Adults with mild intellectual disabilities: can their reading comprehension ability be improved?

Journal of Intellectual Disability Research doi: 10.1111/j.1365-2788.2006.00921.x

pp

–

©

The Authors. Journal Compilation ©

Blackwell Publishing Ltd

835

Blackwell Publishing LtdOxford, UKJIRJournal of Intellectual Disability Research

-

©

The Author. Journal compilation ©


835849

Original Article

Adults’ reading comprehension abilityK. P. van

den Bos

et al.

Correspondence: Kees P. van den Bos, Grote Rozenstraat

, Groningen

TJ, the Netherlands (e-mail: [email protected]).

Adults with mild intellectual disabilities: can their reading comprehension ability be improved?

K. P. van den Bos,

1

H. Nakken,

1

P. G. Nicolay

2

& E. J. van Houten

1

1

Department of Special Education, University of Groningen

2

Stichting NOVO, Groningen, the Netherlands

Abstract

Background

Adults with a mild intellectual disability (ID) often show poor decoding and reading compre-hension skills. The goal of this study was to investigate the effects of teaching text comprehension strategies to these adults. Specific research goals were to deter-mine (

) the effects of two instruction conditions, i.e. strategy instruction to individuals and strategy instruction in small groups in a reciprocal teaching context; (

) intervention programme effects on spe-cific strategy tests (so-called direct effects), and pos-sible differences between strategies; (

) (long-term) transfer effects of the programme on general reading comprehension ability; and (

) the regression of gen-eral text comprehension by the variables of technical reading, IQ, reading comprehension of sentences (RCS), and pretest and posttest scores on the strat-egies taught.

Methods

In total,

adults (age range

–

years; mean age of

years) with ID participated in the study. IQs ranged from

to

with a mean IQ of

. The intervention programme involved

weekly lessons of

h each, taught during

months. Blocks

of lessons included each of Brown and Palincsar’s strategies of summarizing, questioning, clarifying and predicting, as participants read and studied narrative and expository texts.

Results

Results indicated no significant difference between group and individual instruction conditions. Second, direct programme effects – as determined by posttest–pretest contrasts for strategy tests – were substantial, except for the questioning strategy. Third, even more substantial was the transfer effect to general text comprehension. Moreover, the results on this test were well maintained at a follow-up test. Finally, the variance of general reading comprehen-sion ability was best explained by the test of RCS, and only moderately by the strategies trained.

Conclusion

The presently used intervention pro-gramme provides a good starting point for adults with ID to become better readers.

Keywords

adults, cognitive strategies, intervention, reading comprehension

Introduction

The research that will be reported joins western-societal policies that education and support should be directed at bringing human beings and individuals with disabilities especially, to maximum levels of

Journal of Intellectual Disability Research

K. P. van den Bos

et al

. •

Adults’ reading comprehension ability836

©



social, cognitive and emotional independence and self-control (Reid

; Mansell & Ericsson

; Pijl

et al

.

). A certain minimum level of

reading ability

is seen as one of the important prerequisites for inde-pendent functioning.

This article focuses on the group of adults with a mild intellectual disability (ID), and on their rarely studied ability to read and comprehend text (see for a recent exception Jones

et al

.

). In general, read-ing, and reading comprehension in particular, is strongly related to IQ (Stanovich

; Sternberg & Spear

; Bransford

et al

.

), and it is assumed that many individuals with ID will perform poorly at reading tasks. This raises questions as to how their special needs were met in the years that they visited the school, and to training perspectives in the years after school.

As learning to read is not a ‘natural act’ (Gough

), even typically developing children depend on instruction and extended practice to learn to decode. Intertwined with the decoding process is the ‘

reading is thinking

’ or reading comprehension component, and for this aspect as well, ‘for people who are not privileged with a natural disposition or stimulating home surroundings, there is

dependence on the organi-zation of the curriculum

to aid them in learning’ (Cal-fee

, p.

, italics added). This statement applies to children with ID in particular.

There is, however, a dearth in the (Dutch) research literature describing and evaluating the curriculum of reading for children with severe and mild ID (Nicolay

). This particularly applies to the ‘elusive’ component of language comprehension (Aarnoutse

; Curtis

). This component receives much less attention than technical reading (decoding). Therefore, with respect to the population of interest in the present study (i.e.

adults

with ID, who do not attend school anymore), it can be assumed that they were instructed in technical reading/decoding, but much less so in the thinking or comprehension aspect of reading.

Can this capacity be developed in these adults, and if so, what would be an effective approach? These are the central research questions in this study. In order to develop intervention plans, the literature on reading comprehension

intervention

will briefly be reviewed.

In the seminal publications of Borkowski & Cavanaugh (

), Palincsar (

), Palincsar &

Brown (

), Brown & Campione (

), Palincsar & Englert (

) and Brown & Palincsar (

), the focus is on instructing comprehension fostering

strat-egies

such as summarizing, questioning, clarifying word and sentence meanings, and predicting con-tents of later text parts. The populations of interest were normally intelligent children and ‘underachiev-ing adolescents’ (Reid

) with poor reading com-prehension skills and

specific

learning disabilities.A second important feature of Palincsar and

Brown’s publications is the introduction of the con-cept of

reciprocal teaching

of the reading comprehen-sion strategies mentioned. Reciprocal teaching builds upon principles of

direct teaching

of strategies in which a teacher, trainer, or expert explains and models the strategies and provides practice opportunity (Rosenshine & Meister

). However, the addi-tional and ‘unique’ aspects of reciprocal teaching are that strategy

application

develops via dialogues and discussion with and between participants, and that, gradually, the teacher steps back. This means that the responsibility for text processing, thinking and com-menting and also the control of the discussions are transferred to the participants who take turns in these processes.

In numerous studies, Palincsar and Brown’s ideas have been investigated (see, for an important review, Rosenshine & Meister

; and for Dutch studies, e.g. Reitsma

; Brand-Gruwel

; Walraven

; van den Bos

et al

.

). Similar to previous international studies, the Dutch research focus was on populations of

children with poor reading comprehen-sion skills

but from the average IQ range and various performance levels of technical reading. Conclusions were that strategy training is effective

in these popula-tions

, at least in terms of improved scores on compre-hension strategy tests themselves. Another general conclusion was that reciprocal teaching ‘works’. However, despite the practical and theoretical attrac-tiveness of the concept, evaluations of effects have also led to critical questions (Rosenshine & Meister

) to which we will return.The present study differs from the studies men-

tioned in terms of both participants’ chronological ages and intelligence levels, i.e. our target group con-sists of

adults

with ID. Yet, there seemed to be no good reasons for not following the previously applied approaches of strategy instruction and creating liter-acy contexts. Thus, in general, there is a considerable


K. P. van den Bos

et al

. •


©



degree of correspondence between the previously employed instruction conditions, materials and tests, and those of the present study.

Of theoretical interest is the still unresolved ques-tion (Rosenshine & Meister

, pp.

–

) as to whether a reciprocal teaching format (which – as previously mentioned – builds upon direct teaching of strategies but which further involves considerable group interactions and exchange of teacher roles) is more effective than an instructional format that is restricted to direct teaching principles. In the present study, this question will be investigated by comparing two instruction conditions. In the first condition, the reading comprehension strategies of summarizing, questioning, predicting and clarifying are explicitly taught in small groups and there is a gradual change into guided group discussions and exchange of teacher roles. Our second condition involves explicit instruction of these strategies to

individuals

, with some dialogues

but without reciprocal teaching

. See, for further details, the

Method

section and the Appendix

.In the remainder of this introduction, further

aspects and questions in this intervention study with adults with mild ID will be described, followed by hypotheses and predictions.

First, which test types can be distinguished that measure the effects of strategy instruction? Second, which characteristics and psycholinguistic dimen-sions

within

these tests, and also in the content mate-rials of the intervention programme, are relevant for more specific predictions about effects of strategy instruction? Third, which reading and reading-related variables are the best statistical predictors of the adults’ reading comprehension abilities?

As to the first question, in the literature about

text

reading and reciprocal teaching effects (Rosenshine & Meister

), two tests are distinguished, viz., specific ‘experimenter-developed tests’ aimed to measure the ability of learning and applying text comprehension

strategies

, and ‘standardized tests’ of reading comprehension. In intervention research, the specific tests are used to determine the so-called

direct

programme effect, whereas the standardized tests measure the strategy intervention programme’s

trans-fer

effect on

general

reading comprehension ability. Rosenshine & Meister (

) conclude that most intervention studies indicate more substantial pro-gramme effects on specific strategy tests than on gen-

eral tests. This is not quite unexpected because most intervention studies show a high degree of

congruency

between programme contents (e.g. nature of texts employed, questioning and interviewing procedures, etc.) and formats of testing in ‘experimenter-developed’ pretests and posttests. In contrast, the degree of congruency between intervention pro-gramme and standardized or general tests of reading comprehension is usually much lower.

In this study, both test types will be employed, and the prediction will be tested that among adults with ID larger training effects will be found on specific strategy tests than on general reading comprehension ability tests.

In the context of reading tests that will be used in this study, some further issues require elaboration. One is that not all available general tests of reading comprehension involve

texts.

Consider, e.g. the read-ing comprehension sub-test of the Peabody Individ-ual Achievement Test (PIAT) of Markwardt (

). In this test,

sentences

have to be read. Such tests were not used in previous research on transfer effects of intervention programmes involving training of com-prehension strategies. A possible reason is that most comprehension strategies are felt to be meant for texts and passages, and not for isolated sentences. Nevertheless, we thought it might be interesting to include this test in our battery because success of reading comprehension of text will, in part, be deter-mined by ‘local’ skills such as sentence and even word comprehension which together form the basis of texts (Perfetti

). These ideas can be put to a test. If it is true that training of strategies such as summarizing, predicting, etc. is ‘best’ suited for application in texts, then the following prediction seems straightforward: reading comprehension strategy training will show larger effect sizes on tests with texts than on tests at the sentence level.

A second issue concerns the nature of the passages used in reading comprehension tests and in interven-tion programmes. Two general text types can be distinguished, viz., narrative and expository (informative) texts. Narrative texts can be character-ized by a ‘story grammar’ which usually contains formal elements such as setting, goal, attempt, out-come, reaction (Calfee & Calfee

; Calfee

,

,

; van Eldik & Risselada

). Expository texts usually have more argumentative, ‘objective’ and explanatory structures, depending on the writer’s


K. P. van den Bos

et al

. •


© The Authors. Journal Compilation © Blackwell Publishing Ltd

focus on an idea, an event or an object (Calfee & Chambliss ).

Whereas previous intervention research mainly employed expository passages, in the present study (i.e. in the intervention programme, and in the spe-cific and the general reading comprehension tests), a balanced mix of expository and narrative texts will be used. First, we wanted our adult participants to be exposed to a broad array of text types, connected also to their life experiences as adults (see, for further details, the Method section, and Nicolay ). Sec-ond, it is to be expected that the four comprehension strategies of summarizing, questioning, clarifying and predicting are not equally applied. One of the present study’s aims is to explore and interpret whether there are different relationships with narrative and exposi-tory text structures.

Finally, we are interested in generalizing the inter-relationships of reading and reading-related variables assumed in general cognitive developmental theory and reading theory (see e.g. Perfetti ; Hoover & Gough ; Bransford et al. ) to the group of adults with ID. Therefore, generally expected inter-relationships between variables, such as IQ, decoding and comprehension components of reading, and impact of comprehension strategy learning, should also apply to the group of adults with ID.

Examples of predictions concern the following. First, it can be assumed that text processing strategies are ‘naturally’ more connected to the process of read-ing comprehension than to technical reading or decoding. From this it follows that effect size of a strategy training programme should be larger for comprehension measures than for decoding tests. Second, Gough & Tunmer’s () ‘simple view of reading’ assumes that during typical development the impact of the reading component of ‘linguistic com-prehension’ on reading comprehension gradually gets larger than the component of decoding. Further-more, linguistic comprehension is assumed to be closely connected to other ‘higher-order’ cognitive variables, such as IQ and learning by experience, the former variable being more ‘static’ than the latter one which is considered to be more ‘dynamic’ (Brown & French ). Comprehension training programmes are assumed to be an instance of dynamic learning-by-experience settings. If this is true, and if compre-hension strategies can be effectively trained, it follows that the variable of strategies should contribute

‘unique’ variance as a predictor of reading comprehension.

Summarizing

The goal of this study was to investigate the effects of an intervention programme that trains adults with mild ID in learning and applying text comprehension strategies (i.e. summarizing, questioning, clarifying and predicting) in narrative and expository texts. More specifically, we wanted to investigate: effects of two instruction types, i.e. the effect of direct strategy instruction with guided practice to individual participants compared with instruction in small groups of participants followed by guided group discussions and reciprocal teaching elements; effects of the programme on applying text compre-hension strategies; more specifically,a direct programme effects on strategic abilities themselves, separately for expository and narrative texts. For tests of strategies, stronger effects were predicted than for more general tests of reading com-prehension ability;b transfer effects on a word-decoding test, and two general reading comprehension tests, one with sen-tences, and one with texts. The predicted order of effect sizes is: a low to absent effect on word-reading, and a larger effect on text comprehension than for comprehension of sentences;c long-term maintenance effects on general reading comprehension; and predictive values of the variables of word-reading ability (technical reading), reading comprehension of sentences (RCS), IQ and performance on the four specific strategies, as predictors of general reading comprehension of texts.

Method

Participants

In the months before the intervention project started, a letter with information and an enrolment form about the intervention project was presented to adult participants of centres for day activities in the province of Groningen (the Netherlands). The enrol-ment form contained various personal and demo-graphic questions. The preferred training condition (individual training or training in small groups) or


K. P. van den Bos et al. • Adults’ reading comprehension ability839


absence of preference could also be indicated on this form. Ninety-nine participants responded positively by filling out the enrolment form. However, this number exceeded the project’s instructional and test-ing capacity, and a random selection of participants was made to reduce the number to around partic-ipants. In order to be sure that the participants had a reasonable level of technical reading ability, and that they met a formal criterion of ID, the potential participants were tested with a standardized test of word-reading ability, the One Minute Test (OMT) of reading of Brus & Voeten (). Also, with the par-ticipants’ consents, their IQs were obtained from their personal files. Specific criteria for selection were an OMT score of at least correctly read words per minute, and a measured IQ of maximally IQ points (DSM-IV ). These criteria were met by participants. During the project, three persons dropped out because of illness. Descriptive data (age, IQ, gender) of the remaining participants are pre-sented in Table .

Research design and general procedure

This study employs the commonly used research design in which an experimental group is ‘treated’ with an intervention programme (with various instruction conditions), and in which the scores on pretests, posttests and follow-up (‘maintenance’) tests are compared with a ‘non-intervention’ control group. Approximately half of the participants were randomly assigned to either the experimental group or the control group. Assignment to the two instruc-tion conditions of an experimental group, viz., Instruction and Discussion in small Groups (IDG) and Instruction to INdividuals (IN), was not ran-dom, and largely based on the participants’ prefer-ences indicated on the original enrolment form. The three selected participants who had indicated ‘no preference’ were assigned to the IN condition.

However, in the present study, the control group of participants had a special status which is explained as follows. Because we did not want to ‘disappoint’ this group in just using them for test administration, and not offering them the possible learning experi-ences of the intervention programme, it was decided to use the control group twice. In the first period of experimenting, the control group was a control group proper, taking the same general pretests (Measure-

ment ) as the experimental group, but not the inter-vention programme and the specific strategy tests. After completing the pretests, the control group par-ticipants were informed that, after months, there would be a second round of testing (Measurement ), and that soon thereafter, ‘it would be their turn’ to participate in the intervention programme. Thus, the original ‘non-treatment’ control group would yet become an experimental group as well. In order to indicate this, the initial experimental group that par-ticipated in the intervention programme first is referred to as Experimental group I, and the control group from the moment they also participated in the programme (during the months after Measurement ) as Experimental group II (see also the Results section, and Table ).

Implications of this alternative design were the fol-lowing. First, direct programme effects (i.e. on stra-tegic abilities) could be determined for the total group of participants, i.e. across Experimental groups I and II. Second, ‘general’ transfer effects of the interven-tion programme could be investigated by ‘classical’ pretest–posttest comparisons between treatment – no treatment (or experimental I vs. control) groups. Additionally, however, as the control group would finally also become an experimental group, general transfer effects could be evaluated for this group as well. A third implication was that in the final round of posttests (Measurement ) which took place months after group I participants had finished the programme and a few days after Experimental group II had done so, posttests were ‘true’ posttests for Experimental group II, whereas they were long-term follow-up or maintenance tests for Experimental group I.

Intervention programme

The intervention programme was delivered in a quiet room of a centre for day activities for adults with ID in the province of Groningen (the Netherlands). Instructors were the third author and a university student of special education who had ample experi-ence with teaching this kind of training programmes to individuals with ID. These instructors each trained half of the groups and individual participants.

The intervention programme consisted of five blocks of three lessons (one lesson per week during months) which lasted for about h each. Four les-son blocks were devoted to the four specific strategies




Table 1 Descriptive data, means and standard deviations (in brackets) of reading tests of Group (Experimental I, Control, ExperimentalII) × Intervention/instruction condition (IDG, IN) × Measurement (,,) × Text type (narrative, expository)

Experimental group I (n = 19)

IDG: instruction and discussion in three small groups (n = 9)Age (years) 31.1 (14.5)IQ 58.1 (5.8)Gender � � 9 and 0

Tests Measurement 1 Measurement 2 Measurement 3Pretest Posttest Follow-up

Word-reading (OMT) 49.6 (15.4) 48.6 (16.9) 51.6 (16.8)RCS 15.6 (4.7) 18.7 (5.2) 17.3 (4.6)

Narrative Expository Narrative Expository Narrative ExpositoryStrategy

Sum 58.3 (28.0) 37.8 (18.6) 57.8 (29.1) 61.1 (28.3)Quest 44.4 (16.7) 30.6 (24.3) 36.1 (28.3) 27.8 (23.2)Clarif 47.2 (36.3) 50.0 (33.1) 52.8 (36.3) 72.2 (23.2)Predict 38.9 (22.1) 25.0 (21.7) 63.9 (18.2) 55.6 (30.1)

Reading comprehension of text(IRAS)

1.4 (1.1) 1.7 (1.9) 2.8 (1.6) 2.7 (1.5) 3.9 (1.6) 2.7 (1.0)

IN: instruction to individuals (n = 10)Age (years) 31.3 (8.0)IQ 59.6 (5.5)Gender � � 5 and 5

Tests Measurement 1 Measurement 2 Measurement 3Pretest Posttest Follow-up


Narrative Expository Narrative Expository Narrative Expository

Strategy



2.1 (0.7) 2.1 (1.0) 2.8 (1.1) 3.8 (1.1) 4.2 (1.1) 2.5 (1.7)

Experimental group II: control (n = 19)

IDG: instruction and discussion in three small groups (n = 9)Age (years) 39.7 (11.2)IQ 55.7 (5.4)Gender � � 5 and 4

Tests Measurement 1 Measurement 2 Measurement 3Pretest 1 Pretest 2 Posttest

Word-reading (OMT) 48.1 18.7 49.3 (18.3) 47.7 (15.7)RCS 16.6 (5.7) 16.2 (6.9) 20.3 (6.3)

Narrative Expository Narrative Expository Narrative Expository

Strategy





and the fifth and final block was an ‘integration’ block in which each lesson treated the four strategies together. For half of the participants in each of the Experimental groups I and II, lessons occurred in small groups (condition: IDG), and the other half of the group of participants received the lessons in indi-vidual sessions with a trainer (condition: IN). Each lesson had the following instructional phases: () introduction; () instruction; () application; and () round-up. More details are presented in the Appendix . From this appendix, it also becomes clear that the major difference between the two instruction conditions concerned the application phase. For an overview of lesson contents and distri-bution of narrative and expository text types across the lessons, one is referred to tables . and . in Nicolay (, p. ).

Tests

Test for text comprehension strategies

Each of the four comprehension strategies of summa-rizing, questioning, clarifying and predicting was trained in the intervention programme in separate blocks of three lessons. Before and after each of the

four blocks of the intervention programme, profi-ciency in a specific strategy was measured in a pretest and a posttest. Measurement of these four strategies was based on the Dutch test of Four Comprehension Strategies (Brand-Gruwel & Aarnoutse ) which was developed for children at elementary schools, and which contains expository texts only. However, for the present research purposes, texts of a narrative genre were also employed, and most expository texts were rewritten and text themes were adapted to the average level of life experience of adults with mild ID (note e.g. titles such as ‘A bank account’, ‘A contact ad’; Nicolay ). Texts had an average length of words. Most texts were followed by four multiple choice questions with four response alternatives. However, as the numbers of questions accompanying the various texts were not constant, strategy test scores were expressed as percentages of correct answers.

Tests for word-reading ability and general reading comprehension

Across pretest, posttest and follow-up measurements, the following word-reading and general reading com-prehension tests were employed.

Narrative Expository Narrative Expository Narrative ExpositoryReading comprehension of text

(IRAS)2.1 (1.6) 1.6 (1.1) 1.3 (1.3) 1.9 (1.7) 3.3 (1.4) 3.3 (1.6)

IN: instruction to individuals (n = 10)Age (years) 43.0 (15.3)IQ 57.0 (8.9)Gender � � 9 and 1

Tests Measurement 1 Measurement 2 Measurement 3Pretest 1 Pretest 2 Posttest


Strategy



1.6 (1.5) 0.9 (0.9) 1.2 (0.9) 1.8 (1.4) 3.4 (1.3) 3.0 (1.3)

Table 1 Continued

OMT, One Minute Test; RCS, reading comprehension of sentences; Sum, summarizing; Quest, questioning; Clarif, clarifying; Predict, predicting; IRAS, Interactive Reading Assessment System.




Word-reading ability (‘technical reading’)

This ability was measured by the two parallel versions (forms A and B) of the OMT of word-reading (Brus & Voeten ). The (raw) score of this test is the number of words correctly read in min.

General reading comprehension

This ability was measured by two tests. The first one consists of the sub-test RCS which was based (van den Bos & lutje Spelberg, unpublished) on the read-ing comprehension part of Markwardt’s () PIAT. Our version had two parallel forms (A and B) with sentences each. The person must choose the pic-ture (from four pictures) that best reflects the mean-ing of the sentence read. The score is the number of correct responses (max = ).

The second reading comprehension test contained passages and was adapted from a Dutch version (van Eldik & Risselada ) of the part ‘reading compre-hension of text’ from the Interactive Reading Assess-ment System (IRAS) of Calfee & Calfee (). In this study, two parallel sets were employed (forms A and B), with a narrative and an expository passage each. Parallel set A contained the expository text ‘How to make butter’ and the narrative text ‘A boy saved’ (with and words, respectively), and parallel set B contained the expository text ‘How to sell lemonade’ and the narrative text ‘A kitten saved’ (with and words, respectively). Each text consisted of six structural elements, which were the basis for scoring the free-recall productions of the participants and their responses to cues provided by the tester (see, for details, van Eldik & Risselada ). The maximum recall score is six points for each passage.

All tests mentioned were individually adminis-tered, and across the various parallel test or parallel set administrations, ABA orderings were applied. Within an IRAS test session, an expository passage was presented to half of the participants first, fol-lowed by a narrative passage, whereas the reversed order of presentation applied to the other half of the participants. In the next session, these presentation orderings were reversed again.

Results

Table contains means and standard deviations according to the previously described design. As

parametric statistical analyses were planned, all anal-yses were preceded by data inspection and prelimi-nary analyses in order to see whether the two main assumptions of normal distribution and homogeneity of variance were met. In the majority of the analyses, these assumptions were not violated. In case they were, alternative (non-parametric) analyses were applied. These will not be reported, however, as none of these analyses led to different conclusions about significance of differences between means.

Pretest data general

Mean pretest scores and standard deviations of the total group of adults (n = ) on the OMT for word-reading and the two tests for RCS and passages (IRAS) were . (.), . (.) and . (.), respectively. Compared with the present adults, -year-olds from regular schools produce an average OMT score of correctly read words per minute (SD = .) (van den Bos et al. ). On the RCS, the mean score of children from grades through of regular elementary schools is items correct (SD = .) (van den Bos , unpublished data). Apparently, compared with children from regular ele-mentary schools, the present adults score at the low-est reading ability levels.

Effects of the intervention programme

Direct or specific programme effect

Scores on the strategy tests as dependent variables (see grey panels in Table ) were submitted to with Group (Experimental groups I and II), Instruction condition (IDG, IN), Strategy (Summa-rizing, Questioning, Clarifying, Predicting), Text type (Narrative, Expository) and Measurement (Pretest, Posttest) as independent or explanatory variables. Between-subject factors were Group and Instruction condition. Within-subject factors were Strategy, Text type and Measurement.

The between-subject effect Group was not signifi-cant, F1,34 < .. So, in general, Experimental groups I and II participants’ performances did not differ. With regard to Instruction condition, total mean strategy scores for IDG and IN were % and %, respectively. The difference was not significant, F1,34 = ., P = .. Of the interactions with




remaining factors, the only significant effect con-cerned Strategy × Group × Instruction, F3,32 = ., P < .. However, the interpretability of this effect is limited. Differences between mean subtotal scores (Table ) suggest that participants from Experimental group I; IN condition, find Summarizing and Ques-tioning relatively more difficult than participants from Experimental group II; IDG condition, whereas in both groups, the strategy of Predicting in the IDG condition seems somewhat more successful than in the IN condition.

Next, the main effects and interactions of the within-subject factors will be described. All three main effects were significant. The total mean scores of Strategies were %, %, % and %, for Sum-marizing, Questioning, Clarifying and Predicting, respectively, F3,34 = ., P = .. A relatively low score applied to Questioning, and a relatively high one for Clarifying.

The total mean strategy scores on narrative and expository text types were % and %, respec-tively, F1,34 = ., P = .. So, in general, narrative texts seemed somewhat easier than expository texts.

The total mean strategy scores at pretest and post-test measurements were % and %, respectively, F1,34 = ., P = .. This suggests a considerable programme effect.

These main effects were modified by the following interactions. First, the interaction effect Strategies × Text type was significant, F3,32 = ., P = .. The corresponding mean scores indicated that, with narrative texts, Summarizing, Questioning and Predicting were easier than with expository texts, whereas Clarifying seemed easier in expository texts.

Also significant was the interaction Strategies × PrePost measurements, F3,32 = ., P = .. Inspection of means showed, at pretest level, a superior score for Clarifying (about %), whereas the remaining strategies did not differ much from each other (about %). At posttest measure-ment level, gains were apparent for all strategies, except for Questioning. Mean difference scores (posttest scores ‘minus’ pretest scores) were %, %, % and %, for Summarizing, Questioning, Clarifying and Predicting, respectively.

The third significant interaction effect was Text type × PrePost measurements, F1,34 = ., P = .. Inspection of means indicated that narrative texts were easier than expository texts at the pretest level

only (% vs. % correct for narrative and exposi-tory texts, respectively), whereas performances at these text types did not differ at the posttest level (both, % correct). However, this also means that score gains as a result of the intervention programme were largest for expository texts.

Transfer effects

In this analysis, Experimental and ‘original’ Control group’s reading test data, summarized in the first two columns in Table , were compared. Data were pooled across Instruction conditions because of the results from the previous analysis, and because at this phase of the experiment, the assignment of the con-trol group to instruction conditions had no function yet.

Scores on word-reading (OMT), RCS and reading comprehension of text (IRAS), as dependent vari-ables, were submitted to three sets of univariate anal-ysis of variance (), respectively. For OMT and RCS analyses Group (Experimental, Control) and Measurement (Pretest, Posttest) were the indepen-dent variables, and for IRAS the independent vari-ables were Group, Measurement and Text type (Narrative, Expository). Furthermore, in all three analyses, Group was the between-subjects factor, and the remaining factors within-subjects factors.

In the × analysis of the OMT scores, no effects were significant. Accordingly, whether or not persons had participated in the intervention programme, this did not influence their pretest or posttest scores on the OMT. In other words, as expected, the interven-tion programme had no transfer effect on technical reading performance.

Of course, the present study’s question as to a transfer effect to (general) reading comprehension tests was the more important one. The results were as follows. With regard to the RCS test, the interac-tion effect Group × Measurement was significant, F1,36 = ., P < .. This means that the experimen-tal group’s I mean posttest score was significantly higher than their pretest score, whereas the control group’s means did not differ significantly. A similar Group × Measurement interaction effect was found in the Group × Measurement × Text type analysis of IRAS scores, F1,36 = ., P < .. Where the exper-imental group’s PrePost means were . and ., respectively, the control group’s means on these mea-




surements hardly differed (means were . and ., respectively). These results suggest significant trans-fer of comprehension strategy learning to general reading comprehension performance.

Did this transfer effect equally apply to the differ-ent text types of the IRAS test? The results of the previous × × analysis of IRAS scores indicated two additional interaction effects: Text type × Measurement, F1,36 = ., P < ., and the (marginally significant) three-way interaction, F1,36 = ., P = .. Inspection of the correspond-ing means seemed to suggest that, in general, PrePost ‘gains’ applied to expository texts in particular.

In order to see whether these findings remained reliable, Experimental group II persons’ post-intervention IRAS scores (see Table , third measure-ment) were combined with group I data, and submitted to a Text type × PrePost analysis of vari-ance. As before, the main effect PrePost was signifi-cant, F1,37 = ., P = .. Therefore, the previous suggestion of significant transfer of comprehension strategy learning to general reading comprehension performance also applies to the combined experimen-tal groups. However, in this analysis, none of the remaining effects (main effect Text type, and the interaction) were significant. Thus, the suggestion in the previous analysis that the transfer effect (i.e. post-test–pretest difference of IRAS scores) particularly applied to expository texts was not confirmed for the combined experimental groups. We will return to this issue in the paragraphs describing effect size analysis and regression analysis.

Maintenance effect

In order to investigate whether the ‘improved’ scores on the general tests of reading comprehension had been maintained months after termination of the intervention programme, the following analyses were performed. Measurements and scores of the Experimental group I (across instruction conditions) on RCS and IRAS tests were submitted to a paired t-test, and a Measurement × Text type , respec-tively. Measurement data (pretest) were entered as covariate. Although at first impression, the means suggested lower follow-up scores of the RCS test and IRAS-expository texts, the analyses indicated that, controlled for pretest correlations, Measurements and differences were not significant, P > .. On

the IRAS-narrative texts, follow-up scores were even significantly higher than on Measurement . In gen-eral, therefore, we can speak of a positive mainte-nance effect.

Effect sizes

In the introductory section of this article, the follow-ing order from small to larger effect sizes of PrePost differences was predicted: technical reading (OMT) < RCS < reading comprehension of text (IRAS) < mean strategy test score. The obtained effect sizes d (Cohen ) were: OMT = ., RCS = ., Strategies Narrative = ., Strategies Expository = ., IRAS Narrative = . and IRAS Expository = .. These results deviate from the pre-dicted pattern in that the effect size of Strategies is not larger than the improvements of scores on the general tests of reading comprehension. For Strate-gies Narrative, the effect size is even considerably smaller than for the remaining tests, including the RCS test.

Strategy training and other predictors of general text comprehension ability (regression analysis)

Two sets of criterion scores (dependent variables) can be considered. The first set consists of the scores at the third measurement of IRAS which, because of the ABA design, concerned a repetition of the same two texts that had been administered as pretests to all participants half a year before. However, the two ‘halves’ of the total group of participants differed at Measurement in terms of recency of the interven-tion programme, and this might have influenced their Measurement scores. Alternatively, the ‘true’ IRAS posttest scores of the combined experimental groups can be considered (this concerns Measurement for Experimental group I, and, as previously indicated, Measurement , for Experimental group II). How-ever, although the combined groups would, then, be equated in terms of programme recency, their scores were not based (again because of the ABA design) on the same texts. We decided to consider both sets of criterion scores, and these sets will be referred to as IRAS-, and IRAS post (,).

A further goal of the regression analyses was to compare predictive values of the pretest total mean strategy scores with posttest strategy scores, along with




the variables of IQ, technical reading (OMT) and RCS.

The regression results in Table can be interpreted as follows. Comprehension of IRAS-narrative texts (either at Measurement , or at Measurement ) seems to be poorly predicted by the two sets of pre-dictors. The prediction of comprehension of IRAS-expository texts, however, fares better. It is clear, how-ever, that a primary predictive role is played by the test RCS, and that some additional variance is con-tributed by Strategies posttest scores only. With regard to the latter result, it can be noted that in an addi-tional regression analysis in which the four separate strategy scores were entered (instead of employing, as in Table , a ‘combined’ strategy score), the strat-egy of Clarifying turned out to be as adequate a predictor as the combined strategy score.

Discussion

The present study indicated significant direct inter-vention programme effects on adults’ strategies of

reading comprehension, and there were substantial transfer effects to general reading comprehension performance as well.

However, various aspects of the results and the main conclusions need discussion.

One of the central questions concerned a compar-ison of two conditions of strategy instruction. The results did not indicate different effects of these con-ditions. How might this be interpreted?

One reason might be that the two conditions share important principles, based on direct teaching of strategies, and that these principles are both necessary and sufficient for learning. Furthermore, differences between the two conditions might either have ‘com-pensated’ each other, or actually appear smaller than initially assumed. The appendix indicates a difference between the two conditions in discussion time spent in the lessons. Evidently, the IDG condition involved a more ‘lively’ and interactive discussion climate than in individual sessions. On the other hand, this may have resulted into smaller amounts of individual learning time than in the IN condition. After all, in

Table 2 Standard regression analyses of predictor variables on IRAS-2,3 and IRAS-3 scores

IRAS post (2,3)narrative

IRAS post (2,3)expository

IRAS-3narrative

IRAS-3expository

Overall significance F = 1.10adj. R2 = 0.01

F = 6.50**adj. R2 = 0.43

F = 0.83adj. R2 = −0.02

F = 4.92**adj. R2 = 0.35

PredictorsOMT 0.05 0.03 −0.21 −0.13IQ −0.24 0.27 0.08 0.21RCS 0.37* 0.55** 0.29 0.55**Strat Narrat PRE −0.05 0.01 0.02 0.28Strat Expos PRE −0.01 0.11 0.02 −0.04

Overall significance F = 1.25adj. R2 = 0.03

F = 6.80**adj. R2 = 0.44

F = 0.83adj. R2 = −0.02

F = 6.96**adj. R2 = 0.45

PredictorsOMT 0.05 0.09 −0.08 −0.18IQ −0.20 0.27 0.07 0.22RCS 0.32 0.55** 0.24 0.53**Strat Narrat POST 0.15 0.09 0.19 0.11Strat Expos POST −0.07 0.15 0.02 0.36*

*P < ., **P < ..Significance of the whole set of predictors, adjusted R2-values and beta coefficients of the separate predictors have been indicated for the predictor set with pretest scores (upper panel) on strategies applied to narrative and expository texts (abbreviated as Strat Narrat PRE and Strat Expos PRE), and posttest scores (lower panel) on strategies applied to narrative and expository texts (abbreviated as Strat Narrat POST and Strat Expos POST).IRAS, Interactive Reading Assessment System; OMT, One Minute Test; RCS, reading comprehension of sentences.




the latter condition, individuals got all ‘turns’ in the strategy application phases, whereas in the IDG con-dition these turns were equally distributed across the participants.

Besides these differences, that may have compen-sated each other, there are reasons of assuming greater correspondence than difference between the two conditions. In both conditions, the trainer remained the ‘leader’, at least when we compare the status of the trainer in the present study with training characteristics in the typical reports about reciprocal teaching (Palincsar & Brown ). In this literature, the gradual change of the role of instructor and the increasing group responsibility is much more apparent than in the present study. There are therefore reasons to consider the present instruction conditions as hav-ing more in common (viz., direct instruction in apply-ing strategies) than differing in terms of the degree of reciprocal instruction.

A second set of variables that were investigated in this study concerned the psycholinguistic variables of strategy type and text type. The effect of text type was evaluated for the specific strategy tests as well as for the general reading comprehension test (IRAS). With regard to the specific strategies, it could be concluded that the programme worked best for expository texts, in combination with the strategies of predicting and clarifying. Although the analyses of variance of IRAS data did not consistently indicate superior results for expository texts, the remaining analyses did. In the effect size analyses, expository texts showed an effect which was twice as large as on narrative texts. Moreover, the results from the regres-sion analyses to which we will return indicated that Text type is a relevant variable when interpreting effects.

A third point of discussion concerns the result that the programme effect size on strategy training tests (in particular on narrative texts) was somewhat smaller than on the general reading comprehension tests (RCS, IRAS Narrative and IRAS Expository). This finding differs from the opposite prediction in the introductory sections of this study and the previ-ous studies of Rosenshine & Meister () and Brand-Gruwel (), in which larger effect sizes were predicted for strategy tests than for general tests of reading comprehension. An explanation could be that, in contrast to Rosenshine & Meister () suggestions about the greater difficulty degree of gen-

eral (‘standardized’) American tests compared with experimenter-developed strategy tests, our general read-ing comprehension tests had been easier than the strategy tests. For one thing, the presently employed IRAS texts were shorter than the strategy texts. Future studies should employ better ‘matches’ between such text variables. Additionally, differences in the ways in which text comprehension was measured might have influenced research outcomes. In all previously reported studies on comprehension strategy instruc-tion (see e.g. Rosenshine & Meister ; Brand-Gruwel ), ‘standardized’ reading tests as well as the strategy tests had a multiple choice response for-mat. In the present study, this format applied to the strategy test and the RCS, but not to IRAS. IRAS’ response format supplemented free recall with specific cues. We found this interactive measurement proce-dure quite ‘lively’ and attractive for use with adults with mild ID. However, it can be considered a meth-odological limitation of this study that we did not additionally employ a general reading comprehension test with a response format similar to the other tests. Now, this study does not allow an evaluation of such format effects.

A final point of discussion concerns the interpre-tation of effect sizes on general tests of reading com-prehension themselves. Theoretically, various factors could have contributed. One is that the improved scores had been caused by test repetition (according to the ABA design). This argument does not apply for the comparisons of Measurements and , which concerned different sentences and texts (the AB part of the ABA test design involved parallel tests). In order to statistically control for test repetition effects resulting from the AA part, analysis of covariance was applied (see e.g. the maintenance effect analysis in the section Results). Despite such statistical controls, however, test repetition effects cannot be entirely excluded. In future studies, an alternative solution would involve extension and ‘counter balancing’ of the number of parallel test versions (see also the recent study of Jones et al. ).

Was, then, the intervention programme ‘responsi-ble’ for the effects? The answer to this question is fairly complex, and should also involve the difference between text types (narrative and expository). Results from the regression analyses indicated that there was no significant relationship between strate-gies and the narrative texts from the general reading




comprehension test IRAS. Therefore, score improve-ments on narrative IRAS texts cannot be interpreted as a specific programme effect. For expository texts, however, there was one significant relationship with strategies, and therefore, a positive interpreta-tion in terms of a specific programme effect seems warranted.

It is also conceivable that the strategy intervention programme was effective in more general and indirect ways. One general characteristic of the intervention programme was that it worked systematically and according to an instructional plan. It ‘invited’ the participants to read, and take reading comprehension tests during weekly sessions. In total, participants spent extra hours of reading time and practice. The trainers reported the programme to be a positive method to improve reading comprehension. Their impression was that during and after the lessons, participants showed an increasing cognitive alertness, pleasure and pride about their growing mastery of the strategies learned.

For future use of this intervention programme, var-ious adaptations and extensions are possible. To bring people to reading is one issue. Another issue is to motivate them for continuation and further develop-ment of reading proficiency. Learners must learn that they must be active readers (Reid ). Organiza-tions providing services to adults must create perma-nent ‘literacy contexts’ in which activities before, during and after reading (Palincsar & Englert ) are linked to variegated forms of functional reading activities, such as reading for pleasure, reading for study and working situations, and reading for news in articles and periodicals. The role of non-linguistic support forms of text (e.g. the role of illustrations), and the question as to which different feedback and ‘reward’ forms work optimally for adults with an ID should also be given further study. Finally, continu-ation of intervention or literacy support programmes should not persist in the currently used ‘centralized’ instructional format (i.e. with a strongly directive form of trainer instruction and supervision), but aim at a gradual change towards group and individual independence and true instructional reciprocity. Linked to the previously mentioned environmental conditions, such long-term intervention and support programmes would deserve the instructional qualifi-cation of ‘reciprocal teaching’ as designed by its founders Annemarie Palincsar & Ann Brown ().

Acknowledgements

The authors thank all adults who participated in this study. We are also grateful to the NOVO foundation for their continuous support in various stages of the study. Finally, we are grateful to our American col-league Dr Jean Ciborowski Fahey for language cor-rections of the manuscript.

References

Aarnoutse C. A. J. () Begrijpend leesonderwijs aan zwakke lezers [Teaching reading to poor readers]. In: Dyslexie’ [Dyslexia’] (eds K. P. van den Bos & D. R. van Peer), pp. –. Garant, Antwerp/Apeldoorn.

Borkowski J. G. & Cavanaugh J. C. () Maintenance and generalization of skills and strategies by the retarded. In: Handbook of Mental Deficiency (ed. N. R. Ellis), pp. –. Erlbaum, Hillsdale, NJ.

van den Bos K. P., lutje Spelberg H. C., Scheepstra A. J. M. & de Vries J. R. () De Klepel (vorm A en B) [Nonsense Word Test ‘The Clapper’; Form A and B]. Har-court, Amsterdam.

van den Bos K. P., Brand-Gruwel S. & Aarnoutse C. A. J. () Text comprehension strategy instruction with poor readers. Reading and Writing , –.

Brand-Gruwel S. () Onderwijs in tekstbegrip [Teaching listening and text comprehension strategies to poor read-ers]. Doctoral Dissertation. University of Nijmegen, Nijmegen, the Netherlands. Felix, Ubbergen.

Brand-Gruwel S. & Aarnoutse C. A. J. () Begrijpend Leestest Vier Vaardigheden [A Test for Four Strategies of Read-ing Comprehension] Berkhout, Nijmegen.

Bransford J. D., Brown A. L., Cocking R. R., Donovan M. S. & Pellegrino J. W. (eds) () How People Learn Brain, Mind, Experience, and School . National Academy Press, Washington, DC.

Brown A. L. & Campione J. C. () Psychological theory and the study of learning disabilities. American Psycholo-gist , –.

Brown A. L. & French L. A. () The zone of potential development: implications of intelligence testing in the year . Intelligence , –.

Brown A. L. & Palincsar A. S. () Guided, cooperative learning and individual knowledge acquisition. In: Know-ing, Learning and Instruction Essays in Honor of Robert Glaser (ed. L. B. Resnick), pp. –. Erlbaum, Hillsdale, NJ.

Brus B. T. & Voeten M. J. M. () Eén-Minuut-Test Vorm A en B [One-Minute Test. Form A and B]. Harcourt, Amsterdam.

Calfee R. C. () Literacy and illiteracy: teaching the nonreader to survive in the modern world. Annals of Dys-lexia , –.




Calfee R. C. () The mind of the dyslexic. Annals of Dyslexia , –.

Calfee R. C. () Applying cognitive psychology to edu-cational practice: the mind of the reading teacher. Annals of Dyslexia , –.

Calfee R. C. & Calfee K. () Interactive Reading Assessment System, IRAS. Stanford University, Stanford, CA.

Calfee R. C. & Chambliss M. J. () The structural design features of large texts. Educational Psychologist , –.

Cohen J. () Statistical Power Analysis for the Behavioral Sciences, nd edn. Erlbaum, Hillsdale, NJ.

Curtis M. A. () Improving comprehension and com-prehension instruction. International Dyslexia Association: Perspectives , –.

DSM-IV () Diagnostic and Statistical Manual of Mental Disorders, th edn. American Psychiatric Association, Washington, DC.

van Eldik M. & Risselada G. () De ontwikkeling van een leescomponententest [Development of a reading component test]. Unpublished Master Thesis. University of Groningen, Groningen, the Netherlands.

Gough P. B. () How children learn to read and why they fail. Annals of Dyslexia , –.

Gough P. B. & Tunmer W. E. () Decoding, reading, and reading disability. Remedial and Special Education , –.

Hoover W. A. & Gough P. B. () The simple view of reading. Reading and Writing , –.

Jones F. W., Long K. & Finlay W. M. L. () Assessing the reading comprehension of adults with learning dis-abilities. Journal of Intellectual Disability Research , –.

Mansell J. & Ericsson K. (eds) () Deinstitutionalization and Community Living. Chapman & Hall, London.

Markwardt F. C. () Peabody Individual Achievement Test-Revised. Subtest Reading Comprehension. American Guid-ance Service, Circle Pines, MN.

Nicolay P. G. T. () Het bevorderen van begrijpend lezen bij mensen met een verstandelijke handicap [Promoting read-ing comprehension in persons with intellectual disabilities]. Doctoral Dissertation. University of Groningen,

Groningen, the Netherlands. Stichting Kinderstudies, Groningen.

Palincsar A. S. () Improving the reading comprehension of junior high students through the reciprocal teaching of comprehension-monitoring strategies. Unpublished doctoral Dissertation. University of Illinois at Urbana-Champaign.

Palincsar A. S. & Brown A. L. () Reciprocal teaching of comprehension fostering and comprehension monitor-ing activities. Cognition and Instruction , –.

Palincsar A. S. & Englert C. S. () Teaching learning disabled students to read. In: Teaching the Learning Dis-abled. A Cognitive Developmental Approach (ed. D. K. Reid), pp. –. Allyn and Bacon, Inc., Needham, MA.

Perfetti C. A. () Reading Ability. Oxford University Press, New York.

Pijl S. J., Meijer C. & Hegarty S. () Inclusive Education: A Global Agenda. Routledge, London.

Reid D. K. (ed.) () Teaching the Learning Disabled. A Cognitive Developmental Approach. Allyn and Bacon, Inc., Needham, MA.

Reitsma P. () Instructional approaches to problems in reading comprehension of dyslexics. In: Current Directions in Dyslexia Research (eds K. P. van den Bos, L. S. Siegel, D. J. Bakker & D. L. Share), pp. –. Swets & Zeitlinger, Lisse.

Rosenshine B. & Meister C. () Reciprocal teaching: a review of the research. Review of Educational Research , –.

Stanovich K. E. () Cognitive determinants of reading in mentally retarded individuals. International Review of Research in Mental Retardation , –.

Sternberg R. J. & Spear L. C. () A triarchic theory of mental retardation. International Review of Research in Mental Retardation , –.

Walraven M. () In structie in leesstrategieën. Problemen roet begrijpend lezen en het effect van instructie aan zwake lezers. [Instruction in reading strategies. Problems with read-ing comprehension and the effect of instruction to poor readers]. Doctoral dissertation. Free University, Pedological Institute, Amsterdam.

Accepted August




Appendix 1

Instructional phases in a typical lesson of the intervention programme at the conditions of Instruction to INdividuals (IN) and Instruction and Discussion in Groups (IDG)

Instruction to INdividuals (IN)Introduction. Welcome. Trainer names specific strategy. Goal of lesson is explained. Introduction of support sheet. This sheet

contains the name of the strategy (e.g. Clarifying uncertainties) and a commentary question, e.g. ‘Does the passage contain difficult words’?

Instruction. Text parts are marked by bold lines. Participant silently reads the first text part till the bold line. Next, trainer reads this part aloud, and models the strategy (e.g. Summarizing) by announcing and producing aloud ‘in one sentence what this text part is about’. Trainer writes this sentence down on a flap-over. Trainer minimally treats the second text part in a similar way, and sometimes the third text part as well.

Application. Participant reads aloud the support sheet, followed by silently reading the next text part. Participant is asked to ‘think aloud’ while performing the strategy. On the flap-over, trainer writes down the result. Next, the participant’s actions are commented upon by the trainer and trainer provides further instruction if the strategy is not applied optimally. (Note: one could speak of ‘dialogues’ here, but strictly speaking, not of ‘discussion’ as in the IDG condition described below). Trainer adds new information to flap-over contents, and whenever applicable, modifies these. These steps are repeated at each next text part.

Round-up. The whole text has been read. Summary of activities by going through the flap-over information. Prospects. Thank you.

Instruction and Discussion in Groups (IDG)Introduction. Welcome. Trainer names specific strategy. Goal of lesson is explained. Introduction of support sheet. This sheet

contains the name of the strategy (e.g. Clarifying uncertainties) and a commentary question, e.g. ‘Does the passage contain difficult words’?

Instruction. Text parts are marked by bold lines. All group participants silently read the first text part till the bold line. Next, trainer reads this part aloud, and models the strategy (e.g. Summarizing) by announcing and producing aloud ‘in one sentence what this text part is about’. Trainer writes this sentence down on a flap-over. Trainer minimally treats the second text part in a similar way, and sometimes the third text part as well.

Application. At the third or fourth text part, which is read silently by all group members, a volunteer is asked to read the support sheet and, next, to demonstrate strategy performance. After this action, trainer invites remaining group members to discuss the volunteer’s performance. This discussion yields comments and, possibly, alternative (and even ‘faulty’) solutions! Trainer discusses the results. Trainer provides further instruction if the strategy is not performed optimally. Trainer adds new information to flap-over contents, and whenever applicable, modifies these. These steps are repeated at each next text part.

In general, participants practise with two text parts, and then a ‘new’ group member is on turn.Round-up. The whole text has been read. Summary of activities by going through the flap-over information. Prospects. Thank you.Postscriptum. In the third and last lesson of a strategy block (in IN and IDG conditions), trainer provides demonstration at the

first text part only. Participants continue with the remaining parts, and do so relatively independently.

Documents

Adults with mild intellectual disabilities: can their reading comprehension ability be improved?