Upload
tessa
View
213
Download
1
Embed Size (px)
Citation preview
This article was downloaded by: [University of Western Ontario]On: 21 October 2014, At: 12:32Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK
European Journal of CognitivePsychologyPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/pecp20
Beginning adult L2 learners'sensitivity to morphosyntacticviolations: A self-paced readingstudyNatasha Tokowicz a & Tessa Warren aa Departments of Psychology and Linguistics andLearning Research & Development Center , Universityof Pittsburgh , Pittsburgh, PA, USAPublished online: 15 Jul 2010.
To cite this article: Natasha Tokowicz & Tessa Warren (2010) Beginning adult L2learners' sensitivity to morphosyntactic violations: A self-paced reading study, EuropeanJournal of Cognitive Psychology, 22:7, 1092-1106, DOI: 10.1080/09541440903325178
To link to this article: http://dx.doi.org/10.1080/09541440903325178
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, orsuitability for any purpose of the Content. Any opinions and views expressedin this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content shouldnot be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions,claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expresslyforbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
Beginning adult L2 learners’ sensitivity to
morphosyntactic violations: A self-paced reading study
Natasha Tokowicz and Tessa Warren
Departments of Psychology and Linguistics and Learning Research &
Development Center, University of Pittsburgh, Pittsburgh, PA, USA
This study investigated beginning adult second language (L2) learners’ sensitivity toL2 morphosyntactic violations as a function of cross-language similarity. Onlinesensitivity was indexed by self-paced reading times at: (1) the critical word at which theviolation could first be detected, (2) the post-critical word, and (3) the sentence-finalword. In conditions in which morphosyntactic marking systems were similar ordifferent in L1 and L2, reading times on the critical word were slower when it cueda violation than when it did not; however, this sensitivity was not apparent in aconstruction unique to L2. Slower reading times to violations spilled over onto thepost-critical word. Cross-language similarity also influenced sentence-final wordreading times. Despite this online sensitivity, post-sentence grammaticality judge-ments were generally poor. However, these judgements were influenced by morpho-syntactic markings on words after the critical word, suggesting that learners can makeuse of this information.
Keywords: L2 learning; Morphosyntactic violations; Competition Model; Self-
paced reading; Grammaticality judgements.
How do adults learn a second language (L2)? Answering this question requires
understanding how processing changeswith proficiency (e.g., McDonald, 1987).
The present study provides a snapshot of processing at the beginning of
Correspondence should be addressed to Natasha Tokowicz, Learning Research & Development
Center, 3939 O’Hara St., Room 634, University of Pittsburgh, Pittsburgh, PA 15260, USA.
E-mail: [email protected]
During the writing of this manuscript, NT was supported by NSF-BCS 0745372 and TW
was supported by NIH-HD053639. We thank Angels Rusinol for her assistance in designing the
experiment, and for creating the stimuli and programming the experiment. We also thank the
members of the PLUM lab at the University of Pittsburgh, especially Rhonda McClain, Alison
Phillips, and Courtney Smith, for testing participants and coding data. An earlier version of
these findings was reported at the 47th annual meeting of the Psychonomic Society, Houston,
TX, USA. We thank Wouter Duyck and an anonymous reviewer for comments on an earlier
version of this manuscript.
EUROPEAN JOURNAL OF COGNITIVE PSYCHOLOGY
2010, 22 (7), 1092�1106
# 2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/ecp DOI: 10.1080/09541440903325178
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
proficiency development and investigates the following questions: To what kinds
of morphosyntactic information are beginning adult L2 learners sensitive? Are
beginning adult L2 learners sensitive to morphosyntactic violations at the
earliest point at which they can be detected? What is the timecourse of this
sensitivity? What factors influence this timecourse (e.g., cross-language
similarity, amount of agreeing/disagreeing information in the sentence)?
Knowing what kinds of information beginning learners are sensitive to and
when they are sensitive to it is necessary to understand how L2 processing
changes with proficiency, and could have implications for instruction.
Previous research suggests that beginning learners use first language (L1)
cues in L2 comprehension. In particular, according to the Competition
Model (MacWhinney, 1997; see also MacWhinney, 2005), cues in the
language input are more or less available to the learner, and more or less
valid as indicators of function. Availability and validity determine the
strength of individual cues. Because an adult L2 learner’s L1 is entrenched,
initial L2 processing will follow L1 cues. Therefore, the Competition Model
predicts that beginning adult learners should easily process grammatical
patterns that are similar in L1 and L2, but have difficulty with ones that
differ in L1 and L2, because relying on L1 cues will lead to errors in this case.
In an experiment similar to the present one, Tokowicz and MacWhinney
(2005) examined online sensitivity to L2 morphosyntactic violations in
beginning adult learners using event-related potentials (ERPs). They tested
three cross-language similarity conditions: (1) similar, in which both
languages licensed a verb form under similar circumstances (verb progressive
aspect licensing), (2) different, in which one language did and the other did
not mark agreement in the lexical system under investigation, but for which
both languages did mark this kind of agreement in similar lexical systems
(possessive and definite determiner-noun number agreement), and (3)
unique, in which one language had an agreement system the other lacked
(determiner-noun gender agreement). Following the Competition Model,
they predicted that learners would be most sensitive in the similar condition
because of positive transfer and least sensitive in the different condition
because of conflicting cues; predictions for the unique condition depend on
input exposure because this system doesn’t exist in L1.
Similar: Su abuela cocina/*cocinando muy bien.
His/Her grandmother cooks/*cooking very well.
Different: Las/*La gomas estan en la bolsa.
ThePL/*TheSING erasers are in the bag.
MORPHOSYNTACTIC VIOLATIONS IN L2 1093
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
Unique: Mi amiga tiene un/*una apartamento grande.
My friend has aMASC/*aFEM large apartmentMASC.
Tokowicz and MacWhinney (2005) used the P600 ERP component, which
is typically more positive in response to ungrammatical than grammatical
sentences (e.g., Osterhout & Mobley, 1995), to measure online sensitivity.
Post-sentence grammaticality judgements indexed offline sensitivity. Con-
sistent with Competition Model predictions, P600s were more positive to
ungrammatical than grammatical sentences for the similar and unique
constructions, but not for the different construction. Despite this online
sensitivity, grammaticality judgements were at chance.
Other ERP studies have also examined sensitivity to L2 (morpho)syn-
tactic violations in adult learners. Some indicate that moderately-
proficient learners are sensitive to certain violations (e.g., Rossi, Gugler,
Friederici, & Hahne, 2006), whereas others suggest that sensitivity holds
only for more-proficient speakers (e.g., Hahne, Mueller, & Clahsen, 2006;
see van Hell & Tokowicz, in press, for a review). Recent work in this
area is increasingly focused on the influence of cross-language similarity
(e.g., Sabourin & Stowe, 2008; see, for reviews, Kotz, 2009; Tolentino &
Tokowicz, 2009).
The present study extended that of Tokowicz and MacWhinney (2005)
using a converging method: self-paced reading. This is important because
their stimulus presentation rate (300 ms per word; 350 ms blanks between
words) may have hampered beginning learners’ sentence comprehension
because of its speed. Self-paced reading avoids this concern and provides a
more complete measure of processing across the entire sentence, including
immediate disruptions, spillover effects, and/or wrap-up effects at sentence-
final words (Just & Carpenter, 1980). Additionally, the present study
explored an issue that McClain and Tokowicz (2006) raised in a reanalysis
of Tokowicz and MacWhinney, namely that grammaticality judgements were
related to the number of words following the critical word that agreed with it
in number and/or gender. This relationship suggests that beginning L2
learners’ offline judgements may be influenced not only by local agreement
between elements, but also by an accumulation of evidence across the entire
sentence.
We investigated L2 grammatical processing in native English speakers in
the first four semesters of university Spanish. We included four conditions
that represent three levels of similarity between morphosyntactic agreement
or licensing patterns in English and Spanish: Tokowicz and MacWhinney’s
(2005) three conditions and a second similar condition, demonstrative
determiner-noun number agreement.
1094 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
Similar2: Esa/*Esas clase empieza al mediodıa.
This/*These class begins at noon.
Participants read sentences with or without violations in a self-paced
word-by-word paradigm. After each, they judged its grammaticality and
rated their confidence in their judgement. Online sensitivity to violations was
evaluated by comparing reading times on the word at which the violationbecame evident in the ungrammatical conditions (critical word) to reading
times on the analogous word in the grammatical conditions. The word
following the critical word (n�1) was analysed to capture spill-over effects;
the sentence-final word was analysed to investigate sentence wrap-up. In an
exploratory manipulation, participants read either silently or aloud.
We expected participants to show online sensitivity to violations in similar
and unique constructions, given Tokowicz and MacWhinney’s (2005) findings
in a similar population. However, Jiang (2004, 2007) found that proficient L2learners failed to demonstrate sensitivity to morphosyntactic plurality cues in
self-paced reading. This insensitivity seems most likely due to properties of the
learners’ L1 (Chinese), but if learners in general are insensitive to the
morphosyntax of plurality, they should not detect violations in the different
and similar2 conditions. Tokowicz and MacWhinney’s learners did not show
sensitivity to violations in different constructions. If this was because learners
are slow to resolve competition between L1 and L2 cues, our self-paced
methodology might allow time for resolution and we might observe sensitivity.If learners’ processing is slow overall, sensitivity to violations might appear
only in later measures like spillover and final-word reading times. Alterna-
tively, if L2 learners’ processing is not slow, but fragile, such that newly
processed material harms or overwrites previously processed material, later
measures might show less sensitivity to violations than earlier measures.
METHOD
Design
We used a 2 modality (silent, aloud)�4 cross-language similarity (similar1,
similar2, different, unique)�2 grammaticality (grammatical, ungrammatical)
within-participants design.
Participants
Participants were 59 students enrolled in university Spanish courses (semesters
1�4) in Pittsburgh. They were required to be native English speakers with no
MORPHOSYNTACTIC VIOLATIONS IN L2 1095
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
exposure to other languages before age 13. Data from nine participants who
did not meet these criteria were excluded; data from two participants were lost.
Analyses were thus conducted on data from 48 participants. Ten participants
were in the first semester of Spanish, 12 were in each of the second and third
semesters, and 14 were in the fourth. Participants were compensated $15.
Materials
Four Spanish conditions represented three levels of cross-language similarity
(see earlier examples). In ‘‘similar1’’, verbs appeared in either the simple
present tense (grammatical) or in a progressive form that was ungrammatical
because a required auxiliary verb was absent. In ‘‘similar2’’, demonstrative
determiners did or did not agree in number with their head noun. In
‘‘different’’, definite or possessive determiners did or did not agree in
number with their head noun. In ‘‘unique’’, a determiner did or did not agree
in gender with its head noun. Three English conditions served as a baseline.
The first was like the ‘‘similar1’’ Spanish condition, and had either a licensed
simple or unlicensed progressive verb form. In the second English condition,
a reflexive anaphor did or did not agree in number with its antecedent (some
adapted from Osterhout & Mobley, 1995). The third English condition
involved subject-verb person agreement.
Verb aspect licensing: The kitten might scratch/*scratching the child.
Reflexive anaphor agreement: Mark used to talk to himself/*themselves often.
Subject-verb person agreement: Karen likes/*like to go to the zoo.
The position of the critical word varied across items. In conditions
manipulating determiners, the following noun was the critical word, because
that was the point at which the ungrammaticality became evident. For the
conditions manipulating verb aspect, the verb was the critical word. In
addition to the critical conditions, one English and three Spanish filler
conditions were included to increase variability in the grammatical construc-
tions presented. Each condition had a total of 16 items.1 One condition of each
item was assigned to each of four lists in a Latin-square design; these lists were
counterbalanced with respect to grammaticality and modality so that no
participant saw an item in more than one condition.
1 One item in ‘‘unique’’ had a typo and was excluded. Four items in English verb aspect
licensing were eliminated because in the ungrammatical conditions the critical verb could have been
interpreted grammatically, as beginning a reduced relative clause.
1096 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
Procedure
Participants first completed a version of the Waters, Caplan, and Hildebrandt
(1987) working-memory task. On each trial, an English sentence waspresented visually and participants made a binary judgement about its
plausibility. Participants were instructed to attempt to remember the final
word of each sentence. After each set of between two and six sentences, they
typed the final words they remembered from that set (order not being
important); four sets of each size were presented. Performance was indexed by
two measures: (1) set size span, or the set size at which the participant correctly
recalled all of the words on at least two of the four sets of that size, and (2) total
span, or the total number of words correctly recalled (out of 80).Participants then completed four blocks of self-paced reading trials2 in
one of two orders: Spanish-silent, English-silent, Spanish-aloud, English-
aloud; or Spanish-aloud, English-aloud, Spanish-silent, English-silent.
Participants were tape recorded during all blocks to verify silent reading
and to transcribe aloud reading.
A fixation cross appeared at the centre of the screen; the spacebar initiated
each trial and progressed the sentence word-by-word; all words appeared at the
centre of the screen. At the end of each sentence, a screen appeared asking ‘‘Isthis sentence acceptable?’’, to which the participant indicated ‘‘yes’’ or ‘‘no’’ by
pressing a button; participants were instructed to use grammaticality as the basis
for this judgement. Then, a screen appeared asking ‘‘How sure are you about
your decision?’’, to which the participant responded 0, 25, 50, 75, or 100% by
moving the mouse over the response and selecting the appropriate option.
Following this task, participants completed a language history ques-
tionnaire (Tokowicz, Michael, & Kroll, 2004) that gathered demographic
information and self-ratings of L1 and L2 reading, writing, speaking, andcomprehension proficiency. Participants also indicated all languages to
which they had been exposed and under what circumstances (Table 1).
RESULTS
Coding and data trimming
Critical word length was confounded with grammaticality in the similar1
condition, and word length and reading times were significantly correlated in
all blocks, rs�.17, .18, .46, and .32, for the English-aloud, English-silent,
Spanish-aloud, and Spanish-silent blocks, respectively, psB.01. Therefore,
reading times were length-adjusted (Ferreira & Clifton, 1986; Trueswell,
Tanenhaus, & Garnsey, 1994). Specifically, individual regression analyses
2 We used Linger software created by Doug Rohde to present the self-paced reading task.
MORPHOSYNTACTIC VIOLATIONS IN L2 1097
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
were performed for each participant in each block. Word length was regressed
from reading time; residual values were used in all reading time analyses.
Furthermore, we excluded residual reading times longer than 2000 ms from
analyses, accounting for 0.26, 0.56, 1.62, and 2.77% of the data in the English-
aloud, English-silent, Spanish-aloud, and Spanish-silent blocks, respectively.
Coding of aloud responses indicated that participants very rarely corrected
grammatical violations (0.7% of trials in English and 1.1% of trials in Spanish).
These trials were removed from analyses because corrections may affect readingtimes. Data from all other trialswere analysed regardless of judgement accuracy,
because previous studies have demonstrated dissociations between accuracy
and online sensitivity to violations (e.g., Tokowicz & MacWhinney, 2005).
Self-paced reading
English
Critical word. Critical word reading times were faster for grammatical
(M�3.34) than ungrammatical sentences (M�27.50), F1(1, 47)�46.58,
MSE�25506.15, pB.01, F2(1, 41)�30.63, MSE�12331.84, pB.01.
Additionally, there were theoretically uninteresting baseline differences
among conditions, F1(2, 94)�4.78, MSE�20201.17, pB.05, F2(2, 41)�3.28, MSE�7991.83, pB.05. An interaction between condition and
grammaticality was marginally reliable by participants, F1(2, 94)�3.10,
MSE�11154.66, p�.06,3 F2B1; the reflexive anaphor agreement condi-
tion showed the largest grammaticality effect.
TABLE 1Language history questionnaire data
Measure Results
Age (years) 18.9 (1.5)
Age began L2 (years) 14.5 (1.9)
L2 reading ability 5.8 (1.7)
L2 writing ability 5.3 (1.9)
L2 conversation ability 4.7 (1.9)
L2 speech comprehension ability 5.6 (2.1)
Standard deviations are given in parentheses. Reading, writing, conversational, and speech
comprehension ability were rated on a 10-point scale, with 1 indicating the lowest level of ability
and 10 indicating the highest level of ability. Most participants reported having no L2 or Spanish
as an L2, with two exceptions (one Arabic and one Hebrew). All but seven participants indicated
that they were not able to speak, read, write, or understand any language other than English or
Spanish (these participants’ answers were: ASL; French; French and Arabic; Greek; Portuguese
and Latin; Swedish).
3 Reported p and MSE values correspond to the Greenhouse-Geisser nonsphericity correction
for effects that violated the sphericity assumption.
1098 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
Word n�1. The main effect of grammaticality spilled over onto word n�1, with faster reading for grammatical (M��20.88) than ungrammatical
sentences (M�28.39), F1(1, 47)�13.62, MSE�25664.30, pB.01, F2(1,
41)�32.51, MSE�2867.69, pB.01. Word n�1 was also read more quickly
aloud (M��7.65) than silently (M�8.69), F1(1, 47)�4.38, pB.05,
F2(1, 41)�5.77, MSE�3881.12, pB.05.
Final word. Self-paced reading times for sentence-final words showed a
reversed main effect of grammaticality, with slower reading for grammatical
(M�41.99) than ungrammatical sentences (M��56.21), F1(1, 47)�19.29,
MSE�72001.44, pB.01, F2(1, 41)�38.11, MSE�5831.33, pB.01 (Figure 1).
Furthermore, grammaticality and modality interacted, F1(1, 47)�7.89,
MSE�53952.48, pB.01, F2(1, 41)�26.92, MSE�5265.39, pB.01. Simple-
effects tests indicated that the effect of grammaticality was significant in both
analyses for the silent modality, Fs�18.08, psB.01, but only by participants in
the aloud modality, F1�4.16, pB.05, F2B1.
Summary. Grammatical violations in L1 drove immediate disruption
that spilled over onto word n�1. Interestingly, this reading time pattern
reversed for the sentence-final word, suggesting that readers may have
engaged in more wrap-up processing (Just & Carpenter, 1980; Warren,
White, & Reichle, 2009) for grammatical than ungrammatical sentences. We
now turn to self-paced reading analyses for L2 Spanish.
Spanish
Critical word. Reading times on critical words were longer in ungram-
matical than grammatical sentences, F1(1, 46)�20.43, MSE�108723.39,
pB.01, F2(1, 59)�24.21, MSE�29559.99, pB.01. This effect was qualified
Figure 1. Length-adjusted reading times for final words in English sentences as a function of
modality and grammaticality.
MORPHOSYNTACTIC VIOLATIONS IN L2 1099
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
by an interaction with similarity condition, F1(3, 138)�2.93, MSE�82443.00, pB.05, F2(3, 59)�2.79, MSE�29559.99, pB.05 (Figure 2).
Simple-effects tests indicated a grammaticality effect for the similar1,
similar2, and different conditions, Fs�5.07, psB.05, but not for the unique
condition, FsB1. In the item analysis, modality and grammaticality
interacted, F1(1, 46)�2.26, MSE�110904.50, p�.14, F2(1, 59)�5.65,
MSE�14752.07, pB.05. Simple-effects tests on item means demonstrated
a larger grammaticality effect in the silent condition, F2aloud�8.68, pB.01;
F2silent�24.37, pB.01. Finally, there were uninteresting baseline differences
among conditions in the participant analysis, F1(3, 138)�7.95, MSE�103987.15, pB.014, F2(3, 59)�1.62, MSE�182443.62, p�.20.
Word n�1. Word n�1 was read more slowly in ungrammatical (M��26.14) than grammatical sentences (M��62.63), F1(1, 46)�4.68, MSE�53497.45, pB.05, F2(1, 55)�5.78, MSE�14350.72, pB.05. There were
uninteresting baseline differences among conditions in the participant analysis,
F1(3, 138)�4.13, MSE�65132.00, pB.01, F2(3, 55)�1.16, MSE�82911.39,
p�.33.
Final word. On the sentence-final word, similarity and grammaticality
interacted, F1(3, 135)�5.88, MSE�149881.15, pB.01, F2(3, 59)�8.86,
MSE�39980.06, pB.01 (Figure 3). Simple-effects tests indicated that the
grammaticality effect was significant only for the similar2 condition, Fs�
17.41, psB.01 (other FsB1). This interaction qualified a main effect of
grammaticality, F1(1, 45)�6.01, MSE�136982.54, pB.05, F2(1, 59)�5.35, MSE�39980.06, pB.05, and a main effect of similarity that was
uninterpretable given variation among the similarity conditions, F1(3,
135)�11.29, MSE�125222.05, pB.01, F2(3, 59)�3.41, MSE�138570.68, pB.05. Modality and similarity interacted, F1(3, 135)�2.83,
MSE�144108.97, pB.05, F2(3, 59)�3.41, MSE�51450.76, pB.05.
Simple-effects tests indicated that sentences were read more quickly silently
than aloud in only the similar2 condition, Fs�14.92, psB.01 (other FsB1).
This effect qualified a main effect of modality in the item analysis, F1(1, 45)�2.83, MSE�211803.27, p�.10, F2(1, 59)�4.800, MSE�51450.76,
pB.05.
Summary. Learners showed immediate disruption to grammatical
violations in L2 for the similar and different but not unique conditions.
On word n�1, disruption was evident for all conditions. Cross-language
similarity also modulated sensitivity to violations as evidenced by sentence-final
4 See Footnote 3.
1100 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
word reading times, which were longer in grammatical than ungrammatical
sentences in only the similar2 condition. Modality generally had limited
effects, but responses were sometimes faster in the silent conditions. We now
turn to offline responses.
Offline responses
English
Because the behavioural responses in the Spanish blocks were yes-biased,
analyses were performed on d?, a measure of response sensitivity that
corrects for response biases. d? scores range from 0 (no sensitivity/at-chance
performance) to 6.2 (perfect sensitivity). To facilitate comparison, we used
the same measure for English. By participants, responses to English
sentences varied as a function of condition, F1(2, 94)�3.11, MSE�1.22,
Figure 3. Length-adjusted reading times for final words in Spanish sentences as a function of
similarity condition and grammaticality.
Figure 2. Length-adjusted reading times for critical words in Spanish sentences as a function of
similarity condition and grammaticality.
MORPHOSYNTACTIC VIOLATIONS IN L2 1101
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
pB.05, F2B1; sensitivity was highest for the verb aspect licensing condition
(M�5.87), intermediate for subject-verb person agreement (M�5.59), and
lowest for reflexive anaphor agreement (M�5.48); however, these contrasts
did not reach significance in follow-up Duncan’s multiple-range tests. No
other effects were significant.
Spanish
Sensitivity was low overall, but was reliably affected by similarity
condition, F1(3, 141)�16.93, MSE�6.63, pB.01,5 F2(3, 58)�10.59,
MSE�3.55, pB.01. Duncan’s multiple-range tests indicated that perfor-
mance in the unique condition (M�0.55) was reliably worse than in the
other three conditions (Ms�3.23, 2.17, and 2.15, for similar1, similar2, and
different, respectively). Participants had higher confidence in their accurate
judgements (80.4%) than their inaccurate judgements (70.0%), t(48)�6.49,
pB.01. Analyses relating modality and/or similarity condition to confidence
were not possible because of missing cells.
To investigate the relationship between confidence and accuracy and the
amount of agreement information marked on words that followed the
critical word (e.g., McClain & Tokowicz, 2006), we correlated the number of
pieces of agreeing information from the manipulated dimension (number or
gender) with the average accuracy and average confidence rating for that
sentence. The following two examples demonstrate how we calculated pieces
of agreeing information:
El/*Los camion esta en el garaje.
TheSING/*ThePL truck is in the garage.
La/*Las caja esta llena de libros.
TheSING/*ThePL box is full of books.
In both sentences, the verb esta is marked as singular to agree in number
with the critical noun camion and represents one piece of agreeing
information. In the second sentence, the adjective llena is also marked as
singular to agree with the critical noun and represents a second piece of
agreeing information. To increase power, the correlations with pieces of
information collapsed across similarity and modality. However, they were
performed separately for the grammatical and ungrammatical sentences
(thus, we used accuracy rather than d?) because the information may have
been used differently in the presence versus absence of a grammatical
....
.... .....
5 See Footnote 3.
1102 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
violation. Ungrammatical sentences were responded to more accurately
when more agreeing information followed the critical word, r�.33, pB.05.
Ungrammatical sentences showed no relationship between agreement
information and confidence. Grammatical sentences showed no relationship
between agreement information and either accuracy or confidence. Interest-
ingly, of the three constructions with agreement violations, the unique
condition had the least agreeing information, which could explain the low
d? in that condition.
To examine the role of working memory in sensitivity to grammatical
violations, we correlated the two working memory measures with the
difference between critical word RTs to ungrammatical and grammatical
sentences and d?. These correlations were performed separately for the
similarity conditions. None were significant.
DISCUSSION
In this study, beginning adult learners were sensitive to grammatical
violations during reading in both L1 and L2; when a critical word cued a
grammaticality violation, reading times on the critical word and the
immediately following word were longer in both languages. In L2, this
held for conditions with morphosyntactic systems that were also present in
L1, but not for the system unique to L2. In both L1 and L2, sentence-final
reading times were longer for grammatical than ungrammatical sentences; in
L2 this effect was driven by the similar2 (demonstrative determiner-noun
number agreement) condition. Offline judgements indicated that partici-
pants were sensitive to grammaticality in L1 (M�5.64), but relatively
insensitive in L2 (M�2.12), particularly in the unique condition.
These results indicate that learners are immediately sensitive to violations
in morphosyntactic and verb aspect licensing systems common to L1 and
L2, but insensitive to violations in a system unique to L2*at least until word
n�1, although the effect there was only 18 ms for the unique construction.
Interestingly, immediate sensitivity in systems common to L1 and L2 is
present whether the systems implement agreement marking similarly or
differently. Sentence-final word reading times in the similar2 condition, but
no other L2 condition, showed the same distinction between grammaticality
and ungrammaticality present in the L1 conditions. This suggests that
learners process this morphological construction (demonstrative adjective-
noun number agreement) in a more native-like way than the other L2
constructions. However, the fact that offline grammaticality judgement
sensitivity was no better in this condition than two other L2 conditions, and
that judgement accuracy was generally poor, suggests that learners base their
MORPHOSYNTACTIC VIOLATIONS IN L2 1103
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
judgements on different information in the two languages, and that reading
times and end-of-sentence judgements measure different processes.The finding of reliably less sensitivity to L2 unique violations in offline
grammaticality judgements is consistent with previous research (e.g.,
Tokowicz & MacWhinney, 2005). One possible reason is that learners might
try to infer grammatical information from the gist representation available
after the sentence has been read (e.g., Potter & Lombardi, 1990). Because
grammatical gender violations in the unique condition, unlike the number or
aspect violations in all other conditions, have no impact on the sentence’s
semantic/conceptual representation, they should be least available from agist representation.
Learners’ immediate sensitivity to morphosyntactic violations, particu-
larly ones involving plurals, contradicts Jiang’s (2004, 2007) self-paced
reading findings in which native Chinese speakers did not show such
sensitivity in English (L2). Jiang argued that this finding was not related to
the rarity of plural morphology in Chinese, but rather indicated that plural
morphology is not an integratable L2 structure, despite generally early
instruction. The present findings weigh against this interpretation andsuggest that Jiang’s results were likely due to the fact that, for his Chinese
participants, plural morphology is essentially unique to English.
Because of concerns that binary grammaticality judgements may not be
sensitive enough to detect small differences in offline sensitivity to different
kinds of grammatical violations, we gathered learners’ confidence ratings in
those judgements. Although confidence was higher for correct judgements,
neither confidence nor accuracy correlated with the magnitude of sensitivity
to violations at the critical word. Interestingly, accuracy was higher whenthere was additional information in the sentence that agreed with the critical
word, suggesting that early learners used this information to inform their
judgements.
The current experiment included an exploratory manipulation of aloud
versus silent reading. There was no support for the hypothesis that reading
aloud increases monitoring and thus sensitivity to violations. Reading
modality did not affect offline judgement sensitivity, but aloud reading was
slower than silent reading for the similar2 condition at the final word, andgrammaticality effects were larger in the silent than aloud condition at the
critical word. Because reading aloud has an additional vocal component,
and requires coordination between button presses, speaking, and language
processing, it is not entirely surprising that sensitivity was reduced and reading
was slower in this modality, especially for the condition that was most
sensitive to grammaticality at the final word.
It is interesting to compare the findings of the current study to those of
Tokowicz and MacWhinney (2005). Both studies found online sensitivity toviolations in similar L2 conditions and very low offline sensitivity in the
1104 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
unique condition. However, whereas Tokowicz and MacWhinney also found
online sensitivity to the unique condition, the current study found it in thedifferent condition. A possible reason for the discrepancy in the unique
condition is that Tokowicz and MacWhinney’s critical nouns were more likely
to include an overt cue to their gender (the last letter ‘‘o’’ or ‘‘a’’) than those in
the current study. The discrepancy with respect to the different condition
could be related to the speed at which stimuli were presented. In the present
study, there was no external time pressure, perhaps allowing the online
measure to be more sensitive.
The results of the current experiment were not entirely consistent with theCompetition Model. This model predicts that sensitivity to violations should
be highest in the similar conditions, because there would be no competition
between languages, and lowest in the different condition, because competition
would impede processing. In the current study, participants showed online
sensitivity to violations regardless of whether they occurred in morphosyn-
tactic systems that were similar or different in L1 and L2. These findings
highlight the fact that, although broad generalisations about across-language
similarity are useful, other aspects of the Competition Model such as cuevalidity and reliability, exposure, and the kinds of information that cues carry
should be taken into account in concert with cross-language similarity.
Original manuscript received June 2009
Revised manuscript received September 2009
First published online July 2010
REFERENCES
Ferreira, F., & Clifton, C., Jr. (1986). The independence of syntactic processing. Journal of Memory
and Language, 25, 348�368.
Hahne, A., Mueller, J., & Clahsen, H. (2006). Morphological processing in a second language:
Behavioural and event-related potential evidence for storage and decomposition. Journal of
Cognitive Neuroscience, 18, 121�134.
Jiang, N. (2004). Morphological insensitivity in second language processing. Applied Psycholin-
guistics, 25, 603�634.
Jiang, N. (2007). Selective integration of linguistic knowledge in adult second language learning.
Language Learning, 57, 1�33.
Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension.
Psychological Review, 87, 329�354.
Kotz, S. A. (2009). A critical review of ERP and fMRI evidence on L2 syntactic processing. Brain
and Language, 109, 68�74.
MacWhinney, B. (1997). Second language acquisition and the Competition Model. In A. M. B. de
Groot & J. F. Kroll (Eds.), Tutorials in bilingualism: Psycholinguistic perspectives (pp. 113�142).
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
MacWhinney, B. (2005). A unified model of language acquisition. In J. F. Kroll & A. M. B.
De Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 49�67). Oxford,
UK: Oxford University Press.
MORPHOSYNTACTIC VIOLATIONS IN L2 1105
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4
McClain, R., & Tokowicz, N. (2006, July). Measuring sensitivity to Spanish gender and number
agreement: Overt judgments of grammaticality. Poster presented at the Pittsburgh Science of
Learning Center Summer intern poster session, Pittsburgh, PA.
McDonald, J. L. (1987). Sentence interpretation in bilingual speakers of English and Dutch.
Applied Linguistics, 8, 397�413.
Osterhout, L., & Mobley, L. A. (1995). Event-related brain potentials elicited by failure to agree.
Journal of Memory and Language, 34, 739�773.
Potter, M. C., & Lombardi, L. (1990). Regeneration in the short-term recall of sentences. Journal of
Memory and Language, 29, 633�654.
Rossi, S., Gugler, M. F., Friederici, A. D., & Hahne, A. (2006). The impact of proficiency on
syntactic second-language processing of German and Italian: Evidence from event-related
potentials. Journal of Cognitive Neuroscience, 18, 2030�2048.
Sabourin, L., & Stowe, L. A. (2008). Second language processing: When are first and second
languages processed similarly? Second Language Research, 24, 397�430.
Tokowicz, N., & MacWhinney, B. (2005). Implicit and explicit measures of sensitivity to violations
in second language grammar: An event-related potential investigation. Studies in Second
Language Acquisition, 27, 173�204.
Tokowicz, N., Michael, E. B., & Kroll, J. F. (2004). The roles of study-abroad experience and
working-memory capacity in the types of errors made during translation. Bilingualism:
Language and Cognition, 7, 255�272.
Tolentino, L. C., & Tokowicz, N. (2009). Across languages, space, and time: The role of cross-
language similarity in L2 (morpho)syntactic processing as revealed by fMRI and ERP.
Manuscript under review.
Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing: Use
of thematic role information in syntactic disambiguation. Journal of Memory and Language,
33, 285�318.
Van Hell, J. G., & Tokowicz, N. (in press). Event-related brain potentials and second language
learning: Syntactic processing in late L2 learners at different L2 proficiency levels. Second
Language Research.
Warren, T., White, S. J., & Reichle, E. D. (2009). Investigating the causes of wrap-up effects:
Evidence from eye movements and E-Z Reader. Cognition, 111, 132�137.
Waters, G., Caplan, D., & Hildebrandt, N. (1987). Working memory and written sentence
comprehension. In M. Coltheart (Ed.), Attention and performance XII: The psychology of
reading (pp. 531�555). Hove, UK: Lawrence Erlbaum Associates Ltd.
1106 TOKOWICZ AND WARREN
Dow
nloa
ded
by [
Uni
vers
ity o
f W
este
rn O
ntar
io]
at 1
2:32
21
Oct
ober
201
4