Throughout this report, continued use will be made of terms such as rules and grammar, although they may not accurately convey these aspects of the linguistic system.

26 CHAPTER 2

linguistic proficiency as the result of cooperation between a rule-based linguistic system and an exemplar-based system. The argument is that language users do not always compute sentences. In fact, they have a large store of ready-made exemplars or chunks of language at their disposal, and language proficiency is to a greater or lesser extent the result of retrieving and piecing together these ready-made exemplars.

The most pervasive view, however, is that producing language is a computational process. A well-known observation is that of L2 learners attaching the past tense marker to verbs having irregular past tense forms (e.g., he goed instead of he went). Since the L2 learner cannot have observed goed in the input, he or she must have been applying and overgeneralizing a rule. Thus, producing language must be seen as filling out rules with lexical content, irrespective of whether these rules are fully developed or early inaccurate exponents of them. This in turn means that utterance construction involves some quite advanced planning: the speaker has to generate a context-appropriate message; the message must be grammatically encoded in such a way that it expresses the intended meaning; it has to be morphologically and phonologically encoded; and ultimately articulated (e.g., Levelt, 1999). An advantage of such a computational production system is that it allows speakers to be highly creative, because each utterance is generated anew. It also comes with a number of assumptions. Most importantly, given the demands of online communication, such a computational process has to be highly efficient and cheap in terms of cognitive resources.

Although L2 proficiency is likely to be scaffolded by such a computational production system, it probably is not the result of mere computation. In fact, it is commonly acknowledged that particular aspects of language use are exemplar-based or formulaic. This simply means that some phrases are stored as whole constructions rather than generated anew each time they are used. They are phrases so frequent in everyday language that they have become institutionalized. “Roman Catholic” and “red herring” mentioned in the previous section are examples, but short sentences such as: “Have some more!”, “Can I help?” may also be.

The question that rises is where institutionalization stops. There are in fact good arguments supporting the idea that language use leans quite heavily on direct retrieval of exemplars from memory. One important argument supporting this is the observation that people seem to limit themselves to fixed and frequently recurring combinations of lexical elements, rather than exploiting the grammar’s infinite possibilities to be creative. Pawley & Syder (1983) have called this phenomenon native-like selection, and they have proposed lexicalized sentence


stems to account for this phenomenon. Lexicalisation refers to institutionalization of stretches of language; expressions that native speakers would typically use in particular circumstances, social conventions almost. Pawley and Syder provide a list of examples of lexicalized sentence stems with the verb to think: ‘Come to think of it, …’, ‘Think nothing of it’, ‘Think it over’, etc.; but also ‘I think a lot of P’ and ‘P thinks nothing of V-ing NP’ (e.g. walking 50 miles)’ (p. 213). The first three are examples of memorized sentences, and the latter two are what they call lexicalized sentence stems, consisting of a lexicalized nucleus and open elements which afford some coding on behalf of the speaker. The crucial argument is that – rather than using our grammatical knowledge to compute utterances, we revert to “… strings which the speaker or hearer is capable of consciously assembling or analysing, but which on most occasions of use are recalled as wholes or as automatically chained strings.” (Pawley & Syder, 1983: p. 205).

Skehan (1998a) similarly argues for a substantial role of an exemplar-based system in producing language. Taking a processing perspective, he questions several of the assumptions that a primarily computational view on language production entails. For example, he takes issue with the assumption that rule-based language generation would be cheap in terms of cognitive processing. He argues that real-time communication is too demanding to each time generate utterances from scratch, and that language speakers in fact fail the processing capacities to do this. For this reason, they depend on memory-based language production a great deal of the time. In Skehan’s own words: “Producing speech seems to be much more a case of improvising on a clause-by-clause basis, using lexical elements … wherever possible, to minimize processing demands. Then, as ends-of-clauses are approached, improvisation skills allow us to tack one clause on to the next … ” (1998a: p. 37).

Interestingly, Skehan’s views on L2 processing explain how the rule-based and memory-based systems cooperate, and they link nicely to the views on grammatical development put forward in the previous subsection (2.3.2). He sketches three stages: lexicalisation, syntacticalization and relexicalization, which delineate the course of acquisition (1998a). Based on L1 research, research by VanPatten (1996), and Schmidt’s noticing hypothesis (1990), he argues that by default second language learners focus on meaningful aspects of language rather than formal. As a result, second language acquisition in its initial stages is mainly a process of internalizing lexical elements of the L2. However, at some point, these lexical elements are subjected to syntacticalization, a process Skehan does not elaborate upon much, but it seems to refer to an analysis mechanism leading

28 CHAPTER 2

to rule-based linguistic knowledge. The final stage, relexicalization, is more or less equal to a process of automatization. In order to meet the demands of real-time communication, syntacticalized knowledge is relexicalized again, resulting in chunks much like Pawley and Syder’s lexicalised sentence stems.

Thus, like Ellis (2003: see previous subsection), Skehan argues that grammar learning starts with the internalization of exemplars or formulae, that are later ‘syntacticalized’ into grammatical structure. In addition, Skehan seems to argue that there is no end to the institutionalization of chunks of language: chunk learning is an ever continuing process that does not stop once the grammar is fully developed (if one can even ever speak of a fully developed grammar). For efficiency reasons, speakers depend on constructions with lexicalized sentence stems as large as possible. The result is a dual processing system, based on rules and exemplars. “When time is pressing, and contextual support high, memory-based communication is appropriate. When there is more time, and precision is important, the rule-based system can be accessed.” (Skehan, 1998a: p. 90/91). Although language users may have a fully developed grammatical system, in everyday language use, they depend on relexicalized chunks of language that are the result of automatization processes. Speakers will tend to revert to their grammatical knowledge only when the situation calls for it, or when it allows them the time to do so.

2.3.4 Implicit learning processes

The previous subsections have focused on defining the nature and use of implicit grammatical knowledge. This subsection turns to implicit learning processes, the processes that lead to the establishment of associations and the subsequent abstraction of grammatical categories. In the interface debate, a distinction is made between explicit and implicit learning, each leading to qualitatively different types of linguistic knowledge. The goal of this subsection, then, is to shed some light on the learning processes that lead to implicit knowledge. However, as the implicit learning processes themselves are not the focus of this study, they will not be discussed at length, but merely in relation to the nature and use of implicit knowledge as outlined in the previous subsections.

In the last two decades, second language acquisition researchers have increasingly adopted information processing models to explain how second language learners process linguistic input (e.g., Robinson, 2003; Schmidt, 1990; 1994; ed., 1995; Skehan, 1998a; Tomlin & Villa, 1994). The goal of information processing theory is to explain human behaviour through cognition and it is


firmly based in evolution theories of human cognition (Goodson, 2003). Language is in this view not regarded ‘a special case’ which requires innate linguistic universals in order to be learnt. Like all learning, it is rooted in simple and general learning processes, rather than in processes that are unique to language. Information processing can roughly be divided into three general stages (Robinson, 2003). These stages are the perceptual encoding stage, where input is perceived through our senses and encoded, and where certain aspects of the input are mentally registered or selected for further processing. The selected input, often referred to as intake, is the input for the next stage: central processing, where working, short term memory and long-term memory interact in order to achieve comprehension of the message. The last stage is responding, where language users make decisions about how to respond to the input.

The general understanding is that attention is the starting point for all learning (Schmidt, 2001: 12). In the literature, the term attention is actually used to refer to three separate functions (Robinson, 2003), but here it refers to a selection mechanism that takes care of perceptual encoding. As it is impossible to attend to all the input humans perceive, there is a focalising mechanism that focuses on one component at a time, called attention. Attention is a serial mechanism, bringing sensory and/or memory input into a heightened state of awareness on a moment to moment basis. While perceived input decays quickly, attended input actually lasts for a while; and if input is attended to several times it may be written in long-term memory. Thus, attention mediates between input and memory, either encoding new information in long-term memory or enforcing existing encodes, and as such, it is crucial to learning (Goodson, 2003). Schmidt (Schmidt, 1990) was among the first to apply information processing theory to second language acquisition, and introduced noticing as the linguistic equivalent to attention. Noticing specifically refers to paying attention to linguistic form, and occurs, for instance, when you stop to consider the meaning of an unfamiliar word. For Schmidt, noticing operates upon exemplars rather than rules: “the objects of attention and noticing are elements of the surface structure of utterances in the input – instances of language, rather than any abstract rules or principles of which such instances may be exemplars” (2001: p. 5). Noticing or attention, then, is the mechanism that takes care of what Skehan (1998a) referred to as lexicalization: and what N. Ellis (2003) identified as internalizing exemplars. Obviously, the question that rises is how syntacticalization, or the abstraction of grammar, takes place.

Numerous proposals have been put forward to explain the syntacticalization phase, but while there is fairly wide-spread agreement on the construct of

30 CHAPTER 2

noticing, little is known about the syntacticalization process. Doughty (2001b) discusses potential processes for learning in some depth. Following Bialystok (1994b: see also 2.2.2), she argues that a process of analysis is taking place during which unstructured representations become more structured. Doughty supposes that underlying analysis are mapping and restructuring. Mapping refers to learners’ natural tendencies to connect forms with meanings or functions, while restructuring refers to sudden changes in a learner’s interlanguage that lead to more efficient use of particular aspects of the language (Doughty, 2001b). However, the precise workings of restructuring remain unclear. While Doughty stresses that little is known yet about these mechanisms, she does point out that “… mapping and restructuring appear to be both continually in operation and not subject to conscious reflection, although once the insight has occurred, the knowledge itself may become increasingly available for metalinguistic comment (Bialystok, 1994b)”.

In the previous subsection, a parallel was already drawn between relexicalization and automatization. Relexicalization refers to the establishment of large stretches of analyses language that are retrieved as a whole from memory. Automatization processes probably take care of this. DeKeyser (2001) points out that automatization of grammar rules has received little attention in SLA research, but based on the evidence that is available, he argues that automatization involves both the automatization of rules and increased speed of retrieving exemplars from memory. How the first comes about is unclear, but increased retrieval speed results from extensive practice. This implies that importance of producing the second language during L2 acquisition primarily lies in automatizing analyzed knowledge.

Implicit learning, then, can be described as a staged process of noticing, analysis, and automatization. It results from frequent exposure and a subconscious process of figuring language out, and it is the inevitable and uncontrollable result of information processing (Hulstijn, 2002). The question is how these views on implicit knowledge and learning relate to explicit knowledge and learning, as this is central in the interface debate.

2.3.5 Explicit knowledge and learning

While the previous subsections exclusively focused on implicit knowledge and implicit learning, the focus now briefly shifts to explicit knowledge and explicit learning. In the introduction (Chapter 1), explicit knowledge was defined as factual and conscious knowledge about the second language. In essence, this


characterization is correct. However, it is also incomplete. The intention of this subsection is to define both explicit knowledge and learning more precisely. In doing so, it draws gracefully on recently published work by R. Ellis (2004) and Hulstijn (2002).

Reviewing Ellis’s (2004) article on explicit knowledge, it becomes apparent that an important aspect of the definition of explicit knowledge is that it actually involves metalinguistic awareness: the ability to treat the language as an object of thought. This ability requires conscious awareness and knowledge about the language that can be put in factual declarative statements (irrespective of whether these statements are correct or incorrect). If you consider the definition that R. Ellis arrives at, this seems to be one of the main pillars of explicit knowledge:

“Explicit L2 knowledge is the declarative and often anomalous knowledge of the phonological, lexical, grammatical, pragmatic, and sociocritical features of an L2 together with the metalanguage for labeling this knowledge. It is held consciously and is learnable and verbalizable. It is typically accessed through controlled processing when L2 learners experience some kind of linguistic difficulty in the use of the L2. Learners vary in the breadth and depth of their L2 explicit knowledge” (2004: pp. 244/245)

Explicit knowledge, then, is factual knowledge, no different from any other encyclopaedic knowledge, such as knowing that clavicula is Latin for ‘collar bone’. This is also why it is verbalizable: people can express such knowledge about the language in words, although they may not have the proper verbal repertoire to phrase their knowledge accurately. And much like any other factual knowledge, it may be incorrect or incomplete. In short, explicit knowledge involves an explicit understanding about language that can be put into words in as far as the speaker has the vocabulary to do so.

In line with the idea that explicit knowledge always involves meta-linguistic awareness, explicit learning must be a deliberate, wilfully controlled process. It refers to the conscious establishment of declarative knowledge. Obviously, as it involves wilful control, this kind of learning is amenable to instructional efforts. Hulstijn defines explicit learning as follows:

“Explicit learning is a conscious, deliberative process of concept formation and concept linking. This may either take place when learners are being taught concepts and rules by an instructor or textbook, or when they operate in a self-initiated learning mode, trying to develop concepts and rules themselves.” (2002: p. 206)

32 CHAPTER 2

This definition reflects that explicit learning is essentially conscious, and it agrees well with Ellis’s definition of explicit knowledge.

Interestingly, Hulstijn acknowledges that explicit knowledge can come about in two different circumstances. First, it can result from reflection upon implicit grammatical knowledge, either self-initiated or through instruction, as Bialystok (1994b) and Reber (1989) have argued. Second, it can be established through instruction, and this is the reality in most instructed second language acquisition classrooms. More often than not, L2 students have to try to consciously understand the grammar of the L2 without the implicit grammatical knowledge base to draw upon. After all, the reason to teach a particular grammatical form is almost always because teachers suppose or perhaps perceive a gap in the student’s grammatical knowledge. The intention in such cases is that teaching the grammar explicitly ultimately leads to implicit grammatical knowledge. And this is the central concern of this study: does explicit instruction aiming for explicit knowledge positively affect noticing, analysis, or automatization, the processes underlying the development of L2 proficiency?

2.3.6 Reappraising the interface debate

The development of implicit second language grammatical knowledge has been described in this section as a staged process during which chunks of language are internalized, analysed, and automatized. Implicit knowledge should be seen as an ever developing system of associations that ultimately enter into rule-like behaviour, resulting in a grammatical system resembling construction grammar. For reasons of economy, language users do not tend to exploit the full generic power of their grammatical system, but primarily resort to piecing together memorized chunks that may be quite large and require minimal coding, that have been put in place through automatization processes. Explicit knowledge has been defined as conscious, factual, and to some extent verbalizable knowledge, that reflects – although not necessarily correctly – the implicit grammatical system. It can come about in two circumstances: either it is the result of reflection upon the developing grammatical system, or it is the result of instruction, most commonly the case in second language classrooms. The question is whether there is scope for explicit knowledge to play a role in the development of implicit knowledge, given the view outlined in this section.


The interface debate

The interface positions either predict explicit knowledge to convert into implicit knowledge (the strong and the weak interface position), or to facilitate the acquisition of implicit knowledge (the weak and the no interface position). The way implicit and explicit knowledge have been described in this section clearly suggests a dichotomy between the two types of knowledge. In the words of R. Ellis (2004):

“Adopting a connectionist account of implicit linguistic knowledge as an elaborate interconnected network …, it is not easy to see how knowledge as weighted content (i.e., as a set of neural pathways of greater or lesser strength) can be anything other than separate from knowledge of linguistic facts.” (2004: p. 234)

Similarly, both types of knowledge most likely result from two different learning processes. One that is essentially implicit, incidental and the inevitable result of information processing, while the other has been defined as a deliberate and conscious effort to learn rules and concepts. In terms of organisation of knowledge, then, the views on L2 proficiency outlined in this section do not seem to agree with the no interface position and the weak interface position in as far as it predicts conversion of knowledge.

Nevertheless, there are interesting parallels between some of the theoretical notions introduced by proponents of the strong interface position and the views outlined in this section. First of all, Bialystok’s analysis construct, “the process by which mental representations that were loosely organized around meanings (knowledge of the world) become rearranged into explicit representations that are organized around formal structures” (1994a: p. 159; see also 2.2.2), turns out to capture the syntacticalization process quite adequately, if one bears in mind that she uses the term ‘explicit’ in a different sense. But there is also an interesting parallel with DeKeyser’s (1998) tree-step process (knowledge moves from declarative to procedural, to automatized). Arguably, there is but one difference: for DeKeyser, linguistic knowledge starts out as rules, while the views outlined here posit that L2 learning starts with exemplars.

The role of explicit knowledge, then, is probably limited to facilitating implicit learning processes. The development of a hierarchical grammatical system is ultimately a matter of setting the weights properly, normally the result of a frequency-biased process of establishing rule-like associations. Any contribution of explicit instruction to the development of implicit knowledge lies in the provision of exemplars that at some point trigger implicit learning processes. This

34 CHAPTER 2

does not mean that explicit knowledge converts into implicit knowledge; it simply means that incidental implicit learning processes are “concomitant” to deliberate explicit learning processes (Hulstijn, 2002). The value of explicit instruction does not lie in the establishment of explicit knowledge; it simply provides exemplars in much the same way as implicit instruction does.

Developmental readiness, structure complexity, and individual differences

In the previous section, a number of potentially interacting factors were discussed. These were developmental readiness, the nature or complexity of the target structure and individual differences. The question is how such interactions should be understood given the theory of L2 grammar development outlined here.

Developmental readiness was discussed in 2.2.3: it refers to the idea that L2 language development is to some extent subject to fixed developmental orders, which puts constraints on the potential effectiveness of FFI. With regard to explicit knowledge, it is difficult to maintain that anything that is deliberately and consciously learned is constrained by natural orders of acquisition. Like any other type of deliberate learning, it is merely constrained by one’s mental capacities. It should be pointed out, though, that learning explicit grammatical knowledge is probably easier for those that actually already have implicitly acquired knowledge of the rule to be taught. However, as pointed out, the SLA classroom is typically an environment where rules are taught on the basis of perceived knowledge gaps. Implicit grammatical knowledge may well be subject to developmental constraints, though. L2 development has been described as moving from formulae to limited scope patterns to constructions. This development in itself reflects stages, and it does not necessarily apply to language as whole, but to individual rules of language. Thus, in global terms, one can predict that the development of a particular limited scope pattern depends on whether the L2 learner possesses a sufficient amount of related exemplars, and constructions can be developed by the grace of the presence of specific limited scope patterns. How this works out for individual structures depends on the nature of these structures, and perhaps also on the extent to which they are interdependent on other structures.

In 2.2.4, another issue that was put forward as relevant to the interface debate was the idea that the practical value of explicit knowledge may vary between grammar structures. However, from the perspective adopted here, it seems most plausible that there are no fundamental differences in learning different grammar


structures, neither for explicit nor for implicit grammar learning. For explicit grammar learning, learning is no different than any other type of learning: L2 learners will have most difficulty with complex grammar structures. With respect to implicit grammar learning, the constructs of scope and reliability as proposed by Hulstijn and De Graaff (1994) seem particularly important. If language learning is a function of exposure to the target structures, then the evidence that is available in the input must substantially affect implicit learning processes. Differences may also arise, though, because of differences between structures in hierarchical depth, or because some structures appear in many different forms. However, such differences are most likely to lead to differences in rate of learning. Thus, learning to use a morphological rule such as Dutch degrees of comparison may be relatively fast, because it is relatively independent of other grammar structures and relatively shallow in hierarchical terms in that it does not exceed word level.

A final issue that was brought to the fore in 2.2.5 is how individual differences (IDs) affect grammatical development. Aspects of individual differences discussed were attitude, aptitude, age, and L1 background. The perspective outlined here does not have radical consequences for the potential influence of individual differences. An important consequence for posing two separate systems is that the each type of learning may be affected differently by each type of ID. Thus, a L2 learner may have positive attitudes towards explicit learning tasks, but dislike the use of the L2 in more spontaneous situations, leading to faltering implicit learning. It is similarly possible that age and L1 background affect explicit and implicit learning differently. With regard to aptitude, it seems that the distinction made between explicit and implicit learning and the staged implicit learning process outlined in this section may in fact link quite easily to Robinson’s (2002) model of aptitude complexes for different types of learning. Robinson distinguishes aptitudes for explicit learning, noticing, and learning via oral and written input. The first two aptitude complexes link directly to processes described in this section. The latter two may in fact refer to L2 learners’ aptitudes to abstract grammar from exemplars. Obviously, all this is highly speculative, but deserves examination.

In sum, an analysis of the second language learning process suggests that the

role of explicit knowledge is limited to a facilitative role: it may facilitate implicit learning process in that it provides exemplars that trigger implicit learning processes. If this is how explicit instruction affects implicit learning, then there is little reason to assume that it would be superior to implicit types of instruction.

36 CHAPTER 2

In the next section, form-focused instruction studies will be inspected in search of support for this conclusion.

2.4 The interface debate and form-focused instruction research

2.4.1 Introduction

Obviously, one area of research in second language acquisition that can potentially contribute to the interface debate is form-focused instruction research. If explicit instruction can be demonstrated to promote the development of L2 proficiency more than implicit instruction, then that would provide a strong argument in favour of a weak or perhaps even a strong interface position. Despite the wealth of FFI studies that have been conducted in the last three decennia, there are not many that have addressed this issue (DeKeyser, 2003).

In the introduction, it was already pointed out that form-focused instruction research has had little recognition for the idea that explicit and implicit knowledge are separate knowledge systems, and that FFI research has suffered from considerable bias in measuring linguistic knowledge gains. The question that rises, then, is to what extent FFI research provides evidence for an interface between explicit and implicit knowledge. Keeping the need to differentiate between explicit and implicit knowledge in mind, has FFI research effectively demonstrated that explicit instruction promotes the development of implicit knowledge? And what is the importance of related constructs such as developmental readiness, the type of target structure and individual differences if one applies this differentiation to findings reported so far? These issues will be addressed in this section.

The measurement of L2 knowledge gains has been undertaken in many different ways. Before one can adequately appreciate the implications of particular findings, it is important to understand what would be measures of explicit and implicit knowledge. This is briefly discussed in subsection 2.4.2. Subsequently, in 2.4.3, a review of FFI research by R. Ellis is discussed in which the impact of FFI on implicit measures of progress is investigated. Next, in 2.4.4, criteria are laid down for what would constitute evidence for an interface between explicit and implicit knowledge, and studies that meet these criteria are scrutinized in search of such evidence. Then, in 2.4.5, studies are reviewed that


have investigated the effects of explicit and implicit instruction in relation to potentially interacting variables.

2.4.2 Measures of explicit and implicit knowledge

A complicating issue when reviewing research conducted so far is deciding what constitute valid measures of explicit and implicit knowledge. Some guidelines can be found in the literature, though. R. Ellis seems to consider any “…activity that calls for unplanned language use directed at fulfilling some communicative purpose… ” (2002: p. 225) a measure of implicit knowledge, whereas tests that allow for monitoring would measure explicit knowledge. Norris & Ortega make a similar distinction when they contrast ‘free constructed response’ measures with ‘meta-linguistic judgements’, ‘selected responses’, and ‘constrained constructed responses’, the latter measures drawing on “… the application of explicit declarative knowledge under controlled conditions…” (2000: p. 486).

Explicit knowledge tests should call on the learners’ knowledge about the rules of the second language. Response time and the measurement of knowledge of structures in isolated contexts seem to be the most important means to achieve this. A study by Han and Ellis (1998) points out the importance of response time. Using factor analysis on five measures of proficiency – oral proficiency, timed and untimed grammaticality judgements and metalingual comments, they identified two distinct factors: the timed measures (oral proficiency and timed grammaticality judgements) together loaded on one factor, while the untimed measures (untimed grammaticality judgements and metalingual comments) loaded on the other. The former set of measures may be seen to represent implicit knowledge, and the second set explicit knowledge. Therefore, Han and Ellis in fact argue that these results provide evidence for the separateness of explicit and implicit knowledge.

It may be difficult to construct a test of explicit knowledge that prohibits language learners to use their implicit knowledge (Ellis, 2004). For this reason, testing in isolated contexts is of crucial importance. Test takers do not necessarily have to be aware of the fact that particular features of language are in focus, but the problem should be such that they are inclined to search their memories for solutions. Although the possibility of using implicit knowledge cannot be excluded entirely, in the case of L2 acquisition, test takers often do not have implicit knowledge to help them perform well on the explicit knowledge test.

38 CHAPTER 2

Conversely, tests of implicit knowledge have to be unfocused and serving a communicative purpose (R. Ellis, 2002). They should elicit language use in which the use of the features of grammar under investigation is incidental. The study by Han and Ellis (1998) discussed above suggests that oral proficiency and timed grammaticality judgements accomplish this. However, given that timed grammaticality judgement task do test in isolated contexts and do not fulfil a communicative purpose, this type is not considered to be a measure of implicit knowledge in this study.

2.4.3 FFI and its impact on implicit knowledge: Ellis (2002)

Recognizing the need to distinguish between explicit and implicit knowledge, R. Ellis (2002) reviews studies that have assessed how FFI affects the development of implicit knowledge. Although it does not explicitly address differences between explicit and implicit types of FFI, this review is of obvious importance to this study, as it tries to assess the impact of FFI on implicit knowledge measures. The studies included in Ellis’s review are: Harley (1989); Day and Shapson (2001); Lyster (1994); VanPatten and Sanz (1995); Salaberry (1997); Mackey and Philp (1998); Long, Inagaki and Ortega (1998); Mackey (1999); Doughty and Varela (1998); Williams and Evans (1998); and Muranoi (2000); and Ellis analysed these studies on a number of aspects, including the age of the subjects, the kind of structure taught (morphological, syntactical or formulaic), and the type and extent of the instruction. The majority of these studies report success in improving implicit knowledge: seven out of eleven studies.

Ellis’s review suggests that the kind of grammar structure makes a difference. The four studies in which FFI did not lead to gains in implicit knowledge, all targeted syntactic structures. If the targeted structure was morphological (3 studies) or formulaic (1 study), the instruction was always successful. For syntactic structures, success was reported by three out of seven studies. In addition, Ellis points out that one of the determining factors for successful FFI may well be the availability of the targeted structure in everyday input. The structures that were successfully taught were structures that students may have encountered quite frequently outside the classroom environment.

Another factor that Ellis (2002) identifies as important to achieving success, is the extensiveness of the instruction. In seven studies, the instruction was extensive (i.e.: several hours and/or compound tasks), and six of these studies report positive effects. The one exception is Williams and Evans (1998), which


featured the complex passive construction. Ellis identifies a few more factors that may be of importance, although it is difficult to attribute success to these factors unequivocally. For example, in four out of eleven studies, the subjects were young (below 12), and in three of these four, positive effects were reported. However, the targeted grammar rules in these three studies were morphological and formulaic, and the unsuccessful young learners were instructed in a syntactic rule. Thus, age may be of influence, but this cannot be asserted with any certainty.

Ellis concludes that by and large, FFI seems to impact on implicit knowledge, and that “key factors” for successful FFI are “the complexity of target structure, the extent of the instruction, and the availability of the target structure in noninstructional input” (p. 234). However, it is important to recognize that the instruction as realized in the studies included in Ellis’s review all operationalized the instruction in highly meaningful or communicative ways. For example Doughty and Varela (1998), Long, Inagaki and Ortega (1998), Mackey and Philp (1998), and Mackey (1999) all used implicit recasting techniques. Such instruction probably directly affects implicit learning processes. Muranoi (2000) also used recasting techniques, but combined recasting with explicit rule provision. Likewise, the studies of Day and Shapson (2001), Harley (1989), Lyster (1994), and Williams and Evans (1998) also included some explicit instruction, but it was embedded in a compound of communicative tasks, such as linguistic games, role plays, reading and writing activities, etc. These studies, too, provide ample occasion for immediate implicit learning. The instruction in the studies of Salaberry (1997) and VanPatten and Sanz (1995) were the most explicit. Both targeted Spanish preverbal pronouns, but only VanPatten and Sanz found the instruction to impact on written proficiency.

As these studies made little use of explicit types of instruction, no conclusions can be drawn about positive effects of explicit instruction on L2 development. The studies that did use explicit types of instruction, mostly also provided ample occasion for implicit learning. Ellis does suspect, though, that explicit instruction may be more effective. Referring to his stance on the interface between implicit and explicit knowledge and the large amount of studies that have shown that FFI affects explicit knowledge, he speculates that a more effective route to L2 proficiency may be through developing explicit knowledge on behalf of the L2 learner.

40 CHAPTER 2

2.4.4 The interface between explicit and implicit knowledge

Although R. Ellis (2002) has established that FFI impacts on the development of L2 proficiency as measured by implicit knowledge tests, superiority of explicit types of instruction over implicit types of instruction – or vice versa, for that matter – was not demonstrated. This subsection zooms in further on studies that have the potential to address the interface debate. The best evidence for an interface between explicit and implicit knowledge would be provided by studies that monitor implicit grammatical development, and that compare groups with and without explicit knowledge of a particular target structure. This has a number of research design consequences. First of all, explicit and implicit treatments should be compared. Obviously, the explicit treatment is intended to establish explicit knowledge of a particular target structure. However, teaching rules explicitly may not just lead to explicit knowledge; it provides exemplars to the L2 learners that may trigger implicit learning processes as well. For this reason, it is important to contrast the explicit instruction group with an implicit instruction group rather than with a true uninstructed control group. This way, the amount of exposure to the target structure can be kept more or less equal, which neutralizes any effects of concomitant implicit learning. A second requirement is that explicit progress should be measured in addition to implicit progress. This makes it possible to assess whether the groups compared indeed contrast with respect to their explicit knowledge of the target structure.

Just one study was found meeting these requirements: a study by Sanz and Morgan-Short (2004). In addition, the studies of Bienfait (2002), Muranoi (2000), VanPatten and Sanz (1995), and Williams and Evans (1998) meet at least the most essential requirements of comparing explicit instruction with a control group, and using an implicit measure to assess L2 development. These studies will be discussed in this subsection. Table 2.3 summarizes them.

As pointed out, the design of Sanz and Morgan-Short’s (2004) study addresses the interface debate best. In their study, the instruction was delivered by the computer, and involved practice in Spanish preverbal pronouns. Their subjects were university students learning Spanish. In the instruction, the students were faced with sentences containing the target structure, and they had to respond to it appropriately, depending on the task. They compared four different treatment conditions: in one condition, this practice was accompanied by explicit rule explanation, another condition involved explicit feedback during the practice, one condition featured both rule explanation and feedback, and one featured none of these. They measured progress by means of a sentence completion task and a written video retelling task. In the sentence completion task subjects had to use a


particular verb to finish the sentence appropriately. As it did not involve time-pressure, and required subjects to use the rule in a controlled context, it may be accepted as a measure of explicit knowledge. The video retelling task was a free constructed response task, and thus constitutes a measure of implicit progress.

Sanz and Morgan-Short found all four conditions to improve significantly on both measures of progress, but there were no differences between the groups. As there were no differences found between the groups in progress on the implicit measure, this study does not provide evidence for an interface between explicit and implicit knowledge. However, there were also no differences in explicit gain. This could explain why explicit instruction was not found to impact more on L2 proficiency than implicit instruction: the groups compared did not differ in terms of explicit knowledge. This latter finding is important. Apparently, comparing explicitly and implicitly instructed groups is no guarantee for significant explicit knowledge differences. It confirms the importance of the requirement that both explicit and implicit progress be measured in order to asses the interface issue. The goal of Muranoi’s (2000) study was to investigate the effect of interaction enhancement. Using Japanese university students learning English, he compared three different types of instruction: interaction enhancement with form-focused debriefing (IEF), interaction enhancement with meaning-focused debriefing (IEM), and no interaction enhancement with meaning-focused debriefing (NEI). Interaction enhancement involved providing students with scenario’s that were intended to create contexts for using indefinite articles. During the role-play, the teacher would enhance the interaction by providing implicit negative feedback (i.e.: repetition requests and output modification). During the form-focused debriefing, the students received explicit instruction in how to use the indefinite article, while the meaning-focused debriefing discussed the success of the interaction. Progress was assessed by means of oral story retelling, oral and written picture description, and a grammatical judgement task. The results show that the IEF group scored significantly higher on all four measures than the IEM and NEI group. In other words, the group with most explicit knowledge also showed most implicit progress. However, it is important to point out that the IEF group received more exposure to the target structure, because of the differences in debriefing. Thus, the advantage found for the IEF group cannot safely be attributed to the fact that they had more explicit knowledge.

VanPatten and Sanz (1995) intended to assess the effectiveness of Processing Instruction, a specific type of explicit form-focused instruction that hopes to

42 CHAPTER 2

TAB

LE 2

.3:

A su

mm

ary

of F

FI st

udie

s add

ress

ing

the

inte

rfac

e be

twee

n ex

plic

it an

d im

plic

it k

now

ledg

e

Stud

y Su

bjec

ts

Targ

et

stru

ctur

es

Rese

arch

des

ign

Mea

sure

of L

2 de

velo

pmen

t a O

utco

me

Inte

rfac

e ev

iden

ce

Bien

fait

(2

002)

41

Dut

ch S

L su

bjec

ts (a

ge

13-1

7)

8 di

ffer

ent

stru

ctur

es

Exp.

Inst

r.,

cont

rol g

roup

; D

evel

omen

tally

re

ady

(DR

) gr

oup,

D

evel

opm

enta

lly

unre

ady

(DU

) gr

oup.

I.M.:

Ora

l pr

oduc

tion

in

form

al a

nd

info

rmal

task

s;

E.M

.: no

t ass

esse

d

Whe

n D

R,

stud

ents

sho

w

sign

ifica

nt

prog

ress

, irr

esp.

of

the

kind

of

inst

ruct

ion

or

the

kind

of t

arge

t st

ruct

ure

No

adva

ntag

e of

ex

plic

it o

ver

impl

icit

in

stru

ctio

n.

Mur

anoi

(2

000)

91

EFL

un

iver

sity

st

uden

ts

Def

init

e an

d in

defin

ite

arti

cles

Inte

ract

ion

enha

ncem

ent

com

bine

d w

ith

expl

icit

rule

-pr

ovis

ion

or

wit

h m

eani

ng-

focu

sed

inst

ruct

ion;

Co

ntro

l gro

up.

I.M.:

Ora

l sto

ry

desc

ript

ion;

ora

l an

d w

ritt

en p

ictu

re

disc

ript

ion;

E.

M.:

Gra

mm

atic

alit

y ju

dgem

ent t

ask

(GJ)

On

all t

asks

: exp

. Fo

nF >

imp.

Fo

nF >

con

trol

; Ex

p. F

onF

impa

cted

si

gnifi

cant

ly o

n al

l tas

ks, I

mp.

Fo

nF p

rim

arily

on

ora

l tas

ks.

Expl

icit

in

stru

ctio

n si

gnifi

cant

ly

outp

erfo

rmed

im

plic

it

inst

ruct

ion

Sanz

&

Mor

gan-

Shor

t (2

004)

69 S

pani

sh

SL u

nive

rsit

y st

uden

ts

Prev

erba

l pr

onou

ns

Four

con

diti

ons:

gr

oups

dif

fere

d ac

cord

ing

to

prov

isio

n of

ex

plan

atio

n (E

) an

d fe

edba

ck (F

):

[+E,

+F]

[-E,

-F]

[+E,

-F] [

-E, +

F]

I.M. w

ritt

en v

ideo

re

telli

ng ta

sk;

E.M

.: Se

nten

ce

com

plet

ion

c

Sign

ifica

nt

prog

ress

from

pr

e- to

pos

ttes

t fo

r all

four

co

ndit

ions

No

adva

ntag

e of

ru

le p

rovi

sion

an

d/or

exp

licit

fe

edba

ck g

roup

s ov

er im

plic

itly

in

stru

cted

gro

up.

FFI AND THE DEVELOPMENT OF L2 PROFICIENCY 43 V

anPa

tten

&

San

z (1

995)

44 S

pani

sh

SL u

nive

rsit

y st

uden

ts

Prev

erba

l pr

onou

ns

Expl

icit

in

stru

ctio

n(Pr

oce

ssin

g in

stru

ctio

n);

cont

rol g

roup

I.M.:

Vid

eo

rete

lling

task

, bot

h w

ritt

en a

nd o

ral

E.M

.: Se

nten

ce

com

plet

ion

c

Inst

ruct

ion

sign

ifica

ntly

af

fect

ed th

e sc

ores

on

both

I.M

. and

E.M

. Th

e on

ly

exce

ptio

n w

as

the

oral

I.M

. m

easu

re.

Expl

icit

in

stru

ctio

n si

gnifi

cant

ly

impr

oved

wri

tten

L2

pro

ficie

ncy,

bu

t not

ora

l pr

ofic

ienc

y

Will

iam

s an

d Ev

ans

(199

8)

33 E

SL

univ

ersi

ty

stud

ents

Part

icip

ial

adje

ctiv

es;

pass

ives

Expl

icit

in

stru

ctio

n,

Impl

icit

in

stru

ctio

n,

Cont

rol g

roup

I.M.:

Dic

togl

oss

E.M

.: Se

nten

ce

com

plet

ion

b

On

both

st

ruct

ures

, the

ex

peri

men

tal

grou

ps d

id b

ette

r th

an th

e co

ntro

ls

on I.

M. a

nd E

.M

mea

sure

s, e

xcep

t fo

r I.M

. pas

sive

s (n

o gr

oup

diff

eren

ces)

No

adva

ntag

e fo

und

for e

xplic

it

inst

ruct

ion

over

im

plic

it

inst

ruct

ion.

a –

I.M.:

impl

icit

mea

sure

; E.M

.: ex

plic

it m

easu

re

b –

Sent

ence

com

plet

ion

task

s m

ay n

ot b

e a

valid

mea

sure

of e

xplic

it k

now

ledg

e in

that

they

do

not n

eces

sari

ly in

vite

the

L2

lear

ner t

o us

e th

eir

expl

icit

kno

wle

dge.

44 CHAPTER 2

facilitate input processing. They used university students learning Spanish as a foreign language, and compared a processing instruction group with a no instruction group. In this study, the object of study was again the Spanish preverbal pronoun. Four measures of progress were used: an interpretation task, a sentence completion task, a structured interview and a storytelling task. In addition, the latter three were administered in both oral and written mode. They found the processing instruction group to outperform the no instruction group on all measures. Another interesting result was that the effect of their instruction tended to be stronger in the written mode than in the oral mode. However, again, one cannot exclude the possibility that differences in effects between both groups were due to differences in exposure. In fact, in this case, it seems very likely that they are to a considerable extent, as the processing instruction covered two full schooldays focusing on the target form, during which the no instruction group performed regular classroom activities.

Williams and Evans’s (1998) study expressly addressed the issue of whether the kind of grammar structure makes a difference in instruction. Their learners of English as a second language (university students) were divided into three conditions: one that received an input flood, which is an implicit type of instruction; another group received an input flood plus explicit rule explanation; and there was a control group. The instruction targeted participial adjectives of emotive verbs (e.g., interested/interesting) and passive constructions. However, in this case, only the passives are of interest, as they used an explicit and an implicit measure for this structure only (sentence completion and narratives, respectively) It should be noted that the narrative task was not a true free response task, as the subjects were asked to describe pictures and were supplied with a particular phrase that elicits the passive to start off the narrative. Their results show that the control group hardly obtained any progress. Both instructed groups did, but there were no significant differences between the two on the narrative task. The flood plus instruction group did obtain more progress on the sentence completion task, but this difference was not significant. Thus, this study also does not suggest an interface between explicit and implicit knowledge. Interesting, though, is that both experimental groups outperformed the control group on the narrative. Apparently, the type of instruction did not make a difference, as long as students received input. Also, it underscores the importance of keeping the amount of exposure to the target structure equal

Bienfait’s study (2002) was special in that she took great pains to operationalize developmental readiness. She compared progress of two groups: a group that received explicit FFI with a group that continued their normal


classroom activities, and she further differentiated her subjects (13 to 17 year old learners of Dutch as a second language) according to whether they were developmentally ready. Progress was measured by means of a formal and an informal production task at different points in time, both orally administered. The formal task required students to describe a comic strip, and students were expressly warned that their output needed to be as accurate as possible. The informal task consisted of informal conversation. Students were considered developmentally ready for one of the eight target structures monitored, if there were differences in correct use between the two tasks. Bienfait found that there were no differences in success between the two conditions for students that were developmentally ready. In both conditions, significant and equal progress was obtained on both the formal and informal tasks. Students that were developmentally unready did not progress on either task. In a delayed post-test, one month later, still no progress was observed, while the ready students continued to perform well. Because of these findings, Bienfait concluded that there is no added value to teaching grammar explicitly; the stage of development rather than the nature of the instruction is the critical factor for growth of grammatical proficiency (2002, p. 251). This study, then, also did not find explicit instruction to be superior to implicit instruction. It should be pointed out that the study did not include an explicit knowledge test, and the groups compared may simply not have differed in their explicit knowledge of the target structures.

Reviewing the research, then, one must conclude that the evidence for an interface between implicit and explicit knowledge is very slim. Mostly, FFI is found to be effective, but none of the studies allow for the interpretation that explicit instruction is superior in promoting implicit grammatical development. Either there were no differences in explicit knowledge between the groups compared, or there were substantial differences in exposure to the target structure. Three findings are important. The first important finding is Sanz and Morgan-Short’s finding of the lack of difference in explicit knowledge between their conditions, despite differences in the explicitness of the instruction. This means that students can develop explicit knowledge on their own, and underlines the importance of assessing explicit knowledge in addition to implicit knowledge. The second interesting finding is the difference in effects observed between oral and written tasks by VanPatten and Sanz. Apparently, effects of instruction are larger in the domain of writing, or appear first there. And the third finding of importance is the finding by Williams and Evans that both experimental groups outperformed the control group, which indicates that amount of exposure, rather than the nature of the exposure, is of crucial importance.

46 CHAPTER 2

2.4.5 Developmental readiness, structure complexity and ID’s

This subsection reviews studies that have compared explicit and implicit kinds of instruction in relation to the variables that have surfaced in the interface debate as potentially moderating the effectiveness of FFI: developmental readiness, the type of grammar structure taught, and individual differences (IDs). The studies selected for review were those that have assessed any of these aspects in relation to explicit and implicit kinds of instruction, irrespective of the kind of measurement used. The intention is examine how these factors interact with the success of different types of FFI.

FFI studies that expressly address the issue of developmental readiness are rare. The only FFI study in which developmental readiness prominently featured has already been discussed in the previous section: Bienfait (2002). Using two oral production measures that varied in formality – and using differences in performance between these two measures as an indication of readiness, she found that her young learners of Dutch as a second language obtained progress only if developmentally ready. Once they were ready, receiving explicit instruction did not make a difference as compared to students that continued their normal classroom activities. Thus, this study clearly suggests that explicit knowledge about the second language does not offer advantages to second language learners. Bienfait not only addressed the issue of developmental readiness, she also monitored the development of functionally simple structures as opposed to functionally complex structures. Comparing composite scores of all simple structures with composite scores of all complex structures, she found no differences in progress between the two types of structures as tested by the formal task. However, students that were developmentally ready and received explicit instruction obtained more progress in meaningless structures as measured by the informal task.

A number of other studies have investigated how the nature of grammar structures impact upon the effectiveness of different kinds of instruction. Studies by DeKeyser (1995), De Graaff (1997a), and Robinson (1996) have also addressed the issue by comparing the effects of different types of FFI on the acquisition of contrasting grammar structures. Individual differences have not featured prominently in FFI research, although they are commonly hypothesized to affect language learning. Factors such as attitude, age, and L1 background have attracted a considerable amount of attention, but researchers have never investigated these in relation to different kinds of instruction. Attempts are generally made to control for their potential influence, but they do not often figure as explaining variables in FFI research. Aptitude has featured in two


studies: in a study by Robinson (1995; and reprinted in 1997), an in De Graaff’s (1997a) study.

The nature of the grammar structure was investigated by DeKeyser (1995), focusing on structure reliability. He tested the effectiveness of an explicit-deductive approach vs. an implicit-inductive approach for simple categorical (reliable) and probabilistic (unreliable) rules of a miniature artificial language consisting of 98 words. Both structures were morphological rules. The instruction was based on 124 stimulus sentences, and advance rule explanation was provided to the students in the explicit condition. Progress was measured by means of a grammaticality judgement test and a production test (respond in one sentence to a picture). DeKeyser reported that the explicit approach works better for categorical rules in new contexts. In contexts that were also used during the instruction, there were no differences. The implicit approach was found to be more effective for the probabilistic rules. DeKeyser sees theoretical implications: the internalization of unreliable rules may depend on implicit memory-based learning; while reliable rules are analysed and internalized by means of explicit rule-based learning. But the differentiated effect of explicit instruction in old and new contexts also suggests that L2 learners tend to use their exemplar-based knowledge before reverting to their rule-based knowledge.

One of the goals of Robinson (1996) was to investigate Krashen’s (1981) claims that complex rules can only be learned unconsciously. He monitored the learning of a simple and a complex syntactic rule in relation to four different types of instruction. The instruction involved the presentation of 40 stimulus sentences to which the students had to respond to questions about these stimuli appropriately. In the implicit condition, students were asked whether particular words had appeared in the stimulus; the incidental condition required answering text comprehension questions; in the rule-search condition, students were asked whether they already had ideas about the rules sought for; and in the instructed condition, students had to respond to metalinguistic questions about the stimulus. Progress was measured by means of a grammaticality judgement task. Robinson did not find support for Krashen’s claim: there were no differences in performance on the complex rule between the instructed, incidental, and implicit conditions. Students in the rule-search condition did not perform very well. With respect to the simple rule, his findings are similar to those of DeKeyser: the instructed condition was found to outperform the other three conditions.

Contrary to Robinson’s findings, De Graaff (1997a) hypothesised that explicit instruction is especially valuable to the acquisition of complex rules. The argument is that simple rules can be noticed spontaneously, while explicit FFI

48 CHAPTER 2

may help noticing and analysing complex rules. Complexity was operationalized as: “the number of different formal or functional grammatical features that contribute to the specific form of a target structure and the specific function it performs”. De Graaff examined the acquisition of Experanto and Spanish by Dutch university students, contrasting two types of instruction that had a communicative focus. In one condition, students were provided with additional rule explanation and feedback. Using grammaticality judgements and sentence completion tasks, De Graaff found his hypothesis confirmed only for Spanish. Another hypothesis of De Graaff’s study pertained to the potential difference in effectiveness of FFI with respect to morphological and syntactic structures, arguing that morphological rules can be internalized as exemplars, while syntactic rules can only be acquired by means of rule-based learning. The latter type would benefit from explicit instruction. De Graaff found the opposite, though. Another hypothesis of De Graaff’s study is related to IDs and FFI. De Graaff hypothesized that aptitude – measured by means of a Dutch versions of the paired associates test and the grammatical sensitivity test from the MLAT (Carroll & Sapon, 1959) and an additional test assessing the ability to infer word meanings – would affect performance in both conditions equally. Using a composite aptitude score, he found this hypothesis confirmed.

Another study investigating the relation between FFI and aptitude was conducted by Robinson (1995; and reprinted in 1997). This study actually addresses Krashen’s no interface claims. Using the paired associates test and the grammatical sensitivity test from the MLAT (Carroll & Sapon, 1959), Robinson investigated rule-learning of a simple and a complex rule by ESL learners under four different circumstances: implicit, incidental, rule-search and instructed learning. Knowledge of the rules was measured by means of a grammaticality judgement task. Robinson found support for Krashen’s claim that aptitude predicts explicit types of learning. There were no correlations found between the two aptitude measures and the scores obtained by students in the incidental learning condition, while both components were related to both simple and complex rule scores for students in the instructed condition. Other more difficult to interpret findings were the correlations between grammatical sensitivity test scores with both simple and complex rule scores obtained by implicit learners. Robinson speculated that these learners may have been engaged in conscious analysis of the rules.


2.4.6 Conclusion

The goal of this section was to evaluate the extent to which FFI research has provided insights into the interface debate. For a number of reasons, the amount of studies that address this issue is small. First of foremost, FFI studies need to incorporate measures of implicit knowledge, which were defined in 2.4.2 as tests that assess learners’ ability to use the second language in spontaneous situations of use. Most FFI studies have measured progress by means of tests calling on explicit knowledge: tests that assess knowledge of the target structure in isolated contexts and expressly call on knowledge about the language. Another important requirement of FFI studies wanting to address the interface issue is that such studies should compare explicit and implicit types of instruction, preferably in relation to both implicit and explicit progress.

Evidence for an interface between explicit and implicit knowledge has been sought by discussing R. Ellis’s (2002) review of FFI studies that have used implicit measures of progress, and by surveying studies that meet the most important design requirements for addressing the interface issue. Ellis’s review has borne out that FFI mostly promotes implicit grammatical development. He also suspects explicit types of instruction to be more effective than implicit instruction. His review does not warrant that conclusion, though, as most studies evaluated the impact of implicit types of instruction, and when explicit types of instruction were used, these provided ample occasion for implicit learning. An interesting finding is that FFI seemed to be more effective for morphologic and formulaic structures than for syntactic structures. The studies by Bienfait (2002), Muranoi (2000), Sanz and Morgan-Short (2004), VanPatten and Sanz (1995), and Williams and Evans (1998) – studies that in design address the interface issue – also do not provide much evidence for an interface between explicit and implicit knowledge. The studies by Muranoi (2000) and VanPatten and Sanz (1995) did find explicit instruction to be more effective than its control, but this advantage may well be caused by differences in exposure to the target structure. In fact, keeping the amount of exposure equal has proven to be an important design requirement for addressing the interface debate.

The relation between the effectiveness of FFI and developmental readiness, the type of target structure and IDs - issues that all feature in the interface debate, does not provide an overall clear-cut picture. Bienfait’s study clearly suggests that developmental readiness moderates the potential power of FFI. However, even developmentally ready students did not benefit from explicit instruction in her study, except for complex structures. The findings reported by DeKeyser (1995), Robinson (1996), and De Graaff (1997a) with regard to the type

50 CHAPTER 2

of grammar structure taught and using explicit measures progress, seem contradictory. DeKeyser and Robinson report advantages for explicit instruction with reliable or simple structures, while De Graaff finds explicit instruction to promote one of the complex structures in his study. Finally, De Graaff (1997a) and Robinson (1995; 1997) suggest that aptitude as measured by the paired associates test and the grammatical sensitivity test affects explicit learning. Interestingly, as both report correlations between aptitude and progress scores obtained by students in explicit and implicit learning conditions, this conclusion is valid irrespective of the kind of instruction received. However, only explicit measures were used, which means that it remains unclear how these measures of aptitude affect the development of implicit knowledge.

2.5 Summary

The practical value of explicit knowledge to the development of second language proficiency has been the central concern in this chapter. Theoretically, this concern is addressed by the interface debate. Three different positions have been identified, each proposing a different role for explicit knowledge in the course of second language proficiency development. The strong interface position stresses automatization processes, and supposes that explicit knowledge can become implicit through practice and automatization. Consequently, this position attributes a relatively large role to explicit grammatical knowledge, and explicit types of instruction should prove to be more efficient in promoting implicit grammatical knowledge than implicit types. The weak interface position similarly argues that there can be an interface between explicit and implicit knowledge, but posits constraints: those grammatical structures that develop according to a natural order of acquisition can be taught effectively only if the instruction matches the L2 learner’s stage of acquisition. Finally, the no interface position posits that explicit and implicit knowledge are two separate knowledge systems, resulting from two independent mechanisms of learning. In this view, the contribution of explicit grammatical knowledge to the development of implicit knowledge is severely limited: only when teaching explicit knowledge works to promote implicit learning processes may explicit instruction positively affect L2 proficiency development.

The interface issue has been critically evaluated in two ways. First, the nature and use of implicit grammatical knowledge has been discussed, and related to how it is learned. Implicit knowledge has been defined in terms of an associative


network of chunks or exemplars which results from exposure to frequent (co-)occurrence of structures. A grammar emanates when language learners start to replace lexically specific chunks with open class elements that allow slotting in other lexical or schematic elements. Three stages can be identified in this process of developing and learning to use a grammar. The stage of lexicalization refers to the internalization of exemplars, and the mechanism that takes care of lexicalization is noticing. The second stage, syntacticalization, refers to the process of abstracting grammar from exemplars, and occurs entirely subconsciously. It has been argued, though, that the resulting grammar may not be used to compute sentences from scratch in real-time communication, because processing demands may be too high. Rather, larger stretches of language are relexicalized and strung together so that communication requires little coding effort. The development of this network of chunks is slow and gradual, and requires a sufficient amount of exposure and ‘habit-formation’. Strong, rule-like associations are not established overnight, and the ability to use the L2 in an – at least seemingly – algorithmic way is similarly not established overnight

In order to assess the possibility of an interface between explicit and implicit knowledge, the nature, use and learning of implicit grammatical knowledge has been contrasted with explicit grammatical knowledge. The latter was defined as declarative knowledge; and it is seen as the result of a deliberate attempt to understand the rules of the L2. It is best characterized as an explicit, verbalizable, and not necessarily correct, understanding about the second language. If implicit knowledge is indeed best conceived of as an associative network of chunks – as has been argued in this chapter, it is difficult to see how explicit knowledge could convert into implicit knowledge. The conclusion is that there is little theoretical scope for an interface between explicit and implicit knowledge. The value of explicit instruction probably lies in the provision of exemplars triggering implicit learning processes. There is nothing to suggest, however, that exemplars cannot be provided implicitly just as effectively.

The second way this chapter has addressed the interface issue is by investigating the extent to which FFI research has provided valuable insights. A close inspection of the research also does not warrant a theory of L2 proficiency development that allows for the conversion of explicit knowledge into implicit knowledge. Surely, there is plenty of evidence that FFI, especially of the explicit kind, leads to explicit knowledge. Similarly, there is also evidence that FFI leads to implicit knowledge. Not many studies have been conducted that clearly address the interface issue. However, those studies that do contrast explicit and implicit types of instruction and measure implicit progress provide little to suggest that

52 CHAPTER 2

explicit types of FFI have a larger impact upon the development of implicit knowledge than implicit types of instruction, which would be predicted by the weak and strong interface positions. If anything, they provide evidence for the no interface position, in that exposure to the target structure is found to lead to acquisition irrespective of the kind of exposure, and more exposure seems to lead to more acquisition.

An issue that consistently surfaces in the interface debate is the notion of developmental readiness. There is a possibility that explicit knowledge can only convert into implicit knowledge if the instruction is properly timed. The necessity to time would be caused by language developing according to a natural order of acquisition: instruction should match the stage of acquisition in a language learner is at. The views of language acquisition outlined in this chapter do not exclude this possibility. Language acquisition has actually been described as staged acquisition process. Thus, the development of a particular limited scope pattern most likely depends on whether the L2 learner possesses a sufficient amount of related exemplars, and schematic constructions can be developed only by virtue of the presence of particular limited scope patterns. How this works out for individual structures depends on the nature of these structures, and the extent to which they are interdependent on other structures. One study by Bienfait indeed demonstrated the importance of properly timing the instruction. However, even when properly timed, she did not find explicit instruction to be superior to implicit instruction.

An aspect of language acquisition that may similarly obscure the presence of an interface between explicit and implicit knowledge, is that successful FFI may depend on which structure was taught. Differences between the teachability of grammar structures have been hypothesized for a number of reasons, but none of them convincingly. It has been argued in this chapter that acquisition of different structures of grammar should in principle follow the same trajectory from exemplars to constructions. Differences may nevertheless exist because of differences in hierarchical depth and differences in interdependence. There is indeed ample research showing differentiated effects according to the type of grammar structure taught, but a comparison of such findings leads to a fuzzy picture, and pinpointing which characteristics cause differentiation is difficult. Another important issue is that such research has tended to use explicit measures of grammatical development. Consequently, very little can be said about the interaction between the nature of the instruction, the nature of the grammar structure, and L2 proficiency development.


A final issue is that individuals may differ in their ability to successfully exploit explicit knowledge to the benefit of implicit learning. In short, an interface may exist only for some learners. Aptitude is the most important determining candidate. Recent developments in aptitude research suggest that different cognitive abilities underlie different aspects of L2 learning. Robinson (2001) has argued for a distinction between a number of aptitude complexes, one of which being aptitude for explicit rule learning. Other complexes he defines are aptitudes for incidental learning via oral and written content, and aptitude for focus on form. The implication is clear: the cognitive abilities underlying explicit L2 development are different from those underlying implicit L2 development. As such, Robinson’s theory provides another argument for regarding explicit and implicit learning as two separate systems. Studies that have investigated the interaction between aptitude and FFI have only used explicit measures of progress, and found that aptitude predicted progress regardless of the kind of instruction received.

All in all, this chapter has demonstrated that the role that explicit knowledge

should play in second language learning programmes is far from clear. Nevertheless, most agree that explicit instruction should be part of such programmes (Bienfait, 2002: p. 3). One can wonder, though, on what grounds this claim is made. For one thing, our knowledge of how second language proficiency develops is insufficient to make such a claim. In fact, recent developments seem to suggest that learning to use the second language grammar in a rule-like fashion is not a matter of learning to use algorithms correctly. A question of considerable importance is therefore how L2 learners start to make use of a particular grammar structure, and how they develop the ability to use them in a seemingly rule-like way. Another reason to be careful with claims about the role of explicit knowledge is that FFI research shows a clear bias towards measuring progress by means of explicit knowledge tests. As a result, our knowledge of the role of instruction on L2 proficiency development is limited, and research is needed to examine how different types of FFI affect explicit and implicit measures of grammatical progress, and whether instructing explicit grammatical knowledge facilitates implicit grammatical development. Finally, there are other factors that may seriously undermine any positive contribution of explicit knowledge. Developmental readiness, structure complexity, L1 background, and differences between individuals in the ability to make use of explicit knowledge might be constraining factors; the question that rises is how

54 CHAPTER 2

these factors interact with different kinds of instruction and the development of L2 proficiency. These are issues that this study intends to address.

3 Methodology

3.1 Research questions and research design

Developing the ability to use L2 grammar structures proficiently, the role of explicit and implicit instruction therein, and the potential interaction between instruction and a number of potentially intervening variables are issues that this study addresses. In design, this study is both explanatory and descriptive, making use of a quasi-experimental repeated measures research design to investigate the development of two grammatical structures by means of explicit and implicit measures of L2 knowledge. In this section, the research questions and the main features of the research design are presented.

One of the purposes of this study was to provide some insight into how second language learners develop the ability to start using new grammar structures. In the previous chapter, the emergence of grammar was described as a process of abstracting grammar from exemplars, and seemingly algorithmic use of grammar structures is expected develop gradually and slowly. Fine-grained, longitudinal analyses of the emergence of grammatical phenomena during second language acquisition are scarce (DeKeyser, 2001; N. Ellis, 2003), and little is yet known about how L2 learners develop the ability to use grammar structures in spontaneous situations. As this study involved three repeated measures of spontaneous written response data over a period of four months, a descriptive exploration of these data was undertaken. These analyses were intended to gain qualitatitive knowledge of how becoming proficient in the two grammatical phenomena in focus arises. In addition, the outcomes were used to inform subsequent analyses. Thus, the first question to be addressed is:

RQ 1: How do second language learners develop the ability to use the target structures in spontaneous situations of second language use?

The primary goal of this study was to investigate the importance of explicit knowledge to the development of second language proficiency development. As already pointed out in 2.4, FFI research can provide evidence for the existence of

56 CHAPTER 3

an interface between explicit and implicit knowledge, if explicit types of instruction can be found to be more beneficial to the development of L2 proficiency than implicit types of instruction. It was also pointed out that FFI research has not properly addressed the interface debate yet. It has tended to use explicit measures of progress, and studies that did measure implicit progress often did not compare explicit and implicit types of instruction. Comparing explicit and implicit grammar instruction, this study intended to establish whether explicit and implicit grammatical development were differently affected by these two types of instruction, and whether explicit instruction is indeed superior in promoting implicit knowledge development, as many researchers have argued (Bienfait, 2002; Doughty, 2003):

RQ 2: How are explicit and implicit FFI related to the development of explicit and implicit grammatical knowledge?

Another goal of this study was to investigate the interaction of explicit and implicit types of instruction with a number of individual characteristics. The notion that differences between individuals may explain the some extent why a particular kind of instruction is effective is actually common currency in SLA thinking. However, their precise relation to specific types of instruction has not been studied very frequently. This study has explored the relation between a number of defining individual characteristics and effective FFI. Developmental readiness, L1 similarity and Individual Differences such as aptitude, attitude, and age were selected for study:

RQ 3: How do developmental readiness, L1 similarity, and Individual Differences affect the success of explicit and implicit FFI?

Finally, another pervasive notion is that the value of explicit knowledge and FFI varies between different grammar structures. Although FFI research has addressed this question on a number of occasions, no coherent picture has emerged yet. Results seem to be contradictory, and most studies addressing this issue have used explicit measures of grammatical knowledge. Thus, the relationship between effective FFI and the instructed grammar structure is still rather obscure, and an investigation structures complexity while making a distinction between explicit and implicit grammatical development may provide valuable new insights. This issue was investigated by comparing the effect of FFI on two grammar structures varying in complexity, and the following research question was addressed.

METHODOLOGY 57

RQ 4: Does the effectiveness of explicit and implicit FFI depend on the complexity of the instructed grammar structure?

In this study, three different treatment groups have been compared: two experimental groups and one control group. Participants of the experimental conditions took part in a computer-assisted language learning experiment in which they received information – either explicitly or implicitly – about two grammar structures. In 2.4.4, it was asserted that – in order to make for a fair comparison – exposure to the target structures should be kept equal for students in different treatment conditions. For this reason, the amount of input and the linguistic context in which the input was offered was precisely matched, the only difference being the degree of explicitness in presentation of the target structures. However, a true control group was included as well, in order to investigate if and how L2 learners develop their grammatical knowledge without additional focused input. The subject sample of this study consisted of approximately fourteen-year old learners of Dutch as a second language who were enrolled in intensive, often full-time, language learning programmes.

The development of L2 proficiency has been monitored by means of two measures of grammatical knowledge: grammaticality judgements and free written production, tapping into explicit and implicit grammatical knowledge respectively. The review of FFI research has sufficiently demonstrated that too little is known about the effect of FFI on L2 implicit knowledge. However, the use of both an explicit and an implicit measure of grammatical development allows for a deeper investigation of the interface debate. The two tests have been administered at three points in time during the experiment: once before the treatment, once immediately after, and once with a delay of 2 months after the treatment. This way, the participants of this study were followed over a period of three months’ time.

Two contrasting grammar structures have been instructed, and the explicit and implicit tests of grammatical knowledge assessed the students’ knowledge of these two structures. One structure was functionally simple and morphologic, while the other was functionally complex and syntactic. All participants in the experimental conditions received explicit instruction in only one of the two target structures, and they would receive implicit instruction in the other structure. Given this crossed design, this study is best thought of as two separate experiments in one, each focusing on a different grammar structure.

Besides monitoring students’ grammatical knowledge of two grammar structures, information was gathered to enable a rather precise characterization of the participants of this study. They took additional aptitude and proficiency

58 CHAPTER 3

tests; their performance during the instruction was monitored by the computer; teachers were asked to make judgements about their motivation, (verbal) intelligence, and cognitive style; and the participants were asked to fill out background questionnaires about their age, first and second languages, and educational experience. All this information was used both to control for differences between the treatment conditions, and to investigate their relevance to developing grammatical knowledge of the target structures in focus.

The setting in which this research was conducted was the Dutch L2 classroom. Because of the focus on spontaneous second language use, laboratory-like research settings are precluded. In fact, the notion that explicit instruction may be more effective than implicit types of instruction is to a considerable extent based on studies that have taught and tested grammar rules in isolation; and in a number of cases impoverished artificial languages have been used. Such experiments lose out on ecological validity: if too much of the context in which language learning normally takes place is ruled out, variables may be lost that are of crucial importance to language learning, reducing the learners’ task to what De Graaff (1997b) calls a “cognitive puzzle”. His advice: “Only a careful combination of the advantages of both realistic and optimally controlled L2 learning environments can provide real opportunities for studying the effect of instruction on L2 acquisition” (p. 273).

An important design feature of this study was the use of the computer as instructor. This offered a number of advantages. First, the instruction was precisely the same for all subjects in one particular treatment condition. This way, control was exercised over the intervening effects that are introduced if different teachers are involved in explaining the rules of grammar. Second, the amount of exposure to the target structures could be carefully controlled and kept equal between the conditions. A third advantage was that the computer tracked the participants’ performance while they were working with the programme. And the fourth and very important advantage of using the computer was that it offered a convenient way to randomly assign the participants to one of the two treatment conditions. As a result, no set classes were used to populate the individual treatment conditions which would likely have introduced bias. Rather, the two treatment conditions were represented by students from all the classes participating in this study. Only the ‘true’ control group was created on the basis of two set classes, because they did not work with the computer programme.

In the remainder of this chapter, the research design shortly outlined here will

be described in detail, starting with a description of the target structures and how

METHODOLOGY 59

they were instructed (3.2), and followed by an explanation of how grammatical development was assessed (3.3). Next, the participants of this study and their background are introduced (3.4), and they are characterized in terms of potentially decisive individual characteristics (3.5). The chapter is concluded with expounding the organisational and statistical procedures (3.6).

3.2 The instruction

3.2.1 The target structures

The structures selected for instruction and comparison are the degrees of comparison and subordinate clauses. Both theoretical and practical issues have been considered when choosing the target structures, although the practical issues have prevailed.

In 2.3.6, it was argued that there are no fundamental differences in learning different grammar structures: they start as exemplars and gradually become open class structures. However, particular features of the target structure may affect a structure’s learnability and perhaps teachability. The literature mentions a number of such features, such as functional and formal complexity, reliability, scope, and the linguistic nature of the target structure (morphologic, syntactic, and formulaic). In addition, hierarchical depth was put forward as one of these factors. Choosing the proper target structures is not simple though. A practical point is that it is very difficult if not impossible to control for all the features of grammar structures simultaneously. For example, if one decides to compare two grammar structures contrasting with respect to scope, they are likely to differ as well according to one of the other features. This makes it difficult to attribute found differences to one particular feature. A complicating issue is that characterizing a grammar structure according to these features does not constitute categorical choices. Most of these factors are best seen as gliding scales rather than dichotomous choices. Another important practical consideration was the requirement that the participants had to be able to produce a fair amount of Dutch, given the focus on proficiency in this study. Beginning L2 learners simply cannot be expected to be very productive in their L2 yet. Obviously, the target structures had to be more or less new to these somewhat advanced L2 learners. Finally, a design requirement specifically related to this study was that the two grammar structures occur more or less simultaneously in the students’ interlanguage.

60 CHAPTER 3

Several steps have been undertaken to make a well-informed choice of target structures. A first selection of structures was made on the basis of Zebra (1999), a widely used method for teaching Dutch as a second language (see also 3.4.2). The teacher manual belonging to this method provides detailed information about the underlying structural syllabus. A small scale but intensive pilot study was then conducted in order to assess whether the selected structures were indeed emerging in the interlanguage of students that had had approximately one year of Dutch training. This pilot involved inviting students to write short pieces of text, and was also intended to evaluate how well the target structures could be elicited implicitly (see also 3.3.3 on the measurement of implicit grammatical knowledge).

The degrees of comparison and subordinate clauses were found to meet the set requirements best. L2 learners start using these (correctly) rather late; the structures were found to occur more or less simultaneously; and they can be elicited with considerable success. The structures also differ fundamentally from each other: the degrees of comparison are functionally simple and morphological, while subordinate clauses are functionally complex and syntactic. In the remainder of this subsection, both structures will be described in more detail.

Degrees of comparison

The degrees of comparison (DoC) in Dutch are very similar to the English degrees of comparison. The comparative and superlative are formed by attaching –er and –st to the adjective or adverb. Both suffixes carry independent meaning, and thus change the meaning of the adjective or adverb they are attached to. As each suffix carries a straightforward meaning, the structure must be considered functionally simple. However, Dutch also allows for periphrastic DoC by means of the use of the quantifiers meer (more) and meest (most). Periphrastic use is restricted to complex adjectives, though (Aarts & Wekker, 1993); and in the vast majority of cases, periphrastic realization of the DoC will be considered ungrammatical in Dutch. Periphrastic comparison was not subject of the instruction, simply because the kinds of adjectives that are realized periphrastically in Dutch are generally not part of the vocabulary of the L2 learners participating in this study.

As explained in 2.2.4, functional complexity refers to the relationship between meaning and form. If this relationship is straightforward and transparent, then the structure can be considered functionally simple. At first sight, the suffixes that mark the degrees of comparison seem functionally simple in that they have a straightforward meaning. However, Dutch uses the –er suffix also to make nouns

METHODOLOGY 61

TABLE 3.1 The degrees of comparison in Dutch Funny: Nice: Sweet, dear Sick

grappig leuk lief ziek

grappiger leuker liever zieker

grappigst leukst liefst ziekst

of verbs (werken, werker), which may be a complicating factor for L2 learners. Formal complexity refers to the number of operations that have to be performed in order to use the rule correctly. Realization of a form the degrees of comparison requires suffixation only, and can therefore be considered formally simple. Such suffixation does cause changes in spelling of the adjective or adverb sometimes (for example, -f becomes –v when suffixed: lief, liever, liefst). These cannot be heard, and they are in accordance with normal Dutch spelling conventions.

The notions of scope and reliability refer to the extent to which a particular rule is probabilistic. Hulstijn and De Graaff (1994) define scope as the absolute number of cases covered by a rule and reliability as the percentage of instances for which the rule holds (number of exceptions). They also draw borders, which they admit are arbitrary: a rule is large in scope when it covers more than 50 cases, and its reliability is high when it applies to over 90% of all cases. According to these definitions, the degrees of comparison can be considered large in scope, as there are over 50 adjectives that can be inflected. The rule is quite reliable as well, as periphrastic comparison is not very frequent in Dutch, and occurs with adjectives that probably lie outside the L2 learners’ vocabulary.

Subordinate clauses

Subordinate clauses refer to a specific class of embedded sentences in which one clause is subordinate to the other – main – clause. Subordinate clauses are introduced by subordinating conjunctions, which express how the main clause and the subordinate clause relate to each other. For example, subordinating conjunctions can express among others temporal, causal, and conditional relations. Particularly difficult for L2 learners is that the finite verb – by default in second position in main clauses – is positioned at the end of subordinate clauses (see Table 3.2). This repositioning of the finite verb is entirely meaningless, in that it does not change the meaning of the sentence. Neglecting to position the verb at the end of the subclause does not lead to incomprehension or miscomprehension.

62 CHAPTER 3

TABLE 3.2 Subordinate clauses in Dutch (examples taken from the data collected for this study)

1)

Main clause: Subordinate clause:

Mijn vriend vertelt een grap. …, als mijn vriend een grap vertelt. ..., when my friend a joke tells.

2) Main clause: Subordinate clause:

De olifant plukt appels …, omdat de olifant apples plukt. …, because the elephant apples picks.


Hij heeft tegen mij gelogen. …, als hij tegen mij heeft gelogen. ..., when he to me has lied


Zij wil een ijsje eten. …, omdat zij een ijsje wil eten. ..., because she icecream wants to eat.

5) Subordinate clause: Coordinate clause

..., omdat de kleine radio goedkoper is.

..., want de kleine radio is goedkoper. …, because the little radio (is) cheaper (is).

Subordination is somewhat difficult to classify in terms of functional

complexity. As pointed out, the repositioning itself does not carry meaning. However, the meaning of the clause changes, because of the conditionality or causality it expresses by means of the coordinating conjunction. L2 learners therefore might associate the rule with the relationship the clause intends to express. A complicating factor specifically related to expressing causality is that Dutch also has a coordinating conjunction expressing causality: want. In Table 3.2, example 5 illustrates two sentences that are synonymous in meaning, but not in word order. The only valid cue for repositioning the finite verb, then, is the subordinate conjunction itself. In terms of formal complexity subordination must be considered simple, because it requires only one operation.

In terms of scope and reliability, subordination can be considered a reliable rule that is wide in scope. The occurrence of one of the subordinating conjunctions almost always leads to verb repositioning. The only exception for applying the rule may be when the subordinate clause is too complex to move the verb all the way to the end. However, this occurs seldom. The applicability of the rule is virtually endless, and especially conditional and causal subordinate clauses occur rather frequently in everyday language.

METHODOLOGY 63

3.2.2 The instruction

The value of three types of instruction has been evaluated: explicit instruction (EI), implicit instruction (II), and no instruction (NI). In both experimental conditions (EI and II), the participants of this study received focused input of the DoC and SubC by means of a computer-delivered language learning programme. Participants in the NI group simply did not receive any focused instruction; and therefore, no extra input. They continued to do their normal classroom activities while the experimental groups were working with the computer programme. The NI condition was included to investigate how L2 learners develop their grammatical knowledge without input specifically focused on the target structures. This way, the effect of receiving focused instruction – either explicit or implicit – could be evaluated with reference to not receiving any focused instruction.

A crossed treatment design was used (see Table 3.3). This meant that participants in treatment group 1 would receive explicit instruction on the DoC and implicit instruction on SubCs; for participants in treatment group 2, this was the other way around. Consequently, all participants in the two experimental conditions received explicit and implicit instruction, but which target structure they dealt with explicitly depended on the treatment group they were in. This way, the treatment groups were each others controls with respect to exposure to the target structures. Also, all participants did a variety of explicit and implicit kinds of exercises, which made the programme more enjoyable. The instruction was relatively short, in that on average three to four hours were spent on the programme by the participants.

The instruction was delivered by means of a computer programme called Taal in Themas, which was developed for the occasion. As already pointed out in 3.1, the use of the computer offered some important advantages. Using the computer as a teacher ensured that the instruction was precisely the same for all subjects. But also the amount of exposure to the target structures could be carefully matched, which is considered to be an important design requirement. Input matching will be discussed in the next subsection. Detailed information about the

TABLE 3.3 Organisation of the experimental treatment groups Condition Group 1 Group 2 Explicit instruction Degrees of Comparison Subordinate clauses Implicit instruction Subordinate clauses Degrees of Comparison

64 CHAPTER 3

development of the programme and in-class procedures is given in 3.6. In addition, some screenshots of the programme can be found in Appendix A1.

Most would agree that any of FFI should be integrated in functional and communicative contexts. Using the computer for the instruction is to some extent at odds with this requirement, simply because computers do not communicate. The kinds of activities that can be offered by means of the computer are limited, and such activities do mostly not serve a functional purpose comparable to those of everyday communication. This is not problematic, though, as long as the explicit instruction leads to explicit knowledge, and as long as the input offered during the explicit instruction is matched with implicit input. The instruction was nevertheless set in a functional context. A series of eight lessons was developed centred around three different themes: Advertising, the Olympic Games, and The Netherlands and Water. Each lesson consisted of short texts about these themes, and students would do a number of unfocused exercises related to the texts. These texts and exercises would be interlaced with focused exercises, either implicitly or explicitly offering FFI in the DoC or SubC. As there were eight lessons in total, four focused on the Degrees of Comparison, and four focused on Subordinate clauses. Whether this focus was explicit or implicit depended on the treatment group the students were in (see Table 3.3). In addition, a detailed outline of the instruction is provided in Appendix A2.

In line with the definition of explicit knowledge given in 2.4.2 and the purpose of this study, the aim of the explicit instruction was to create an explicit understanding about the target structures. This aim was explicitly stated from the outset, this way establishing the intention on behalf of the learner to deliberately learn a particular rule. Focused exercises would be clearly marked by a special logo, caption, and background colour. In addition, the exercises would be introduced by rule explanation, either introducing new aspects of the rule, or repeating earlier discussed aspects. In the subsequent exercises, the participants were continuously reminded to apply the rules just explained.

Following recommendations by Kuiken (1999), and the example of Zebra grammatica (Andringa et al., 2000), rule explanation focused primarily on functional use of the structure. The instruction would be accompanied by examples, and if possible, visual aids. Care was taken not to make the explanation too technical. Given that the participants’ command of specific linguistic terms was limited, such terms were used sparingly; and when used, they would be carefully defined. With respect to subordinate clauses, it should be pointed out that the instruction focused primarily on conditional and causal types of subordinate clauses. These are by far the most frequent in Dutch, and the use of

METHODOLOGY 65

other subordinate clauses proved difficult to elicit (see 3.3.3). Nevertheless, other types of subordinate clauses were not ignored: they were simply less frequent. The Appendices A3 and A4 list all the explicit instruction as it was provided to the participants.

Although using the computer as instructor does limit the possibilities, several kinds of exercises were used to strengthen the students’ explicit knowledge. In the receptive phase of the instruction, students had to recognize the target structure in example sentences by means of multiple choice or yes/no questions. Input enhancement was also used in dialogues in which the target structure would be underlined. As the students progressed, more productive exercises would be used, such as gap-filling and sentence completion. As indicated, Appendix A2 provides a precise outline of all the exercises students had to do.

The implicit instruction was created to invite the students to process the target structure for meaning. However, the students were never made aware of this: the focus was entirely on meaning in the perception of the participants; and there were no explicit attempts to get the language learner to notice the target structure that would be abundantly present in the comprehension exercises. An implicit focus-on-form was achieved by presenting the target structures in the context of text comprehension exercises. These exercises were always based on one of the texts just read, and providing the proper answer simply required remembering or checking the facts in the text. The participants would have to read the instruction and possible answers carefully; and thus they were forced to process the target structure for meaning. In later stages of the programma, the participants were also asked to type their answers to questions that were designed to elicit the target structures.

During the focused exercises, immediate feedback was provided. In the beginning stages of the instruction, the feedback function was often used to restate the rule (explicit feedback), or repeat the correct answer (implicit feedback). In later stages, feedback might simply consist of conveying whether the answer given was right or wrong.

3.2.3 Input matching

As explicit instruction may lead to concomitant implicit learning effects, care was taken to keep the amount of input and the nature of the input equal. The input in the experimental conditions was precisely matched, meaning that the same input sentences were used to create either an explicit or implicit focus on the target structures. Given that explicit and implicit instruction are very

66 CHAPTER 3

different in nature, this was not an easy task. In practice, input matching was realized by modifying the text comprehension exercises into grammar exercises. This way, the original implicit focus on the target structures was changed into an explicit focus, while the input itself remained unchanged. Thus, the subjects in both conditions were exposed to the same input, meaning that they had more or less equal chance to learn implicitly from the instruction.

Despite the efforts to keep the input between the conditions similar, there were nevertheless some differences. One obvious difference involves the rule explanation that was part of the explicit instruction. Also, students in the EI condition were given the opportunity to return to the rule explanation screens during the grammar exercises. Providing this opportunity was considered an important aid to effectively develop explicit knowledge, but it may have led to unequal exposure. In order to make up for these inequalities, students in the implicit condition received extra input during the exercises themselves, for example, by means of feedback screens. This way, the number of times that participants in the EI and II conditions encountered the target structures was more or less equal.

Table 3.4 provides examples of how this input matching has been realized throughout the instruction.

3.3 Assessing grammatical development

3.3.1 Introduction

The development of grammatical proficiency has been monitored by means of two tests of grammatical knowledge, one tapping into explicit knowledge, and one tapping into implicit knowledge. Several authors have cautioned that the effect of instructional efforts may not be immediate, but delayed (Gass, 1991, 1997; Lightbown, 1994; Gass en Varonis, 1994). The argument is that teaching explicit knowledge may not impact immediately on the development of implicit knowledge, but will facilitate implicit learning once the L2 learners starts acquiring the particular structure. For this reason, many FFI studies have included delayed knowledge tests. In this study, the L2 learners’ grammatical knowledge has been assessed at three different points in time: once before the instruction (T0), once immediately after the instruction (T1), and once with a delay of two months after the instruction (T2).

METHODOLOGY 67

TABLE 3.4 Examples of input matching (translated from Dutch) Explicit instruction Implicit instruction DoC: Chapter 2, ex. 1; (see App. A2):

Which word is a form of the DoC? When a commercial is funny, people think it is nicer.

This statement is about the text you just read. Is it true or false?

When a commercial is funny, people think it is nicer.

DoC: Chapter 2, ex. 2; (see App. A2): These 3 sentences are taken from advertisements. Which sentence does not contain a form of the DoC

A – X now washes even cleaner. B – With X, colours stay nicer. C – X is not expensive.

What does this advertisement try to tell you?

A – X now washes even cleaner. B – With X, colours stay nicer. C – X is not expensive.

DoC: Chapter 7, ex. 1; (see App. A2): Fill in one of the (three) words in the sentence below. Use a form of the DoC

X is one of Holland’s [high/fast/small] skaters.

Fill in of the (three) words in the sentence below. Create a logical sentence.

X is one of Holland’s [high/fast/small] skaters.

SubC: Chapter 1, ex. 3; (see App. A2): Which sentence is does not have a subordinate clause?

A – X like the Sony, because it is very small.

B – x likes the Sony, because it looks smart.

C – X like the Sony: it is very small and it looks smart.

Why does X like the Sony discman?

A – X like the Sony, because it is very small.

B – X likes the Sony, because it looks smart.

C – X like the Sony, because it is very small and it looks smart.

SubC: Chapter 8, ex 1; (see App. A2): Make from two sentences one complex sentence:

1 - You can earn a lot of money. 2 - You are good in sports.

You can earn a lof of money if …

Read the text (excerpt given) and finish the sentence You can earn a lot of money if …

68 CHAPTER 3

Both the explicit and the implicit test were presented in precisely the same form at each time of measurement. A number of researchers have adopted the technique of changing words in their tests just to give the language learner the impression that their subjects are doing a different test (i.e., Sanz & Morgan-Short, 2004). However, it was felt that even changing words could lead to differences in test difficulty. Therefore, no modifications to the tests were made in order avoid all possibility of introducing such differences. The test takers were explicitly told that the tests were precisely the same, so their teachers could see they had improved their knowledge of Dutch.

In the remainder of this section, both tests are discussed in detail.

3.3.2 Explicit grammatical knowledge

In 2.4.2, a number of requirements were put forward that explicit knowledge tests should meet. Most importantly, the task should be such that it expressly calls on one’s knowledge about a particular rule of grammar, and consequently, it should allow the test taker the time to do so. For these reasons, an untimed paper-and-pencil grammaticality judgement task was used. Also, grammaticality judgement tasks call on explicit knowledge without making explicit reference to the target structures in focus, which would offer an advantage to students who had explicit instruction. Grammaticality judgement tasks require the L2 learner to judge whether a particular stimulus is grammatically correct or incorrect. In this particular test, the participants were given target sentences containing incorrect realizations of the target structure, and were instructed to indicate whether they recognized any errors in the sentence. In addition, they were asked to underline the error if they saw one. For the degrees of comparison, subjects had to see that a particular adjective was not marked for comparison while the context required this. For subordination, default word order was used in the subordinate clause. Thus, the target structures were always presented incorrectly in the test. Examples are provided in Table 3.5. The entire test can be found in Appendix B1.

TABLE 3.5: Grammaticality judgement examples. 1) Anne is best lang, maar Lien is nog veel lang. correct / incorrect 2) Ik wil graag een ringetje in mijn neus, omdat ik vind

dat ontzettend mooi. correct / incorrect

METHODOLOGY 69

The number of items of which this test consisted was small: it consisted of 20 items in total, six focusing on the degrees of comparison, and eight focusing on subordinate clauses. Five sentences contained errors unrelated to the target structures, and one sentence was correct. The number of test items focusing on the target structures must be considered small, then. In general, test reliability is positively improved as the number of test items increases. However, a small scale pilot (eighteen subjects) with a 40-item GJ test revealed practical problems. As the subjects of this study were selected to have little knowledge of the target structures in focus, taking this test proved simply too difficult. It took too much time, and some students became quite anxious. In light of this experience, it was decided to keep the number of test items small. In addition, the five sentences that contained errors not focusing on one of the target structures were kept quite simple.

The explicit knowledge score expresses the number of times students identified incorrect use of the target structure in the stimulus sentences. Table 3.6 provides the reliability coefficients for both the degrees of comparison and subordinate clauses. Although the test was administered as one task, these coefficients are reported for each target structure independently. The reason is that all analyses in this study will be presented independently for each target structure. The reliabilities for both target structures and at each time of measurement are sufficiently high, especially given the small amount of test items.

By way of validating the explicit knowledge test, a second test was administered (to be found in Appendix B2). This test measured the participants’ explicit knowledge of the target structures by means of controlled production tasks, and because of the more explicit nature of this test, it was administered at T1 only. There were ten items in the test for each structure. For the degrees of comparison, gap sentences were given in which students had to fill in a

TABLE 3.6: Reliability coefficients (KR-20) of the grammaticality judgement test and the controlled production test (T1 only)

Test T0 (n=101) T1 (n=101) T2 (n=76) Grammaticality judgement

Degrees of comparison .70 .80 .81 Subordination .78 .86 .78

Controlled production Degrees of comparison - .84 - Subordination - .93 -

70 CHAPTER 3

comparative form. The adjective to be used was given. For SubC, students had to combine two clauses by means of a given subordinate conjunction. The subjects had to do this task on the basis of an example sentence. Although no explicit reference was made to the target structures, this test more explicitly focuses the students towards the target structures. The reliability coefficients can be found in Table 3.6. After correcting for attenuation, high correlations were found between the two tests (.81 for the DoC, and .73 for SubC). This indicates that the two tests measure the same construct to a considerable extent; and therefore, that the students’ explicit knowledge of the target structures was quite adequately measured by the grammaticality judgement task. The controlled production test was not used for purposes other than validating the grammaticality judgement task.

3.3.3 Implicit grammatical knowledge

The primary requirement of implicit grammatical knowledge tests is that they assess the ability to use the target structures in spontaneous situations of language use. Spontaneous use of the target structures means that their use was unplanned or incidental, and that the structures were used to fulfil a clear communicative purpose. In addition, although the notion of proficiency actually covers accuracy, fluency, and pragmatic aspects of language use; in this study the focus is on the ability to use the grammar structures accurately.

Implicit grammatical knowledge has been measured by means of a free written response test. A number of considerations underlie the choice for a written production test. Most importantly, the effects of instruction are expected to affect written proficiency before they do oral proficiency. The reason for this is that conversation – being more time-pressured – requires higher degrees of automatization (Bialystok 1989). Therefore, it is likely that L2 learners have more opportunity to use their explicit knowledge as a monitor in written proficiency tests. A number of FFI studies confirm this. Day and Shapson (2001) and VanPatten and Sanz (1995), for example, found no effect of their instruction on oral proficiency tests, but it did affect written proficiency (see also 2.4.4). Thus, a written proficiency test probably maximizes chances of finding an effect. A more practical consideration to choose for a written proficiency test was that collecting and processing written data is simply faster and less work-intensive, thus allowing for a larger subject pool.

The free written response test was designed to elicit incidental use of the target structures. Students were required to read a short text describing a

METHODOLOGY 71

particular situation, or to look at a particular comic strip or picture. Questions were asked about these, and the students had to respond to these appropriately in one or two sentences. The situations were designed in such a way that the L2 learner would likely use the target structures, but there was no explicit invitation to do so. The students were expressly invited to use their imagination when replying to the questions, as long as their response was appropriate. Consequently, there was no guarantee that the L2 learners would actually use the target structures in their response. Students were kept unaware of the fact that they were tested for their knowledge of and ability to use a particular grammar structure. The test was presented as a writing proficiency test, and the instruction simply asked them to write short sentences in reply to the questions, and to write carefully and correctly.

Test items that intended to elicit forms of the degrees of comparison placed the test taker in a situation in which they had to make comparisons. For example, they were asked to compare two radios, or to choose between two pairs of jeans. No attempt was made to specifically elicit the use of either comparative or superlative forms, as comparatives and superlatives can be used more or less interchangeably depending on the perspective taken. For example, in response to the question which radio they would buy, students might answer: “I buy the cheapest” or “the one that is cheaper”. Both are equally appropriate responses to the question.

Elicitation of subordinate clauses aimed primarily for conditional and causal SubCs. These are by far the most frequent in Dutch. In addition, the pilot study revealed that SubCs expressing relations other then conditional or causal relations proved to be very difficult to elicit. Conditional SubCs can be elicited very effectively by means of questions asking when (under what condition) something may occur. For example, questions starting with the Dutch question word wanneer (when, under what circumstances) resulted in a very high rates of conditional SubC use. Situations aiming for the use of causal SubCs proved to be less compelling. Students were asked to explain what caused a particular thing to happen in their opinion, or why they would make a particular choice. For example, when asking why they would choose a particular radio, students might answer: “because it is cheaper”. However, they also frequently omitted the subordinate conjunction: “it is cheaper”; or avoided the use of the subordinate conjunction by means of the synonymous coordinate conjunction want: “[want] it is cheaper”. For this reason, the test contained only three questions aiming for conditional SubCs, and eighteen questions aiming for causal SubCs.

72 CHAPTER 3

In total, 29 test items intended to elicit one of the target structures. Eight focused uniquely on the DoC; eleven elicited SubC; and ten items invited the use of both DoC and SubC simultaneously. Consequently, there were 18 DoC items and 21 SubC items. The test items were selected on the basis of small-scale pilots with twelve L2 learners that had already acquired the target structures. If an item elicited the use of one of the target structures in at least 50 percent of the cases from these advanced learners, the item would be considered fit for the test. During the pilot phase, it also became clear that elicitation of incidental use of subordinate clauses other than conditional or causal subordinate clauses proved extremely difficult. For this reason, only conditional and causal SubCs were elicited. The entire test can be found in Appendix B3, and the DoC and SubC items are marked.

The implicit knowledge score calculated on the basis of the free written response task simply consisted of the amount of times the target structures were used correctly in the test: the frequency of occurrence. Each occurrence of the target structure was counted, so if one test item generated multiple uses of a target structure, each use would be counted. For this reason, percentages of correct use could not be calculated: they would in some cases exceed a hundred percent. However, for calculating test reliability coefficients, a test item was scored ‘correct’ if it elicited at least one correct instance of the target structure. Incorrect use, no use or avoided use were scored as ‘incorrect’. Table 3.7 provides the reliabilities for both grammar structures. As can be seen, they are considerably high.

TABLE 3.7 Reliability coefficients (KR-20) of the implicit knowledge tests Free written response task pre (n=101) post (n=101) 2nd post (n=76) Degrees of comparison .83 .83 .77 Subordination .91 .90 .89

METHODOLOGY 73

3.4 The participants

3.4.1 Introduction

The L2 learners participating in this study were students from ten different schools from the North and West of The Netherlands. The participants in the experimental conditions (EI and II) came from eight different schools, and they were randomly assigned to either treatment condition by means of the computer programme. There was no matching of any kind. As the subjects in the NI condition did not work with the computer programme, this group consisted of two intact classes from two different schools. In this section, the participants are introduced. In 3.4.2, their current educational background is discussed, followed by a discussion of the mortality that this study suffered from in 3.4.3. Subsequently, the students are characterized in terms of general Dutch proficiency (3.4.4), day-to-day exposure to the Dutch language (3.4.5), and performance during the instruction (3.4.6).

3.4.2 The subjects and their educational background

The subject pool for this study consisted of 101 students from so-called ‘ISK’ classes. ISK is short for ‘Internationale SchakelKlas, and refers to a special secondary school educational facility for immigrant students. Mostly, the ISK is an integrated part of a secondary school, but sometimes local authorities install a separate ISK school for the entire district (Berenst, Bienfait, Hofstede & Van der Schaaf 1999). ISK classes provide intensive Dutch courses for students ranging between 12 and 18 years old, with the intention to enable them to follow regular secondary programmes as soon as possible. The students of ISK classes are a diverse group. They came from many different countries, and came to the Netherlands for various reasons. Sometimes they fled their home country because of unstable political regimes or poor economic conditions, but they also came because of parents working in the Netherlands. Often, they are involved in lengthy legal procedures in order to obtain a ‘green card’ that would allow them to stay in The Netherlands. While awaiting the outcome of such procedures, they often live in special housing facilities, and have the legal right to education (Berenst et al., 1999). Generally speaking, these students have a strong motivation to learn Dutch, because for most their future depends on knowing

74 CHAPTER 3

Dutch. The subjects of this study came from 10 different ISK schools from the North and West of the country. These schools were selected based on proximity, availability of students, and willingness to cooperate.

Generally speaking, grammar is not very prominent in the curricula of ISK classes. As pointed out, the majority of ISK’s use Zebra (Alons, Bienfait, & et al., 1999) as the basis of their Dutch instruction, which prepares immigrant children for the regular secondary educational system. Grammar instruction in Zebra is mainly implicit. Although each lesson does target particular grammatical structures, there is no explicit reference to these structures. Zebra is accompanied by a reference grammar, Zebra Grammatica (Andringa, Bienfait, Kuiken & Van der Schaaf 2000) , but it is not introduced until lesson twenty-six. In practice, this means that students do not work with grammar until their second year of Dutch.

Some ISK classes do organise additional grammar instruction based on other methods of Dutch as a second language (mostly for adults), or on self-made materials. Of the ten participating ISK schools, three schools indicated providing such additional grammar hours. In one school, the amount of time spent on grammar exceeded one hour per week. In addition, all participating teachers indicated providing on demand grammar instruction, i.e. when one of their students asked for it. None of the teachers indicated having taught one of the structures focused on in this study.

3.4.3 Mortality

This study suffered from considerable mortality in the subject sample. From pre-test to post-test, 101 students were tested, but a quarter of the subjects was lost between the post-test and the second post-test. Ultimately, the grammatical development of 76 L2 learners was followed over a period of three months. A number of reasons underlie this loss. First, this study was conducted in a time of changing national admittance policies. This led to a sharp decrease in the number of refugees allowed into The Netherlands. In less than two years’ time, the number of ISK students decimated and many ISK schools had to close their doors, or were threatened to do so within the near future. One group was lost because of this: the second post-test could not be organised in the school’s last weeks of class. Another reason for loss was that some students simply left The Netherlands, or were relocated to other schools or to other areas in The Netherlands. Finally, some students were lost because of absence on the days of testing. Table 3.8 provides an overview of the subjects per condition. Especially

METHODOLOGY 75

TABLE 3.8 Number of participants per treatment condition Treatment condition N (pre post) N (pre post delayed) Condition 1: DoC expl – Sub impl 41 32 Condition 2: Doc impl – Sub expl 40 35 Control: no instruction (NI) 20 9 Total 101 76

the NI group – already the smallest group – suffered from considerable loss, which makes this group rather small. There was another problem with the NI group. One class inadvertently received some explicit instruction in the DoC. For this reason, they had to be excluded from some of the analyses.

A point of concern is whether the loss was random. This concern will be addressed throughout this study by investigating whether the mortality somehow affected the nature of the subject sample and the results presented in this report.

3.4.4 General L2 proficiency

The subjects were selected for participation in this study on the basis of their general L2 proficiency. The intention was to put together a subject sample that was homogeneous with respect to their proficiency. In fact, a number of different indicators of proficiency were gathered. They will be described here shortly. Each measures proficiency in a different way, and may tap into different aspects of a students’ proficiency. These measures were merely used to gain insight into the participants’ general level of proficiency.

De ISK-toetsen

De ISK-toetsen (Schuurs, 1999) were primarily used for subject selection. They are a series of standardized tests especially developed for assessing general proficiency of Dutch of ISK students. The test series is used widely by ISK’s in The Netherlands to monitor progress in Dutch, and it contains listening, speaking, writing, and reading components. As most schools use this test, it provided a good starting point for subject selection. Subjects scoring approximately fifty points on this test were considered fit for participation. For 85 subjects, ISK test scores were obtained. Their mean score was 53.1 (SD = 5.2), and observations ranged between 41 and 67. No significant differences were found between the three conditions on ISK test scores, irrespective of mortality.

76 CHAPTER 3

The C-test

The C-test provides another indication of overall proficiency in the second language (Klein-Braley, 1997). C-tests are a type of cloze tests, in which parts of words are deleted rather than entire words. In C-tests, every n-th word is partially deleted. Students have to complete the word on the basis of the context and the first half of the word: the first and last sentences of the text are left untouched. The advantage of the C-test is that it generates many test items on a small text. In addition, if multiple texts are used with different topics, knowledge-of-the-world effects can be minimized. Also, it can be scored quickly and objectively (Klein-Braley, 1997; Weir, 1988).

The test used for this study was developed for the occasion. The test can be found in Appendix C1. It consisted of three short and simple texts with different topics. On the basis of small-scale pilots, it was decided to delete every fourth word of the text. In total, the test consisted of 64 items. It was administered to all students at T0 together with a number of other tests. Test reliability was found to be adequate (Crohnbach’s alpha = .85). The mean score was 64 percent correct (SD = 12.9), and observations ranged between 36 and 97 percent correct. No significant differences were found between the conditions, irrespective of mortality.

Written proficiency

The free written response data gathered to measure implicit knowledge (see 3.3.3) have also been used to determine three different components of written proficiency: fluency, grammatical accuracy and spelling ability. Fluency expresses how comfortable second language learners are in producing written text, and grammatical accuracy refers to freedom from error in relation to native-like use (Wolfe-Quintero, Inagaki, & Kim, 1998). Finally, spelling refers to the ability to form words correctly in the L2. These three measures were included as measures expressing specifically the students’ ability to write in their L2.

The three measures have been calculated on the basis of all free written response task utterances at T0. Following recommendations by Wolfe-Quintero, Inagaki, and Kim (1998), fluency was calculated by counting and averaging the number of words per clause realized per student. The mean fluency score was 5.5 words per clause (SD = 0.5), and scores ranged from 3.8 to 6.9. Grammatical accuracy was obtained by counting the number of grammatical errors per clause. For example, if there was no subject-verb agreement, this would count as one error. Spelling errors were not included in the count; nor were errors in using the

METHODOLOGY 77

target structures included, as these were intentionally elicited. The mean number of errors per clause was 0.84 (SD = 0.5); scores ranged from 0.015 to 2.04. Finally, spelling was calculated as the number of spelling errors per clauses. On average, 0.31 (SD = 0.21) spelling errors per clause were made, and scores ranged between 0.02 and 0.9. Again, no significant differences were observed between the conditions, irrespective of mortality.

3.4.5 Exposure to the second language

In order to gain insight into the educational background of the participants and the amount of exposure they had had to the Dutch language, a questionnaire was submitted to their teachers. The teachers were asked how much contact the students might have with the Dutch language, and how long their students had been in The Netherlands. There were significant differences between the three conditions.

The amount of exposure outside the ISK was assessed by asking teacher to estimate how much they thought their students would use Dutch outside school. They were given the following options to choose from: often, fairly often, fairly little, and hardly. Data were obtained for 92 participants. According to their teachers, eleven participants used Dutch often; 44 percent used Dutch fairly often; 38 percent used Dutch fairly little; and 6 percent were said to hardly use any Dutch outside school. The Kruskal-Wallis statistic was used to assess how the observations were distributed across the three treatment conditions. A significant difference was observed: X2 = 11.99, df = 3, p. = .00, N = 92. This significant difference is caused by the NI group: they used Dutch less than the other two groups, and if they were excluded from the analysis, no significant differences were observed.

Another indication of exposure to Dutch is the length of the students’ stay in the Netherlands. Teachers were asked to give the date of arrival in the Netherlands. Using this date, the number months until the pre-test date was calculated. Data were obtained from 84 participants. Their mean length of stay was 21.2 months (SD = 7.8). The observations ranged substantially: one student had been in The Netherlands for just 8 months, while another had been in The Netherlands for 46 months already. A one-way ANOVA was run to test for significant between-group differences: there were none for the total sample. However, for the sample that was followed for three months, there were significant between-group differences: F(2,63) = 5.24, p. = .01. Again, the NI

78 CHAPTER 3

group was the odd one out, its participants having spent considerably more time in The Netherlands than the participants from the experimental conditions.

3.5 Assessment of individual characteristics

3.5.1 Introduction

One of the goals of this study is to evaluate how differences between learners affect explicit and implicit learning. A number of individual difference (ID) variables have been found to affect L2 learning, and will therefore be evaluated in this study. They are: developmental readiness, L1 similarity, age, aptitude, learning style, and motivation. Each of these factors has been assessed; this section explains how. It starts with developmental readiness (3.5.2), then continues with L1 similarity (3.5.3), Age (3.5.4), Aptitude (3.5.5), and ends with the affective variables learning style and motivation (3.5.6).

3.5.2 Developmental readiness

Developmental readiness refers to the phenomenon that grammar may only be effectively instructed if the L2 learner is in a particular stage of grammatical development. A complicating issue is that there is no evident way to operationalize the notion of ‘developmental readiness’ without extensive prior research. An important reason is that it is simply unclear what underlies the notion of developmental readiness (see 2.2.3): constraints specific to linguistic processing, general processing capacity limitations, or input features. As there is no clear theoretical motivation to operationalize developmental readiness, determining sensitivity to instruction has to be based on the participants’ interlanguage. In FFI research, several approaches have been used, but all are based on emergence of correct use. Some have used previous research to define stages in the acquisition of their target structure (e.g., Pienemann, 1989; White, 1998). Another approach has been to monitor L2 use in formal and informal situations, where differences in use of the target structure between the two situations indicated readiness (e.g., Bienfait, 2002: see also 2.4.4 and 2.4.5). Finally, emergence of correct use in the pre-test has been taken to be an indication of developmental readiness (e.g., Williams & Evans, 1998). As there is no previous research to base stages of acquisition on, and because monitoring

METHODOLOGY 79

formal and informal language use was considered to be too labour-intensive, the latter approach has been adopted.

Emergence of correct forms of the target structure in the participants’ use of Dutch will be taken to indicate developmental readiness. Naturally, readiness has to be determined for each target structure independently. A two-level nominal variable will be created for each target structure based on whether correct forms of the structures were used at T0. The group of students that already do use correct forms of the target structures are considered developmentally ready, while non-use is taken to indicate unreadiness. When selecting and assigning students to the treatment conditions, developmental readiness was not taken into consideration. In other words, there was no matching, and possible effects of developmental readiness are assessed post hoc.

3.5.3 First (and second) language background

As pointed out, the students of ISK schools are quite diverse with respect to their first language (L1) background. Due to a rapid decline in student population of ISK schools, the intention to select students with similar L1 backgrounds could not be fulfilled. In fact, this study now includes speakers of 33 different languages. The most frequently occurring language was Portuguese as spoken in Angola: it was the L1 of 23 students. There were no other languages with more than ten speakers present in the subject pool. Also, only five out of 101 participants spoke a Germanic language as L1: English. The speakers were almost exclusively originating from Asian or African countries. Eleven participants were from Europe: four of them spoke Russian, four spoke Turkish, there were two speakers of Polish, and participant spoke Bulgarian.

One of the goals of this study is to assess whether effective FFI depends on a student’s L1. To enable answering this question, particular features of the L1 were coded that might affect the students’ L2 proficiency. For the degrees of comparison, the participants’ L1 was coded for how they express comparison: morphologically or periphrastically. If the language exclusively allowed for periphrastic degrees of comparison, it would be coded as marked. Otherwise, it was coded as unmarked. A Chi-square analysis revealed no significant differences in the distribution of marked and unmarked L1s between the three conditions: X2 = 4.60, df = 2, p. = .10, n = 98. The trend towards significance here is caused by the control group (for the delayed group: X2 = 3.42, df = 2, p. = .18). Because Dutch requires inversion in subordinate clauses, the L1 was coded for their default word order (svo or sov), and whether it has similar inversion in

80 CHAPTER 3

subordinate clauses. One language – Armenian – seems to have similar inversion in subordinate clauses: there are three speakers of this language in the subject pool. Given the small number of speakers, this similarity was ignored. A Chi-square analysis on the distribution of svo and sov languages did not reveal significant differences between the three conditions: X2 = .74, df = 2, p. = .69, n = 98. (for the delayed group: X2 = 7.35, df = 2, p. = .03

The participants were also asked if they spoke any other second languages. Forty-nine students reported speaking an L2 besides Dutch. These languages were mostly regional languages or French or Spanish as spoken in former colonies. The participants were also asked to assess their command of this L2 as compared to their current level of Dutch. Twenty subjects reported speaking their L2 better than they did Dutch. None of the thirteen students that reported a Germanic language as their L2 (all English) spoke it better than Dutch. Three reported it being equal; the other ten indicated being worse in English than in Dutch. Effects of the students’ L2 have not been investigated.

3.5.4 Age

Age is another factor that has been found to affect L2 acquisition, and may influence the effectiveness of FFI. Age has been measured in years, and the participants of this study range somewhere between 12 and 18 years old. In fact, 85% of all participants fall within the 13-16 interval: five subjects were twelve years old, seven were 17, and three subjects were 18 years old. Treating age as an ordinal variable, the Kruskal-Wallis statistic for independent samples was used to test how age is distributed across the three conditions. There were no significant differences: X2 = .69, df = 2, p. = .71, N = 101.

3.5.5 Aptitude

For this study, two components of aptitude were assessed, both based on tests from the Modern Language Aptitude Test (MLAT) battery developed by Carroll and Sapon (1959). In this case, rote memory and grammatical sensitivity (GS) were assessed. Both aptitude components were assessed by means of tests developed for the occasion, as the standardized Dutch form of the MLAT – the Verbale Aanleg Test (Drenth & Van Wieringen, 1969) – was considered to be too difficult for the present subject sample.

In the original, the rote memory test requires the language learner to rote-learn Kurdish words. Since it could not be excluded that speakers of Kurdish or related languages might participate in this study, the test was administered using

METHODOLOGY 81

Swedish words to learn. None of the participants had to be excluded because of prior knowledge of Swedish. In addition, the test procedure was somewhat simplified as compared to the original: it involved a learning session of 2 minutes, in which the participants were asked to memorize the Dutch meaning of sixteen Swedish words. Immediately afterwards, they were given a sheet containing the Swedish words, and they had to give the Dutch meanings. An item was scored correct if the student reproduced the correct meaning: spelling errors were ignored. Ultimately, memory scores were obtained from 97 of 101 subjects. Reliability was found to be sufficient (Cronbach’s alpha = .80, N = 97). The mean score was 8.9 (SD = 3.7) , and the scores ranged from 0 to 16. The test can be found in Appendix C2.

The grammatical sensitivity (GS) test assesses the ability to match phrases that have the same grammatical functions in pairs of sentences. In its original form, the GS component is administered in the test taker’s L1. However, in the light of the different L1’s of the participants of this study, it was administered in Dutch. Robinson (1997) similarly employed the MLAT, and did not find significant correlations between the MLAT and general proficiency, which is an indication that the MLAT scores were not influenced by L2 proficiency. The test consisted of fifteen items, and an item was scored correct if the grammatical function of the example sentence phrase and the test sentence phrase matched. For this test, there were quite some missing values, as it was the last test to be administered in a row of tests: some participants did not take this test due to lack of time. Scores were obtained from 75 students, and the test was found to sufficiently reliable (Cronbach’s alpha = .81, N = 75). The mean score was 6.7 (SD = 3.8), and the scores ranged from 0 to 15. The test can be found in Appendix C3.

One way ANOVAs were run to assess whether there were differences between the conditions for either memory or GS. No significant differences were found, irrespective of mortality.

3.5.6 Affective variables

A number of personality factors or affective variables may influence the ability to learn. An attempt was made to gain insight into the participants’ personalities by means of a teacher questionnaire. Teachers were asked to judge their students on seventeen aspects relating to attitude, aptitude, motivation, structure-dependence and precision by means of five-point Likert scale items. The questionnaire can be found in Appendix C4.

82 CHAPTER 3

TABLE 3.9 Factors, eigenvalues and explained variance

Factor Factor 1 Factor 2 Factor 3 Eigenvalue Explained variance Cumulative variance

5.1342.7242.72

1.8615.5258.24

1.31 10.94 69.17

Contributing items

Factor loading

precise vs. chaotic .60 Well-considered (vs. impulsive) .82

Intellectual (vs. emotional) .65

Factor 1: cognitive style

Independent of others .74

Enjoys school .86 Motivated to learn Dutch .71

Factor 2: Motivation

Likes to speak Dutch .84

Has a feeling for language .91 Understands grammar instruction easily .90

Factor 3: Aptitude

Remembers things easily .80 The results of this questionnaire were subjected to a principle components

factor analysis to investigate whether particular sets of items could be combined to represent personality factors such as motivation, aptitude, etc. The data resulting from this questionnaire turned out to be suitable for such an analysis (KMO measure of sampling adequacy = .84, p. < 0.01). A number of items were removed first, though, because they did not correlate sufficiently with other items, which is a requirement for factor analysis. In Appendix C4, the questions marked with an asterisk were removed from the analysis. Once these items had been removed, all the conditions for performing a factor analysis were met. Oblique rotation – to be used when the factors to be extracted may be related to each other – was used to extract the factors. In this case, there are indeed no theoretical grounds to state that the factors should be uncorrelated. The results of the factor analysis can be found in Table 3.9. The outcome was a solution involving three factors with eigenvalues greater than one, and explaining in total 69 percent of the variance. Table 3.9 provides the details.

The resulting factors turned out to be well-interpretable. The first factor has been called ‘cognitive style’, and it expresses one’s operating style; whether one

METHODOLOGY 83

adopts a precise and well-considered approach to things to be undertaken, or a more chaotic and impulsive approach. Independence and intellect (which was contrasted with emotion) also loaded on this factor, which seems plausible. The second factor clearly represents the students’ motivation to be in school and learn Dutch. Finally, the three questions that relate to aptitude were found to load on the third factor.

One way ANOVAs were used to investigate whether the three conditions differed significantly with respect to either one of these factors. For none of the three factors, however, significant differences were observed, again irrespective of mortality.

3.6 Procedures and data processing

3.6.1 Introduction

In this section, matters of research organisation are discussed. The procedures and steps that were taken with regard to the collection of the data are described in 3.6.2. Next, in 3.6.3, a description is given of how the instruction was practically organised at schools. Issues of coding and data processing are dealt with in 3.6.4; and the section is concluded with a discussion of the statistical procedures followed, in 3.6.5.

3.6.2 Data collection

All data have been collected over a period of three to four months. The first test session took place at T0, immediately preceding the instruction. The second test session (T1) would immediately follow the instruction, and the third session (T2) took place approximately two months after T1. It should be pointed out that the pre-test was not always immediately followed by the instruction. In some cases, technical problems with the computer programme led to a delay of two or three weeks before the students started with the instruction. Table 3.10 provides an overview of the tests and times of administration. As can be seen, the free written response task and the grammaticality judgement test were administered at all three times of testing. At T0, three additional tests were administered: the paired associates test, the grammatical sensitivity test, and the c-test. At T1, the controlled production test was administered as an extra (see 3.3.2).

84 CHAPTER 3

All tests were administered in class by the researcher, mostly in absence of the students’ regular teacher. The students were told that they participated in a study into the acquisition of Dutch as a second language, and that their Dutch would be monitored for this purpose. In addition, they were reassured that the information would be used for research purposes only, and that they did not serve any additional purposes related to their school career. No mention was made of grammar or the specific grammar structures in focus. Nor was the computer programme mentioned that was part of the research. In a number of cases, however, their teacher had explained that the testing and their computer work were related.

At T0, a total of five tests were administered. These tests were administered in a fixed order. The first test would always be the paired associates test, because it involved strict timing. The students were challenged to try and remember as many words as they could in two minutes time. After this time, the students had to turn around their paper, and they were asked to give the correct Dutch meanings of sixteen Swedish words. They were allowed to think for as long as they needed, but it generally did not take much more time than five minutes to fill in the meanings. The next test was the free written response task, which was always given before the grammaticality judgement test. The test was introduced as an assessment of their writing proficiency, and the students were instructed to respond appropriately to the situations given to them. They were told explicitly that they were free to be creative in their response: all answers would be considered correct, as long as they took care of their Dutch.

The free written response test would be followed by the grammaticality judgement test, the c-test, and the grammatical sensitivity test. Because there were substantial differences between students in the time it took them to finish the free written response task, students did not have to wait for the rest to finish. As soon as they finished the test, they would continue with the grammaticality judgement test. This way, no students ever had to wait and perhaps take the opportunity to distract their classmates. As a result, the grammaticality judgement test, the c-test and the grammaticality judgement test were not explained in class, but individually. Explaining the grammaticality judgement test and the c-test proved rather straightforward, and did not lead to problems. Care was taken, though, to make sure students would underline the error in the grammaticality judgement sentences. The last test to be taken was the grammatical sensitivity test. This test was most difficult to explain. It was presented as a puzzle that students needed to solve on the basis of the examples

METHODOLOGY 85

TABLE 3.10 Overview of administered tests Type of test T0 T1 T2 Assessment of gram. development

Free written response task X X X Grammaticality judgement X X X Controlled production X

Individual differences Paired associates X Grammatical sensitivity X General language proficiency X

(see Appendix C3). The examples were all read out and discussed, but no clues were given as to why phrases matched. All in all, administering these five tests would take approximately two hours.

At T1 and T2, testing took much less time, as the participants only took three (T1), or two (T2) tests. The order of the tests was the same: so the free written response task was administered before the GJ task and the controlled production task. At this time, students were told that they would do the same tests as before. In addition, they were told that the results would be compared in order to investigate how much they learned, and they were encouraged to show they had learned. They understood why the tests were the same, and did not object to this.

Finally, it should be pointed out that a strong dissociation was created between testing and instruction. All the tests taken by the students were paper and pencil tests, and they were administered by the researcher at all three times of testing. The instruction was delivered by the computer, and the students’ regular teacher would accompany them during their computer work. In addition, although the computer instruction did contain exercises resembling grammaticality judgements, the instruction exercises and the test exercises were different. This way, effects of task familiarity were hoped to be reduced to a minimum.

3.6.3 Organisation of the instruction

As indicated, all instruction was delivered by means of a computer programme called Taal in Themas. The programme was created for the occasion by means of Authorware (1999) software especially designed to develop e-learning applications. The programme was designed to work on stand-alone computers, but has also been used successfully on school networks. It was distributed by

86 CHAPTER 3

means of a CD-rom, and would be installed by the schools’ system managers on demand of the participating teachers.

A special feature of the programme was its database connectivity. This enabled random assignment to the treatment groups. Microsoft Access databases were used for this. Before the start of the instruction, the teacher would provide the names of the participating students, which were then entered into the database. At this point, the students were randomly assigned to either treatment groups one or two (see 3.2.2). Also, they would be assigned their own login number. During the instruction, the programme would interface with this database and record student performance. When all students had finished the instruction, the school’s system manager would return the databases by email.

Once the programme was installed and the students had been pre-tested, the instruction started. Students would work for one hour per week with the computer programme. At some schools, they worked two hours per week. However, it also occurred that students skipped a week because of previous engagements in the school curriculum. The students worked individually on separate pc’s in computer classrooms, and their regular teacher would accompany them. Students were not told that there were two versions of the programme, and their neighbour might actually be doing grammar exercises about a different grammar structure. Most never found out, but when they did, the teacher would explain. To start the programme, students had to enter their login number. On first time login, students received an introduction about the programme and its workings. As all progress was recorded, students could stop at any point during the programme: on subsequent login, the programme automatically returned to the point the student stopped. On average, three one-hour sessions were needed to pass through the entire programme. As soon as all students had finished, the immediate post test was administered.

3.6.4 Coding

Each item of the free written response test was coded for a number of different aspects. First of all, actual use was coded, and whether this use was correct, incorrect or clearly avoided. In addition, the way the structure was realized in the utterance was also coded in order to assess whether correct use depended on the structure’s realization.

For the DoC, three different aspects of structure realization have been coded. First, an obvious feature of use that was coded was the actual degree used for

METHODOLOGY 87

TABLE 3.11 Examples of DoC use Use Examples Correct use Omdat nog kleiner is.

De meneer wordt steeds dikker Ik kies de goedkoopste en mooiste De ene is duner dan de tweede Ik vind deze mooier dan die andere Ik kies de mooiste broek, met de mooiste kleur Ik ga eerst kijk welk is de mooist Hij heeft het snelste tijd

Incorrect use Dan hij ga steeds dik worden hij is een beetje oude omdat die veel groot en beter is Bril 1 is mooi dan bril 2 Want op de bank is veilig dan op de vloer Ik kies de nieuw en de mooi broek Want is de goedkoopster auto Omdat is de mooie Want is de sportief auto

Avoided use Hij wordt meer dik Hij is meer groot geworden en sterk

No use Omdat hij gaat snel Omdat het is groot voor ik en mijn 3 kinderen Die heeft meer graintie [garantie] dan radio 2. The man can become sick.

comparison: either comparative or superlative degree. Second, the syntactic function was coded. Adjectives and their inflected forms can appear either attributively or predicatively (Aarts & Wekker, 1993). When used attributively, adjectives modify a noun phrase (de snelste auto: ‘the fastest car’), while predicative adjectives function as subject or object attributes (die auto is het snelst: ‘that car is the fastest’). If comparatives were used predicatively, the presence or absence of a comparative clause would also be coded (Deze radio is groter dan de andere: ‘This radio is larger than the other’). The third and last aspect of realization to be coded was the adjective itself.

88 CHAPTER 3

Coding decisions for the degrees of comparison were rather straightforward. If the participants used the degrees of comparison, they would mostly use them correctly. In Table 3.11, examples of correct use can be found. As soon as –er or –st was properly attached to the adjective to express comparison, an item would coded as correct. The schwa-inflection that Dutch adjectives receive in particular circumstances (een groter huis; het grotere huis) was ignored, as this inflection is not related to the use of the DoC. Errors in spelling were coded, but they did not affect the decision of whether a form would be coded as correct or incorrect.

Incorrect use was infrequent. Rather than producing incorrect forms, students would simply neglect to use the DoC. Sometimes, though, they would provide a context in which the use of DoC is obligatory. For example, the Dutch adverb steeds requires a comparative form. An error that was actually quite frequent was the omission of the comparative marker in combination with comparative clauses. Omission of a form of the DoC while the context clearly demands this would be marked as incorrect. Examples can be found in Table 3.11. A special case of omission of DoC markers is avoidance of use by means of the adverbs meer and meest (more and most). Because these adverbs betray a clear intention to express comparison, such cases would be coded as avoidance (see Table 3.11 for examples). They occurred infrequently. Items would be coded as no use in case no form of the DoC was used. In these cases, there is simply no telling as to whether the student intended to express comparison or not. If the question was left blank, it was also coded as no use.

A final remark is required with respect to coding the use of the DoC. It is important to point out that incorrect use of the DoC may go unnoticed. Quite frequently, students were found to omit the comparative suffix in phrases containing a comparative clause: “one is small than the other”. In the process of acquiring the DoC, L2 learners apparently fail to mark the adjective for comparison. When comparative clauses are used, such omissions can be noticed and marked as incorrect. However, if one does not use a comparative clause, there is no telling what someone writing “one is small” actually intended to say. Such phrases were coded as ‘no use’. Given that omission was quite frequent with comparative clauses, it is likely that ‘incorrect use’ has gone unnoticed, and that incorrect use scores are therefore an unfair reflection of real incorrect use.

For SubC, three different aspects of the structure’s realization have been coded. First, the type of SubC has been coded; or, in practice, the subordinate conjunction used. The second aspect pertained to verb phrase complexity. A distinction was made between single-verb phrases consisting of be or have

METHODOLOGY 89

TABLE 3.12 Examples of SubC use Use Examples Correct use Als ik iets grappigs zie of hoor

Als er een grapje gezegd wordt Wanneer je blij bent Omdat die mooi is Omdat het nat wordt door de regen Omdat die in minder tijd heeft gewonnen Omdat de trein zonder te stoppen gaat Zodat de olifant niet meer kan storen Doordat ze op een ijs staan

Incorrect use Als je doet een sport Als ik wil iets kopen Als ze rodelt over een andere vrind van mij Omdat is goedkoper is Omdat op de grond heb je meer plaats dan op de bank Omdat de olifant kan niet meer de fruiten pakken Omdat Jan heeft 7 minut gedaan Omdat hij kan goed luisteren

Avoided use Want is goedkoopste Want hij wil fruit uit de boom halen Want zij is bang

Not analysable

Als je rent Omdat ze willen slapen Omdat ze wil dat iederen haar hoort

No use De ene vogel valt door de liefde De olifant eet de appels Deze auto is iets grooter dan de andere twee autos Boes wil mol slaan

auxiliaries; single-verb phrases other than consisting of be or have auxiliaries; and multiple-verb phrases. Finally, clause complexity was also considered by way of the verb phrase complements realized. Three types of clauses have been distinguished: adjectives; direct objects, either with or without attributive adjective; and all other, more complex types of complements.

90 CHAPTER 3

For SubCs, coding decisions were also not too difficult. When the verb was properly placed at the end of the subordinate clause, it would be considered correct. Table 3.12 provides examples. Incorrect SubCs were very frequent, and errors always consisted of using the default main clause word order. Causality was regularly expressed by means of the coordinate conjunction want instead of the subordinate conjunction omdat, which would be coded as avoidance. It should be noted that although the Dutch may have preferences to use either want of omdat in particular contexts, the use of want is perfectly grammatical. Utterances can only be judged for their correctness if the predicate contains a complement. In cases there was no complement realized or when the complement was a relative clause, the utterance would be coded as not analysable. Finally, if no subordinate conjunction was used, or if the question was left blank, the utterance was coded as no use.

3.6.5 Statistical procedures

A number of different statistical methods have been used to answer the research questions posited in 3.1. The first research question pertaining to the nature and development of implicit grammatical knowledge was addressed by examining the free written response task data. In the previous subsection, the contexts in which the target structures may appear have been identified. By means of descriptive statistics, the use of the target structures in each of these contexts has been characterized. In addition, a comparison of percentages of correct use provided insights into the relationship between correct use and context of appearance. No inferential statistics were used.

The second research question addresses the effect of different kinds of instruction on the development of explicit and implicit grammatical knowledge. The independent variable ‘instruction’ was operationalized as a three-level between-subjects variable (EI, II, and NI). Two dependent variables were used: the grammaticality judgement scores and the free written response task scores, both obtained at three different points in time. Univariate analysis of variance with repeated measures was used to test the hypotheses related to the third research question. The analyses were performed separately for both target structures and both dependent variables.

To answer the third research question, two different statistical techniques were used, depending on which variable was addressed. In addition, rather than using three-level within-subjects dependent variable for grammatical

METHODOLOGY 91

development, gain scores were computed for explicit and implicit grammatical progress in both target structures. Developmental readiness and L1 similarity were both defined as two-level between-subjects variables. Therefore, univariate analysis of variance was used to explore for each of these two variables separately how they related to explicit and implicit gain. The remaining ID variables (the aptitude measures, cognitive style, motivation, and age) were all operationalized as interval variables. Therefore, correlation analyses were used to explore the relationship between these ID variables and explicit and implicit gain.

The fourth and final research question has not been addressed statistically, because the nature of the grammar structure has not been operationalized as a statistical factor. The research design actually involved two parallel experiments and does not logically allow for statistical comparison of the two structures In addition, the unequal scales of measurement for both the grammaticality judgement task and the free written response task were another reason not to introduce the kind of grammar structure as an additional statistical factor. Thus, the question was simply answered by accumulating and comparing the results on each of the previous research questions.

Statistical assumptions

Normally, analyses of variance require that the assumption of homogeneity of variance be met. However, in repeated measures designs, which involve comparing data from the same subjects over time, not only variances across conditions should be equal, covariances between pairs of conditions need to be as well This is referred to as compound symmetry or sphericity (Field, 2000). If the assumption of sphericity is violated, there are a number of corrections that can be applied to the data. In this study, the most conservative correction will be reported in case of a violation of sphericity: the Greenhouse-Geisser estimate.

Analysing skewness and kurtosis values for each measure demonstrated that the data were not always normally distributed. In fact, at T0, all distributions were positively skewed. Skewness diminished at T1; and at T2, there were no significantly skewed distributions. As the participants had been selected to have little knowledge of the target structures, a positively skewed distribution at T0 is hardly surprising. Violations of kurtosis occurred as well, but not very often. Although these violations were unfortunate, analysis of variance is robust against such violations (Van den Bercken & Voeten, 2002), especially if these violations are not too grave, as was the case here.

92 CHAPTER 3

Post hoc analyses

If necessary, post hoc tests were be carried out for a deeper investigation of found effects. For these analyses, the post hoc tests available in SPSS’s GLM repeated measures procedure were not used, as they do not compare differences between group means at each point of measurement. Group means are conflated over time, which makes for an uninteresting comparison between groups. In addition, there are technical issues. The error term used in post hoc repeated measures analyses to compute significance is based on the mean error of all cells in the design. This is problematic in case of violations of equality of variance. For these reasons, post hoc analyses were performed for each point of measurement separately, using one-way ANOVA’s comparing the groups one-by-one. In case of significant differences, effect sizes (Cohen’s d) were reported.

4 Findings

4.1 Introduction

This chapter explores the answers this study provides to the four research questions put forward in 3.1. Each section deals with one of the questions. First, in 4.2, the free written response task data are explored to shed light on how second language learners start to use the two grammatical phenomena in focus. The analyses presented in 4.3 are intended to demonstrate how explicit and implicit instruction affect the learners’ explicit and implicit knowledge of the target structures. Then, the focus shifts towards potentially interacting variables. In 4.4, the influence of developmental readiness, L1 similarity, and ID variables such as aptitude, motivation, learning style and age is examined. Finally, in 4.5, an exploration is provided of the potential influence of structure complexity.

4.2 Learning to use grammar

4.2.1 Introduction

RQ 1: How do second language learners develop the ability to use the target structures in spontaneous situations of second language use?

The analyses presented in this section are intended to describe how L2 learners develop the ability from ‘not-knowing-how-to-use’ to ‘knowing-how-to-use’ the degrees of comparison (DoC) and subordinate clauses (SubC) in contexts of free use. This will be investigated irrespective of potential effects of instruction. There are two main avenues of exploration to do with the development of more efficient use of the structure and rate of learning. In 4.2.2, efficiency of use will be examined by means of the emergence of correct use of the target structures in relation to how they are realized in the utterance. To this end, descriptive statistics are presented that provide a static description of the use of the target structures at each time of measurement (T0, T1, and T2), differentiated

94 CHAPTER 4

according to their realizations. The realizations that will be distinguished have already been introduced in 3.6.4. Subsection 4.2.3 focuses on how fast the ability to use the target structure develops. The analyses will be presented for each target structure separately, and all participants are included.

4.2.2 Structure realizations and correct use


Before correct use of the DoC will be considered in relation to their realizations, it is helpful to discuss some general characteristics of use. Table 4.1 describes the use of the DoC at T0, T1, and T2 without further differentiation. In total, the free written response task provided 18 contexts that invited the participants to use the DoC. The table shows that overall use increases from an average of 6.4 instances of use in the pre-test to 8.2 in the second post-test. Decomposing this figure into correct and incorrect use shows that the DoC were used correctly more often than incorrectly. Correct use increases from 4.7 times at T0 to 7.0 times at T2. Incorrect use occurred on average just over one time per test. Sometimes, the participants avoided the use of the DoC by means of circumscription (for example by using meer dik (more fat) instead of dikker (fatter); see 3.6.4). Avoidance did not occur very often: it was used on average 0.7 times at T0, and its use dropped to 0.2 times at T2. Finally, productivity expresses the number of different adjectives with which a form of the DoC was used. The average was 3.3 at T0, and it increased to 4.7 at T2.

The question explored here is whether correct use depends on the structure’s realization. In 3.6.4, the following realizations have been put forward as relevant to the use of the DoC: 1) the degree of comparison expressed (comparative or superlative); 2) the syntactic form of the comparison: attributive, predicative, or predicative including a comparative clause; 3) and the adjective used. Correct use of the DoC will be considered in relation to each these realizations.

First, differences in correct use of comparative and superlative realizations will be considered. The free written response task intended to elicit both forms, without expressly intending to elicit the use of either comparative or superlative forms in particular situations. Table 4.2 provides mean use figures for each point of measurement. In addition, percentages are presented that express how often the particular form was used correctly. The numbers clearly show that the subjects had a clear preference for using the comparative form in this test. At

FINDINGS 95

TABLE 4.1 The use of the degrees of comparison in the free written response task. Characteristics of use Statistic

Use at T0 (N=101)

Use at T1 (N=101)

Use at T2 (N=76)

Total use Mean 6.4 8.1 8.2 SD 4.2 4.5 4.1 Range 0 – 21 0 – 22 1 – 19 Correct use Mean 4.7 6.6 7.0 SD 4.3 4.9 4.4 Range 0 – 18 0 – 22 0 – 17 Incorrect use Mean 1.7 1.5 1.2 SD 1.9 2.0 1.7 Range 0 – 8 0 – 9 0 – 8 Avoided use Mean 0.7 0.4 0.1 SD 1.1 0.9 0.4 Range 0 – 5 0 – 4 0 – 2 Productivity Mean 3.3 4.3 4.7 SD 2.8 2.8 2.7 Range 0 – 12 0 – 10 0 – 10

TABLE 4.2 Mean use and correct use in percentages of comparative and superlative Doc in the free written response task.

Use at T0 Use at T1 Use at T2 Realization M (SD) % cor M (SD) % cor M (SD) % cor

Comparative 5.4 (3.9) 71 6.9 (4.5) 80 7.0 (3.9) 84 Superlative 1.0 (1.3) 77 1.3 (1.6) 83 1.1 (1.3) 90

T0, it was used 5.4 times, while the superlative was used on average only once. There was a clear rise in mean use of comparative forms from pre-test to second post-test; however, mean use of superlative forms remained more or less constant. This may indicate either that there was no learning effect for superlatives, or that the task simply did not invite the use of superlatives sufficiently. If one considers the percentages of correct use, there do not seem to be any differences between comparatives and superlatives. Both were used correctly in more than 70 percent of the cases at T0, and both demonstrate an increase in correct use.

96 CHAPTER 4

Another realization that may have led to differentiation in correct use is the syntactic form in which instances of the DoC appear. In 3.6.4, these realizations have already been introduced: for comparative DoCs, predicative use (Pred) and predicative use including a comparative clause (Pred+) have been distinguished. For superlatives, attributive (Att) and predicative (Pred) use have been distinguished. Again, mean use figures at each time of measurement and correct use in percentages are presented (Table 4.3). Both Pred and Pred+ types of comparatives occurred regularly at T0: Pred use occurred 2.9 times, and Pred+ occurred 2.6 times. For both, mean use increased from T0 to T2. Percentages of correct use also increased substantially for both. However, there is a clear difference in correct use of Pred and Pred+. At T0, the former type was used correctly in 83 percent of the instances of occurrence, while 50 percent of the latter type was used correctly: a difference of 33 percentage points. At T2, this difference was reduced to 16 percentage points.

Further differentiation of superlative instances of use led to very low frequencies of use. In addition, there was no noticeable increase in use in either attributive or predicative realizations. If percentages of correct use are considered, however, there does seem to be increased performance. In fact, at T2, the L2 learners made very few mistakes when using a superlative form. There are no indications of differentiated correct use between attributive and predicative superlatives.

The final realization to be considered is the adjective itself. In total, the free written response task elicited the use of the DoC with 29 different adjectives. Some of these occurred rather frequently, while others occurred only once or twice. Obviously, the situations used to elicit the DoC favour the use of particular adjectives, while others will not tend to occur as much. Therefore, frequencies of use are not informative. Table 4.4 presents percentages of correct use of the most frequently occurring adjectives (those that occur at least 25 times at each point of measurement) in the free written response task data.

Again, the question is whether one can speak of differentiated correct use depending on the adjective used. What stands out from Table 4.4 is that there did not seem to be major differences in correct use at each time of measurement. Groot and oud were the most difficult to use correctly, judging from their scores at T0. The difference between the highest and lowest score at T0 was 22 percentage points (the difference between dik and groot). This difference levelled out somewhat at T1 and T2; and at T2, all scores fell within a margin of 14 percentage points. For all adjectives, percentages of correct use increased from T0 to T2, although there was one exception to the rule: dik showed a slight decrease.

FINDINGS 97

TABLE 4.3 Mean use and correct use in percentages of the DoC in the free written response task, differentiated according to syntactic realizations.

T0 (N=101) T1 (N=101) T2 (N=76) Realization M (SD) % cor M (SD) % cor M (SD) % cor Comparative a

Pred. 2.9 (2.7) 83 3.8 (2.8) 89 3.6 (2.7) 92 Pred+CompC 2.6 (2.0) 50 3.2 (2.6) 68 3.3 (2.5) 76

Superlative Att. 0.3 (0.6) 81 0.4 (0.7) 86 0.2 (0.4) 94 Pred. 0.8 (1.5) 83 1.0 (1.2) 82 0.9 (1.1) 87

a – In principle, attributive comparatives can also occur in Dutch, but they hardly did in these data. Therefore, they were not included in this table.

TABLE 4.4 Correct use of the degrees of comparison in percentages for eight frequently occurring adjectives.

Correct use in % Realization T0 (N=101) T1 (N=101) T2 (N=76)

dik (fat) 84 76 79 duur (expensive) 69 67 78 goedkoop (cheap) 76 86 88 groot (big) 62 76 84 klein (small) 76 81 89 mooi (beautiful) 76 77 92 oud (old) 64 75 80 snel (fast) 82 81 91 All in all, the analyses presented here do not provide many indications of

differentiated correct use across different realizations of the DoC. The clearest indication of differentiated use was observed between Pred and Pred+ comparatives. There may have been differentiated use between some of the adjectives, too.

Subordinate Clauses

Table 4.5 displays the general characteristics of use of the subordinate clauses. It provides mean use, standard deviation and the range at each time of

98 CHAPTER 4

TABLE 4.5 The use of subordinate clauses in the free written response task. Characteristics of use Statistic

Use at T0 (N=101)

Use at T1 (N=101)

Use at T2 (N=76)

Total Mean 13.6 13.5 14.1 SD 5.9 5.8 5.9 Range 1 – 25 0 – 22 0 – 24

Correct Mean 5.0 6.4 8.2 SD 5.2 5.4 5.5 Range 0 - 20 0 – 18 0 – 20

Incorrect Mean 7.6 6.0 4.8 SD 5.4 4.9 4.7 Range 0 – 18 0 – 18 0 – 17

Avoided Mean 2.7 2.8 2.7 SD 4.4 4.6 4.6 Range 0 – 15 0 – 16 0 – 15

TABLE 4.6 Mean use and correct use in percentages of conditional and causal SubCs in the free written response task.

T0 (N=101) T1 (N=101) T2 (N=76) Realization M (SD) % cor M (SD) % cor M (SD) % cor Conditionals 3.0 (1.4) 64 3.0 (1.2) 72 3.2 (1.2) 85 Causals 9.6 (5.3) 28 9.4 (5.3) 40 9.7 (5.5) 49

measurement, and broken down according to total, correct, incorrect, and avoided use. In total, the free written response task provided 21 opportunities to use SubCs. On average, the participants actually took this opportunity 13.6 times at T0. As can be seen, total mean use of SubCs remained more or less constant from T0 to T2. However, if one considers correct use, there was an increase in mean use from 5.0 at T0 to 8.2 at T2. Conversely, incorrect use decreased from 7.6 at T0 to 4.8 at T2. As the total use figure is made up of correct and incorrect use, the lack of increase in total use must be due to these opposite trends in correct and incorrect use. In 3.6.4, it was pointed out that Dutch causal subordinate clauses can easily be avoided by means of a coordinate construction with want, which does not affect word order. The average use of avoidance was 2.7 times at T0. There was no noticeable increase or decrease in the use of this construction.

FINDINGS 99

Again, correct use of subordinate clauses will be examined in relation to how the structure was realized in the utterance. The realizations to be considered have been identified in 3.6.4: for SubC, the contexts of use are defined by: 1) the type of subordinate clause (conditional or causal); 2) verb phrase complexity; and 3) clause complexity. Each will be discussed below.

The first type of realization to be considered is the subordinate relation the clause expresses. It is important to remember that the test primarily invited the use of conditional and causal relationships, and only these two types are contrasted here (see 3.3.3). In addition, it should be kept in mind that the test contained only three situations eliciting the use of conditionals, while it contained 18 situations targeting causal SubCs because of expected difficulties in eliciting their use. Table 4.6 shows that average use of conditional SubCs was approximately three times per test at each time of measurement. Causal SubCs were used somewhere between 9.4 and 9.7 times. There was no increase in mean use. If one considers the percentages of correct use, a rather large difference in correct use between conditionals and causals shows up. At T0, 64 percent of the conditionals were used correctly, a number that increased to 85 percent at T2. In contrast, no more than 28 percent of the causal SubCs were used correctly at T0; and despite an increase in correct use, this difference still amounted to 36 percentage points at T2.

Another realization to be considered pertains to the complexity of the verb phrase used in SubCs. The verb phrases distinguished were those consisting of 1) single auxiliaries (Aux); 2) single verbs (SV); and multiple verbs (MV) (see also 3.6.4). Table 4.7 displays the frequencies of use and percentages of correct use from T0 to T2 for each type of verb phrase, and – given the previous analyses – differentiated according to type of SubC. The differences in mean use were substantial, but – as pointed out before – one should not attach too much value to these, as these may well be influenced by the nature of the test. The percentages show that the students obtained progress, irrespective of verb phrase complexity. However, for both conditional and causal SubCs, auxiliaries were used correctly most often, although the difference with single-verb phrases was not very large. For both conditional and causal SubCs, the students clearly had most difficulties with verb phrases containing multiple verbs.

Finally, the complexity of the subordinate clause may also affect the L2 learner’s ability to properly place the verb. Three different types of clause structure were coded: those consisting of adjectives (Adj); direct objects (Do); and complex constituent structures (CompC) (also see 3.4). Again, the statistics will

100 CHAPTER 4

TABLE 4.7 Mean use and correct use in percentages of conditional and causal SubCs in the free written response task, differentiated for verb phrase complexity.

T0 (N=101) T1 (N=101) T2 (N=76) Realization M (SD) % cor M (SD) % cor M (SD) % cor Conditionals

Aux 0.8 (0.8) 74 0.7 (0.7) 90 0.9 (0.8) 90 SV 1.7 (1.3) 79 1.6 (1.1) 83 1.7 (1.1) 90 MV 0.5 (0.7) 50 0.6 (0.7) 68 0.4 (0.5) 68

Causals Aux 5.4 (2.9) 42 5.0 (2.6) 56 4.8 (3.0) 63 SV 2.5 (1.8) 34 2.4 (1.6) 43 2.5 (2.2) 50 MV 0.9 (0.8) 15 1.7 (1.5) 27 1.5 (1.8) 38

TABLE 4.8 Mean use and correct use in percentages of conditional and causal SubCs in the free written response task, differentiated for clause complexity.

T0 (N=101) T1 (N=101) T2 (N=76) Realization M (SD) % cor M (SD) % cor M (SD) % cor Conditionals

Adj 0.9 (0.7) 74 1.0 (0.8) 85 1.2 (1.0) 93 Do 1.6 (1.1) 74 1.3 (0.9) 83 1.4 (1.1) 84 CompC. 0.5 (0.5) 55 0.5 (0.5) 73 0.5 (0.4) 76

Causals Adj 5.1 (2.8) 45 5.2 (2.9) 54 5.0 (2.7) 59 Do 3.1 (2.0) 28 2.8 (1.9) 41 2.5 (1.8) 53 CompC. 1.0 (1.0) 16 1.2 (1.0) 27 1.3 (1.2) 37

be presented for conditional and causal SubCs separately (see Table 4.8). There will be no further differentiation according to the kinds of verb phrases used, because frequencies of occurrence would be too low for such a fine-grained differentiation. The mean use figures serve to provide an indication of frequency of use of the different types of SubCs. What stands out is that mean use remained more or less constant throughout the time. However, the percentages of correct use again rose without exception. For both conditional and causal SubCs, Adj clauses were easiest to realize correctly. In fact, at T2, 93 percent of these were

FINDINGS 101

TABLE 4.9 Co-occurrence of conditional and causal subordinate clauses.

Characteristics of use

T0 % of N

(N = 101)

T1 % of N

(N = 101)

T2 % of N

(N = 76) No correct conditionals, no correct causals 16 11 3 Correct conditionals, no correct causal 30 22 15 No correct conditionals, correct causals 3 3 4 Correct conditionals and causals 51 64 78

used correctly in case of conditional SubCs. The differences in correct use between Adj clauses and Do clauses were marginal. As to be expected, placing the verbs correctly clearly proved most difficult in CompC clauses.

For SubCs, one additional analysis was performed. Given the clear differences that were found in correct use between conditional and causal SubCs, it is interesting to compare and contrast simultaneous use of each type. Table 4.9 describes the appearance of correct use of causals and conditionals. At T0, 16 percent of the subjects did not use any correct conditionals and causals. As would be expected, this number dropped to three percent at T2. The table also shows that correct use of verb placement in conditional clauses is by no means a guarantee that the L2 learner will also place the verb correctly in clauses expressing causal relationships. At T0, 30 percent used correct instances of conditionals, but no correct causals. This number decreased to 15 percent at T2. However, correct verb placement in causals almost always entailed correct verb placement in conditionals. Only 3 to 4 percent of the subjects used causals correctly without using conditionals correctly. The ability to use verb placement correctly in both contexts is demonstrated by 51 percent of the subjects, and increasing to 78 percent at T2.

In sum, these analyses provide a number of indications that correct use of subordination depends on its realization. There was a substantial difference in correct use of conditional and causal SubCs, and the ability to use conditionals correctly clearly preceded the ability to causals correctly. The complexity of the verb phrase and the sentence structure also led to differentiated use.

4.2.3 Developing the ability to use the target structure

This subsection focuses on how fast L2 learners move from not-knowing-to-use a particular grammar structure to knowing-to-use it. This will be explored by investigating the use of the target structures with reference to four stages of

102 CHAPTER 4

development, which are based on the developmental stages White (1998) used to monitor progress in the use of possessive determiners. On the basis of previous research, White was able to define a rather fine-grained sequence of eight stages, pertaining specifically to the development of possessive determiners. These eight stages roughly reflect four overarching stages: preemergence, emergence, postemergence and targetlike performance (White, 1998: p. 105). Preemergence simply refers to non-use or avoidance of the target structure. In the emergence stage, learners start using the target structure, but incorrectly. White interprets this as a phase in which the language learner does not yet apply the rule. In the postemergence stage, learners slowly start to use the rule correctly; and in the last stage, the rule is used in a targetlike way. These stages, then, are not structure-specific, and they are concrete and observable in data. The ability to use a target structure in spontaneous situations moves from: a stage of 1) no use; to 2) incorrect use; to 3) variable use; and 4) all correct use. The expectation that the ability to use the target structures correctly develops slowly will be addressed by means of these stages. There will be no further differentiation according to contexts of use for the DoC. For the subordinate clauses, the analyses will be performed separately for conditional and causal SubCs.


If the participants of this study are categorized according to how they use the target structure at each time of measurement, it is possible to see how they move from one stage to the next. The top half of Table 4.10 depicts how the learners use the DoC at T0 and T1. At T0, for instance, seven students did not demonstrate any use of the DoC, but at T1, five of them actually started to use the DoC in the free written response task: three show variable use, and two use all forms of the DoC correctly. Out of the eleven students that used the DoC only incorrectly at T0, six demonstrated variable use at T1. Most students showed variable use at T0: 50 out of 101. At T1, 20 started to use the DoC without error, and only two students ‘moved backwards’, in that they did not use any correct forms anymore at T1, while they did at T0. The same kind of ‘deterioration’ can be observed for the 33 students that did not make any errors at T0. At T1, 12 showed signs of variable use. If one considers moving from one stage to the next an indication of progress, then 31 students can be said to demonstrate increased performance. Fourteen students showed a decrease in performance: this number is almost entirely caused by students that demonstrated variable use at T1 while

FINDINGS 103

TABLE 4.10 Changes in patterns of use of the degrees of comparison in the free written response task.

Patterns of use at T1 in absolute numbers (N = 101)

Patterns of use Use at

T0 No use

incorrect use

Variable use

Correct use

No use 7 2 - 3 2 Only incorrect use 11 - 5 6 - Variable use 50 - 2 28 20 Only correct use 33 - - 12 21

Use at T1


No use 1 - - 1 - Only incorrect use 2 - 2 - - Variable use 39 - 1 25 13 Only correct use 34 - - 10 24

making no errors at T0. A large majority, 56 students out of 101, remained in the same stage of development. Although cell frequencies are lower due to mortality in the research sample, the lower half of Table 4.10 yields the same picture. There was some progression, primarily from variable use to correct use; there was also regression from correct use to variable use; but the majority again did not change in how they used the degrees of comparison.

Subordinate clauses

The same stages of development can be applied to students’ use of subordinate clauses. Tables 4.11 and 4.12 present the changes in patterns of use for respectively conditional and causal SubCs. For both types of SubCs, the picture that emerges is very similar to the one described for the DoC. The number of students that showed progression, either from T0 to T1 or from T1 to T2, was always substantially larger than the number of students that showed regression. But the majority of the students – invariably more than 50 percent – did not show any difference in the way they used subordinate clauses. Finally, the category that again was least stable was the ‘correct use’ category: although no errors were made at T0, a substantial amount of students returned to variable use at T1.

104 CHAPTER 4

TABLE 4.11 Changes in patterns of use of conditional subordinate clauses in the free written response task.



T0 No use

incorrect use

Variable use

Correct use

No use 5 3 2 1 - Only incorrect use 41 1 4 6 3 Variable use 45 - 4 15 16 Only correct use 10 - - 11 35

Use at T1


No use 4 2 - - - Only incorrect use 27 1 1 4 2 Variable use 37 - 2 6 15 Only correct use 8 - - 6 38

TABLE 4.12 Changes in patterns of use of causal subordinate clauses in the free written response task.



T0 No use

incorrect use

Variable use

Correct use

No use 6 1 2 - 2 Only incorrect use 14 2 22 16 1 Variable use 35 1 3 35 6 Only correct use 46 2 - 4 4

Use at T1


No use 2 3 - - 1 Only incorrect use 7 1 5 12 - Variable use 23 1 1 36 6 Only correct use 44 3 - 2 5

FINDINGS 105

4.2.4 Conclusion

Does correct use depend on the realization of the target structure? The question wase addressed for each target structure separately. Characteristic of the use of the DoC was that the frequency of use in the free written response task increased from T0 to T2. When total use scores were broken down according to correct and incorrect use, then incorrect use of the DoC turned out to quite infrequent. Correct use was more frequent, and a clear increase in correct use was observed. Productivity, the number of different adjectives the DoC were used with, also showed a clear increase from T0 to T2. Correct use of the DoC was investigated in three different contexts. No differences have been observed in correct use of comparatives or superlatives, nor have clear differences been found based on the adjectives with which the DoC were used. However, correct use of comparatives did seem to be affected by the presence of comparative clauses. With respect to this latter finding, it is important to point out that omission of the comparative suffix in the presence of a comparative clause can be clearly marked as incorrect. In contrast, omission of this suffix in all other contexts would go unnoticed, because the grammaticality of the sentence is not affected (see also 3.6.4). For this reason, the difference in the percentages of correct use of comparatives with and without a comparative clause may be the result of unnoticed errors coded as no use. The conclusion must therefore be that it has not been demonstrated beyond doubt that there is differentiation in correct use according to the structure’s realization.

The mean use of SubCs remained constant from T0 to T2. However, there was a clear increase in mean correct use and a decrease in mean incorrect use. For SubCs, the structure’s realization does seem to make a difference. Especially the difference in percentages of correct use between conditional and causal SubCs is remarkable: conditionals were used correctly much more. Further differentiation according to verb phrase complexity also revealed differences in correct use between phrases consisting of auxiliaries or single verbs on the one hand and phrases consisting of multiple verbs on the other. Auxiliaries were easiest to use; phrases containing multiple verbs were clearly more difficult to use correctly. Similarly, when differentiating for clause complexity, a difference in correct use was observed between relatively simple and more complex clauses. Percentages of correct use were lower as clause complexity increased. The most pervasive difference was observed between conditional and causal SubCs, though. For this reason, their co-occurrence was investigated as well. This clearly indicated that L2 learners started using conditionals correctly before causals, and correct use of

106 CHAPTER 4

causal SubCs seemed to preclude the possibility that conditionals were used incorrectly.

In order to gain insight into how fast the participants of this study developed the ability to use the target structure, students were categorized according to four developmental stages: preemergence, emergence, postemergence and targetlike behaviour. The analyses have shown that the participants of this study did not show much progress in terms of these stages. The participants were found to be very persistent in how they used the target structures, and most did not change their behaviour from one point of time to the next. As the stages used were not structure-specific, they may simply have been too crude to capture progress. Nevertheless, as this study spanned a period of three to four months, it seems justified to conclude that the ability to use the two grammatical phenomena under study in spontaneous situations develops slowly.

4.3 FFI, explicit knowledge, and grammatical development

4.3.1 Introduction

The analyses presented in this section pertain to the second research question:

RQ 2: How are explicit and implicit FFI related to the development of explicit and implicit grammatical knowledge?

The expectation was that explicit instruction is superior only in promoting the development of explicit grammatical knowledge. For the development of implicit grammatical knowledge, no differences were expected for the two types of instruction. These hypotheses were tested by means of univariate repeated measures ANOVA’s for two different measures of grammatical progress. The untimed grammaticality judgement (GJ) task scores were used as an indication of explicit progress, while correct use of the target structures in the free written response task was taken to be an indicator of implicit progress. An interaction was sought between instruction (Instr), operationalized as a three-level between subjects factor (EI, II and NI); and progress, also a three-level within subjects factor (T0, T1, and T2). The analyses were performed separately for each target structure, and each measure of progress.

FINDINGS 107

4.3.2 FFI and the development of the degrees of comparison

Preliminary analyses

For each target structure at each time of measurement, means and standard deviations have been computed. Table 4.13 depicts these. As pointed out in 3.4.3, for the degrees of comparison, there were insufficient control group data because a number of control group participants received explicit instruction in the degrees of comparison. For this reason, the analyses for the degrees of comparison were limited to a comparison of the EI and II conditions. The pre-test data of both measures were submitted to a one-way analysis of variance to test for significant differences between the groups at the start of the experiment. No significant differences were found. In addition, the three groups were tested for significant differences on any of the general proficiency measures (see 3.4) and ID variables (see 3.5). No differences were found, which is an indication that the participants were randomly distributed across the conditions.

As no significant differences are expected between instruction and correct use in the free written response, sufficient statistical power to detect interaction effects is an important design requirement. Assuming equal group sizes of 32 per group for the sake of power analysis, the 2 by 3 repeated measures design tested in this subsection reaches a power of 1.0 to detect medium within-subjects effects and medium interaction effects (f = .25, according to Cohen's (1988) standards).

Main analyses

For each measure, a 2 by 3 ANOVA was run, comparing only the EI and II groups. The results of these analyses can be found in Table 4.14. The table shows that progress over time was significant for both the explicit and the implicit

TABLE 4.13 Degrees of comparison: means and standard deviations for GJ task scores and correct use in the free written response task.

Measure Instr n T0

M (SD) T1

M (SD) T2

M (SD) EI 32 2.44 (2.08) 3.50 (1.93) 3.78 (1.98) GJ task II 35 1.94 (1.78) 2.06 (2.18) 2.49 (2.02)

EI 32 5.44 (4.79) 7.53 (5.91) 7.25 (4.79) Correct use II 35 5.09 (4.72) 6.66 (4.54) 7.14 (4.37)

108 CHAPTER 4

TABLE 4.14 Degrees of comparison; Repeated measures ANOVA for progress and

instruction Source SS df MS F p. GJ task a

progress 30.36 1.65 18.45 13.38 .00 progress x Instr 8.71 1.65 5.29 3.39 .03 Error 147.45 106.96 1.38

Correct Use progress 158.52 2 79.26 12.84 .00 progress x Intr 5.13 2 2.57 .42 .66 Error 802.42 130 6.17

a – Because the assumption of sphericity was violated, the Greenhouse-Geisser estimate is reported for this measure.

knowledge measure. However, for Correct Use, there was no interaction between progress and the type of instruction received. Only the grammaticality judgement (GJ) task scores were affected by the instruction provided: explicit instruction led to higher scores on the GJ task. These results will be discussed in more detail based on Figure 4.1, which illustrates per measure how the groups scored at each time of measurement. In addition, the main and interaction effects found will be explored further by means of post hoc analyses.

As pointed out in 3.4.3, this study suffered from considerable mortality in the subject sample from T1 to T2. To assess whether this loss of subjects may have affected the outcome, a 2 by 2 ANOVA has been run as well, in which the within subjects factor progress only had two levels, T0 and T1. The results were the same: significant overall progress was observed for both the GJ task and Correct Use, while an interaction between progress and instruction was found only for the GJ task scores. Thus, mortality does not seem to have affected these results.

Figure 4.1A indicates that the main effect for progress on the GJ task is mainly due to the progress obtained by the EI group: the II group showed marginal progress only, which would probably not differ much from a NI group. Post hoc analyses investigating progress for each group individually support this observation. Paired samples t-tests revealed that the only significant progress was obtained by the EI group between T0 and T1(t(31) = -3.14, p < .01). Post hoc analyses have also been used to investigate the interaction between progress and instruction. One-way ANOVA’s revealed significant differences between the two

FINDINGS 109

FIGURE 4.1 Mean scores for the grammaticality judgement task and correct use in the free written response task at each time of measurement.

A: GJ task mean scores: B: Correct use mean scores:

TABLE 4.15 Explicit DoC knowledge: post hoc pairwise comparisons at each time of measurement.

Time of measurement

Observed difference a

F

Effect size (in d)

T0 EI = II - - T1 EI > II F(1,65) = 8.14 .70 T2 EI > II F(1,65) = 7.01 .64 a – significant differences at alpha = .05 and using the Bonferroni adjustment are

indicated by > or <.

groups at both T1 and T2, the EI significantly outperforming the II group at both times. In Table 4.15, these analyses are reported along with the effect sizes in case of significant differences.

Figure 4.1B shows that progress in correct use was similar for the EI and II groups. It also suggests that progress was obtained primarily between T0 and T1. From T1 to T2, which actually spans a longer time period, the students obtained only marginal progress. This observation was tested and confirmed by means of paired samples t-tests. Progress was significant from T0 to T1 (correct use: t(66) = -3.98), but no significant progress was obtained between T1 and T2. Overall, the size of the effect from T0 to T2 was moderate (d = .41 ) in terms of Cohen’s (1988) standards.

0

1

2

3

4

5

6

EI II

T0

T1

T2

0

2

4

6

8

10

EI II

T0

T1

T2

110 CHAPTER 4

4.3.3 FFI and the development of subordinate clauses


In this subsection, the same analyses are presented as in the previous subsection, but this time measures of SubC progress are used. Table 4.16 summarizes the means and standard deviations for each measure at each time of measurement. One-way ANOVAs were used to assess for both measures whether the groups differed significantly at T0. No significant differences were observed. In addition, the three groups were tested for significant differences on any of the general proficiency measures (see 3.4) and ID variables (see 3.5). Again, no differences were found. Although the NI group is small, the current 3 by 3 repeated measures design still reaches sufficient power of to detect medium within-subjects and interaction effects: if all groups consisted of nine participants, the power to detect an interaction effect size of f = .25 would still be .98.

Main analyses

For each measure, 3 by 3 ANOVA was run in order to compare the progress obtained by the three different instruction groups. The results are displayed in Table 4.17. The picture is similar to that of the DoC: main effects were observed for both the explicit and implicit measures, but an interaction between progress and instruction was found only for the GJ task. Thus, the instruction does not TABLE 4.16 Subordinate clauses: means and standard deviations for GJ task scores

and correct use in the free written response task.

Measure Instr N T0

M (SD) T1

M (SD) T2

M (SD) EI 35 2.17 (2.41) 3.86 (2.64) 3.63 (2.69) II 32 1.44 (1.72) 2.44 (2.26) 3.19 (2.47)

GJ task

NI 9 1.00 (0.71) 0.56 (0.88) 0.67 (0.87)

EI 35 5.94 (5.75) 7.03 (5.52) 7.54 (4.90) II 32 6.09 (5.42) 8.19 (5.39) 9.25 (5.94)

Correct use

NI 9 4.11 (4.96) 3.56 (4.28) 6.00 (6.12)

FINDINGS 111

TABLE 4.17 Subordinate clauses; Repeated measures ANOVA for progress and

instruction

Source SS df MS F p. GJ task

progress 26.68 2 13.34 6.60 .00 progress x Instr 26.93 4 6.73 3.33 .01 Error 295.17 146 2.02

Correct use a progress 131.07 1.75 75.04 9.33 .00 progress x Instr 40.66 3.49 11.64 1.45 .23 Error 1026.01 127.51 8.05


seem to affect performance on the free written response task. Only the GJ task scores are affected by the instruction. Mortality was again not found to have affected the results presented here. Using a two-level factor for progress, thus increasing the subject sample to 101, the same outcome was obtained: the loss of subjects between T1 and T2 is not likely to have affected the results substantially.

The analyses presented in section 4.2.2 have shown that the participants of this study made a differentiation in correct use of conditional and causal SubCs. In order to test the possibility that the instruction affected the use of either conditional or causal SubC development separately, the analyses have also been conducted for correct use of each type of SubC. The results yielded the same picture: for both types of SubCs, main effects were found, but no interaction between correct use and instruction. The results presented in Table 4.17 will be further explored by means of figure 4.2 and post hoc analyses, without making the differentiation between conditional and causal SubCs.

For the GJ task, a main effect and an interaction between progress and instruction were observed. Figure 4.2A clearly suggests that the overall main effect for time was caused by both experimental groups. The control group simply did not show progress on grammaticality judgement performance. Again, paired samples t-tests have been used to examine progress for each group individually. Significant progress was obtained by the experimental groups progress between T0 and T1 (for the EI group: t(34) = -4.89, p < .01, and for the II group (t(31) = -3.09, p < .01). This is an indication that little additional progress was obtained after the subjects stopped receiving instruction.

112 CHAPTER 4

FIGURE 4.2 Mean scores for the grammaticality judgement task and correct use in the free written response task at each time of measurement.

A: GJ task mean scores: B: Correct use mean scores:

TABLE 4.18 Explicit SubC knowledge: post hoc pairwise comparisons at each time of measurement.

Time of measurement

Observed difference a

F

Effect size (in d)

T0 EI = II = NI - - T1 EI = II

EI > NI II = NI

- F (1,42) = 13.52

-

- 1.37

- T2 EI = II

EI > NI II > NI

- F (1,42) = 10.46 F (1,39) = 8.93

- 1.21 1.13

a – significant differences at alpha = .05 and using the Bonferroni adjustment are indicated by > or <.

The interaction found between progress and instruction for the GJ task

indicates differentiated progress across the conditions. Figure 4.2A also clearly demonstrates that this interaction was to a large extent caused by the control group behaving differently. Indeed, post hoc pairwise comparisons using the bonferroni adjustment reveal a number of significant differences. Table 4.18 summarizes the outcomes. Before the instruction, no significant differences were

0

2

4

6

8

EI II NI

T0

T1

T2

0

2

4

6

8

10

12

EI II NI

T0

T1

T2

FINDINGS 113

observed between the groups. Immediately after the instruction, the EI group significantly outperformed the NI group. At the third time of measurement, the II group significantly outperformed the NI group too.

Main effects were observed for Correct Use. These effects were further explored by means of paired samples t-tests. From T0 to T1, the progress that was overall obtained was significant (correct use: t(75) = -3.29, p < .01). This also applied to progress between T1 and T2: (correct use: t(75) = -2.55, p < .01). Thus, subjects increased their implicit knowledge during the instruction, and continued to do so in the two months following. The size of the effect from T0 to T2 was again moderate (d = .42).

4.3.4 The interface between explicit and implicit knowledge

Introduction

The analyses presented in the previous subsections (4.3.2 and 4.3.3) have clearly demonstrated for both target structures that received explicit instruction did not present an advantage to students for developing implicit grammatical knowledge. This was notwithstanding the fact that the groups differed with respect to their explicit knowledge of the target structures. Thus, this study does not lend support to an interface between explicit and implicit knowledge; and similarly, it does not provide arguments in favour of explicit FFI. However, it may be that L2 learners differ in their ability to handle and benefit from explicit instruction, and that this is only given to some, not all L2 learners. In other words, it is possible that the presence of an interface between explicit and implicit knowledge is obscured in these data by those L2 learners that did not benefit from the explicit instruction. To investigate this, the analyses were rerun, comparing an ‘explicit knowledge’ (EK) group with a ‘no explicit knowledge’ (NEK) group.

The EK and NEK groups were created taking the EI and the II groups as a starting point. From the EI group, all participants were excluded that did not show an explicit knowledge increase of over 1 standard deviation, creating a group of explicitly instructed participants that actually benefited substantially from this instruction. In doing so, 14 out of the original 32 subjects that received explicit DoC instruction were lost. This means that 44 percent of the subjects was not able to benefit much from the instruction. For SubC, even fewer participants were able to benefit from the explicit instruction: 20 out of 35 subjects had to be removed from the subject sample: a loss of 57 percent of the subjects. There were

114 CHAPTER 4

no reasons to eliminate subjects from the II conditions: they were not changed, but simply renamed NEK.

One-way ANOVAs were performed to test whether the groups differed significantly at T0, which would introduce a bias in the analyses. No significant differences were found, and the data were considered ready for the intended analyses. Also, despite the loss of subjects for these analyses, the statistical power is still sufficient to detect medium effect sizes.

Interfaces of knowledge

In Table 4.19, the mean scores and standard deviations are presented for both the explicit and implicit measure of grammatical knowledge and broken down according to time of measurement and target structure. Then, two by three univariate analyses of variance were again run for each measure of grammatical development using Condition as the between-subjects factor and progress as the within-subjects factor. Obviously, as the two contrasted groups were set to differ with respect to explicit knowledge, significant interactions were found for the GJ task for both target structures (for DoC: F(2, 98) = 20.21, p < 0.01; and for SubC: F(2, 82) = 24.73, p < 0.01; see also Table 4.20). Quite remarkable, though, was that the explicit knowledge of SubCs for the EK group of learners was not stable, given the deteriorated performance on the GJ task at T2 (see Table 4.19).

TABLE 4.19 Means and standard deviations for both Doc and SubC GJ task scores

and correct use.

Measure Condition N T0

M (SD) T1

M (SD) T2

M (SD) Degrees of comparison

EK 18 2.06 (1.92) 4.33 (1.53) 4.33 (1.78) GJ task NEK 35 1.94 (1.78) 2.06 (2.18) 2.49 (2.02)

EK 18 6.17 (3.85) 9.06 (5.70) 8.06 (4.53) Correct use NEK 35 5.09 (4.72) 6.66 (4.54) 7.14 (4.37)

Subordinate clauses GJ task EK 15 1.53 (1.46) 5.07 (1.79) 3.73 (2.89) NEK 32 1.44 (1.72) 2.44 (2.26) 3.19 (2.47)

Correct use EK 15 5.53 (4.22) 7.93 (4.95) 7.80 (4.78) NEK 32 6.09 (5.42) 8.19 (5.38) 9.25 (5.94)

FINDINGS 115

TABLE 4.20 Repeated measures ANOVA for progress and instruction Measure and source of variation SS df MS F

p.

Degrees of comparison GJ task a

progress 54.92 1.75 31.47 26.93 .00 progress x cond. 31.20 1.75 17.88 15.30 .00 Error 104.01 89.00 1.17

Correct use progress 141.57 2 70.79 10.53 .00 progress x cond. 15.74 2 7.87 1.17 .31 Error 685.59 102 6.72

Subordinate clauses GJ task

progress 124.24 2 62.12 26.39 .00 progress x cond. 37.31 2 18.66 7.93 .00 Error 211.82 90 2.35

Correct use progress 171.80 2 85.90 11.40 .00 progress x cond. 7.88 2 3.94 0.52 .60 Error 678.36 90 7.54


For the Correct Use measure, significant interactions between progress and

condition were not observed for either target structure (see Table 4.20). The analyses did turn out significant main effects for all measures. The conclusion must be, though, that even the L2 learners that benefited substantially from the explicit instruction did not demonstrate any advantage in their implicit grammatical performance.

4.3.5 Conclusion

The analyses presented here were intended to test two expectations: (1) explicit instruction is superior only in promoting the development of explicit grammatical knowledge; (2) there will be no differences between the two types of

116 CHAPTER 4

instruction when it comes to the development of implicit grammatical knowledge. Both hypotheses were confirmed.

Explicit knowledge as measured by the grammaticality judgement task was found to be promoted most by explicit instruction: this finding applied to both target structures. However, there were some differences in explicit development between the two structures. The first difference pertained to the effect of the implicit instruction. According to these data, implicit instruction did not affect explicit DoC development, while it did effectively promote explicit SubC development. Post hoc analyses have demonstrated for the DoC that, at T1 and T2, the EI group outperformed the II group. However, for SubCs, the difference between the EI and II conditions on the GJ task scores was never significant, but both significantly outperformed the control group at T2. A second difference in explicit knowledge gain between the two target structures pertained to the size of the effect. According to Cohen’s standards, moderate to large effect sizes were observed between the EI and II groups on the explicit DoC scores. As pointed out, for the SubCs, there were no differences between the two experimental groups, but both outperformed the NI group, and the effect sizes are large. The results suggest that the instruction more effectively promoted explicit knowledge of the SubC than that of the DoC. However, these effect sizes need to be interpreted with some caution, as there was no control group reference for the DoC.

A finding that applied to both explicit and implicit knowledge and to both target structures is that most progress was obtained between T0 to T1. Mostly, the progress obtained between T1 and T2 was small and insignificant, and sometimes slight regression could be observed.

All in all, the hypothesis that explicit knowledge of a particular structure promotes the ability to use that structure in spontaneous situations of language use has to be rejected. No advantage was observed for either structure on the implicit measure. Even when the data where manipulated in order to compare groups that differed substantially in explicit knowledge, no effect of explicit instruction was found. Consequently, this study provides no evidence of an interface between explicit and implicit knowledge.

FINDINGS 117

4.4 FFI and learner characteristics

4.4.1 Introduction

RQ 3: How do developmental readiness, L1 similarity, and Individual Differences affect the success explicit and implicit FFI?

This section is concerned with how learner characteristics affect explicit and implicit progress, and whether they interact with the type of instruction received. A number of individual traits have been considered: developmental readiness, L1 similarity, aptitude, cognitive style, motivation and age. For developmental readiness and L1 background, operationalized as two-level between-subjects factors, 2 by 2 univariate ANOVAs will be used to test how they affect progress individually and in interaction with instruction. Correlation analyses will be used to assess the degree of relationship between the remaining ID variables and explicit and implicit grammatical progress. The overall aim of this section is to find out which variables are related to explicit and implicit progress, possibly in interaction with the kind of instruction received.

Gain scores for the GJ task and correct use in free written response task will be used as indicators of explicit and implicit grammatical progress, respectively. These gain scores are effectively operationalized as the standardized residuals of pairs of observations. Using standardized residuals instead of raw gain scores (or fitted residuals) offers the advantage of having a measure of gain that is standardized. In addition, it is independent of the unit of measurement and differences in scale (which are different for the all the measures used). For explicit knowledge, the gain scores reflect how much progress students obtained from T0 to T1, because explicit knowledge was found to be affected by the instruction. Any progress obtained after the instruction stopped may therefore not be the result of the explicit learning mechanisms intended to monitor. For implicit knowledge, progress from T0 to T2 is investigated. Table 4.21 provides an overview of the different gain scores used as dependent variables in this section.

Subsections 4.3.2 and 4.3.3 focus on developmental readiness and L1 similarity, respectively. In 4.3.4, the relation between knowledge gain and aptitude, cognitive style, motivation and age is investigated.

118 CHAPTER 4

TABLE 4.21 The composition of the explicit (EK) and implicit (IK) knowledge gain

scores. GAIN SCORE EXPRESSES: N EK DoC gain Standardized gain score obtained on the

grammaticality judgement task for the DoC from T0 to T1.

101

EK SubC gain Standardized gain score obtained on the grammaticality judgement task for SubCs from T0 to T1.

101

IK DoC gain Standardized gain score obtained on the free written response task for the DoC from T0 to T2.

76

IK SubC Gain Standardized gain score obtained on the free written response task for SubCs from T0 to T2.

76

IK cond. SubC Gain Standardized gain score obtained on the free written response task for conditional SubCs from T0 to T2.

76

IK causal SubC Gain Standardized gain score obtained on the free written response task for causal SubCs from T0 to T2.

76

Overall EK gain Composite standardized gain score of the DoC and SubC GJ task gain scores: (EK DoC + EK SubC) / 2 from T0 to T1.

101

Overall IK gain Composite standardized gain score of the DoC and SubC GJ task gain scores: (IK DoC + IK SubC) / 2 from T0 to T2.

76

4.4.2 Developmental readiness


This section investigates by means of univariate ANOVA whether explicit and implicit gain depend on developmental readiness (DR); and – if possible – whether there is an interaction with the type of instruction received. In 3.5.2, developmental readiness has been operationalized: L2 learners may be considered

FINDINGS 119

TABLE 4.22 Developmental readiness; number of observations, mean gain scores (standardized, see 4.4.1), and standard deviations for explicit and implicit knowledge, and for both target structures.

Factor EI II (+ NI) a Total DR DoC N M (SD) N M (SD) N M (SD) EK DoC gain

Ready 33 0.50 (1.05) 34 -0.25 (0.96) 67 0.12 (1.07) Unready 8 -0.46 (0.37) 6 -0.72 (0.53) 14 -0.57 (0.45)

IK DoC gain Ready 27 -0.06 (0.93) 37 0.47 (1.05) 64 0.18 (0.97) Unready 5 -0.64 (0.61) 7 -0.17 (0.48) 12 -0.96 (0.36)

DR conditional SubC EK SubC gain

Ready 32 0.42 (1.08) 34 0.02 (0.98) 66 0.21 (1.04) Unready 8 0.04 (1.35) 7 -0.39 (0.62) 15 -0.16 (1.06)

IK cond. SubC gain Ready 29 -0.07 (0.72) 36 0.08 (1.09) 65 0.01 (0.94) Unready 6 0.55 (0.98) 5 -0.83 (1.38) 11 -0.08 (1.33)

DR causal SubC EK SubC gain

Ready 23 0.50 (1.07) 22 0.05 (1.16) 45 0.28 (1.12) Unready 17 0.12 (1.21) 19 -0.17 (0.59) 36 -0.03 (0.93)

IK causal. SubC gain Ready 21 -0.27 (1.08) 24 -0.02 (0.97) 45 0.14 (1.07) Unready 14 0.27 (1.02) 17 -0.02 (0.84) 31 -0.02 (0.89)

a – For implicit gain scores, the II and NI groups were merged. For explicit gain scores, the NI group was not included.

developmentally ready when they start using the target structure correctly. Naturally, the analyses will be performed separately for the degrees of comparison and subordinate clauses. In addition, given the findings in 4.2, a differentiation will be made for conditional and causal subordinate clauses. Thus, three DR factors have been defined, related to the use of the DoC, conditional SubCs, and causal SubCs: DR DoC, DR cond. SubC, and DR causal SubC.

Making such a division, however, leads to a very uneven distribution of observations over the unready and ready groups. Table 4.22 provides the number of observations for both explicit gain and implicit gain when a division according

120 CHAPTER 4

to readiness is made. For explicit gain, the students in the NI condition were not included in the analyses, simply because the number of observations per cell would be too small. Given that instruction has not been found to affect the development of implicit knowledge (see 4.3), the II and NI groups were merged for the analyses pertaining to implicit gain. What stands out is that the number of observations for the unready groups is quite small, especially for the factors DR DoC and DR cond. SubC. The question that rises is whether there is sufficient statistical power to detect main and interaction effects. In order to obtain sufficient power (.80) to detect large main effect sizes (f = .40 according to Cohen’s (1988) standards), a total sample size of 52 is needed, or 26 per group. This means that the total sample size for the investigation of both explicit and implicit gain is adequate, but the total unready group sizes for both DR DoC and DR cond. SubC are too small (see Table 4.22, last column). The same conclusion must be drawn for testing the interaction between developmental readiness and instruction. Detecting large interaction effects (f = .40) at a power of .80 requires a total sample size of 52. In a 2 by 2 design as employed here, this means that each group should contain 13 subjects, a requirement that is frequently not met. Despite the small unready group sizes, the intended analyses will be performed. One should keep in mind, though, that if significance fails to occur, this may simply be due to a lack of statistical power.

One question that was addressed before the model proposed here was tested, was whether developmental readiness may be considered an independent predictor of grammatical development. In order to assess this, one-way ANOVAs were conducted for each of the three DR factors with a number of ID and control subject variables. Table 4.23 summarizes the findings. The results show that dividing students into groups according to beginning correct use of the target structure introduced differences between the groups compared. For all three factors, the ready L2 learners outperformed the unready learners on Grammatical Accuracy and the C-test. For Dr DoC, the ready students also did better on Fluency and the Cito ISK test. In addition, L2 learners that were ready for conditional SubCs were significantly younger; and the learners that were ready for causal SubCs did better on the grammatical sensitivity task.

It seems, then, that dividing students into groups according to developmental readiness introduces differences between the groups compared. But perhaps such differences are to be expected, as the construct of developmental readiness expresses ‘advancedness’ of the L2 learners’ linguistic systems. Ready learners were indeed consistently found to outperform unready learners. As these

FINDINGS 121

TABLE 4.23 One-way ANOVAs for the developmental readiness factors and a number of subject variables.

Developmental readiness factor

Factor DR DoC

F (df b,w), p. DR Cond. SubC

F (df b,w), p. DR Causal SubC

F (df b,w), p. ID variables -

Memory - - - Gram. sens. - - 5.91 (1,97) * Age - 6.01 (1,98) * - TJ cogn. style - - - TJ motivation - - -

Proficiency variables

Fluency 9.30 (1,99) ** - - Gram. Acc. 9.85 (1,99) ** 30.93 (1,99) ** 15.70 (1,99) ** Spelling - - - Cito ISK 6.46 (1,99) * - - C-test 12.03 (1,97) ** 8.68 (1,97) ** 5.58 (1,97) *

* significant at alpha < .05 ** significant at alpha < .01

differences may well be inherent to the construct of DR, they will not be corrected for in the analyses presented here.

Main analyses

Using developmental readiness and instruction as two-level between-subjects factors, univariate 2 by 2 ANOVAs have been run to evaluate how DR affects overall explicit and implicit gain, and how DR interacts with instruction. Table 4.23 shows that, for both explicit and implicit DoC gain, the model predicts a significant amount of between-group variance. For explicit DoC gain, a significant main effect has been found for developmental readiness, which indicates that students that were developmentally ready have obtained more gain than students that were unready. The factor instruction just misses significance in this analysis. The interaction between developmental readiness and instruction is also not

122 CHAPTER 4

significant. For implicit DoC gain, developmentally ready L2 learners also significantly outperform the unready learners. As to be expected, no significance is observed for instruction. The interaction between DR and instruction is also not significant.

The same univariate ANOVAs have been performed for developmental readiness in relation to conditional and causal SubCs, but none of the models was found to predict a significant amount of between-group variance. For explicit SubC gain, irrespective of which DR factor was used, the model was not found to predict any of the variance between the groups: there were no main effects found, and there was also no interaction between developmental readiness and instruction. Because the NI group was not included in these analyses, the lack of significance for the factor instruction was expected (For SubCs, no differences were observed between EI and II; see 4.3.3). The same results were obtained for

TABLE 4.23 Degrees of comparison, univariate 2 by 2 ANOVA for instruction and developmental readiness

Measure and source of variation SS df MS

F

p.

EK DoC gain Corrected model 15.25 3 5.08 5.76 0.00

Intercept 2.41 1 2.42 2.74 0.10 DR DoC 5.87 1 5.87 6.64 0.01

Instruction 2.94 1 2.94 3.33 0.07 DR DoC * Instr. 0.67 1 0.67 0.76 0.38

Error 67.98 77 0.88 Total 83.24 81

Corrected total 83.24 80 IK DoC gain

Corrected model 13.82 3 4.61 5.51 0.00 Intercept 6.04 1 6.04 7.22 0.00

DR DoC 13.52 1 13.52 16.18 0.00 Instruction 0.00 1 0.00 0.00 0.96

DR DoC * Instr. 0.39 1 0.39 0.47 0.50 Error 60.18 72 0.84 Total 74 76

Corrected total 74 75

FINDINGS 123

both conditional and causal implicit SubC gain: irrespective of the DR factor used, the models did not predict the variance between the groups.

4.4.3 L1 similarity


Another individual characteristic that may interact with instruction is the learner’s L1. This section investigates whether learners with similar L1s have an advantage over learners with L1’s that are rather different from the L2. L1 similarity has been narrowly defined in specific relation to the two target structures used in this study (see 3.5.3), resulting in two two-level between-subjects factors: L1 sim DoC and L1 sim SubC. They express whether the target structures are similar or different from in the participants’ L1. Univariate ANOVAs will be used to investigate the interaction of these factors with the instruction received for both explicit and implicit gain. As before, the NI group will not be included in these analyses, due to the relatively small number of observations. Table 4.24 describes the input for these analyses. Both L1 similarity factors lead to even distributions across the groups. In order to achieve sufficient statistical power, the same requirements apply as in the previous subsection: 26 subjects per group are needed to detect large main effects, and 13 subjects per group to detect large interaction effects. As there is only one cell frequency below 13 (see Table 4.24), these requirements are sufficiently met.

One-way ANOVAs were performed to investigate whether the groups resulting from a division according to the similarity factors were still sufficiently comparable. Some significant differences were found: for L1 sim DoC, the groups differed significantly on the C-test (F(1,72 = 10.90; p < .01) and on grammatical accuracy (F(1,74) = 10.84; p < .01); for L1 sim SubC, significant differences were observed for aptitude (F(1,74) = 7.15; p < .01), the Cito ISK test (F(1,74) = 5.18; p < .05); and spelling (F(1,74) = 6.96; p < .01). These variables will be introduced as covariates in the following analyses in order to reduce the variance they introduce.

124 CHAPTER 4

TABLE 4.24 L1 similarity: number of observations, mean gain scores (standardized,

see 4.4.1), and standard deviations for explicit and implicit knowledge, and for both target structures.

Factor EI II Total L1 sim DoC N M (SD) N M (SD) N M (SD) EK DoC gain

Similar 17 0.39 (1.02) 15 -0.39 (1.19) 32 0.03 (1.15) Different 20 0.12 (0.94) 24 -0.28 (0.75) 44 -0.01 (0.85)

IK DoC gain Similar 15 0.63 (1.22) 12 -0.16 (0.89) 27 0.28 (1.14) Different 13 -0.55 (0.78) 22 0.04 (0.96) 35 -0.18 (0.93)

L1 sim SubC

EK SubC gain Similar 21 0.24 (0.98) 24 0.14 (1.13) 45 0.19 (1.04) Different 16 0.54 (1.37) 15 -0.32 (0.62) 31 0.09 (1.13)

IK SubC gain Similar 14 -0.30 (0.93) 21 0.44 (1.06) 35 -0.01 (0.97) Different 14 0.11 (1.06) 13 -0.11 (0.90) 27 -0.01 (1.00)

TABLE 4.25 IK DoC gain: univariate 2 by 2 ANOVA for instruction and L1 similarity Measure and source of variation SS df MS

F

p.

Corrected model 20.28 5 4.06 4.93 0.00 Intercept 1.47 1 1.47 1.78 0.19

C-test 4.34 1 4.34 5.27 0.02 Gram. accuracy 1.88 1 1.88 2.28 0.14

L1 sim DoC 0.12 1 0.12 0.14 0.71 Instruction 0.36 1 0.36 0.44 0.51

L1 sim DoC * Instr. 6.43 1 6.43 7.81 0.01 Error 46.10 56 0.82 Total 66.40 62

Corrected total 66.38 61

FINDINGS 125

Main analyses

For both explicit and implicit DoC gain, the overall corrected model was found to explain a significant amount of variation. For explicit DoC gain, a main effect was found for instruction; L1 similarity did not explain a significant amount of variance, nor did the interaction between instruction and L1 similarity. The significant main effect for instruction confirms the findings reported in .3.2. For implicit DoC gain, the results are presented in Table 4.25. The table shows that no significance was found for the individual factors, but the interaction between L1 similarity and instruction was significant: students with L1s that also realize the degrees of comparison by means of morphological suffixation do significantly better on the free written response task when they receive explicit instruction (see Table 4.25). The results did not change when the analysis was run without introducing the covariates.

For both explicit and implicit SubC gain, no significance was observed, irrespective of whether the covariates were introduced, or whether a differentiation was made for conditional and causal SubCs. Nevertheless, these analyses do provide evidence that explicit instruction can be more effective than implicit instruction in promoting implicit knowledge; but only for the DoC, and only if the L1 is similar.

4.4.4 Aptitude, cognitive style, motivation and age


This section focuses on how individual differences affect explicit and implicit progress, and whether there is an interaction with the kind of instruction received. Five relevant ID variables have been identified: two aptitude components – memory and grammatical sensitivity (GS), cognitive style, motivation, and age. In addition, overall aptitude was included, which is a composite score of memory and GS. The degree of relationship between these ID variables and explicit and implicit gain has been explored by means of correlation analyses. Again, gain scores operationalized as standardized residuals have been used as dependent variables. As the effect of the ID variables is evaluated in relation to the kind of instruction received, the NI subjects were excluded for the majority of the analyses presented here: the number of observations is too small to obtain trustworthy correlation coefficients.

126 CHAPTER 4

Preliminary analyses indicated that there were no significant differences between the groups on any of the ID variables at T0. Also, the degree of interrelationship between the ID variables has been assessed in order to evaluate the extent to which they measure the same constructs. Hardly any significant correlations were found. The two aptitude scores – memory and grammatical sensitivity – were found to be interrelated (R = .41, N = 81, p < .01). The only other correlation found was between grammatical sensitivity and cognitive style (R = -.23, N = 78, p < .05): this latter correlation indicates that students that are considered precise and independent by their teachers, do better on the GS task.

Main analyses

Correlation analyses have been used to investigate the degree of relationship between the ID variables and explicit and implicit gain obtained in the two instruction conditions. In Table 4.26, the Pearson product-moment correlation values are presented for both explicit and implicit DoC gain. For explicit gain, there is a significant negative correlation between gain in the EI condition and age, which means that younger students tend to profit more from the explicit instruction as measured by grammaticality judgements. If gain is measured by means of the free written response task, the EI scores correlate significantly with memory as measured by the paired associates test and overall aptitude. Given that instruction was not found to affect implicit knowledge in previous analyses, a difference between the EI and II conditions is unexpected. However, the correlations between the II scores and memory and aptitude are also rather high (.33 and .32): they just miss out on significance. In addition, due to the small number of observations, the differences between the EI and II correlations would need to be quite large to safely conclude that a variable is differently related to EI and II gain. No significant correlations have been found for grammatical sensitivity, cognitive style and motivation.

The same analyses have been performed for subordinate clauses. The picture that emerges for SubC gain is quite different from that of DoC gain. What stands out in Table 4.27 are the significant correlations of memory, grammatical sensitivity, and overall aptitude with the explicit EI gain scores. In addition, there is again a significant negative correlation between gain in the EI condition and age. However, this time explicit gain in the II condition also correlates negatively with age. Cognitive style shows a significant positive correlation with explicit II gain; and interestingly, there is a negative, although not significant, correlation

FINDINGS 127

TABLE 4.26 Pearson product-moment correlation coefficients for the degrees of comparison

Explicit Doc gain Implicit DoC gain ID variables

EI (N=41)

II (N=40)

EI (N=32)

II (N=35)

Memory .10 .12 .48** .33 Gram. sensitivity -.01 .30 .28 .20 Aptitude .05 .23 .45* .32 TJ Cognitive style .22 -.23 .06 .16 TJ Motivation .08 -.10 -.15 .16 Age -.36* -.18 .01 -.21

* Significant at p < .05 ** Significant at p < .01

TABLE 4.27 Pearson product-moment correlation coefficients for subordinate clauses Explicit SubC gain Implicit SubC gain ID variables

EI (N=40)

II (N=41)

EI (N=35)

II (N=32)

Memory .34* .29 -.08 -.16 Gram. sensitivity .42* .18 .19 -.12 Aptitude .45** .06 .05 -.16 TJ Cognitive style -.32 .36* -.16 -.06 TJ Motivation -.27 .02 -.27 -.07 Age -.41** -.32* -.25 -.17


with EI gain. The distances between the EI and II correlations are also quite large; at least for GS, Aptitude and Cognitive style, they may fall outside each others confidence intervals. Thus, for these three factors, there may be an interaction with the kind of instruction received. The implicit SubC gain scores do not correlate with any of the ID variables. Further differentiation according to conditional and causal SubCs also did not lead to any significant correlations.

It is difficult to abstract a coherent picture from the correlations presented in

tables 4.26 and 4.27. It seems fair to conclude that age is related to explicit gain, irrespective of the kind of instruction provided. Memory and aptitude seem to be

128 CHAPTER 4

related to DoC gain if measured by means of the free written response task. However, for SubC, the aptitude measures relate to explicit gain in the EI condition.

Individual differences and overall linguistic proficiency

An additional perspective can be obtained by relating the ID variables to a number of other L2 proficiency measures. Five measures are available: the Cito ISK test, the C-test test, Fluency, Grammatical accuracy, and spelling. The Cito ISK test and the C-test test proficiency in controlled contexts (see 3.4.4), while fluency, grammatical accuracy and spelling scores were calculated on the basis of the free written response task data. Table 4.28 shows the correlations. What stands out is that both the Cito ISK test, the C-test, and spelling correlated significantly with all three aptitude measures, while fluency and grammatical accuracy did not (except for a weak correlation between GS and grammatical accuracy). This suggests that aptitude as measured in this study is particularly important in contexts of controlled L2 use. The only other correlation that appeared is the one between cognitive style and spelling: students that were considered structured and precise by their teachers did better on spelling. Motivation and age have not been found to correlate with any of the proficiency measures.

TABLE 4.28 Pearson product-moment correlation coefficients for subordinate clauses Proficiency measure ID variables

Cito ISK (N=40)

C-test (N=41)

Fluency (N=35)

Gram. ac. (N=32)

Spelling (N=35)

Memory .28* .39** .01 -.13 -.52** Gram. sensitivity .41** .52** -.07 -.22* -.25* Aptitude .41** .53** -.03 -.21 -.47** TJ Cognitive style -.12 -.07 .02 -.20 .33** TJ Motivation -.07 -.15 -.15 -.02 .13 Age .16 -.06 -.18 .14 .11


FINDINGS 129

4.4.5 Conclusion

The analyses in this section have focused on how differences between learners affect the development of explicit and implicit grammatical development, and whether this effect is related to the kind of instruction received.

The results presented here lend some support to the hypothesis that developmental readiness is related to effective FFI, in that developmental readiness was found to affect both explicit and implicit progress in the degrees of comparison. Students who already use correct forms of the DoC obtained significantly more progress than students who used no or incorrect forms of the DoC. No interaction could be demonstrated with the kind of instruction students received, either because there is no interaction, or because of the small unready groups causing insufficient statistical power to even detect large effects. For SubC, no evidence was found to suggest that developmental readiness is related to progress. Again, for conditional SubCs, this failure of significance to occur may be due to a lack of statistical power. It should also be noted that developmental readiness was found to be related to a number of proficiency measures, most notably grammatical accuracy and the C-test.

The relation between FFI and L1 background was investigated in a similar way. L1 similarity was found to interact with instruction for implicit DoC gain. Students whose L1 also realizes the degrees of comparison by means of morphological suffixation and who received explicit instruction were found to outperform students that received implicit instruction and students whose L1 is different. This finding is remarkable, because it actually constitutes evidence in favour of an interface between explicit and implicit knowledge. No other significant effects were found for L1 similarity.

In the last subsection, the relation of a number of ID variables with explicit and implicit progress was explored, again in relation to the kind of instruction received. The variables memory, grammatical sensitivity, overall aptitude, cognitive style, motivation and age have been considered. Progress in the DoC was found to be related to memory and overall aptitude, but only if measured implicitly. Explicit progress in the explicit condition was related to age: the younger the students, the more they benefited from the explicit instruction. For SubC, the aptitude measures were related to explicit progress in the EI condition. Cognitive style was related to progress in the II condition, and age was related to progress in both conditions. There were no correlations with implicit progress. In short, age seems to be related to explicit progress of both target structures. For the remainder, progress in the DoC and SubC did not have correlations in

130 CHAPTER 4

common. All in all, the analyses regarding the ID variables did not provide a coherent picture.

4.5 Summary: differences in the use of DoC and SubC

4.5.1 Introduction

The final research question to be addressed is:

RQ 4: Does the effectiveness of explicit and implicit FFI depend on the complexity of the instructed grammar structure?

In design, this study incorporated two identical, parallel experiments in which two contrasting grammar structures were taught. The motivation for this setup was to put the idea to the test that instruction may have different effects, depending on the complexity of the structure to be taught. Based on previous research, which seemed to suggest a contrast between morphological and syntactic structures, the degrees of comparison and subordination were chosen as targets. As noted before, this contrast was not operationalized as a statistical factor in this research. Rather, all analyses have been conducted and presented separately. This section explores to what extent differences were observed between the two structures with respect to the development of use (4.5.2), the effect of FFI on grammatical development (4.5.3), and the relation between grammatical progress and learner characteristics (4.5.4). In this section, then, an answer to the above mentioned research question will be sought by means of a review of the previously presented findings. As such, this section reads as a summary.

4.5.2 Differences in characteristics of use

In 4.2, a more descriptive analysis of the free written response data was presented, with the specific purpose to explore how second language learners start to make use of the two grammar structures in focus. For both structures, a number of relevant types of realizations were defined, and correct use of the structure was considered in relation to the realizations. For both structures, substantial progress over time was observed in terms of correct use: this observation applied to all contexts of use that were considered. However, if correct use was considered in relation to the different realizations, then there did

FINDINGS 131

seem to be an important contrast. For the DoC, the percentages of correct use were more or less equal across all the different realizations, except when a comparative clause was realized. For the SubCs, though, clear differentiation was observed between the different realizations. The percentages were substantially higher for conditional subordinate clauses as compared to causal subordinate. Similar contrasts were observed for subordination realized with SV (single verb) verb phrases and MV (multiple verbs) verb phrases, and for simple constituent structure clauses and complex constituent structure clauses. Correct use of the SubCs, then, was affected by their realizations, while correct use of the DoC was more or less unaffected by their realizations.

In order to gain insight into how L2 learners move from not using the two target structures to correct use, they were considered in relation to four patterns of use: no use, incorrect use, variable use, and correct use. All participants were classified according to these patterns at each time of measurement. The picture that emerged was very similar for both structures: the participants were found to be quite persistent in how they used the target structures. Over fifty percent of the students did not demonstrate progress in terms of moving from one stage to the next. Also similar for both structures was that a relatively large number of students who demonstrated only correct use at one point, returned to variable use. The conclusion was that for both structures, learning to them in spontaneous situations involves extended periods of incorrect use.

4.5.3 FFI, grammatical development, and structure complexity

In 4.3, two hypotheses were tested by means of univariate repeated measures ANOVA’s: explicit instruction (EI) is more effective in promoting explicit knowledge; and explicit and implicit instruction (II) will be equally effective in promoting implicit knowledge. For subordination, the NI condition was also included in the analysis. Comparing the outcomes the analyses, minor differences were observed for the structures.

The analyses were aimed at investigating interactions between instruction (EI, II (and NI)) and progress over time (T0, T1, and T2). They were performed for both measures of progress: the grammaticality judgement task (GJ task) and correct use in the free written response task. For the GJ task, significant overall progress was observed for both structures, as well as an interaction between progress and instruction. However, there were differences between the two structures, which are illustrated by means of Table 4.29. The table shows that at

132 CHAPTER 4

TABLE 4.29 Explicit knowledge: post hoc pairwise comparisons at each time of measurement.

Time of measurement Degrees of Comparison

Subordination

T0 EI = II EI = II = NI T1 EI > II EI = II

EI > NI II = NI

T2 EI > II EI = II > NI

T0, there were no significant differences between the groups. Some differences surface at T1: for the DoC, the EI group was found to have more explicit knowledge than the II group, and this difference persisted at T2. These findings are according to expectation. However, for the SubCs, there are never significant differences between the EI and II group. The progress that was obtained by the EI group was all obtained between T0 and T1, which is why only the EI group differs significant from the NI group at this point. The II group obtained slow but gradual progress, and by T2, they had caught up with the EI group, and both groups significantly outperform the NI group. In sum, implicit instruction led to significant gains in explicit SubC knowledge, but not in explicit DoC knowledge.

The results concerning the interaction between instruction and the development of implicit knowledge are easily summarized: there was none. For both target structures, significant overall progress was observed, and both kinds of instruction were found to be effective. Although the NI group clearly lagged behind the experimental groups, the difference between this group and the experimental conditions was not significant. Even when the EI group was cleared of those students that did not gain explicit knowledge, no interaction was observed. The conclusion was therefore that having explicit knowledge does not constitute an advantage for the acquisition of implicit knowledge.

4.5.4 Individual characteristics and structure complexity

Section 4.4 explored how sources of individual difference may have influenced the effectiveness of instruction. The interaction of explicit and implicit progress with developmental readiness and L1 similarity was explored by means of univariate ANOVAs. For aptitude, attitude and age, correlation analyses were used.

FINDINGS 133

Developmental readiness (DR) was operationalized as emergence of correct use (see 3.5.2). The analyses pertaining to DR revealed clear differences between the two grammar structures, in that DR was found to be related to DoC progress, but not to SubC progress. It was found that students who already used correct forms of the DoC benefit more from the instruction than students who did not yet use correct forms. This effect was observed for both explicit and implicit progress. However, no interaction was found between instruction and developmental readiness, meaning that the beneficial effect of DR applied irrespective of the kind of instruction received. It should be pointed out that the categorization according to whether they had already used correct forms at T0 lead to significant differences between the resulting groups on a number of proficiency measures. Apparently, the construct of DR as it was operationalized in this study expresses to some degree ‘advancedness’ of the learners interlanguage system. It should also be pointed out that for some of the analyses, insufficient statistical power was achieved, which may have obscured the presence of existing differences.

The analyses pertaining to L1 similarity also revealed a difference between the two structures. L1 similarity was defined in specific relation to the target structures. For the DoC, it divided the first languages of the participating students according to whether they realize comparison morphologically or periphrastically; for subordination, it encoded the default word order of the participants’ L1. L1similarity was found to be related only to implicit DoC, and only in interaction with instruction: students with L1’s that also express comparison morphologically demonstrated more implicit progress, provided they had had explicit instruction. This result actually constitutes a little bit of evidence in favour of an interface between explicit and implicit knowledge. The effect was only observed for the DoC.

By means of correlation analyses, the relation between explicit and implicit progress in the two target structures was related to aptitude, attitude, and age. For the attitude measures – cognitive style and motivation, and age, no differences were observed with respect to the two grammar structures. The attitude measures were not found to be related to any type of grammatical progress. Age was related to explicit progress of both the DoC and SubCs. There did not seem to be an interaction with age and instruction, although for both the DoC and SubCs the correlation was strongest with EI progress. The analyses pertaining to aptitude did reveal some differences between the two structures. Both memory and grammatical sensitivity (GS) were found to be related to explicit SubC progress, while this relation was absent for explicit DoC progress.

134 CHAPTER 4

Another finding was that memory was associated to implicit DoC progress, while none of the aptitude measures was implicated in implicit SubC progress. To both observations applied that the associations were strongest for progress in the EI condition. However, the confidence intervals for the correlations presented were quite large, which means that little can be said with certainty about potential interactions with instruction.

4.5.5 Conclusion

In two parallel experiments, the effects of instruction in two contrasting grammar structures were tested. If structure complexity does not affect instruction and acquisition, no differences in outcome should be observed for the two experiments. However, differences occurred frequently. Differentiated use of subordination was observed across different realizations, while such differentiation was largely absent for the DoC. Also, there were differences related to the effectiveness of implicit instruction, developmental readiness, L1 similarity, and aptitude. The question is to what extent these differences are in fact caused by or related to structure complexity. This issue will be taken up in the next chapter, where all findings will be discussed in relation to the theories and earlier research discussed in Chapter 2.

5 Discussion

5.1 Introduction

In this chapter, the findings reported in the previous chapter will be evaluated in relation to the theories and earlier FFI research discussed in Chapter 2. First, the research design will be evaluated in 5.2, highlighting both strengths and weaknesses. Next, in 5.3, the findings will be discussed in relation to the issues pertaining to the interface debate, SLA theory as outlined in 2.3, and previous FFI research findings (2.4). In 5.4, the implications for pedagogy will be discussed, and the book will be concluded in 5.5. Recommendations for future research will be provided along the way.

5.2 Evaluation of the research design

This study has sought to combine the advantages of laboratory research with the advantages of classroom research. In a review of SLA laboratory research, Hulstijn (1997) points out that the reason to employ laboratory research designs has been to exercise greater control over exposure to the target structures and to ensure equal conditions during the instruction: effects of instruction simply stand more chance of being spotted in tightly controlled settings. He illustrates this by quoting Carrol and Swain: “If feedback does not work in an experimental situation, it is highly unlikely that it would work elsewhere (1993, pp. 361-362)” (p. 134). Nevertheless, field research is needed to assess to what extent laboratory research findings may be generalized. After all, there is no guarantee that treatments that were found to be effective in the lab, will also work in classroom settings. In fact, this study has illustrated the limitations, as instruction was found to affect performance on decontextualized and more natural language tasks differently.

This study has controlled for a number of variables that traditionally may have caused interference in SLA classroom research. First of all, the computer was used as the means to present the instruction in order to exercise control over the input. It allowed for large numbers of students in ten different schools to receive

136 CHAPTER 5

precisely the same instruction. The use of the computer also incorporates disadvantages, though. The instruction may be less effective, simply because the participants had to learn about structures in their L2 from written sources without any help of the teacher. Nevertheless, although the instruction was relatively brief (approximately three hours), significant explicit knowledge gains were obtained, which means that the instruction must have been effective. It may seem like a risk was taken to assume that the instruction would be effective, but the materials were tested in advance, and many studies have made use of even briefer treatments and still observed substantial progress (e.g., Day & Shapson, 2001; Jourdenais et al., 1995; Muranoi, 2000; Robinson, 1996; Salaberry, 1997). It should be pointed out, though, that although the computer was used, the experiment was not taken out of class. Being enrolled in an intensive Dutch course, the participants will have encountered the target structures in natural settings as well. In addition, there is a chance that teachers responded to questions by students, and thus that some students received explicit feedback about one of the target structures on an individual basis. In fact, one of the control group classes was excluded from the analysis for this reason.

Classroom research has seldom used true randomization to compare the effects of instruction (Norris & Ortega, 2000). Rather, intact classes have been used, and sometimes also different teachers have been used to present the instruction. In this study, the computer was used to randomly assign the participants to either of the two treatment groups. This way, both experimental conditions were equally represented by students from all participating schools. This method of randomization proved to be successful, as no significant differences were observed between the two groups on the dependent variables at the start of the experiment, on the ID variables nor on any of the background variables assessed. At the same time, this study underscores the importance of randomization: the no-treatment control group was made up of intact classes, and this group was found to differ from both treatment groups in previous exposure to the L2 (see 3.4.5). In retrospect, the control group should have been sampled in the same way as the two experimental groups, by means of the computer. However, this would have required the development of additional ‘dummy’ material for those students to work with.

An aspect of research design that has often featured in laboratory research is the use of (semi-) artificial languages (Hulstijn, 1997). Obviously, this is a design feature that was not adopted in this study. The focus on second language proficiency simply precluded this option. A source of variation that is introduced by using natural languages is a lack of control over the participants’ history with

DISCUSSION 137

the target language. The participants of this study were all advanced learners, but some had needed more time than others to achieve their current level of proficiency. This variation cannot but be accepted, as it comes with the very purpose of this study. In fact, the use of impoverished linguistic environments may be one of the key factors for why lab research lacks ecological validity. The use of an artificial language allows for carefully crafting grammar structures so that they differ only with respect to the feature under investigation. However, the risk of abstracting away too much from natural languages is quite real. Hulstijn (1997) observes: “One might wonder whether some of the lab studies … investigated purely cognitive concept formation rather than (second) language learning.” (p. 139).

A point of improvement would be the relative heterogeneity of the subject sample. It was already pointed out that there were differences in previous exposure to Dutch. The role of the Dutch language varied from being (one of) the language(s) used at home to being used at school only. The participants were also heterogeneous with respect to their first languages. However, some control was exercised by means of careful selection procedures with respect to L2 proficiency (3.4.4) and a classification of first languages in specific reference to the structures under investigation (see 3.5.3). The technique used to gain control over first language influence is unorthodox, yet promising. Indeed, in the past, quite a number of FFI studies have used a sample of mixed L1 backgrounds, mostly operating under the silent assumption that the L1 did not affect the results of their studies. However, given that transfer has frequently been demonstrated, and in the light of the assumption that L1 transfer is structure-specific rather than language specific (Odlin, 2003), this technique may on the one hand be an important requirement for dealing with mixed L1 backgrounds, and on the other hand go a long way in controlling for transfer effects.

The dependent measures that were used to assess progress were motivated by the constructs of explicit and implicit knowledge (see 3.3). An important reason to choose the untimed grammaticality judgement test to assess explicit knowledge was that it is the most widely used test in FFI research to measure gain. An additional advantage was that it calls upon explicit knowledge quite unobtrusively, thus minimizing potential retest effects. Because the participants were selected on the basis of not having knowledge of the target structures, they felt the test was quite difficult and for this reason the number of test items had to be kept small. Nevertheless, the test was found to be quite reliable (see 3.3.2, Table 3.6). Also, earlier findings were replicated, which is an indication of the test’s validity. There is certainly room for improvement, though, and Ellis (2004)

138 CHAPTER 5

provides a number of recommendations to enhance the validity of explicit knowledge tests. Most importantly, in addition to identifying the ungrammatical feature in the stimulus sentence, Ellis recommends to ask students to indicate how certain they are of their judgement, and whether they made their judgement on the basis of knowledge or intuition.

The most important purpose of the implicit knowledge test was that it obtained spontaneous written text from the participants without focusing them on the target structures. It can safely be concluded that the implicit elicitation technique that was used in this study was successful. In the beginning, the participants were quite concerned with providing correct answers, but once it was made clear that any answer was correct as long as they focused on the correctness of their formulations, most students actually enjoyed responding to the situations they were presented with. Even students that seemed to have adopted the strategy to give short answers (less scope for errors) were still found to use the target structures in their answers. Because of the elicitation techniques, data processing was fairly cost-efficient, in that a relatively small amount of data needed to be gathered as compared to sampling in real-life spontaneous settings. However, this testing method does require extensive piloting and fine-tuning in order to be certain that it elicits appropriately. Also, it should be noted that as soon as such kinds of elicitation techniques are used, particular contexts of use may be emphasized that are perhaps not as frequent in real time use. Reliabilities were calculated based on whether a correct form of the target structure was provided for each of the defined situations, and these reliabilities were found to be good (see 3.3.3, Table 3.7).

Because sampling and data processing were fairly cost-efficient, group sizes could be quite large. As a result, the statistical power to detect significance was considerable: effect sizes of f = .15 and higher could be detected. Sufficient statistical power is important in order to exclude the possibility that absence of significant effects – which were hypothesized – is due to an insufficient number of participants. Only for the analyses with regard to developmental readiness, the obtained statistical power turned out to be too small.

In short, this study has controlled for a number of features that may have introduced variation in earlier classroom FFI studies. Most importantly, the instruction was equal for all participants, and the exposure to the target structures during the instruction was highly regulated. Because of the random assignment to the experimental conditions, effects of potentially interfering variables were most likely mitigating rather than distorting.

DISCUSSION 139

5.3 Evaluation of theory

5.3.1 Introduction

In this section, the outcomes of this study will be evaluated in relation to the interface debate (2.2), the ideas on SLA expounded in 2.3, and earlier FFI research findings (2.4). This discussion will typically first recall into memory the theoretical framework and earlier findings, followed by a short summary of this study’s findings, and a discussion of the meaning of these findings. This discussion is roughly structured according to the research questions. Thus, 5.3.2 discusses the findings in relation to the exploration of how L2 learners started to make use of the two target structures. Next, in 5.3.3, the evidence in relation to the interface debate is discussed. 5.3.4 discusses the different sources of individual difference: how they interact with instruction and how they impact upon grammatical development. Finally, in 5.3.5, discusses the importance of structure complexity in relation to grammatical development and instruction.

5.3.2 The nature of implicit L2 knowledge

In Chapter 2, the constructivist views on implicit second language knowledge were presented. The main tenet of the constructivist view is that grammatical knowledge of the L2 emerges in the course of acquisition. The development of L2 grammatical knowledge starts with the internalization of exemplars: words and phrases that contain grammar structures are stored as wholes in memory. L2 learners may use these chunks when producing language, but not in a computational way. Rather, they are retrieved from memory as such. Through subconscious reflection on the internalized exemplars, a process referred to as syntacticalization, L2 learners ultimately develop an open class grammatical system much like construction grammar. Learners may observe regular patterns in the intake, and start using such structures accordingly. Such patterns may still be limited in scope, in that they are related to specific lexical contexts. The constructivist view on the nature of L2 knowledge is closely related to the views on representation of L2 knowledge. Implicit knowledge is seen as represented in a complex and dynamic associative network, and implicit learning is essentially the frequency-based establishment of associations and the subsequent abstraction of grammatical categories. These do not come into existence overnight, but result from frequent exposure and a subconscious process of figuring language out. In

140 CHAPTER 5

short, implicit learning is the inevitable and uncontrollable result of information processing (N. Ellis, 2002; Hulstijn, 2002).

One goal of this study was to investigate how L2 learners start to make use of developing L2 grammar structures. To this end, the use of the two target structures, the degrees of comparison and subordinate clauses, was investigated by means of the free written response data that was gathered over a period of three to four months. One aspect of use that was investigated was the notion of differentiated correct use across different types of realizations. Another aspect of use under investigation was the time it took L2 learners to start using the target structures in a flawless manner. The analyses were intended to be explorative and descriptive, and no hypotheses were formulated.

Differentiation in correctly using the target structure across different realizations was clearly demonstrated for SubCs. For this structure, substantial differences were found in correct use of conditional and causal subordinate clauses, amounting to 36 percentage points at T0, and remaining as large at T2. Conditional SubCs were used correctly more often than causal SubCs, and they were also used before causal SubCs. In addition, there were differences in correct use related to clause and verb phrase complexity: application of the structure in complex contexts meant higher rates of incorrect use. For the DoC, the findings were less clear. At T0, 80 percent of the forms were used correctly. This applied to all but one type of use: if comparatives were used in combination with comparative clauses, no more than 50 percent of the form was used correctly.

The findings with regard to SubCs suggest the application of a different word order in subordinate clauses to be lexically specific. Especially the difference in correct use between conditional and causal subordinate clauses was too substantial to be coincidental, and it clearly suggests that, in beginning stages of acquiring the ability to use SubCs, L2 learners relate the word order change that subordinate clauses require to the use of a particular subordinate conjunction. The question that rises is at what point – if at all – L2 learners do in fact connect the various types of subordinate clauses as belonging to one particular class of clauses with common formal characteristics. In other words, at what point is a more general grammar structure abstracted. The only relation that was found between the two types of SubCs studied, is that correct use of conditionals clearly preceded correct use of causals. If L2 learners had already recognized both types of SubCs as one class of structures, this difference should not have occurred. One can only assume that at one point, the connection is made. However, an account that sees the different types of SubCs as separately represented is tenable as well.

DISCUSSION 141

The absence of clear findings of lexical specificity in the use of the DoC may be due to a data analysis problem. Especially the differentiated use of comparatives in combination with comparative clauses is indicative of this. As pointed out in 3.3.3, L2 learners’ intentions to express comparison may go unnoticed, because of a tendency to fail to mark adjectives for comparison while they do intend to express comparison. Because comparative clauses clearly betray the intention to compare, omissions can be – and were – counted as incorrect. However, for all other uses of the DoC such contextual clues are often not present, possibly resulting in unnoticed errors. For example, L2 learners writing “Deze radio is mooi” (This radio is nice) may actually have meant to write “Deze radio is mooier” (This radio is nicer). Because the L2 learners’ true intentions could not be determined, such phrases were coded as ‘no use’. Given the fact that the test items implicitly elicited comparison, and the fact that higher rates of elicitation were obtained in the pilot phase for more advanced students, it is likely that numerous instances of incorrect use were missed. As a result, the percentages of correct use of the DoC were overestimated, and conclusions about differentiation of use of the DoC in different contexts of appearance cannot safely be drawn.

For both target structures, progress was found to be gradual and slow.

Progress was investigated by means of the stages of development proposed by White (1998). Four patterns of use were distinguished: no use, only incorrect use, variable (both correct and incorrect) use, and only correct use. Despite the average increase of correct forms in the free written response task, the participants were found to largely continue their ways of using the target structure: over 50 percent of the students demonstrated the same pattern of use from one testing moment to the other. If they did change their ways, they were found primarily to progress. For example, if students had at one point used correct forms of the target structures, they were hardly ever found to return to the use of only incorrect forms in subsequent tests. These results demonstrate that learning to use a grammar structure is not an instantaneous affair. There do not seem to be clear turning points that mark the acquisition of a particular structure. Rather, learning to use a structure correctly involves an extended period of time during which incorrect use and correct use co-occur, until at one point incorrect use disappears. Correct use seems to start in the more simple or salient contexts, and is gradually extended to more complex contexts.

This gradual learning process of replacing incorrect use for correct use agrees with the frequency-based language learning account as outlined in 2.3. Based on exposure to the target structures in the input, ‘correct’ representations of the

142 CHAPTER 5

structure are established that gradually outweigh the ‘incorrect’ representations. It is important to understand that ‘incorrect’ here does not refer to faulty application of a structure, but to non-use. For the DoC, virtually all forms that were found to be and coded incorrect were examples of non-use: comparative markers were failed to use while contextual clues clearly indicated the intention to express comparison. Similarly, for SubCs, incorrect use hardly ever meant that L2 learners realized incorrect, non-existent sentence structure. Rather, sentence structure was left unaffected, and default main clause word order was used. The image that rises is that L2 learners have to replace a default structure that is being applied to express a broad set of meanings with specific structures associated with specific meanings. This process of replacement takes time, and during this process, L2 learners seem to have both correct and incorrect forms represented and available, alternating in a seemingly spurious way between them.

5.3.3 FFI and the development of explicit and implicit knowledge

The primary goal of this study was to investigate whether explicit instruction contributes to becoming proficient in using the second language grammar. Theoretically, this concern is addressed by the interface debate. Three different positions assign a different role to explicit knowledge in the course of developing implicit knowledge, the kind of knowledge that allows speakers to be proficient in the L2. In essence, the controversy in the interface debate evolves around the organisation of linguistic knowledge. All three positions make a distinction between explicit and implicit knowledge. For the strong interface position, knowledge of the L2 starts out as explicit and converts into implicit knowledge: one learning process is posited, covering several stages. Thus, there is a direct relationship between explicit and implicit knowledge. The weak interface position and the no interface position both argue explicit and implicit knowledge to be completely separate. The weak interface position still allows for the conversion of explicit knowledge into implicit knowledge, but argues that this depends on the kind of grammar structure. It also allows for delayed effects. The no interface position claims the two knowledge systems to be completely separate, and no conversion is possible. Potential effects of explicit knowledge are indirect, in that it may affect implicit learning processes.

In Chapter 2, this theoretical problem was approached from two different perspectives. On the one hand, recent SLA developments were evaluated in order to arrive at a better understanding of the constructs of explicit and implicit

DISCUSSION 143

knowledge. On the other hand, FFI research was scrutinized in search of evidence that explicit knowledge promotes the development of implicit knowledge. The outcome of this undertaking was that neither of the two approaches provides arguments in favour of a weak or strong interface. Based on this, it was hypothesized that there is indeed no interface between explicit and implicit knowledge: explicit instruction should not be more effective than implicit instruction in promoting the ability to use the language in spontaneous language use. Nor should it be less effective, because concomitant implicit learning effects will arise during the explicit instruction. These hypotheses were tested by comparing how explicit and implicit instruction affected the development of both explicit and implicit knowledge.

FFI and explicit knowledge

The results of this study replicate the predominant finding in FFI research that explicit instruction is more effective in promoting L2 knowledge as measured by explicit knowledge tests. For both target structures, significant immediate (i.e., from T0 to T1) explicit progress was observed when the rules of grammar were explicitly taught. Not much additional progress was observed once the instruction stopped. Implicit instruction was found to be effective only for one of the two target structures, subordinate clauses, from T0 to T2. Similar results have been reported in other studies using measures of progress that tap explicit knowledge. In such studies, implicit instruction is hardly ever found to be superior to explicit instruction: there are either no differences between explicit and implicit instruction, or explicit instruction works better. This is true, for example, for the studies reviewed in 2.4.5, but also Norris and Ortega (2000) – though not differentiating for the kinds of tests used – reported substantially higher effect sizes for explicit types of instruction over implicit types of instruction in their meta-analysis.

The results of this study suggest that there is an interaction between the kind of grammar structure taught and the kind of instruction received. For both target structures, explicit instruction was found to promote explicit knowledge quite effectively. Implicit instruction proved effect only for one of the two structures. For the simple morphological structure (the degrees of comparison), no effect of implicit instruction was observed. For the complex syntactic structure (subordination), on the other hand, implicit instruction ultimately (by T2) led to equal amounts of explicit knowledge. In previous studies, similar findings have been reported. DeKeyser and Robinson, for example, also found explicit

144 CHAPTER 5

instruction to be more effective than implicit instruction for simple structures. In a similar vein, De Graaff found explicit instruction to work better for morphological structures, and found no differences between explicit and implicit instruction for syntactic structures. Based on this study and previous studies, then, the conclusion must be that EI is more effective in promoting explicit knowledge of simple structures. Implicit instruction promotes explicit knowledge only when the structure is complex.

Explicit instruction promoting explicit knowledge is according to expectation in the light of the definition of explicit knowledge as being declarative in nature and requiring deliberate effort to be learned. That implicit instruction may be just as effective is somewhat puzzling, and it is even more puzzling that this should hold only for complex grammar structures. Transfer of implicit knowledge probably explains why students in the implicit condition also improve their explicit knowledge: in 2.4.2, it was already pointed out that it is difficult to prevent students from using their implicit knowledge in explicit knowledge tests. Thus, the long term progress in explicit knowledge of the SubCs observed for students in the implicit instruction condition may reflect growth in implicit knowledge which students may revert to in absence of explicit knowledge. However, why this should have occurred only for the complex structure remains difficult to answer.

The growth of explicit knowledge was found to be durable. At T2, performance was equal to performance at T1 for both target structures, which means that the knowledge was not lost nor had any further knowledge been acquired once the instruction had stopped. This finding is in line with the definition of explicit knowledge proposed by R. Ellis (2004; see 2.3.5). Conscious, declarative knowledge is learned deliberately, and further growth without deliberate learning would be unexpected. In fact, if not regularly attended to, explicit knowledge is probably lost. A number of FFI studies have indeed reported loss of explicit knowledge at delayed post testing.

Implicit knowledge

No differences in development of implicit knowledge were found between explicit instruction and implicit instruction, irrespective of the kind of grammar structure concerned. Overall, significant progress was obtained by all students in both experimental conditions. In terms of effects sizes, progress was moderate. It is difficult to ascertain whether the instruction gave rise to much implicit learning as opposed to students who did not receive any focused instruction. For the

DISCUSSION 145

degrees of comparison, the control group data were contaminated. For the SubCs, both experimental conditions demonstrated better performance than the control group, but because the control group did gain some progress between T1 and T2, differences between the control group and the experimental groups missed significance.

The finding that there was no difference between the two experimental conditions concerning implicit progress was according to expectation. Based on an analysis of SLA theory and FFI research, this study departed from the no interface hypothesis. Actually, the no interface position suggests that methods stimulating implicit learning should outperform methods stimulating explicit learning: if the two knowledge systems are separate, then instruction targeting explicit knowledge cannot readily be expected to promote implicit knowledge. Nevertheless, the expected outcome was that there would be no differences between the two experimental conditions concerning implicit knowledge growth. This expectation was based on Hulstijn’s suggestion that explicit instruction leads to concomitant implicit learning. The results for both target structures confirm this notion, in that both experimental conditions demonstrated equal implicit progress.

The findings of this study are also in agreement with previous findings. In 2.4.4, FFI research was scrutinized in search of evidence for an interface between explicit and implicit knowledge. The outcome of this analysis was that in two studies an advantage was found for explicit instruction over implicit instruction (Muranoi and VanPatten & Sanz). However, in both studies, the explicit instruction involved quantatively more input in comparison to implicit instruction, and it was argued that the findings cannot safely be taken as evidence of explicit instruction being more effective in promoting implicit knowledge. In the other studies mentioned in 2.4.4, no differences were observed between explicit and implicit types of instruction. This study replicates these latter findings, and underscores the importance of keeping the amount of exposure to the target structures during the instruction equal.

An important implication of these findings is that implicit learning will occur as long as relevant information is processed, irrespective of how the information is processed. In this study, care was taken to keep the exposure to the target structures in both experimental groups equal: even the sentences in which the structures appeared were literally the same across the conditions, although they were used either to explain and practise the target structure (EI), or to do text comprehension exercises (II) (see also 3.2). Despite the fact that the structures were processed for completely different purposes, there were no differences in

146 CHAPTER 5

implicit learning results. This not only agrees with the notion of concomitant implicit learning during explicit instruction, it also confirms the idea of information processing being an uncontrollable process (see 2.3), which inevitably leads to learning. Without wanting to devaluate the importance of quality (i.e.: how the input is processed), quantity of exposure seems to be a vital aspect of learning.

The growth of implicit knowledge was found to be durable. The progress that was obtained between T0 and T1 was not lost between T1 and T2. However, a difference was observed between the two target structures. No further implicit knowledge of the DoC was obtained between T1 and T2, while implicit knowledge of SubCs continued to grow once the instruction stopped. Possible reasons for this difference will be dealt with in 5.3.5, where structure complexity is discussed.

5.3.4 Individual differences and the effectiveness of instruction

Another goal of this study has been to relate explicit and implicit learning to particular individual/personal qualities. On the one hand, such an investigation could uncover dependencies between the ability to use explicit knowledge to improve L2 performance and particular ID variables. On the other hand, such an investigation could provide further evidence for the separateness of the explicit and implicit learning and knowledge systems if ID’s are found to be related differently to explicit and implicit grammatical development. On a theoretical level, ID’s are quite commonly held to moderate the effectiveness of instruction. One of Krashen’s claims when he put forward his no interface position was that only some learners are able to use their explicit knowledge as a monitor. In addition, he claimed that explicit and implicit learning are referenced by two different individual factors: attitude and aptitude. Similarly, Ellis argued that the effectiveness of instruction interacts with the L2 learners’ stage of development. But also age and L1 background have been related to effective FFI (see Norris & Ortega, 2000, p. 422). Nevertheless, FFI studies relating the effectiveness of instruction to individual characteristics have been rather scarce, and basing predictions on earlier research is difficult. In this study, a number of potentially moderating learner variables have been included (developmental readiness, L1 similarity, attitude, aptitude, age, and cognitive style), with the objective to explore the relation between these variables and explicit and implicit progress.

DISCUSSION 147

Developmental readiness

Developmental readiness refers to the notion that L2 development is to some extent subject to fixed orders of acquisition. The implications for instruction are that instruction can only be effective when L2 learners are sensitive (i.e., developmentally ready) to the instruction because it matches their stage of acquisition. This study provides some evidence that FFI is more effective when L2 learners are developmentally ready. More precisely, developmental readiness (DR) was not found to be related to SubC progress, but it was found to be related to both explicit and implicit progress in the DoC. There were no interactions between DR and the type of instruction received. The meaning of these findings is that instruction in the DoC was more effective once the L2 learner is developmentally ready; how the instruction is realized – explicit or implicit – is inconsequential. It should be pointed out that for some of the analyses (those pertaining to the DoC and conditional SubCs) insufficient statistical power was reached, meaning that if significance failed to occur, this may have been caused by an insufficient number of subjects.

Two aspects about these findings are difficult to interpret. The first unexpected finding is that DR was found to influence the development of the DoC only. One may argue, though, that this is due to the differences between the target structures, and these will be discussed in detail in the following section. The second puzzling finding is that both explicit and implicit DoC gain were affected by DR. Given the definition of explicit knowledge (see 2.3.5), there is no limit to what can be learned explicitly at any given time, and explicit learning should not be constrained by DR. Nevertheless, the participants in this study that were developmentally ready improved their explicit knowledge more than students who were not yet ready.

Practical rather than theoretical grounds motivated the operationalization of DR. DR was defined in terms of emergence of correct use. The advantages are that emergence of correct use is not structure-specific and that it is clearly observable in the language of learners, even in classroom settings. In 3.5.2, it was already pointed out, though, that this operationalization is crude in the light of the underlying theory: knowledge of grammar structures develop according to fixed stages, and instruction should match these stages. This suggests that as learners move through the stages, instruction should be continuously adapted to match each stage. Such instruction is practically unattainable, which is why emergence of correct use was chosen. From a theoretical perspective, however, this is quite arbitrary, and because no relation was found between the development of implicit

148 CHAPTER 5

SubC knowledge and DR, emergence of correct use does not seem to be a valid operationalization for all structures.

Another point needs to be made concerning the operationalization of DR. In this study, developmental readiness expressed ‘advancedness’ of the L2 learners’ grammatical system in that two groups were created according to whether or not correct forms had emerged in the participants’ production (see 3.5.2). The preliminary analyses have shown that the two groups that were thus created also differ significantly on a number of other variables: most notably grammatical accuracy and the C-test (see 4.4.2, Table 4.23). This is perfectly logical, as grammatical accuracy and the C-test can be considered expressions of ‘advancedness of the linguistic system too, though not related to specific target structures. The question is what this means? Either, DR is simply a structure-specific indicator of proficiency that signifies at what point the input should be intensified, or DR groups the more talented and the less talented language learners together. This too would explain why the ‘ready learners’ obtained more progress. In addition, it explains the puzzling relation between DR and explicit grammatical progress.

This study does not provide evidence that the ability to use explicit knowledge to monitor implicit knowledge depends on developmental readiness. Perhaps instruction needs to be tuned more precisely to stages in the acquisition of grammar structures. The study does provide some evidence that instruction is more effective for those that have already demonstrated beginning correct use, and thus confirms findings by Williams and Evans (1998), who found that the greatest gains were obtained by those demonstrating knowledge of the target structures in the pre-test. The precise involvement of DR is difficult to determine. Either the construct successfully signifies a point at which instruction would be beneficial, or it merely groups the more talented and the less talented together.

L1 similarity

The only evidence of a positive effect of explicit knowledge on implicit grammatical performance was found in interaction with first language similarity, and only for the DoC. This finding will be further discussed in 5.3.6. None of the first languages were actually similar to Dutch, as there were no subjects with Germanic first languages present in the sample. Similarity in this study meant that the first language realizes the meaning expressed by the structures in focus in a similar way (see 3.5.3). The question is why L1 influence only observed for the DoC? The cause might lie in the differences between the two structures (to be

DISCUSSION 149

discussed in 5.3.5), but it more likely lies in the uniqueness of the inversion rule in Dutch subordination.

Most, if not all, languages have means to express the degrees of comparison. In Dutch, they are realized morphologically and sometimes periphrastically. Other languages similarly either use morphological marking, periphrasis or both, which makes comparison straightforward. In contrast, the inversion rule required in Dutch subordinate clauses is quite unique. Some other languages, such as German, have it too, but in many languages inversion in subordinate clauses simply does not exist. Only Armenian was found to have something similar, but there were only three speakers in the sample. Consequently, there was nothing similar in the participants’ first language; and therefore, nothing to help understand the instruction. This may explain why no effects were found for the SubCs. Because subordination affects word order, default L1 word order was coded instead of inversion, but it was not found to affect the results.

Aptitude and attitude

The effects of a number of individual factors were studied in relation to explicit and implicit gain. To this end, explicit and implicit gain scores for both instruction conditions were related to the variables memory, grammatical sensitivity (GS), cognitive style, and motivation by means of correlation analyses. Research about the involvement of aptitude and attitude has often been triggered by Krashen’s argument that attitude affects implicit learning, while aptitude affects explicit learning. Robinson and others have already pointed out that aptitude as it was measured by most aptitude test batteries was operationalized in a restricted sense, excluding components of aptitude that might relate to implicit learning. In this study, Krashen’s claims have also been addressed. For aptitude, memory and GS were used, which both should correlate with explicit progress only. For attitude, motivation and cognitive style were used, as these two measures surfaced from a factor analysis on the teacher questionnaire.

Hardly any evidence was provided that these individual factors interact with instruction: if significant correlations were observed, they were quite low. In addition, the distances between the correlations observed for either instruction condition were never very large; they would fall within each others confidence intervals, which is why one cannot maintain that particular factors were differently involved in each type of instruction. However, some of the individual factors were found to be specifically related to either explicit or implicit DoC or SubC gain.

150 CHAPTER 5

The absence of an interaction between instruction and aptitude confirms some of De Graaff’’s findings. In his study, he found that aptitude predicted progress in both instruction conditions equally. However, this study suggests a more complex relation between aptitude and progress. The findings suggest that the workings of aptitude are structure-specific: memory was related to correct use of the DoC in the free written production task, but not to correct use of SubCs. Conversely, Memory and GS were related to SubC grammaticality judgement performance, but not to DoC grammaticality judgement performance (see 4.4.4, Table 4.26 and 4.27). Such differences are unexpected, and can only be explained in relation to the nature of the target structures. This will be taken up in the following subsection (5.3.5).

On a theoretical note, the expectation that aptitude (as measured in this study) is positively correlated to performance in explicit instruction conditions is based on the notion that information is processed differently in the two conditions (see 2.3). If explicit and implicit instruction call upon different learning mechanisms, then that would constitute further evidence for the separateness of the two systems. The question that rises, though, is to what extent differences in processing during testing have been assessed, rather than differences in processing during the instruction. A grammaticality judgement task is bound to call upon mechanisms involving explicit handling of linguistic information, at least more than spontaneous production would. The relation between types of testing and aptitude has been examined in this study by relating memory and GS to five proficiency measures that were used as control variables. Both memory and GS were found to be significantly related to those measures that tested language proficiency in isolated contexts (the Cito ISK test and the C-test). However, no correlations were found between the aptitude measures and those proficiency measures based on spontaneous production (fluency and grammatical accuracy). It indicates that memory and GS indeed assess ed the involvement of explicit processing mechanisms. However, it is difficult to tease apart at what point these processes were involved: during the instruction or during testing.

With respect to motivation and cognitive style, no correlations were observed with either explicit or implicit progress. This study provides no evidence for positive affects of motivation or cognitive style on implicit learning. Given these results, the conclusion has to be that either the role of motivation and cognitive are limited, or their effects were too small to be captured, which is not unlikely in the light of the power limitations of the correlation analyses. Another possibility is that the teacher judgements that were used to assess motivation and cognitive

DISCUSSION 151

style were not adequately reflections of the learners’ true motivation and cognitive style. The results, then, do not support Krashen’s claim with regard to the role of attitude in implicit learning. However, this would have been unlikely in the light of the definitions of explicit and implicit knowledge. In fact, one can argue exactly the opposite to Krashen’s claim; implicit learning has been characterized as an inescapable, uncontrollable process, and for this reason it seems unlikely that implicit learning would be restricted by aspects of attitude. Explicit learning, on the other hand, requires deliberation and conscious effort, and low motivation is likely to hinder such efforts. This study does not provide support for these claims either, but this may be due to insufficient statistical power or inaccurate assessment of attitude.

Age

Finally, the effect of age was assessed. Age was found to be related to explicit progress, in that the younger participants tended to obtain more explicit progress than the somewhat older learners. After a period of intensified input – either explicit or implicit, younger learners demonstrated more progress on the grammaticality judgement task than older learners. Two questions come to the fore: why is only explicit learning affected by age, and what explains the advantage of younger learners. Actually, a tremendous amount of literature has been devoted to the latter question (e.g., DeKeyser, 2000; Doughty, 2003; Robinson, 1997). It has been suggested that there may be a fundamental difference between learning at a young age and learning at an older age due to changes related to maturation or to the disengagement of an inborn language learning facility. These results actually suggest that a fundamental difference applies only to explicit learning. Although these results need replication, the repercussions for the fundamental difference discussion can be quite substantial. It would be particularly interesting to evaluate on what kinds of tests claims about a fundamental difference are based.

5.3.5 Structure complexity and the effectiveness of FFI

Fundamental differences in how different types of grammar structures are learned would seem unlikely according to constructivist theories of second language acquisition. The emergence of grammar being a frequency-based abstraction process applies universally to all grammar structures. In this light, the suggestion that different modes of learning (memory-based versus rule-based) take care of different types of grammar structures must be incorrect. However,

152 CHAPTER 5

complexity in some form probably does affect the success of instruction. In fact, numerous complexity-related features of grammar structures have been proposed to affect the success of instruction: formal and functional complexity; scope and reliability; whether or not structures are developmentally constrained. Mostly, there is some evidence to support each of these proposals, but scientific comparison is difficult because if structures are classified according to all these distinctions, each structure is unique. Nevertheless, it seems that, if progress is assessed with measures calling upon explicit knowledge, simpler and more reliable structures can be taught explicitly with more success than complex and unreliable structures (see 2.4.5). It should be pointed out that simple structures in these studies were mostly also morphological structures. The findings for the impact of FFI on implicit progress are similar; the tendency is that morphological and formulaic structures are taught with more success than syntactic structures (R. Ellis 2002; see 2.4.3).

In design, this study involved two identical, parallel experiments in which two contrasting grammar structures featured. The goal was to discover to how instruction affects the acquisition of these structures. The target structures were chosen on the basis of practical and theoretical considerations. The most important difference is that the DoC structure is morphological, while the SubC structure is syntactic. In this study, differences were observed for the two structures a number of times. And for some of these, it seems that the differences are due to the difference in nature of the target structures. The question is to what extent these differences can be attributed to differences in the nature of the two structures.

One of the findings was that for the SubCs, differentiated use was demonstrated. Percentages of correct use were substantially higher for conditional SubCs in comparison to causal SubCs, and the same differentiation was observed for clause complexity and verb phrase complexity. In 5.3.3, this difference was already discussed and related to the possibility that incorrect use of the DoC may go unnoticed. However, another or an additional reason may lie in structure complexity. In fact, these findings provide suggestions towards how to define complexity. The very fact that the use of the DoC was not affected by the context suggests the structure has no impact upon the utterance beyond the adjective it is used with (except when comparative clauses are used). The differentiated use of the SubCs, on the other hand, suggests that the structure can be subdivided into different realizations of use that have to be learned more or less individually. These realizations are likely to be defined by the number of verbs the verb phrase consists of, and the constituents realized. Complexity, then,

DISCUSSION 153

may be a direct function of the number of different realizations a structure can appear in. This definition of complexity agrees with the persistent finding that morphological structures are taught with more ease than syntactic structures. Longitudinal research studying the development of using structures in spontaneous situations would be needed to gain further insights.

Another finding that may be better understood in the light of the differences in structure complexity is the finding that developmental readiness (DR) was found to affect DoC progress only. In 5.3.4, where DR was discussed, doubts were already expressed as to the validity of the operationalization. The construct of DR is actually based upon the notion of a staged acquisition process, and it was already pointed out that instruction may need to be continuously adapted to match the learners’ stage of acquisition. However, perhaps sensitivity to instruction is better understood in terms complexity, meaning that DR should be determined in specific relation the acquisition of specific realizations of a structure rather than the structure as a whole. Assuming there are only few different realizations of the DoC, emergence of correct use accurately signifies sensitivity to instruction. For subordination, emergence of correct use was simply too crude in its operationalization, as no differentiation was made for the different ways the structure can be realized.

5.3.6 The interface issue

The findings of this study agree best with the no interface hypothesis. According to both the weak and the strong interface positions one would expect explicit instruction to be more effective; the no interface position as posited in this study would not expect any diffences between the two types of instruction, because both equally lead to implicit learning effects. Most compelling is the finding that explicit instruction was indeed not found to promote the ability to use the target structures in free writing any more than implicit instruction. The analyses clearly demonstrated that the participants who received explicit instruction had more explicit knowledge, yet this supposed advantage according to the strong and weak interface positions did not translate into higher rates of implicit learning. Even when the subject sample was cleared of those students that had not developed any explicit knowledge on the basis of the explicit instruction they received – after all, students may differ in their ability to learn explicit knowledge – no significant differences were found between students with and without explicit knowledge. In the light of these findings, it seems unlikely that explicit knowledge converted into implicit knowledge. Thus, it seems

154 CHAPTER 5

justified to conclude that the strong interface position is incorrect; both types of knowledge are separately represented, and they result from separate learning mechanisms. The weak interface position is similarly not supported in as far as it supports the conversion of explicit knowledge into implicit knowledge.

Another finding that may be construed as evidence for the absence of an interface between explicit and implicit knowledge is the finding that the development of explicit and implicit knowledge was differently affected by the two types of instruction. For the development of linguistic knowledge as measured by grammaticality judgements, explicit instruction was found to lead to more knowledge gains than implicit instruction. If linguistic knowledge was measured implicitly, however, no differences were observed between the two types of instruction. For both target structures, significant implicit progress was observed, but there was no advantage of one type of instruction over the other. If the grammaticality judgements and free written response task had assessed the same kind of linguistic knowledge, this difference would be unexpected. Both tasks must therefore be considered to be measures of different kinds of linguistic knowledge.

Both the weak interface position and the no interface position allow for indirect effects of explicit knowledge upon implicit knowledge, meaning that having explicit knowledge may affect implicit learning processes. Ellis (1994a) also sees explicit knowledge as a facilitator of implicit knowledge in that it may help the L2 learner to notice and to notice-the-gap. Similarly, Krashen (1982) admits to potential indirect positive effects of explicit knowledge, because it may sometimes be used as a monitor or it may make the input more comprehensible (see also 2.2.2). In fact, this study provides a little bit of evidence of such an indirect positive effect. For the degrees of comparison only, L1 similarity was found to interact with the kind of instruction received. Students who received explicit instruction in the degrees of comparison and who had first languages that also realize comparison morphologically were found to significantly outperform implicitly instructed students and students with different L1’s. An explanation for this result may be that explicit instruction made the participants aware of the similarities between their first and second language, reducing the perceived complexity of the structure, and consequently enabling them to apply first language processing strategies to their written L2 performance.

The findings with regard to L1 similarity may also suggest that the possibility of indirect positive effects of explicit knowledge upon the development of implicit knowledge may depend on the quality of the explicit knowledge. Making students aware of this similarity, although this was not done intentionally in this study,

DISCUSSION 155

may have improved the quality of the L2 learners’ explicit knowledge to such an extent, that successful monitoring became possible. In this study, explicit knowledge was taught by means of the computer, and although great care was taken to teach the target structures effectively, the explicit knowledge these participants had may not have been good enough. It is true that teaching via the computer does limit the ways in which rules can be practised, and it does lack the interaction with a teacher who may perceive inadequacies in the students’ knowledge where the computer would miss these.

All in all, this study lends credence to a no interface position, but it should be acknowledged that this study was not water-proof. There is a chance that this study has failed to establish the presence of an interface because of the target structures that were chosen. Another distinct possibility is that different types of explicit instruction aimed at facilitating input processing are more effective. Also, the delay in effects that the weak interface position expects may not have been long enough in this study. And finally, factors that were not included in this study, or that may have been inappropriately operationalized, may have covered up the existence of an interface. Nevertheless, the limited influence of explicit knowledge on the development of proficiency is in agreement with SLA theory and previous FFI findings. Is should be pointed out that these no interface claims are domain specific: they apply to the domain of syntax only, and its use in spontaneous settings. In addition, it remains to be seen to what extent these results can be generalized to older learners.

5.4 Implications for second language pedagogy

The ultimate goal of second language teaching is proficiency. This report has argued that, when trying to become proficient in their second language, L2 learners most likely stand very little to gain from having explicit knowledge of the rules of grammar. The question that rises is what role should be attributed to explicit instruction in second language acquisition curricula, especially in the light of the fact that the no interface position has been a reason to plea in favour of dismissing explicit types of instruction. It is important to stress that these findings do no support such pleas. This does study has not demonstrated the superiority of explicit types of instruction, but it also has not demonstrated the superiority of implicit types of instruction. Explicit instruction can be a valuable pedagogical tool, because it triggers concomitant implicit learning effects.

156 CHAPTER 5

First of all, some researchers say the value of explicit instruction does not lie in the establishment of explicit knowledge, but in that it can promote input processing mechanisms. As already pointed out in the introduction (Chapter 1), So-called ‘focus-on-form’ research has concerned itself with the provision of instruction intended to shortly shift the focus to form during communication, arguing that the moment at which errors occur in communication is the moment at which (explicit) instruction would be most effective (e.g., Doughty, 2001b; Doughty & Williams, eds., 1998). Similarly, Processing Instruction (PI) aims at facilitating input processing mechanisms by pointing out to learners that their default processing strategies put them on the wrong track. As a note on the side, it should be remarked that this study contradicts this latter claim. The findings with regard to L1 similarity, however, suggest that explaining when default processing strategies would be effective is a good approach to explicit instruction, which seems to clash with VanPatten’s claims. Nevertheless, this study has not addressed such claims, and leaves open the possibility that explicit instruction is indeed superior to implicit instruction in such cases.

Explicit instruction can be a valuable added ingredient in second language learning programmes provided they target implicit learning. There are in fact a number of reasons to resort to more explicit types of instruction every now and then. First of all, lessons in which explicit information is provided may form a good pretext to offer intensified form-focused input. Some structures may actually be quite difficult to supply in communicative settings of language use, and more explicitly focused language practice may then successfully lead to intensified input provision. Another reason to incorporate explicit teaching in the curriculum is simply that it meets the desires of many learners. As Krashen puts it: “There is no denying that that there is a certain satisfaction, for some of us, in knowing a conscious rule.” (Krashen, 1981: p. 113). And this statement does not only apply to second language learners, but also to their teachers. Thus, explicit instruction may be valuable simply to keep class interesting and varied for both teachers and students. One complaint that was put forward by users of Zebra (Alons, Bienfait, & et al., 1999; and see 3.4.2) – in which grammar is offered implicitly only in the first two years – is that the exclusive focus on communication actually becomes tedious. A form-focused exercise every-now-and-then would be a welcome change for those who work with this textbook nearly everyday for at least a year.

It should also be acknowledged that explicit knowledge is not necessarily worthless. It may be valuable in particular circumstances. This study investigated two grammatical phenomena, and perhaps explicit knowledge of other

DISCUSSION 157

grammatical structure is useful to have. Also, explicit knowledge may not aid the acquisition of implicit knowledge, but once L2 learners have implicit knowledge of a particular grammatical phenomenon, explicit instruction be useful to have for situations of language use that allow for preparation. Also, the focus in this study was exclusively on the domain of grammar. No conclusions should be drawn about the value of explicit knowledge about other aspects of language, such as pronunciation, vocabulary, and communication strategies.

5.5 Conclusion

This study has contributed to the area of FFI research in a number of ways. First, counter to current thinking, this study has come to the conclusion that the value of explicit knowledge is limited. In an attempt to demonstrate the presence of an interface between explicit and implicit knowledge, no evidence of such an interface has been provided. In the experiment that was conducted, one group of students received explicit instruction in two grammar structures while another group received implicit instruction. In addition, a no-treatment control group was included. Having explicit knowledge of the target structures has not proved to be an advantage to L2 learners in that no differences in implicit knowledge were observed between the implicit and explicit treatment groups. Because the strong interface position supposes unconditional conversion of explicit knowledge into implicit knowledge, it is falsified by these outcomes. Indirect effects of explicit knowledge upon implicit knowledge cannot be ruled out; in fact, this study has found that explicit instruction effectively promoted implicit knowledge of the simple morphological target structure, if the L2 had a similar structure. The parallelism between the first and second language either simplifies the explicit knowledge to such an extent that monitoring becomes a possibility, or it allows for successful transfer to occur. No further interactions between instruction and grammatical progress were observed.

This study also gives rise to reconsider earlier FFI findings. In part, this study was motivated by a bias in FFI research towards the use of decontextualized knowledge tests at the expense of tests assessing the ability to use grammatical knowledge in natural settings. The outcomes clearly suggest that measures assessing grammatical knowledge in isolated contexts cannot readily be taken as representative of the learner’s implicit grammatical knowledge. FFI findings need to be reconsidered according to how progress was measured. Given that the use of measures of natural language have been rare, replication studies will be welcome.

158 CHAPTER 5

For example, as this study has shown, but also pointed out by the DeKeyser (2003), there are surprisingly few studies that actually compared explicit and implicit types of instruction. Most FFI studies have laboured under the assumption that explicit types of FFI are necessary, and have moved on to investigate the effects of different kinds of instruction in interaction with numerous other variables. Such studies, too, need to be considered in the light of the kind of measures used.

Unusual to FFI studies, even to those that have used measures of spontaneous language use, was the qualitatively oriented exploration of the free written response data. This analysis has proved to be quite rewarding, though. Perhaps the most important outcome was the clear differentiation in the ability to correctly use conditional and causal subordinate clauses. The finding suggests that language learners do not see the parallelism between the two manifestations of the structure, and they seem to learn each manifestation independently. On the basis of these findings, and in as far as possible, the analyses for subordination have been conducted separately for each type. These findings have also provided suggestions as to how to define structure complexity. Complexity might be directly related to the number of different manifestations a structure can appear in. Although it is not entirely clear what would constitute a single manifestation, it seems fair to conclude that syntactic structures generally have more than morphological structures. Thus, this explanation would explain why instruction has primarily been found to be effective with morphological structures.

This study emphasizes that L2 research is in desperate need of longitudinal studies that investigate how L2 learners start to use their developing L2 knowledge. Not only would this inform theories of knowledge and representation, it is probably also an important precursor for the kind of study conducted here. If one wants to make claims about the effect of instruction on grammatical development, it is important to understand precisely how particular structures emerge in the L2 learner’s language. With simple frequency counts and with points of measurement that were quite far apart, this study has been able to uncover interesting features of learner language. Much can probably be learned if learner language is more systematically examined.

This study underscores Doughty’s claim that: “the case for explicit instruction

has been overstated.” (2003, p. 274). However, these results should not be taken to mean that second language learning can do without instruction, as the no interface position has often been interpreted. Although in this study the effects

DISCUSSION 159

of instruction were not spectacular, it has been very clearly demonstrated the instruction is an important catalyst of learning (Larsen-Freeman & Long, 1991; Long, 1983). The purpose of instruction should be to intensify the input. The focus should be on meaningful interaction, but some explicit FFI would probably do more good than harm. Grammar instruction should not aim for the establishment of explicit knowledge. Rather, for grammar instruction to be optimally effective, it should be geared towards the provision of structures in realistic and functional settings, and teaching the structure in isolated contexts should be avoided. After all, the goal of second language grammar teaching is not conscious knowledge of grammar, but proficiency.

Appendices

Appendix A1: Screenshots of the computer programme

162 APPENDICES

Appendix A2: A detailed outline of the instruction

Treatment group 1: EI in the degrees of comparison; II in subordinate clauses

Treatment group 2: EI in subordinate clauses; II in the degrees of comparison

Advertising: Chapter 1, What is advertising? Target structure: Subordinate clauses - Rule explanation (screen 1 and 2;

App. A4)

- Ex. 1: 6 yn; ‘Is this statement correct?’

- Ex. 1: 6 yn; ‘Is this a complex sentence?’

- Ex. 2: 4 mc; ‘Which sentence is correct?’

- Ex. 2: 4 mc; ‘Is this a complex sentence?’

- Rule explanation (screen 3; App. A4)

- Dialogue flooded with target structure, unenhanced.

- Dialogue flooded with target structure, underlined.

- Ex. 3: 5 mc; ‘What does X say about Y?’

- Ex. 3: 5 mc; ‘Which sentence does not contain a SubC?’

Advertising: Chapter 2, Tricks of the trade Target structure: Degrees of comparison

- Rule explanation (screen 1 and 2; App. A3)

- Ex. 1: 6 mc; ‘Which word in the sentence is a form of DoC?

- Ex. 1: 6 tf. Is this sentence true or false?

- Ex. 2: 6 mc; Choose the proper form from the list?

- Ex. 2: 6 mc; ‘What is true according to the text?

- Dialogue flooded with target structure, underlined.

- Dialogue flooded with target structure, unenhanced.

- Ex. 3: 5 mc; which sentence does not contain a form of the DoC?

- Ex. 3: 5 mc; Questions about dialogue: What does X say about Y?

APPENDICES 163



Advertising: Chapter 3, Making commercials Target structure: Subordinate clauses


- Ex. 1: 4 mc.; text comprehension - Ex. 1: 4 mc; ‘Identify the sub. conjunction’


- Ex. 2: 7 mc.; Questions about commercials

- Ex. 2: 7 mc; ‘which sentence is correct’, focuses on place of verb.

- Ex. 3: 8 gf; verbs in sub. clauses

- Ex. 3: 8 se; indentifying sub. conjunctions in dialogue

- Ex. 4: 5 mc; questions about written dialogue

- Ex. 4: 4 mc; ‘which sentence is correct’, place of verb.

Advertising: Chapter 4, Rules of advertising Target structure: Degrees of comparison

- Rule explanation (screen 3 and 4, App. A3)

- Ex. 1 and 2 combined; 14 gf and mc; first identify correct structure; then identify as comparative or superlative

- Ex. 1: 4 mc; ‘Which statement is true?’

- Ex. 2: 10 yn; ‘Do you agree with the statement?’

- Ex. 3: 9 se; ‘identify forms of the DoC in written dialogue’ (9)

- Ex. 3: 9 gf; choose between two given adjectives (marked for DoC)

Note. EI=explicit instruction; II=implicit instruction; dd=drag and drop; mc=multiple choice; oq=open questions; se= search of target; tf=true or false statements; yn= yes or no questions.

164 APPENDICES



The Netherlands and water: Chapter 5, Below sea level Target structure: Degrees of comparison


- Ex. 1: 3 mc; Which sentence contains a comparative or superlative?

- Ex. 1: 3 mc; text comprehension


- Ex. 2: 6 gf; Fill in a form of the DoC. - Ex. 2: 6 gf; Fill in a word to create a logical sentence.

- Ex. 3: gf; Gap text; fill in a form of the DoC

- Ex. 3: 11 dd; to create a logical text.

The Netherlands and water: Chapter 6, Flood disaster Target structure: Subordinate clauses - Rule explanation (screen 6; App.

A4) - Ex. 1: 6 tf. Is the statement about

the text true of false/ - Ex. 2: 5 oq. Finishing sentences on

the basis of the text. ‘This could happen, because … ‘

- Ex. 1 and 2 combined: 11 oq. Combining 2 given sentences into 1. Subordinate conjunction to be used was given.

- Ex. 3: 9 dd; drag and drop verbs of sub. clauses to create logical dialogue.

- Ex. 3: 9 se; search subordinate conjunction in written dialogue

- Ex. 4: 4 mc; text comprehension - Ex. 4: 4 mc; which sentence is incorrect

APPENDICES 165



Olympic games: Chapter 7, Everyone wants to compete Target structure: Degrees of comparison

- Ex. 1: 4 gf; Choose the proper adjective, and make a form of the DoC.

- Ex. 1: 4 gf; choose between three adjectives to make a logical sentence






- Ex. 4: 11 gf; choose between four adjectives to create a logical text

Olympic games: Chapter 8, Peace and friendship Target structure: Subordinate clauses

- Ex. 1: 13 oq; text comprehension ‘This is, because … ‘

- Ex. 1: 13 oq; Combine two sentences into one. Subordinate conjunction to be used was given

- Ex. 2: 12 gf; choose between four adjectives to create a logical text

- Ex. 2: 12 mc; identifying which is the correct verb phrase location.

Note. EI=explicit instruction; II=implicit instruction; dd=drag and drop; mc=multiple choice; oq=open questions; se= search of target; tf=true or false statements; yn= yes or no questions.

166 APPENDICES

Appendix A3: Rule explanation of the degrees of comparison

Screen 1 Heel vaak wil je mensen of dingen met elkaar vergelijken. Hiervoor kun je de trappen van vergelijking gebruiken. Kijk maar eens: [ images of objects in different sizes ] Groot – groter - grootst [ images of differently priced objects] Goedkoop – goedkoper - goedkoopst

Screen 2 In Nederlandse zinnen kun je de trappen van vergelijking heel vaak tegenkomen. Kijk maar: Van deze shampoo wordt je haar nog zachter Onze patatjes zijn de lekkerste patatjes De goedkoopste radio’s vind je bij Mediamarkt De onderstreepte woorden zijn voorbeelden van de trappen van vergelijking.

Screen 3 Herinner je je de trappen van vergelijking nog? [ images of objects in different sizes ] Goedkoop – goedkoper – goedkoopst Met de trappen van vergelijking kun je mensen of dingen met elkaar vergelijken. Kijk maar: Ik vind aardrijkskunde een leuk vak. Geschiedenis is nog leuker, maar ik

vind economie het leukst.

APPENDICES 167

Screen 4 De trappen van vergelijking hebben twee vormen:

- De vergrotende trap (leuker) - De overtreffende trap (leukst)

De vergrotende trap maak je met de letters –er. De overtreffende trap maak je met de letters –st. Kijk maar:

Vergrotende trap Overtreffende trap Leuk Mooi Goedkoop gek

Leuker Mooier Goedkoper gekker

Leukst Mooist Goedkoopst gekst

Screen 5 Kijk eens naar deze trappen van vergelijking:

Lange klinker: Hoog Laag

Hoger Lager

Hoogst laagst

Bij woorden met een lange klinker schrijf je de vergrotende trap met één klinker en één medeklinker. Kijk eens naar deze trappen van vergelijking:

korte klinker: Nat Kort

Natter Korter

Natst kortst

Bij woorden met een korte klinker schrijf je de vergrotende trap met één klinker en twee medeklinkers

168 APPENDICES

Appendix A4: Rule explanation of subordination

Screen 1 Grammatica: Zinnen met ‘omdat’, ‘zodat’… Met woorden zoals: omdat, zodat, als, hoewel, terwijl, enz… kun je van 2 zinnen 1 zin maken. 2 zinnen 1 - Vroeger werd veel reclame voor huisvrouwen gemaakt. 2 - Zij kochten meestal de boodschappen 1 zin: Vroeger werd veel reclame voor huisvrouwen gemaakt, omdat zij

meestal de boodschappen kochten. Als je zo van 2 zinnen 1 zin maakt, noem je dat een samengestelde zin.

Screen 2 Belangrijk: Zinnen met woorden zoals: Omdat, zodat, als toen terwijl hoewel Noem je: Samengestelde zinnen.

APPENDICES 169

Screen 3 Een zin met ‘omdat’ of ‘zodat’ bestaat altijd uit een hoofdzin en een ondergeschikte zin. De busreis gaat veel sneller, als je naar muziek luistert. Hoofdzin ondergeschikte zin De ondergeschikte zin begint met ‘als’, ‘omdat’, ‘zodat’, enz. Let op! De hoofdzin staat niet altijd vooraan! Kijk maar eens. Hier is de zin omgedraaid: Als je naar muziek luistert, gaat de busreis veel sneller, ondergeschikte zin Hoofdzin

Screen 4 In hoofdstuk 1 heb je geleerd wat een samengestelde zin is. Weet je het nog? Als je van 2 zinnen 1 zin maakt met woordjes zoals omdat, als, zodat, dan heb je een samengestelde zin. Een voorbeeld van een samengestelde zin: Reclamemakers maken meerdere storyboards, zodat een bedrijf kan

kiezen. Voegwoord Let op! De woordjes waarmee je een samengestelde zin maakt, noem je een voegwoord. Dit zijn dus allemaal voegwoorden: Omdat – doordat – nadat – zodat – terwijl – als – toen – hoewel

170 APPENDICES

Screen 5 Je weet nu wat een voegwoord is. Maar let op, er is iets bijzonders aan de hand. De volgorde van de woorden in de zin met het voegwoord wordt anders. Kijk maar: Het maken van een reclamespotje is moelijk. …, omdat het maken van een reclamespotje moeilijk is. In de zin met het voegwoord staan de werkwoorden allemaal aan het eind van de zin! Nog een voorbeeld: Mensen werken bij een reclamebureau …, als mensen bij een reclameburea u werken.

Screen 6 Als je een samengestelde zin moet maken, let dan op de werkwoorden. Kijk maar eens naar het voorbeeld. De dijken braken midden in de nacht door. …, omdat de dijken midden in de nacht door braken. In de zin met het voegwoord staan de werkwoorden allemaal aan het eind van de zin. Nog een voorbeeld: Het water stroomde het land in. …, zodat het water het land in stroomde.

APPENDICES 171

Appendix B1: Grammaticality judgement test

Goed of fout? Lees de zinnen. In veel zinnen staat een fout. Zie jij de fout? Zet een steep onder het foute woord. Let op:

- Veel zinnen, maar niet alle zinnen zijn fout. - Zet alleen een streep als de zin fout is. - Er is steeds 1 woord fout. Dus ook maar 1 streep zetten. - Je hoeft de zin niet te verbeteren.

Voorbeeld:

Ahmed heb zijn haar geverfd, want het was blond en nu is het rood!

GOED/FOUT

Fatimah heeft geen tijd om met zijn vrienden te gaan skaten.

GOED/FOUT

Nu jij!

1 Weet je niet dat planten dood gaan, als je geeft ze geen water?

GOED/FOUT

2 De jurk van Alida is mooi dan de jurk van Roos. GOED/FOUT

3 Samira ging morgen even naar de dokter, want ze heeft een zere knie.

GOED/FOUT

4 Ik wil graag een ringetje in mijn neus, omdat ik vind dat ontzettend mooi.

GOED/FOUT

5 Vanavond zijn er een hele leuke show op de televisie. GOED/FOUT

6 Van alle leraren vind ik onze leraar geschiedenis het leuk.

GOED/FOUT

7 De mobiele telefoon ging drie keer over, voordat Ali kon hem opnemen.

GOED/FOUT

172 APPENDICES

8 De klas geven de leraar een hele grote bos bloemen. GOED/FOUT

9 Ik vind het saai op school, dus ik hopen dat het snel vakantie is.

GOED/FOUT

10 Fatima moet rennen voor de bus, zodat ze komt niet te laat op school.

GOED/FOUT

11 Ik vind Donald Duck de saaie strip die er is. GOED/FOUT

12 Ik ga naar huis, omdat ik heb thuis nog heel veel te doen.

GOED/FOUT

13 Henk gaat heel veel geld sparen, zodat hij kan een scooter kopen.

GOED/FOUT

14 De film van gisteravond was de slechte die ik ooit gezien heb!

GOED/FOUT

15 Gisteren heb ik lekker in het park wandelen. GOED/FOUT

16 Anne is best lang, maar Lien is nog veel lang. GOED/FOUT

17 De leraar is kwaad, als wij weer eens komen te laat. GOED/FOUT

18 Ik hou niet van boeken lezen, maar stripboeken vind ik wel leuk.

GOED/FOUT

19 Ruth heeft drie kwartier gewacht, voordat ze eindelijk was aan de beurt.

GOED/FOUT

20 Deze koffer is wel groot, maar ik heb een veel grote koffer nodig.

GOED/FOUT

APPENDICES 173

Appendix B2: Controlled production

Samengestelde zinnen Maak van 2 zinnen 1 zin, gebruik steeds het woord tussen haakjes. Voorbeeld: (omdat) Farcha moet heel erg lachen. Erik doet raar. …………………………………………………………………………………… Nu jij! (als) Ik ga met je mee. Het is mooi weer. …………………………………………………………………………… (omdat) Anna gaat op de fiets vandaag. De bussen rijden niet. …………………………………………………………………………… (doordat) Ik was laat voor de les. Het regende ontzettend lang en hard. …………………………………………………………………………… (hoewel) Wij gaan morgen naar het strand. Het weer is morgen niet goed. …………………………………………………………………………... (omdat) We gaan samen naar huis. We wonen in dezelfde straat. …………………………………………………………………………… (als) Sammy zingt graag. Hij staat onder de douche. …………………………………………………………………………… (zodat) Erik werkt heel hard Hij heeft vanavond tijd voor zijn vrienden. …………………………………………………………………………… (omdat) Suna geeft Masha bloemen. Masha heeft haar goed geholpen. …………………………………………………………………………… (toen) Ik heb de politie gebeld. Ik hoorde een harde klap. ………………………………………………………………………… (als) Ik kan naar huis. De auto is eindelijk gerepareerd. ………………………………………………………………………………

174 APPENDICES

De trappen van vergelijking Vul het woord tussen haakjes in, zodat de zin weer klopt. Voorbeeld: (goedkoop) Een krant is …………….. dan een boek. (leuk) Van alle sporten is voetbal het ………………… (lang) Met de bus duurt de reis ………………… dan met de trein. (grappig) Vanavond is er een ………………… film op televisie. (aardig) David is de ………………… jongen die ik ken. (mooi) Ahmed heeft een mooie mobiele telefoon, maar die van Jana is ……………… (oud) Die meneer is de ………………… die ik ooit gezien heb. (zacht) Deze bank is wel zacht, maar de bank die ik wil kopen moet nog ………… zijn. (gezellig) Ik vond het feest van gisteravond heel ………………… (groot) Ik heb grote handen, maar Ahmeds handen zijn het ………………… (duur) Een auto is ………………… dan een fiets.

APPENDICES 175

Appendix B3: The free written response task

Vraag 1: (SubC 1)

Kijk naar deze strip. Er zitten beertjes in de boom.

Waarom gaan de oren van de beertjes hangen? ………………………………………………………………………..........................

Vraag 2: (DoC 1)

Je bent in een winkel om een nieuwe broek te kopen. Je hebt twee broeken gepast. Ze passen allebei goed, dus je moet kiezen. Welke kies je? ………………………………………………………………………..........................

Vraag 3: (DoC 2) (DoC 3) (DoC 4) / (SubC 2)

Je gaat een radio kopen. Kijk goed naar deze radio’s.

Wat valt je op? ……………………………………………………………………….......................... Wat zijn de verschillen? ……………………………………………………………………….......................... Welke zou jij kopen? ……………………………………………………………………….......................... Waarom? ………………………………………………………………………..........................

176 APPENDICES

Vraag 4: (SubC 3)

Wanneer wordt je boos op je vriend/vriendin? ………………………………………………………………………..........................

Vraag 5: (DoC 5)

Een meneer is heel dik. Hij zou minder moeten eten, maar hij eet juist steeds meer. Hij eet, en hij eet, en hij eet. Wat gebeurt er met de meneer? ………………………………………………………………………..........................

Vraag 6: (SubC 4) (SubC 5)

Kijk naar deze strip. Boes staat op de trap. Hij plukt appels.

Waarom is Boes boos? ……………………………………………………………………….......................... Waarom maakt Boes planken op de slagtanden van de olifant? ………………………………………………………………………..........................

Vraag 7: (DoC 6)

Kijk naar de foto’s. Op de foto’s staat dezelfde man. Hij is wel een beetje veranderd.

Wat valt je op? ………………………………………………………………………..........................

Vraag 8: (SubC 6)

Wanneer moet je lachen? ………………………………………………………………………..........................

APPENDICES 177

Vraag 9: (DoC 7) / (SubC 7)

Je gaat bij iemand logeren. Maar er is geen bed. Je moet kiezen: je mag op de bank slapen of op de vloer. Wat kies je? ……………………………………………………………………….......................... Waarom? ………………………………………………………………………..........................

Vraag 10: (SubC 8)

Kijk naar deze strip:

Waarom rent de mol weg? ………………………………………………………………………..........................

Vraag 11: (DoC 8)

Kijk naar de mannetjes.

Wat valt je op? ………………………………………………………………………..........................

178 APPENDICES

Vraag 12: (DoC 9) / (SubC 9) (DoC 10) / (SubC 10) (DoC 11) / (SubC 11)

Jij gaat een auto kopen. Je kunt kiezen uit deze 3 auto’s.

Stel je voor: Je hebt 3 kinderen. Welke auto neem je dan? ……………………………………………………………………….......................... Waarom? ……………………………………………………………………….......................... Stel je voor: Je hebt niet zo veel geld. Welke auto neem je dan? ……………………………………………………………………….......................... Waarom? ……………………………………………………………………….......................... Stel je voor: Je houdt van snel rijden. Welke auto neem je dan? ……………………………………………………………………….......................... Waarom? ………………………………………………………………………..........................

Vraag 13: (SubC 12)

Maria stampt de kamer uit, en ze slaat heel hard met de deur. Wat denk jij? Waarom doet zij dat? ………………………………………………………………………..........................

Vraag 14: (DoC 12) (DoC 13) / (SubC 13)

Kijk goed naar deze twee brillen.

Wat zijn de verschillen tussen de twee brillen? ……………………………………………………………………….......................... Welke bril zou jij liever dragen? ……………………………………………………………………….......................... Waarom? ………………………………………………………………………..........................

APPENDICES 179

Vraag 15: (SubC 14)

Kijk naar deze strip over de pinguïns:

Waardoor vallen de pinguïns in het water? ………………………………………………………………………..........................


Kijk naar de tabel: Jan, Piet en Kees hebben aan een fietswedstrijd meegedaan.

Deelnemer Tijd Jan 2:21:07 Piet 2:19:58 Kees 2:23:34

Wie heeft gewonnen? ……………………………………………………………………….......................... Hoe weet je dat? ………………………………………………………………………..........................


Jan wil naar huis. Hij kan met de bus en met de trein. De trein gaat zonder stoppen naar huis. De bus stopt nog op allerlei plaatsen onderweg. Jan kiest de trein. Waarom kiest Jan de trein? ………………………………………………………………………..........................

180 APPENDICES

Vraag 18: (SubC 17)

Wanneer krijg je dorst? ………………………………………………………………………..........................

Vraag 19: (SubC 18)

Kijk naar deze strip over de twee vogels:

Waardoor valt de ene vogel om? ………………………………………………………………………..........................

Vraag 20: (DoC 16) (DoC 17) / (SubC 19)

Kijk naar de mobiele telefoons:

Wat is het verschil? ……………………………………………………………………….......................... Welke koop je? ……………………………………………………………………….......................... Waarom koop je die? ………………………………………………………………………..........................

APPENDICES 181

Vraag 21: (SubC 20)

Kijk naar deze strip over de zeehond:

Waarom doet de zeehond laarzen aan? ………………………………………………………………………..........................


Je vriend/vriendin is jarig. Je wilt hem/haar een bloem geven.

Welke bloem kies je? ……………………………………………………………………….......................... Waarom? ………………………………………………………………………..........................

182 APPENDICES

Appendix C1: The C-test

INVULTOETS

Maak de woorden af. Het woord is altijd de helft of de helft +1 langer.

De leer_____ wordt sn___ weer be___. Sla het woord over als je het woord niet weet. Een voorbeeld:

Slapeloosheid Van iedere 100 Nederlanders slapen er 39 erg slecht. Mensen die aan slapel_______ lijden, gebruiken heel va___ slaapmiddelen. Van slaapmiddelen wo___ je rustig en j__ gaat beter slapen. Ma___ ze hebben ook nad____. Je lichaam went er___. Daardoor moet je ste____ meer pillen slikken o__ in slaap te ko___. Als je ermee stopt, krijg je ontwenningsverschijnselen: je gaat zweten, je wordt misselijk, en je moet braken.

Nu jij! Zweet Je lichaam bestaat voor het grootste deel uit water. Water krijg je bin____ door te drinken e__ door te eten. H__ water gaat het lic____ uit via plassen, poe____, ademen, en zweten. Hoe____ een mens zweet i__ verschillend. Iemand die bijvoo______ zwaar werk doet, o__ fanatiek sport, zweet ve___. Maar je zweet o__ als je niks do___ of slaapt: zweten ga___ dag en nacht do___. Zweet bestaat voor 99 pro____ uit water. De re___ is zout. Lik maar eens over je bezwete bovenlip, dan proef je dat wel.

APPENDICES 183

Post Een postzegel is misschien wel het kleinste stukje papier dat geld waard is. Het geld d__ je voor e__ postzegel betaalt, ga___ naar TPG Post. D__ zorgt ervoor d__ alle brieven e__ pakjes die vers_____ worden bij d__ geadresseerde aankomen. El__ dag worden i__ Nederland meer d__ 22 miljoen bri____, kaarten en pak___ verstuurd. Als j__ deze allemaal ach____ elkaar legt, kr___ je een r__ van drieduizend kilo_____. Dat is onge_____ van Amsterdam na___ het noorden v__ Afrika. Je ku___ postzegels ook verza_____. Miljoenen mensen op de wereld doen dat. Het internet De laaste jaren is het internet heel erg populair geworden. Heel veel men____ zijn aangesloten, e__ ze zetten infor______ op het inte____ over de din____ die ze inter______ vinden. Daardoor k__ je op h__ internet veel vin____. Ben je o__ zoek naar lek____ recepten, het laa____ nieuws, de sne____ weg naar j__ vakantieadres, informatie ov___ je hobby? O__ het internet i__ altijd een webs____ met leuke o__ handige informatie t__ vinden. Het inte____ lijkt op e__ bibliotheek die j__ gewoon thuis ku__ bezoeken. Handig to__? En tegenwoordig kun je ook al winkelen op internet.

184 APPENDICES

Appendix C2: the rote memory test

WOORDJESTOETS

- Leer de woorden uit je hoofd! - Je hebt 2 minuten de tijd! Zweeds – Nederlands Samtal - Gesprek Pussa - Zoenen Mormor - Oma Stel - Stijf Byxor - Broek Elak - Gemeen Och - En Ljuvlig - Lekker Bro - Brug Liten - Klein Barn - Kind Ny - Nieuw Slang - Fietsband Tretton - Dertien Pris - Prijs Bakom - Achter

APPENDICES 185

Appendix C3: the grammatical sensitivity test

EEN PUZZELTOETS Dit is een moeilijke toest, want je krijgt geen uitleg, maar je moet zelf uitzoeken hoe het moet. Je moet dus puzzelen! Je krijgt wel 4 voorbeelden. Dus, kijk goed naar de voorbeelden, en doe het dan zelf. Wat moet je doen? In zin A is een zinsdeel onderstreept. In zin B zijn 4 zinsdelen onderstreept. Welk zinsdeel in B is hetzelfde als in A? Is dat 1, 2, 3 of 4. Zet een rondje om het goeie antwoord. Voorbeelden:

A: Janneke loopt naar huis. B: De oude mevrouw eet elke dag een appel.

1 2 3 4 A: Miriam geeft mij een appel. B: Farcha slaapt lekker op de bank.

1 2 3 4 A: Aan de boom hangen lekkere appels. B: Er staat een bankje in de tuin.

1 2 3 4 A: Moeder bakt zelf een brood. B: Hij koopt mooie bloemen voor zijn vriendin.

1 2 3 4 Nu jij!

1. A: Mevrouw Jansen is snel naar het Ziekenhuis gebracht. B: Het kleine kind zit op de bank een koekje te eten. 1 2 3 4

2. A: Het meisje loopt snel naar huis. B: Hans gebruikt suiker in de thee. 1 2 3 4

3. A: In Nederland leven 17 miljoen mensen. B: In dit huis ligt de kamer naast de keuken. 1 2 3 4

186 APPENDICES

4. A: Alle kinderen lopen snel naar de grote school. B: De hond heeft een mooi hok. 1 2 3 4

5. A: Zij heeft met haar auto een ongeluk gehad. B: Ik wil dat je Anna mijn boek terug geeft. 1 2 3 4

6. A: Zij schaatst met haar vriendinnen op de ijsbaan. B: Pieter let niet op en zakt op zijn schaatsen door het ijs. 1 2 3 4

7. A: Gaat Jan met de auto naar zijn werk? B: De trein is drie kwartier te laat op het station aangekomen. 1 2 3 4

8. A: Harry Potter heeft een grappige bril op zijn neus staan. B: De leraar gaf de leerling een paar bandjes om thuis te luisteren. 1 2 3 4

9. A: Ricardo en Steven doen de afwas niet graag. B: Ik heb een kadootje gekregen van Henk. 1 2 3 4

10. A: Mijn radio heb ik van hem gekregen. B: De zwerver heeft mij zijn jas gegeven. 1 2 3 4

11. A: Mario kreeg vorige week een heel groot kado. B: Waarom heb jij die boeken niet aan hem gegeven? 1 2 3 4

12. A: De kinderen spelen In het enge bos. B: In het gezellige huis is het licht, maar op straat zijn de lichten uit. 1 2 3 4

13. A: Weet Marie misschien iets meer over dit goeie nieuws? B: Velen hebben de finish van de race niet gehaald. 1 2 3 4

14. A: De oude man sloeg de hond met zijn wandelstok. B: Op de piano kan mijn buurman de mooiste muziek maken. 1 2 3 4

15. A: Het publiek gaf de zanger een warm applaus. B: Eet jij mijn taartje dan maar op. 1 2 3 4

APPENDICES 187

Appendix C4: the teacher questionnaire

Geef een gevoelsmatig oordeel op onderstaande stellingen. Vergelijk de leerling steeds met de groep.

1 of 5 betekent: is goed van toepassing op deze leerling. 2 of 4 betekent: is redelijk van toepassing 3 betekent: neutraal

Deze leerling… 1. vindt het leuk op school 1 2 3 4 5 vindt het niet leuk op school

2. is heel gemotiveerd om

Nederlands te leren 1 2 3 4 5

is niet gemotiveerd om Nederlands te leren

3. spreekt graag Nederlands 1 2 3 4 5 spreekt niet graag Nederlands

4. heeft veel aanmoediging

nodig 1 2 3 4 5

heeft niet veel aanmoediging nodig

5. houdt van structuur 1 2 3 4 5 houdt niet van structuur

6. heeft gevoel voor taal 1 2 3 4 5 heeft weinig gevoel voor taal

7. heeft weinig moeite met

het begrijpen van grammatica-uitleg

1 2 3 4 5heeft veel moeite met het

begrijpen van grammatica-uitleg

8. kan goed dingen

onthouden 1 2 3 4 5 kan slecht dingen onthouden

9.* is gericht op details 1 2 3 4 5 is gericht op globale kenmerken

10. is zorgvuldig 1 2 3 4 5 is warrig

11. is bedachtzaam 1 2 3 4 5 is impulsief

12. is intellectueel 1 2 3 4 5 is emotioneel

13. is onafhankelijk van

anderen 1 2 3 4 5 is afhankelijk van anderen

14.* is gereserveerd 1 2 3 4 5 is sociaalvoelend

15.* is dominant 1 2 3 4 5 is volgzaam

16.* is introvert 1 2 3 4 5 is extravert

17.* is zeker 1 2 3 4 5 is onzeker

References

Authorware (1999). San Francisco: Macromedia. Aarts, F. G. A. M & H.C. Wekker (1993). A contrastive grammar of English and

Dutch. Groningen: Martinus Nijhoff. Alons, L., N. Bienfait & et al. (1999). Zebra: Nederlands als tweede taal voor

anderstaligen in het voortgezet onderwijs. Amsterdam: Meulenhoff Educatief. Anderson, J. (1995). Learning and memory: An integrated approach. New York:

Wiley. Andringa, S. J., N. Bienfait, F. Kuiken & W. van der Schaaf (2000). Zebra

grammatica. Amsterdam: Meulenhoff Educatief. Bercken, J. H. L. van den & M.J.M. Voeten (2002). Variantianalyse: De GLM

benadering. Groningen: Stenfert Kroese. Berenst, J., N. Bienfait, D. Hofstede & N. van der Schaaf (1999). Organisatie van

de eerste opvang. In: M. van de Laarschot & G. Lemmens (eds.). Nederlands als tweede taal in het voortgezet onderwijs. Amsterdam: Meulenhoff Educatief.

Bialystok, E. (1989). Psycholinguistic dimensions of second language proficiency. In: W. Rutherford & M. A. Sharwood Smith (eds.). Grammar and second language teaching: A book of readings. New York: Newbury House.

Bialystok, E. (1994a). Analysis and control in the development of second language proficiency. Studies in Second Language Acquisition, 16, 2: 157-168.

Bialystok, E. (1994b). Representation and ways of knowing: Three issues in second language acquisition. In: N. C. Ellis (eds.). Implicit and explicit learning of languages. London: Academic Press.

Bialystok, E. (2002). Cognitive processes of L2 users. In: V. J. Cook (eds.). Portraits of the L2 user. Clevedon: Multilingual Matters.

Bienfait, N. (2002). Grammatica-onderwijs aan allochtone jongeren. Dissertation: Rijksuniversiteit Groningen

Bley-Vroman, R. (1988). The fundamental character of foreign language learning. In: W. Rutherford & M. A. Sharwood Smith (eds.). Grammar and second language teaching: A book of readings. Rowley, MA: Newbury House.

Brumfit, C. J. (1994). The linguist and the language teaching profession: Ghost in a machine? In: R. Barasch & C. Vaughan James (eds.). Beyond the monitor

190 REFERENCES

model: Current theory and practice in second language acquisition. Boston, MA: Heinle and Heinle.

Carr, T.H. & T. Curran (1994). Learning about structure. Studies in Second Language Acquisition, 16, 2: 205-230.

Carrol, S. & M. Swain (1993). Explicit and implicit negative feedback: An empirical study of the learning of linguistic generalizations. Studies in Second Language Acquisition, 15, 357-386.

Carroll, J. B. & S.M. Sapon (1959). Modern languages aptitude test. New York: Psychological Corporation.

Carswell Pennington, M. (1995). New ways in teaching grammar. Alexandria: Teachers of English to Speakers of Other Languages.

Chomsky, N. (1965). Aspects of theory and syntax. Cambridge: MIT Press. Cohen.J. (1988). Statistical power analysis for the behavioral sciences. Hillsdale, NJ:

Lawrence Erlbaum. Day, E.M. & S.M. Shapson (2001). Integrating formal and functional apporaches

to language teaching in French immersion: An experimental study. Language Learning, 51, supplement 1: 47-80.

DeKeyser, R. (1995). Learning second language grammar rules: An experiment with a miniature linguistic system. Studies in Second Language Acquisition, 17, 379-410.

DeKeyser, R. (1997). Beyond explicit learning: Automatizing second language morphosyntax. Studies in Second Language Acquisition, 19, 195-221.

DeKeyser, R. (1998). Beyond focus on form: Cognitive perspectives on learning and practicing in L2 grammar. In: C. Doughty & J. Williams (eds.). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

DeKeyser, R. (2000). The robustness of critical period effects in second language acquisition. Studies in Second Language Acquisition, 22, 4: 499-533.

DeKeyser, R. (2001). Automaticity and automatization. In: P. Robinson (eds.). Cognition and second language instruction. Cambridge: Cambridge University Press.

DeKeyser, R. (2003). Implicit and explicit learning. In: C. Doughty & M. H. Long (eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing.

Doughty, C. (2001a). Cognitive underpinning of focus on form. In: P. Robinson (eds.). Cognition and second language instruction. Cambridge: Cambridge University Press.

REFERENCES 191

Doughty, C. (2001b). Cognitive underpinnings of focus on form. In: P. Robinson (eds.). Cognition and second language instruction. Cambridge: Cambridge University Press.

Doughty, C. (2003). Instructed SLA: Constraints, compensation and enhancement. In: C. Doughty & M. H. Long (eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing.

Doughty, C. & E. Varela (1998). Communicative focus on form. In: C. Doughty & J. Williams (eds.). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

Doughty, C. & J. Williams (1998). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

Doughty, C. & J. Williams (1998). Pedagogical choices in focus on form. In: C. Doughty & J. Williams (eds.). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

Drenth, P. & J. van Wieringen (1969). Verbale aanleg test '69. Lisse: Swets & Zeitlinger.

Ellis, N.C. (2002). Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition, 24, 143-188.

Ellis, N. C. (2003). Constructions, chunking, and connectionism: The emergence of second language structure. In: C. Doughty & M. H. Long (eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing.

Ellis, R. (1990). Instructed second language acquisition. Oxford: Basil Blackwell. Ellis, R. (1994a). A theory of instructed second language acquisition. In: N. C. Ellis

(eds.). Implicit and explicit learning of languages. London: Academic Press. Ellis, R. (1994b). The study of second language acquisition. Oxford: Oxford

University Press. Ellis, R. (1995). Appraising second language acquisition theory in relation to

language pedagogy. In: G. Cook & B. Seidlhofer (eds.). Principle in practice in applied linguistics. Oxford: Oxford University Press.

Ellis, R. (1997). SLA research and language teaching. Oxford: Oxford University Press.

Ellis, R. (2001). Investigating form-focused instruction. Language Learning, 51, supplement 1: 1-46.

Ellis, R. (2002). Does form-focused instruction affect the acquisition of implicit knowledge? A review of the research. Studies in Second Language Acquisition, 24, 223-236.

192 REFERENCES

Ellis, R. (2004). The definition and measurement of L2 explicit knowledge. Language Learning, 54, 2: 227-275.

Field, A. (2000). Discovering statistics using SPSS for Windows : advanced techniques for the beginner. London: Sage.

Gass, S.M. (1988). Integrating research areas: A framework for second language studies. Applied Linguistics, 9, 198-217.

Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press.

Goldschneider, J.M. & R. DeKeyser (2001). Explaining the "Natural order of L2 morpheme acquisition" in English: A meta-analysis of multiple determinants. Language Learning, 51, 1: 1-50.

Goodson, F. E. (2003). The evolution and function of cognition. Mahwah, NJ: Lawrence Erlbaum Associates.

Graaff, R. de (1997a). Differential effects of explicit instruction on second language acquisition. Dissertation: Vrije Universiteit Amsterdam

Graaff, R. de (1997b). The experanto experiment: Effects of explicit instruction on second language acquisition. Studies in Second Language Acquisition, 19, 2: 249-276.

Gregg, K. R. (2001). Learnability and second language acquisition theory. In: P. Robinson (eds.). Cognition and second language instruction. Cambridge: Cambridge University Press.

Han, Y. & R. Ellis (1998). Implicit knowledge, explicit knowledge and general proficiency. Language Teaching Research, 2, 1: 1-23.

Harley, B. (1989). Functional grammar in French immersion: A classroom experiment. Applied Linguistics, 10, 3: 331-359.

Harley, B. (1993). Instructional strategies and SLA in early French immersion. Studies in Second Language Acquisition, 15, 2: 245-260.

Hulstijn, J.H. (1997). Second language acquisition research in the laboratory: Possibilities and limitations. Studies in Second Language Acquisition, 19, 131-143.

Hulstijn, J.H. (2002). Towards a unified account of the representation, processing and acquisition of second language knowledge. Second Language Research, 18, 3: 193-223.

Hulstijn, J.H. & R. de Graaff (1994). Under what conditions does explicit knowledge of a second language facilitate the acquisition of implicit knowledge? A research proposal. AILA Review, 11, 97-112.

Jourdenais, R., M. Ota, S. Stauffer, B. Boyson & C. Doughty (1995). Does textual enhancement promote noticing? A think-aloud protocol analysis. In: R.

REFERENCES 193

Schmidt (eds.). Attention and awareness in foreign language learning. Honolulu: University of Hawai'í, Second Language Teaching & Curriculum Center.

Klein-Braley, C. (1997). C-tests in the context of reduced redundancy testing: an appraisal. Language Testing, 14, 1: 47-84.

Krashen, S. (1981). Second language acquisition and second language learning. Oxford: Pergamon.

Krashen, S. (1982). Principles and practice in second language acquisition. Oxford: Pergamon Press.

Krashen, S. (1985). The input hypothesis: Issues and implications. Lincolnwood, Il.: Laredo Publishing.

Krashen, S. (1994). The input hypothesis and its rivals. In: N. C. Ellis (eds.). Implicit and explicit learning of languages. London: Academic Press.

Krashen, S. & T. Terrell (1983). The natural approach: Language acquisition in the classroom. Oxford: Pergamon.

Kuiken, F. (1999). Grammatica. In: M. van de Laarschot & G. Lemmens (eds.). Nederlands als tweede taal in het voortgezet onderwijs. Amsterdam: Meulenhoff Educatief.

Kwakernaak, E. (1995). Grammatica in het VTO: vernieuwingskansen en –grenzen. Levende Talen, 505, 592-597.

Larsen-Freeman, D. & M.H. Long (1991). An introduction to second language acquisition research. New York: Longman.

Levelt, W. J. M. (1999). Producing spoken language: A blueprint of the speaker. In: C. M. Brown & P. Hagoort (eds.). The neurocognition of language. Oxford: Oxford University Press.

Lightbown, P. M. (1998). The importance of timing in focus on form. In: C. Doughty & J. Williams (eds.). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

Lightbown, P.M. (2001). Classroom SLA research and second language teaching. Applied Linguistics, 21, 4: 431-462.

Long, M.H. (1983). Does second language instruction make a difference? A review of the research. Tesol Quarterly, 17, 359-382.

Long, M.H., S. Inagaki & L. Ortega (1998). The role of implicit negative feedback in SLA: Models and recasts in Japanese and Spanish. The Modern Language Journal, 82, 3: 357-371.

Lyster, R. (1994). The effect of functional-analytic teaching on aspects of French immersion students' sociolinguistic competence. Applied Linguistics, 15, 3: 263-287.

194 REFERENCES

Mackey, A. (1999). Input, interaction, and second language development: An empirical study of question formation in ESL. Studies in Second Language Acquisition, 21, 557-587.

Mackey, A. & J. Philp (1998). Conversational interaction and second language development: Recasts, responses, and red herrings? The Modern Language Journal, 82, 3: 338-356.

Macrory, G. & V. Stone (2000). Pupil progress in the acquisition of the perfect tense in French: The relationship between knowledge and use. Language Teaching Research, 4, 54-82.

Meisel, J., H. Clahsen & M. Pienemann (1981). On determining developmental stages in natural second language acquisition. Studies in Second Language Acquisition, 3, 1: 109-135.

Muranoi, H. (2000). Focus on form through interaction enhancement: Integrating formal instruction into a communicative task in EFL classrooms. Language Learning, 50, 4: 617-673.

Nattinger, J. R. & J.S. DeCarrico (1992). Lexical phrases and language teaching. Oxford: Oxford University Press.

Norris, J.M. & L. Ortega (2000). Effectiveness of L2 instruction: a research synthesis and quantative meta-analysis. Language Learning, 50, 3: 417-528.

O'Malley, J.M., A.U. Chamot & C. Walker (1987). Some applications of cognitive theory to second language acquisition. Studies in Second Language Acquisition, 9, 287-306.

Odlin, T. (2003). Cross-linguistic influence. In: C. Doughty & M. H. Long (eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing.

Pawley, A. & F. Syder (1983). Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In: J. C. Richards & R. Schmidt (eds.). Language and communication. London: Longman.

Pienemann, M. (1988). Determining the influence of instruction on L2 speech processing. AILA Review, 5, 40-72.

Pienemann, M. (1989). Is language teachable? Psycholinguistic experiments and hypothesis. Applied Linguistics, 10, 59-79.

Pine, J.M. & E.V.M. Lieven (1997). Slot and frame patterns and the development of the determiner category. Applied pscyholinguistics, 18, 123-138.

Reber, A.S. (1989). Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118, 3: 219-235.

Robinson, P. (1995). Aptitude, awareness, and the fundamental similarity of implicit and explicit second language learning. In: R. Schmidt (eds.).

REFERENCES 195

Attention and awareness in foreign language learning. Honolulu: Second language teaching and curriculum center.

Robinson, P. (1996). Learning simple and complex second language rules under implicit, incidental, rule-search and instructed conditions. Studies in Second Language Acquisition, 18, 233-247.

Robinson, P. (1997). Individual differences and fundamental similarity of implicit and explicit adult second language learning. Language Learning, 47, 1: 45-99.

Robinson, P. (2001). Individual differences, cognitive abilities, aptitude complexes and learning conditions in second language acquisition. Second Language Research, 17, 4: 368-392.

Robinson, P. (2002). Learning conditions, aptitude complexes and SLA: a framework for research and pedagogy. In: P. Robinson (eds.). Individual differences and instructed language learning. Amsterdam: John Benjamins.

Robinson, P. (2003). Attention and memory during SLA. In: C. Doughty & M. H. Long (eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing.

Salaberry, M.R. (1997). The role of input and output practice in second language acquisition. The Canadian Modern Language Review, 53, 2: 422-451.

Sanz, C. & K. Morgan-Short (2004). Positive evidence versus explicit rule presentation and explicit negative feedback: A computer-assisted study. Language Learning, 54, 1: 35-78.

Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 11, 2: 129-158.

Schmidt, R. (1994). Deconstructing consciousness in search of useful definitions for applied linguistics. AILA Review, 11, 11-26.

Schmidt, R. (1995). Attention and awareness in foreign language learning. Honolulu: University of Hawaii Second Language Teaching and Curriculum Center.

Schmidt, R. (2001). Attention. In: P. Robinson (eds.). Cognition and second language instruction. Cambridge: Cambridge University Press.

Schuurs, U. (1999). De ISK-toetsen. Arnhem: Cito. Sharwood Smith, M. A. (1988). Consciousness raising and the second language

learner. In: W. Rutherford & M. A. Sharwood Smith (eds.). Grammar and second language teaching: A book of readings. New York: Newbury House.

Sharwood Smith, M.A. (1993). Input enhancement in instructed SLA. Studies in Second Language Acquisition, 15, 165-179.

Sinclair, J. (1991). Corpus, concordance, collocation. Oxford: Oxford University Press.

196 REFERENCES

Skehan, P. (1998a). A cognitive approach to language learning. Oxford: Oxford University Press.

Skehan, P. (1998b). A framework for the implementation of task-based instruction. Applied Linguistics, 17, 1: 38-64.

Tomlin, R.S. & V. Villa (1994). Attention in cognitive science and second language acquisition. Studies in Second Language Acquisition, 16, 2: 183-203.

Trahey, M. & L. White (1993). Positive evidence and preemption in the second language classroom. Studies in Second Language Acquisition, 15, 2: 181-204.

VanPatten, B. (1987). On babies and bathwater: Input in foreign language learning. Modern Language Journal, 71, 156-164.

VanPatten, B. (1996). Input processing and grammar instruction. New York: Ablex. VanPatten, B. & T. Cadierno (1993). Explicit instruction and input processing.

Studies in Second Language Acquisition, 15, 2: 225-241. VanPatten, B. & C. Sanz (1995). From input to output: Processing instruction and

communicative tasks. In: F. R. Eckman et al. (eds.). Second language acquisition theory and pedagogy. Mahwah, NJ: Lawrence Erlbaum.

Weir, C. (1988). Communicative language testing with special reference to English as a foreign language. Exeter: University of Exeter.

White, J. (1998). Getting the learners attention: A typographical input enhancement study. In: C. Doughty & J. Williams (eds.). Focus on form in classroom SLA. Cambridge: Cambridge University Press.

Williams, J. & J. Evans (1998). What kind of focus and on which forms? In: C. Doughty & J. Williams (eds.). Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.

Wolfe-Quintero, K., S. Inagaki & H. Kim (1998). Second language development in writing: Measures of fluency, accuracy & complexity. Honolulu: University of Hawaii Second Language Teaching and Curriculum Center.

Nederlandse samenvatting

Als grammaticaonderwijs wordt ingezet in tweedetaalleerprogramma’s, is kennis van de grammatica zelden een op zichzelfstaand doel. Het beoogde doel is meestal taalvaardigheid: het kunnen communiceren in de tweede taal. De vraag is dan wat de meerwaarde is van bewuste kennis van de regels van een tweede taal. Deze vraag is al vaak behandeld in onderzoek naar vormgericht tweedetaalonderwijs, en het beeld dat daaruit spreekt wijst erop dat meer leerwinst behaald kan worden, als grammaticaregels expliciet aangeboden worden in tweedetaalleerprogramma’s. Tegelijkertijd moet geconstateerd worden dat dit beeld gebaseerd is op onderzoek dat haken en ogen vertoont. Zo zijn er veel verschillende methodes van onderzoek toegepast, is er weinig replicerend onderzoek gedaan, en zijn mogelijk interfererende variabelen vaak onvoldoende gecontroleerd. Daarbij komt dat de effecten van grammaticaonderwijs veelal zijn gemeten met toetsen die kennis van grammatica in een geïsoleerde context meten, dat wil zeggen, aan de hand van taaltaken waarin grammaticastructuren in afzondering worden aangeboden. Of grammaticaonderwijs ook leidt tot het kunnen gebruiken van de onderwezen regels in realistische gebruikssituaties is weinig onderzocht, terwijl dit toch het ultieme doel is. Vanwege deze beperkingen in onderzoeksmethodiek is de vraag die in dit onderzoek gesteld wordt – of leerlingen baat hebben bij bewuste kennis van de grammatica van een tweede taal voor de ontwikkeling van hun tweedetaalvaardigheid – nog steeds moeilijk te beantwoorden.

Taalverwervingstheorieën die zich richten op instructie maken over het algemeen onderscheid tussen expliciete en impliciete linguïstische kennis. Het eerste type kennis verwijst naar declaratieve, bewuste (en meestal ook bewust geleerde) kennis van een tweede taal, terwijl het tweede type kennis verwijst naar procedurele kennis: het soort kennis dat wordt toegepast in alledaagse, spontane gebruikssituaties. Zo bekeken wordt het nut van bewuste regelkennis – en dus van expliciet grammaticaonderwijs – bepaald door de mate waarin expliciete en impliciete kennis interageren. Het onderwijzen van en oefenen met de regels van de tweede taal leidt tot expliciete grammatica kennis, maar is alleen dan zinvol als het ook leidt tot impliciete kennis. Er kunnen in de literatuur drie hypotheses onderscheiden worden die uitspraken doen over het belang van expliciete kennis bij het verwerven van impliciete kennis: de non-interfacehypothese, de sterke-

198 NEDERLANDSE SAMENVATTING

interfacehypothese en de zwakke-interfacehypothese. Dit onderzoek moet gezien worden als een evaluatie van deze hypotheses.

De non-interfacehypothese veronderstelt twee volstrekt gescheiden

kennissystemen, en gaat ervan uit dat de expliciete kennis die leerlingen bewust moeten leren nauwelijks van waarde is voor de impliciete kennis die leerlingen onbewust verwerven. Deze hypothese is gebaseerd op de idee dat tweedetaalverwerving niet anders is dan eerstetaalverwerving, dat ook geheel onbewust plaatsvindt. De motor voor taalverwerving is een speciale taalleermodule, het language acquisition device, die genoeg heeft aan een ‘begrijpelijk’ taalaanbod. Een leerling in fase i zal op basis van i + 1 input een stap verder komen in de verwerving, tenminste als het ‘affectieve filter’, oftewel de mate waarin de leerling openstaat voor het leren, dat toelaat. In sommige gevallen kan expliciet geleerde kennis een positieve invloed hebben op de verwerving van impliciete kennis, bijvoorbeeld doordat expliciet leren het affectieve filter kan verlagen, maar dat laat onverlet de gescheidenheid van het expliciete en impliciete kennissysteem. De snelste weg naar taalvaardigheid is een begrijpelijk en motiverend taalaanbod, en expliciete instructie is hoogstens een omweg.

De sterke-interfacehypothese veronderstelt dat expliciete en impliciete kennis de uitersten zijn van een continu kennissysteem. Eén variant van de sterke-interfacehypothese ziet tweedetaalleren min of meer als een proces van automatisering, niet anders dan bijvoorbeeld leren autorijden. Oefening baart kunst: leerlingen beginnen typisch met expliciete kennis, en in de beginfase moeten leerlingen oefenen door deze kennis bewust toe te passen bij hun taalproductie. Dit leidt tot procedurele en uiteindelijk tot geautomatiseerde kennis. Een andere variant van de sterke-interfacehypothese stelt dat tweedetaalleerders beginnen met impliciete kennis, en dat zij zich geleidelijk bewust worden van de regels van de taal die zij leren. Hun kennis wordt dus al maar explicieter, en dit stelt hen in staat de regels te gebruiken in nieuwe gebruikscontexten. De verschillen tussen beide varianten van de sterke-interfacehypothese zijn aanzienlijk, maar voor allebei geldt dat expliciete instructie een bepalende factor is in het taalleerproces.

De zwakke-interfacehypothese, ten slotte, veronderstelt net als de non-interfacehypothese twee gescheiden kennissystemen die op verschillende manieren tot stand komen. De zwakke-interfacehypothese ziet niettemin een substantiële rol weggelegd voor expliciete kennis. De aanname is dat expliciete kennis de impliciete verwerving van grammatica vergemakkelijkt doordat het leerders beter in staat stelt structuren op te merken in het taalaanbod en te

NEDERLANDSE SAMENVATTING 199

vergelijken met de eigen productie. De zwakke-interfacehypothese probeert ook rekening te houden met de bevinding dat bepaalde aspecten van taal gehouden zijn aan vaste verwervingsvolgordes. De consequenties hiervan zouden kunnen zijn dat grammaticaonderwijs alleen effectief is als de instructie precies getimed is op het moment dat leerders in het juiste ontwikkelingsstadium zijn, of als de onderwezen structuren niet onderhevig zijn aan vaste verwervingsvolgordes. Het zou ook kunnen zijn dat de effecten van expliciete instructie vertraagd optreden, namelijk op het moment dat de leerder het juiste ontwikkelingsstadium bereikt.

Een aantal factoren dat mogelijk bepalend kan zijn voor de effectiviteit van grammaticaonderwijs doorkruist het interfacedebat nog. Aansluiten bij het ontwikkelingsstadium is al genoemd als een potentieel beslissende factor, maar ook de complexiteit van grammaticastructuren en cognitieve verschillen tussen individuele leerders zouden van invloed kunnen zijn. Zo is wel beweerd dat expliciete kennis van complexe structuren weinig zinvol is omdat tweedetaalleerders niet in staat zijn zulke kennis toe te passen, of omgekeerd dat tweedetaalleerders zelf wel in staat zijn expliciete kennis van simpele structuren op te doen, en juist baat hebben bij expliciete instructie van complexe structuren. Ook de betrouwbaarheid van de structuur (kent hij veel uitzonderingen) en de hoeveelheid toepassingscontexten zijn wel genoemd als succesfactoren. Wat betreft cognitieve verschillen is het argument dat sommige leerders simpelweg beter in staat zijn om te gaan met expliciete informatie ten behoeve van de ontwikkeling van hun tweedetaalvaardigheid.

Om meer inzicht te verkrijgen in het interfacedebat, en dan met name in de

aard en de gebruikswaarde van expliciete en impliciete kennis, zijn recente theorieën van tweedetaalverwerving geraadpleegd om zo de theoretische mogelijkheid van een interface tussen de twee soorten kennis opnieuw te kunnen beoordelen. Een belangrijk kenmerk van constructivistische theorieën van tweedetaalverwerving, die meer en meer aan invloed winnen, is dat zij er niet vanuit gaan dat bepaalde linguïstische principes zijn aangeboren. Grammaticale kennis is niet het product van taalspecifieke leermechanismen, maar van algemene associatieprocessen die kenmerkend zijn voor elk leerproces. In dit licht moet impliciete kennis gezien worden als een immens netwerk van ‘chunks’, losse woorden en frasen, die in de beginfase weinig onderling verband houden. Een hiërarchisch georganiseerd systeem ontstaat wanneer leerders frasen gaan herkennen als zijnde van hetzelfde type en lexicaal specifieke elementen vervangen door open categorieën. Zo zou bijvoorbeeld uit chunks als (veel)( plezier), (veel)(lawaai), (veel)(geld), etc. de chunk (veel X) kunnen ontstaan, waar X


vervangen kan worden door een groot aantal lexicaal specifieke chunks, maar ook door chunks die zelf open elementen bevatten, zoals (X auto’s). Linguïstische kennis begint dus als ongeanalyseerde kennis opgeslagen als chunks in het geheugen. Het grammaticale systeem ontstaat doordat leerders structuren steeds meer los kunnen zien van hun specifieke lexicale realisaties.

Het toepassen van het grammaticale systeem is tot op zekere hoogte een computationeel proces. Dit wordt bewezen door het gebruik van overgegeneraliseerde vormen als loopte door tweedetaalleerders, een vorm die ze niet opgepikt kunnen hebben uit hun omgeving. Toch is moeilijk vol te houden dat taalproductie strikt computationeel is. Regels kennen te vaak uitzonderingen, en corpusonderzoek heeft laten zien dat taalgebruik in hoge mate geïnstitutionaliseerd is. Bovendien zou computatie te veel geheugencapaciteit vergen. Dit alles wijst erop dat taalproductie en dus taalvaardigheid in belangrijke mate steunt op het direct ophalen en aaneen rijgen van chunks uit het geheugen. Aldus kunnen er drie fasen onderscheiden worden in het taalverwervingsproces. De eerste fase is de lexicalisatie fase waarin door middel van ‘noticing’ en ‘perceptual encoding’ chunks worden opgeslagen in het geheugen. Deze chunks vormen de input voor het syntacticalisatie proces, waarbij door middel van ‘analyse’ grammaticale structuur ontstaat. Relexicalisatie ten slotte, is het automatiseringsproces waarbij geanalyseerde stukken taal als chunks worden opgeslagen, zodat ze bruikbaar worden in communicatie. Op computatie wordt alleen teruggevallen als de situatie daarom vraagt en als de tijd het toestaat.

Als deze kenschets van impliciete kennis en het leren daarvan wordt teruggevoerd op het interfacedebat, dan lijkt het waarschijnlijk dat impliciete en expliciete kennis gescheiden kennissystemen zijn. Impliciete grammaticakennis is feitelijk het resultaat van sterke associatie; expliciete grammaticakennis daarentegen is niet anders dan enige andere feitenkennis, zoals weten dat ‘clavicula’ Latijn is voor ‘sleutelbeen’. Het is moeilijk voor te stellen dat expliciete kennis over zou kunnen gaan in impliciete kennis, en daarmee lijkt de sterke-interfacehypothese uitgesloten. Expliciete instructie zou niettemin een positief effect kunnen hebben op de verwerving van impliciete kennis, niet zozeer omdat zulke instructie leidt tot expliciete kennis, maar omdat het bijkomende impliciete leereffecten veroorzaakt. Het hier beschreven taalleerproces heeft niet onmiddellijk consequenties voor mogelijke effecten van ontwikkelingsstadia, de complexiteit van de grammaticastructuur en individuele cognitieve verschillen, maar het is moeilijk vast te stellen hoe zij interageren met vormgerichte instructie.


Empirische evidentie voor een interface tussen expliciete en impliciete kennis is gezocht in studies naar vormgericht tweedetaalonderwijs. In een overzichtsartikel bespreekt Ellis (2002) elf studies die het effect van vormgericht grammaticaonderwijs op impliciete kennis hebben onderzocht, maar het merendeel van deze studies maakte alleen gebruik van impliciete vormgerichte instructie. Overigens rapporteerden de meeste studies positieve effecten van zulke instructie, met name als vaste ‘formules’ of morfologische structuren onderwezen werden. Impliciet grammaticaonderwijs bleek weinig effectief voor syntactische structuren. Hoe dan ook, het artikel levert geen bewijzen voor een interface tussen expliciete en impliciete kennis, omdat niet aangenomen mag worden dat de vergeleken groepen van elkaar verschilden wat betreft expliciete kennis. Juist dit is de belangrijkste voorwaarde waaraan voldaan moet worden om een interface aan te tonen.

Duidelijke bewijzen kunnen geleverd worden door studies die expliciete en impliciete varianten van vormgerichte instructie met elkaar vergelijken, zodat de blootstelling aan de onderwezen structuur gelijk is. Bovendien is het van belang dat zowel expliciete als impliciete kennistoetsen zijn afgenomen. Er zijn weinig studies die aan deze criteria voldoen. Er is één studie gevonden die qua design voldoet aan bovengenoemde criteria. Door de criteria enigszins op te rekken door studies toe te staan die geen expliciete kennistoetsen hebben gebruikt, of die een expliciete instructiegroep hebben vergeleken met een ware-controlegroep (zij ontvingen dus geen vormgerichte instructie), kunnen vijf studies aangewezen worden die informatief kunnen zijn voor het interfacedebat. Drie van deze studies vonden geen verschillen tussen de expliciete instructiegroep en de impliciete instructiegroep of controlegroep. Twee studies lieten zien dat de expliciet onderwezen groepen het beter deden dan de controles, waarbij dat voor één van deze twee alleen gold voor het geschreven domein. Voor beide studies gold bovendien dat er duidelijke verschillen waren tussen de vergeleken groepen in blootstelling aan de onderwezen structuren, altijd in het voordeel van de expliciete instructie groepen. Hierdoor kan niet aangenomen worden dat de leerlingen in de expliciete condities voordeel hadden vanwege hun expliciete kennis, en meer in het algemeen moet geconcludeerd worden dat het bewijs voor een interface uiterst mager is.

Een aantal onderzoeken heeft gekeken naar de interactie tussen vormgerichte instructie en het aansluiten bij het ontwikkelingsstadium van leerlingen, de complexiteit van de grammaticastructuur of individuele cognitieve verschillen. Aansluiten bij het ontwikkelingsstadium van leerders lijkt van belang voor effectief vormgericht onderwijs, maar er is geen reden aan te nemen dat expliciete


instructie effectiever is dan impliciete instructie áls deze aansluiting bereikt is. Evenzo geldt dat er aanwijzingen zijn dat sommige structuren effectiever onderwezen kunnen worden dan andere (al is het beeld diffuus), maar niets suggereert dat een interface alleen geldt voor bepaalde structuren. Wat betreft de betrokkenheid van bepaalde cognitieve vaardigheden kan gezegd worden dat specifieke vaardigheden verband lijken te houden met expliciet dan wel impliciet leren, maar er is geen reden aan te nemen dat een interface alleen geldt voor leerders met een specifiek cognitief profiel.

Al met al is er reden genoeg om het belang van expliciete kennis, en daarmee

expliciete instructie, in twijfel te trekken. Daarom is in het onderhavige onderzoek gekeken naar de ontwikkeling van grammatica in spontane gebruikssituaties, de rol die expliciete en impliciete instructie hierbij spelen, en de interactie tussen de effectiviteit van de instructie en mogelijk interfererende variabelen. Het onderzoek was zowel beschrijvend als verklarend van aard, en er is gebruik gemaakt van een quasi-experimenteel onderzoeksdesign met herhaalde metingen, waarbij de ontwikkeling van twee grammaticastructuren is gevolgd door middel van zowel expliciete als impliciete kennismaten. De doelgroep voor dit onderzoek waren leerders van het Nederlands als tweede taal (gemiddeld 14 jaar oud). Zij ontvingen door middel van een computerprogramma expliciete of impliciete vormgerichte instructie; de controlegroep ontving geen instructie. De volgende onderzoeksvragen zijn gesteld:

1. Hoe ontwikkelen tweedetaalleerders de vaardigheid om de doelstructuren toe te passen in spontane gebruikssituaties?

2. Hoe hangen expliciete en impliciete vormgerichte instructie samen met de ontwikkeling van zowel expliciete als impliciete grammaticakennis?

3. Hoe beïnvloeden ontwikkelingsstadia, de moedertaal en individuele cognitieve verschillen het succes van expliciete en impliciete vormgerichte instructie?

4. Hangt de effectiviteit van expliciete en impliciete instructie af van de complexiteit van de onderwezen structuur?

De twee onderwezen grammaticastructuren waren de trappen van vergelijking

en onderschikking. Deze structuren zijn gekozen op basis van intensief en kleinschalig vooronderzoek, waaruit bleek dat het gebruik van beide structuren vrij goed kan worden uitgelokt, en dat ze min of meer gelijktijdig worden verworven door de doelgroep. Bovendien zijn de structuren contrasterend: de trappen van vergelijking is een morfologische en functioneel simpele structuur,


terwijl onderschikking een syntactische en complexe structuur is. Als gezegd werd de vormgerichte instructie via de computer aangeboden. Daarbij is geprobeerd functionele contexten te creëren. Leerlingen lazen teksten over thema’s zoals Reclame, de Olympische Spelen en Water. De oefeningen die leerlingen kregen bij deze teksten waren vaak expliciet dan wel impliciet gefocust op de doelstructuren, al waren er ook oefeningen bij die zich niet op een van de twee doelstructuren richtten. De expliciete instructie had puur het bijbrengen van en oefenen met bewuste regelkennis ten doel; de impliciete instructie was bedoeld om leerlingen bloot te stellen aan de doelstructuren in betekenisvolle contexten. De instructie was zo georganiseerd dat de twee experimentele groepen elkaars controle waren: leerlingen in groep 1 kregen expliciete instructie over de trappen van vergelijking en impliciete instructie over onderschikking; leerlingen in groep 2 kregen impliciete instructie over de trappen van vergelijking en expliciete instructie over onderschikking. Om de blootstelling aan de structuren zo gelijk mogelijk te houden is de inhoud van de expliciete en impliciete instructie zo precies mogelijk gematcht. Dit wil zeggen dat het aanbod van de doelstructuren qua hoeveelheid en context zo veel mogelijk hetzelfde waren. De duur van de instructie was relatief kort: gemiddeld ongeveer drie lesuren.

In dit onderzoek zijn twee grammaticale kennistoetsen afgenomen die respectievelijk expliciete en impliciete kennis maten. Deze toetsen zijn op drie momenten afgenomen, voor en onmiddellijk na de instructie, en nog eens twee maanden na het aflopen van het experiment. De expliciete kennistoets bestond uit grammaticaliteitsoordelen waarbij de leerlingen steeds incorrecte realisaties van de doelstructuur als incorrect moesten kunnen herkennen. De toets was vooral vóór de instructie moeilijk voor de leerlingen omdat ze de structuren nog niet kenden. Om die reden is het aantal items laag gehouden (zes voor de trappen van vergelijking en acht voor onderschikking), en zijn er makkelijke items toegevoegd over andere grammaticastructuren. De betrouwbaarheden waren behoorlijk. De impliciete kennistoets was een schrijftoets. Leerlingen moesten reageren op informele situaties die zo waren gekozen dat ze het gebruik van de doelstructuren op een onopvallende manier uitlokten. Er is gekozen voor een schrijftoets omdat de verwachting is dat een interface hier eerder zichtbaar zal worden dan bij mondelinge toetsen. De schrijfproducten werden per item gecodeerd voor correct, incorrect en geen gebruik. Ook is er gekeken naar de manier waarop en de context waarin de doelstructuren werden gerealiseerd. De toets onderscheidde achttien items voor de trappen van vergelijking en eenentwintig voor onderschikking. Voor deze toets waren de betrouwbaarheden


goed te noemen. Beide testen zijn op elk meetmoment steeds onveranderd afgenomen, waarbij de impliciete kennistoets altijd eerst werd afgenomen.

In totaal hebben 101 leerlingen van zogenaamde ‘internationale schakelklassen’ (ISK) als proefpersonen aan dit onderzoek deelgenomen. ISK leerlingen zijn afkomstig uit veel verschillende landen, en zijn vaak met of zonder ouders gevlucht uit hun land van herkomst. Zolang ze in afwachting zijn van de uitslag van vaak jaren lopende juridische procedures, krijgen ze in ISK’s intensieve Nederlandse les. De lesprogramma’s in ISK’s kenmerken zich door een sterke focus op functioneel en betekenisgericht onderwijs, waarbij expliciet grammaticaonderwijs een ondergeschikte rol speelt, en soms zelfs geheel afwezig is. De deelnemers aan dit onderzoek waren afkomstig van scholen uit het noorden en westen van Nederland, die werden gekozen op basis van nabijheid, de beschikbaarheid van leerlingen, en bereidheid tot deelname. De leerlingen werden geselecteerd voor deelname aan de hand van door ISK’s veelgebruikte, speciaal voor de doelgroep ontwikkelde, en gestandaardiseerde taalvaardigheidstoetsen. Tussen het tweede en derde meetmoment is door allerlei oorzaken een aanzienlijk deel van de leerlingen afgevallen. Zodoende waren er uiteindelijk van 76 leerlingen voldoende gegevens beschikbaar.

Er is geprobeerd om vast te stellen of de effectiviteit van de instructie nog is beïnvloed door andere, potentieel interfererende variabelen, te weten het ontwikkelingsstadium, de moedertaal, leeftijd, taalaanleg, motivatie en leerstijl. Het ontwikkelingsstadium van de leerling is geoperationaliseerd als een nominale variabele waarbij leerlingen aan de hand van het al dan niet voorkomen van correcte vormen van de doelstructuur in de impliciete grammaticale kennistoets zijn aangemerkt als ‘gevoelig’ of ‘ongevoelig’ voor de instructie. Voor het bepalen van de invloed van de moedertaal is ook een nominale variabele gecreëerd door vast te stellen of de doelstructuren al dan niet op een vergelijkbare wijze in de moedertaal van de leerlingen werden gerealiseerd. Daarom is voor de trappen van vergelijking gekeken naar morfologische dan wel perifrastische realisatie; voor onderschikking is gekeken naar de zinsstructuur in ondergeschikte zinnen. Leeftijd is gemeten in jaren. Taalaanleg is gemeten door middel van twee speciaal ontwikkelde in het Nederlands afgenomen toetsen: voor grammaticale gevoeligheid en geheugen. Motivatie en leerstijl, ten slotte, zijn bepaald met behulp van een docentenenquête.

De instructie werd door middel van een speciaal ontwikkeld computerprogramma aangeboden. De leerlingen werkten in klassikaal verband met de computer, en dit werd georganiseerd door hun docenten. Het gebruik van de computer had een aantal voordelen. Zo kon voorkomen worden dat de


resultaten werden beïnvloed door verschillen tussen docenten, en de blootstelling aan de doelstructuren kon heel precies gematched worden. Een belangrijk bijkomend voordeel was de mogelijkheid van volstrekt willekeurige toewijzingen tot de experimentele groepen, waardoor deze groepen werden gerepresenteerd door leerlingen van alle scholen. De controlegroep moest wel gevormd worden uit vaste klassen, omdat de leerlingen in deze conditie niet met het computerprogramma werkten. Het gebruik van de computer had bovendien nog het voordeel dat er een dissociatie ontstond tussen de instructie en de testafname: dit laatste geschiedde aan de hand van pen en papier, en in aanwezigheid van de onderzoeker.

Om een antwoord te krijgen op de eerste onderzoeksvraag hoe

tweedetaalleerders de doelstructuren leren gebruiken is gekeken naar correct gebruik van de doelstructuren in relatie tot specifieke realisatiekenmerken. Voor de trappen van vergelijking is onderscheid gemaakt tussen comparatieve en superlatieve vormen; predicatief en attributief gebruik; het gebruik van comparatieve bijzinnen; en het gebruikte adjectief. De trappen van vergelijking werden van het eerste naar het laatste meetmoment steeds vaker correct gebruikt, en steeds minder vaak incorrect. Er werden geen verschillen gevonden in correct gebruik voor een van de genoemde realisatiekenmerken, behalve als comparatieven werden gebruikt in combinatie met een comparatieve bijzin. In die gevallen werd ongeveer dertig procent minder correct gebruik gerealiseerd. Voor onderschikking is onderscheid gemaakt tussen conditionele en causale onderschikking; de gebruikte werkwoorden; en de complexiteit van de ondergeschikte bijzin. Hier viel op dat leerlingen veel beter in staat waren conditionele onderschikking correct te realiseren dan causale onderschikking. En als verder gedifferentieerd werd, dan bleek dat er ook meer fouten werden gemaakt als er meerdere werkwoorden in de ondergeschikte bijzin voorkwamen, of als de ondergeschikte bijzin complex was. Er is ook voor beide structuren geconstateerd dat er niet veel veranderde in de gebruikspatronen over drie maanden tijd. Leerlingen die alleen incorrect gebruik vertoonden op het eerste meetmoment, bleven dat vaak doen, en dat gold ook voor leerlingen die correct en incorrect gebruik afwisselden of die alleen correcte vormen gebruikten. De gemiddelde vooruitgang was langzaam en geleidelijk.

De analyses met betrekking tot de effecten van de vormgerichte instructie konden geen bewijs leveren voor de hypothese dat expliciete instructie, en daarmee expliciete kennis, een voordeel vormt bij de ontwikkeling van impliciete kennis. Zoals verwacht werd de ontwikkeling van expliciete kennis positief


beïnvloed door expliciete instructie, met name op de korte termijn; voor onderschikking gold dat op de lange termijn impliciete instructie even effectief was. Wat betreft de ontwikkeling van impliciete kennis is gebleken dat de leerlingen in de experimentele condities significante vooruitgang lieten zien, maar er waren geen verschillen tussen de instructie condities: ze bleken beide even effectief. Ook als die leerlingen werden uitgesloten die niet in staat waren geweest te profiteren van de expliciete instructie, en dus wellicht de expliciete kennis misten om daarvan te profiteren, bleven significante verschillen tussen de condities uit. De controlegroep, voor zover deze opgenomen kon worden in de analyses, bleef over het algemeen achter bij de experimentele groepen.

De vraag of de effectiviteit van de vormgerichte instructie nog werd beïnvloed door andere factoren is onderzocht voor een aantal variabelen. Zo is aangetoond dat het ontwikkelingsstadium van de leerling inderdaad van invloed is op de effectiviteit van de instructie. Voor beide doelstructuren gold dat leerlingen in het gevoelige ontwikkelingsstadium meer leerwinst behaalden, ongeacht of deze leerwinst expliciet of impliciet werd gemeten, en ongeacht het soort instructie. Er moet wel opgemerkt worden dat de groepen soms te klein werden om significante te kunnen constateren, waardoor mogelijke interacties met bijvoorbeeld instructie onopgemerkt gebleven kunnen zijn. De resultaten wat betreft de moedertaal hebben voor beide structuren geen significante effecten laten zien op zowel expliciete als impliciete grammaticale ontwikkeling. Maar voor de trappen van vergelijking is wel een interactie aangetoond tussen moedertaal en instructie: leerlingen die expliciete instructie ontvingen vertoonden meer impliciete leerwinst als de moedertaal de trappen van vergelijking ook morfologisch realiseerde. Dit is opmerkelijk, want het betekent dat expliciete instructie een voordeel biedt bij de impliciete verwerving van simpele of morfologische grammaticastructuren. Dit zou kunnen wijzen op een ‘interface onder voorwaarden’ tussen expliciete en impliciete kennis. Tot slot is nog gevonden dat leeftijd verband houdt met de ontwikkeling van expliciete kennis. Ongeacht de instructie bleken jongere leerlingen beter in staat expliciete kennis te ontwikkelen dan oudere. Voor de overige variabelen kon geen relatie aangetoond worden, of was het beeld diffuus.

Voor de vierde onderzoeksvraag betreffende de relatie tussen vormgerichte instructie en de complexiteit van de onderwezen structuur zijn geen aanvullende analyses gedaan. Er is alleen een vergelijking getrokken aan de hand van de hierboven beschreven analyses, en die hebben laten zien dat de resultaten voor beide structuren regelmatig verschilden.


De resultaten met betrekking tot de vraag hoe leerlingen grammaticastructuren leren gebruiken zijn in overeenstemming met de hypothese dat tweedetaalverwerving begint met het leren van chunks, en dat beginnend gebruik van grammaticastructuren afhankelijk is van specifieke lexicale realisaties. Het correct kunnen toepassen van de regels voor onderschikking bleek inderdaad af te hangen van specifieke lexicale vormen: veel leerlingen in dit onderzoek associeerden de verandering van zinsstructuur in ondergeschikte zinnen specifiek met het onderschikkende voegwoord als, en lieten de zinsstructuur ongemoeid in andere typen onderschikking. Dit onderscheid zou men niet verwachten als onderschikking als één structuur in het geheugen gerepresenteerd zou zijn. Voor de trappen van vergelijking zijn geen duidelijke gevallen gevonden van verbanden tussen correct gebruik en specifieke lexicale realisaties, maar dit is hoogstwaarschijnlijk omdat zulke verbanden niet zichtbaar te maken waren. Uit de data bleek ook dat leerlingen correcte en incorrecte vormen van de doelstructuren vrij lang naast elkaar gebruiken. Het verwervingsproces zou men kunnen karakteriseren als een proces waarbij correcte vormen zich langzamerhand verspreiden over verschillende gebruikscontexten en zo de overhand krijgen over incorrecte vormen. Het is overigens belangrijk op te merken dat incorrect dan niet betekent dat een regel foutief wordt toegepast, maar simpelweg niet. Incorrect gebruik van de trappen van vergelijking betekende dat er geen comparatief of superlatief suffix werd gebruikt terwijl de context die vereiste; incorrect gebruik van onderschikking betekende bijna altijd dat hoofdzinstructuur werd toegepast.

Zoals verwacht op basis van de besproken theorie en eerder onderzoek naar vormgerichte instructie geeft ook dit onderzoek weinig aanleiding tot het veronderstellen van een interface tussen expliciete en impliciete kennis. Het hebben van expliciete kennis is geen voordeel gebleken bij de verwerving van impliciete kennis, maar het is nadrukkelijk ook geen nadeel gebleken. De beste verklaring hiervoor is dat er weliswaar gescheiden kennissystemen bestaan, maar dat expliciete instructie, mits op een betekenisvolle wijze aangeboden, niet alleen leidt tot expliciete kennis, maar evengoed ook impliciete leerprocessen los kan maken. Op basis van zowel de sterke- als zwakke-interfacehypothese zou men echter een voordeel verwachten voor leerlingen die expliciete kennis bezitten. Omdat dit niet het geval is gebleken, ondersteunen de uitkomsten van deze studie de non-interfacehypothese. Wel moet opgemerkt worden dat de zwakke-interfacehypothese een vertraagd voordeel voorspelt. De vertraging van twee maanden die in deze studie is gehanteerd kan te kort geweest zijn. Bovendien kunnen op basis van dit onderzoek indirecte effecten van expliciete kennis op de


verwerving van impliciete kennis niet helemaal worden uitgesloten; deze studie suggereert zelfs dat expliciete instructie een voordeel oplevert voor eenvoudige structuren, mits die structuur op min of meer dezelfde wijze in de moedertaal wordt gerealiseerd. Echter, deze laatste bevinding wijst met meer waarschijnlijkheid op transfer vanuit de moedertaal dan op een interface tussen expliciete en impliciete kennis. Tot slot is het nog van belang op te merken dat de uitkomsten van dit onderzoek belangrijke implicaties hebben voor de interpretatie van eerder uitgevoerd vormgerichte instructieonderzoek: men kan instructie-effecten gevonden op basis van gedecontextualiseerde kennistoetsen niet generaliseren naar tweedetaalvaardigheid.

Afgezien van de bevinding dat expliciete instructie voordelig was voor die leerlingen die een vergelijkbare structuur als de trappen van vergelijking in hun moedertaal hadden, zijn er geen aanwijzingen gevonden dat de interface tussen expliciete en impliciete kennis afhangt van andere factoren. Met betrekking tot de ontwikkelingsstadia is wel gebleken dat vomgerichte instructie in zijn algemeenheid effectiever is als leerlingen zich in een gevoelig stadium bevinden, maar hierbij moet aangetekend worden dat de bepaling van gevoeligheid gebaseerd is op een vrij grof en mogelijkerwijs arbitrair criterium. Toekomstig onderzoek zou eerst verwervingsvolgordes moeten vaststellen (zo die daar zijn), en die gebruiken voor de operationalisatie van deze variabele. Ook leeftijd bleek gerelateerd te zijn aan de effectiviteit van de instructie, maar dan alleen voor expliciete kennis. Deze bevinding is opmerkelijk in het licht van de ‘fundamental difference’ hypothese, die stelt dat taalverwerving na een bepaalde leeftijd fundamenteel anders is. De uitkomsten met betrekking tot leeftijd suggeren dat dat alleen geldt voor expliciet leren.

In het verleden is het non-interfacestandpunt nogal eens gebruikt bij

pleidooien voor het afschaffen van grammaticaonderwijs. De resultaten van dit onderzoek moeten nadrukkelijk niet zo geïnterpreteerd worden. Integendeel, het feit dat expliciete instructie vanwege bijkomende impliciete leereffecten even effectief is gebleken als impliciete instructie is juist een reden om af en toe expliciete instructie op te nemen in tweedetaalleerprogramma’s. Het belang van vormgerichte instructie schuilt met name in het feit dat leerlingen intensiever blootgesteld kunnen worden aan de doelstructuren dan zou kunnen met alleen een begrijpelijk taalaanbod. Het is niet altijd eenvoudig om een grammaticastructuur op een zinvolle wijze en intensief te laten figuren in het taalaanbod: expliciete instructie zou juist een uitstekend voorwendsel kunnen zijn om dat te bereiken. Ook bestaat bij zowel leerlingen als docenten vaak de


behoefte om grammatica expliciet te behandelen, waardoor het taalleren voor beiden een aangenamere en gevarieerdere ervaring is. Wel moet de instructie zo vormgegeven zijn dat deze volop de gelegenheid biedt tot het op een betekenisvolle manier verwerken van de structuur. Want uiteindelijk is niet bewuste regelkennis, maar tweedetaalvaardigheid het doel.

Groningen dissertations in linguistics (GRODIL)

1. Henriëtte de Swart (1991). Adverbs of Quantification: A Generalized Quantifier Approach.

2. Eric Hoekstra (1991). Licensing Conditions on Phrase Structure. 3. Dicky Gilbers (1992). Phonological Networks. A Theory of Segment Representation. 4. Helen de Hoop (1992). Case Configuration and Noun Phrase Interpretation. 5. Gosse Bouma (1993). Nonmonotonicity and Categorial Unification Grammar. 6. Peter I. Blok (1993). The Interpretation of Focus. 7. Roelien Bastiaanse (1993). Studies in Aphasia. 8. Bert Bos (1993). Rapid User Interface Development with the Script Language Gist. 9. Wim Kosmeijer (1993). Barriers and Licensing. 10. Jan-Wouter Zwart (1993). Dutch Syntax: A Minimalist Approach. 11. Mark Kas (1993). Essays on Boolean Functions and Negative Polarity. 12. Ton van der Wouden (1994). Negative Contexts. 13. Joop Houtman (1994). Coordination and Constituency: A Study in Categorial

Grammar. 14. Petra Hendriks (1995). Comparatives and Categorial Grammar. 15. Maarten de Wind (1995). Inversion in French. 16. Jelly Julia de Jong (1996). The Case of Bound Pronouns in Peripheral Romance. 17. Sjoukje van der Wal (1996). Negative Polarity Items and Negation: Tandem

Acquisition. 18. Anastasia Giannakidou (1997). The Landscape of Polarity Items. 19. Karen Lattewitz (1997). Adjacency in Dutch and German. 20. Edith Kaan (1997). Processing Subject-Object Ambiguities in Dutch. 21. Henny Klein (1997). Adverbs of Degree in Dutch. 22. Leonie Bosveld-de Smet (1998). On Mass and Plural Quantification: The case of

French ‘des’/‘du’-NPs. 23. Rita Landeweerd (1998). Discourse semantics of perspective and temporal

structure. 24. Mettina Veenstra (1998). Formalizing the Minimalist Program. 25. Roel Jonkers (1998). Comprehension and Production of Verbs in aphasic Speakers.

26. Erik F. Tjong Kim Sang (1998). Machine Learning of Phonotactics. 27. Paulien Rijkhoek (1998). On Degree Phrases and Result Clauses. 28. Jan de Jong (1999). Specific Language Impairment in Dutch: Inflectional

Morphology and Argument Structure. 29. H. Wee (1999). Definite Focus. 30. Eun-Hee Lee (2000). Dynamic and Stative Information in Temporal Reasoning:

Korean tense and aspect in discourse. 31. Ivilin P. Stoianov (2001). Connectionist Lexical Processing. 32. Klarien van der Linde (2001). Sonority substitutions. 33. Monique Lamers (2001). Sentence processing: using syntactic, semantic, and

thematic information. 34. Shalom Zuckerman (2001). The Acquisition of "Optional" Movement. 35. Rob Koeling (2001). Dialogue-Based Disambiguation: Using Dialogue Status to

Improve Speech Understanding. 36. Esther Ruigendijk (2002). Case assignment in Agrammatism: a cross-linguistic

study. 37. Tony Mullen (2002). An Investigation into Compositional Features and Feature

Merging for Maximum Entropy-Based Parse Selection. 38. Nanette Bienfait (2002). Grammatica-onderwijs aan allochtone jongeren. 39. Dirk-Bart den Ouden (2002). Phonology in Aphasia: Syllables and segments in

level-specific deficits. 40. Rienk Withaar (2002). The Role of the Phonological Loop in Sentence

Comprehension. 41. Kim Sauter (2002). Transfer and Access to Universal Grammar in Adult Second

Language Acquisition. 42. Laura Sabourin (2003). Grammatical Gender and Second Language Processing: An

ERP Study. 43. Hein van Schie (2003). Visual Semantics. 44. Lilia Schürcks-Grozeva (2003). Binding and Bulgarian. 45. Stasinos Konstantopoulos (2003). Using ILP to Learn Local Linguistic

Structures. 46. Wilbert Heeringa (2004). Measuring Dialect Pronunciation Differences using

Levenshtein Distance. 47. Wouter Jansen (2004). Laryngeal Contrast and Phonetic Voicing: A Laboratory

Phonology.

48. Judith Rispens (2004). Syntactic and phonological processing in developmental dyslexia.

49. Danielle Bougaïré (2004). L'approche communicative des campagnes de sensibilisation en santé publique au Burkina Faso: Les cas de la planification familiale, du sida et de l'excision.

50. Tanja Gaustad (2004). Linguistic Knowledge and Word Sense Disambiguation. 51. Susanne Schoof (2004). An HPSG Account of Nonfinite Verbal Complements in

Latin. 52. M. Begoña Villada Moirón (2005). Data-driven identification of fixed

expressions and their modifiability. 53. Robbert Prins (2005). Finite-State Pre-Processing for Natural Language

Analysis. 54. Leonoor van der Beek (2005) Topics in Corpus-Based Dutch Syntax 55. Keiko Yoshioka (2005). Linguistic and gestural introduction and tracking of

referents in L1 and L2 discourse. 56. Sible Andringa (2005) Form-focused instruction and the development of second

language proficiency. GRODIL Secretary of the Department of General Linguistics P.O. Box 716 9700 AS Groningen The Netherlands

Documents

Form-focused instruction and the development of … than logical that teaching gramma ... either overtly or covertly, ... pressing in the light of the oft heard complaint in pedagogy