Children s Difficulties in Text Comprehension: Assessing Causal Issues

Transcription

1 Children s Difficulties in Text Comprehension: Assessing Causal Issues Jane Oakhill Kate Cain University of Sussex In this article we consider the difficulties of children who have a specific reading comprehension problem. Our earlier work has shown that good and poor comprehenders differ, in particular, in their ability to make inferences, integrate information in text, understand story structure, and monitor their understanding. We outline some studies that illustrate the poor comprehenders problems and present two studies that use a comprehension-age match design to explore the direction of causality between comprehension skill and other abilities. We also present data from the first and second stages of a longitudinal study, when the children were 7 to 8 and 8 to 9 years old. Multiple regression analyses show that a number of factors predict significant variance in comprehension skill even after general ability factors such as IQ and vocabulary have been taken into account. These findings suggest that, not only can children have comprehension problems in the absence of word recognition problems, but that distinctly different skills predict variance in word recognition and variance in comprehension. The data support the view that single-word reading skills and the ability to build integrated text representations make independent contributions to overall reading ability. We discuss the implications of these findings for our understanding of children s problems in text comprehension, for deaf readers, and for remediation. A number of children have problems with reading comprehension, even when they are competent at singleword recognition and have adequate vocabularies. Although single-word reading and comprehension skill The research reported here was supported by an Economic and Social Research Council grant (R ) awarded to Jane Oakhill and Peter Bryant. We thank the staff and pupils from all of the Brighton and Hove schools who participated in this work. Correspondence should be sent to Jane Oakhill, Laboratory of Experimental Psychology, University of Sussex, Falmer, Brighton BN1 9QG, England Oxford University Press are, in general, highly correlated, a substantial minority of children develops the former but not the latter skill. The article is organized in two parts. In the first, we state what we mean by poor comprehension and explain how we select our subjects and the characteristics of poor comprehenders. In the second part, we explore the interrelation and interdependence of the various abilities we have found to be related to comprehension skill. We also address issues of causality: how causality can be explored, and which, if any, subskills might be causally linked to reading comprehension and its development. Characteristics of Poor Comprehenders In this part of the article, we differentiate good and poor comprehenders. We do not wish to imply that these are the only ways in which they might differ they almost certainly are not. However, these are the particular skills that we consider further in the second part of the article. Selection of Comprehension Skill Groups The hearing participants in all our studies have been selected from mainstream schools. The children are given two assessments of reading: the Gates- MacGinitie Vocabulary subtest (MacGinitie & Mac- Ginitie, 1989) and the Neale Analysis of Reading Ability: Revised (Neale, 1989). The Neale Analysis is an individually administered test, comprised of a series of

2 52 Journal of Deaf Studies and Deaf Education 5:1 Winter 2000 Table 1 Characteristics of good and poor comprehenders Chronological age Neale accuracy age Gates-MacGinitie vocabulary Neale comprehension age Less-skilled 7;9 8; ;3 Skilled 7;9 8; ;1 Ages are given as years; months. The reading accuracy and reading comprehension scores are the age equivalent scores provided by the Neale Analysis. The maximum score possible on the Gates-MacGinitie test is 48. short stories, graded in difficulty. Children read each story aloud, and any word reading errors are corrected. After each story, they are asked a series of questions. Testing stops after the child has made a prescribed number of reading accuracy errors on a given story. The Neale Analysis provides separate age equivalent scores for reading accuracy, based on the number of word pronunciation errors that a child makes, and reading comprehension, based on the number of questions about the stories that the child answers correctly. The children were also tested on the Gates-MacGinitie Primary Two Vocabulary Test (Gates & MacGinitie, 1965). This is a group-administered task in which children have to select one of four words that goes best with an accompanying picture. Thus, it acts as a measure of silent word recognition and provides an index of the child s vocabulary. Performance on this test was used to exclude both very poor and exceptional readers. In most of the studies to be described here, we compare groups who are matched on their ability to read words both in and out of context, but who differ in comprehension skill. Typical groups of subjects are shown in Table 1. Comparable Reading-Related Skills for Good and Poor Comprehenders The lack of a difference between groups is always subject to the problems associated with confirmation of the null hypothesis, but we feel it is important to point out those reading-related skills that do not cause difficulties for the type of poor comprehender that we are interested in (those with age-appropriate reading skills). We have found that good and poor comprehenders do not differ on accuracy, speed, or automaticity of singleword decoding (Yuill & Oakhill, 1991). Neither do they differ on traditional assessments of short-term memory (digit- or word-span) or on verbatim memory for sen- tences or short texts (Yuill & Oakhill, 1991). In addition, their comprehension problems seem to be at the text, rather than the word, level. We have not found differences in performance between groups on Bishop s (1982) TROG, though other researchers (Stothard & Hulme, 1992) have found a difference on this test, so we remain open to the possibility that poor comprehenders text-level problems might, at least to some extent, be constrained by their comprehension at the sentence level. Skills and Abilities That Differentiate Good and Poor Comprehenders First, our earlier work has shown that good and poor comprehenders differ in their propensity to make two different types of inference. Poor comprehenders make ineffective text connecting inferences (Baker & Stein, 1981). That is, they cannot easily integrate information explicitly provided by the text to establish cohesion between different sentences (Oakhill, 1982; Oakhill, Yuill, & Parkin, 1986). Neither can they easily incorporate information outside of the text, that is, general knowledge, with information in the text to fill in missing details (Cain & Oakhill, 1998; Oakhill, 1984). Baker and Stein term these gap-filling inferences. We outline a study that illustrates differences between the groups in inference making later in the article. Second, we have found that good and poor comprehenders differ in their ability to derive the structure and main point of a story (Yuill & Oakhill, 1991). In our ongoing longitudinal study, we have used a sentence anagram task to explore children s understanding of story structure. In this task, the children are required to reconstruct a story from a set of randomly ordered sentences, given to them on strips of paper. Their responses are scored according to how well their ordering

3 Children s Difficulties in Text Comprehension 53 correlates with the correct ordering. Performance on this task is a predictor of comprehension skill (Oakhill, Cain, & Bryant, 1999). We have also shown differences between groups in their ability to produce causally coherent and well-structured stories (Cain & Oakhill, 1996) and we discuss these findings in part 2. Third, we have found that poor comprehenders are worse at monitoring their ongoing comprehension of a text. The way in which we have measured comprehension monitoring is to ask children to read stories and to indicate any problems in those stories (for instance, blatantly contradictory pieces of information in the text). Children who are not effectively monitoring their understanding should find it more difficult to detect inconsistencies between pieces of information in text, and we have found this to be the case, particularly when the items of information that need to be integrated are not in adjacent sentences (Oakhill, Hartt, & Samols, 1996). Fourth, although we have not found differences between good and poor comprehenders on standard tests of short-term memory, such as digit span, we have found differences on tasks of working memory: tasks that make demands on both storage and processing. One such task, for instance, is a sentence span, which we have adapted from Siegel and Ryan (1989). In this task, the experimenter reads out a set of sentences, each of which is missing a highly predictable final word. The children are required to complete each sentence (processing component) and to remember the final word of each (storage component). After a series of such sentences (up to four), the child recalls the final words, in order. We have found that such tasks differentiate between good and poor comprehenders and are highly predictive of variance in comprehension skill, even after verbal IQ, single-word reading ability, and vocabulary knowledge have been partialled out. Further details of such tasks, and their relation to comprehension skill can be found in Seigneuric, Ehrlich, Oakhill, and Yuill (in press). Thus, we have shown that good and poor comprehenders exhibit difficulties with a variety of different comprehension subskills. However, these data provide us with no more than a sort of catalogue of what poor comprehenders can and cannot do. Relations Among Comprehension Skills Our more recent studies move beyond this listing of strengths and weaknesses, to look in more detail at the interrelations between various skills and abilities. For instance, good inference making may be fundamental to comprehension ability (i.e., inference skill may be causally implicated in the development of reading comprehension). Alternatively, good comprehension (and, presumably, lots of reading for comprehension) may lead to good inference skills (i.e., comprehension skill may be the driving influence on inference making). It is obviously essential to address such issues if one wants to design effective remediation procedures for poor comprehenders. One can explore causal hypotheses in various ways. First, one could use longitudinal or training studies, but these are both time-consuming and expensive and, therefore, usually only worth pursuing once candidate causes have been identified. One way to assess which possible causal candidates are the most plausible is to adopt a design analogous to the reading-level match design advocated by Bryant and colleagues: The comprehension-age match design (CAM; Cain, Oakhill, & Bryant, in press). Although this design cannot prove the direction of causality between reading comprehension and a subskill, it can help to establish whether or not proficiency in a particular skill is more likely a cause or a consequence of good reading comprehension. As outlined at the start of this section, if the poor comprehenders are impaired on a particular measure, such as inference-making ability, the poor comprehenders difficulties may be either the result of their poor reading comprehension or an underlying cause of their general reading comprehension difficulties. In order to establish which of these two alternatives in the more plausible, we compare the poor comprehenders performance with that of younger children who are at the same level of comprehension ability, but who are comprehending at an age-appropriate level: the CAM group. Example groups are shown in Table 2. If the CAM group does better than the poor comprehenders on a given task, the difference cannot result from the difference between the two groups in comprehension skill, because they are at the same absolute

4 54 Journal of Deaf Studies and Deaf Education 5:1 Winter 2000 Table 2 Group characteristics in comprehension-age match design Comprehension-age Good comprehenders Poor comprehenders Measurement match (n 27) (n 24) (n 29) Reading accuracy 6;8 7;10 7;11 (4.67) (5.28) (7.44) Reading comprehension 6;8 8;2 6;7 (3.45) (5.97) (3.94) Gates-MacGinitie sight vocabulary (3.60) (2.92) (3.46) Ages are given as years; months with standard deviations in months. The reading accuracy and reading comprehension scores are the age equivalent scores provided by the Neale Analysis. The maximum score possible on the Gates-MacGinitie test is 48. level of comprehension. It is, therefore, more likely that the skill difference is associated with the cause of the poor comprehenders delayed reading comprehension. We now outline two studies the first on inferences, and the second on story structure that have used this design. Inference Study With a CAM Group The main aim of the inference study was to explore the direction of the relation between comprehension skill and the ability to make inferences. Specifically, it asked whether there is any evidence that poor inferential skill might be causally linked to comprehension problems. The study also had a further subsidiary aim: to explore the reasons why good and poor comprehenders might differ in their propensity to make inferences, but those will not be covered in any detail here. The interested reader is referred to Cain and Oakhill (in press). Participants. The groups of skilled and less skilled comprehenders were selected from populations of 7 8- year-olds as described earlier. On the basis of performance on the Neale test, 29 less-skilled comprehenders were selected to take part in further testing. The lessskilled comprehenders all had age-appropriate reading accuracy, but their comprehension ages were below their chronological ages, and at least 6 months below their reading accuracy age. The mean difference between reading comprehension and reading accuracy was approximately 16 months. A comparison group of skilled comprehenders was selected from the original sample. These children had also obtained ageappropriate reading accuracy levels, but their comprehension scores were at or above that predicted by their reading accuracy age. The skilled and less-skilled comprehender groups were matched for reading accuracy, but differed in reading comprehension. The two groups were also matched for chronological age, Gates- MacGinitie vocabulary scores, and the number of stories that they had read from the Neale Analysis. The same tests were used for the selection of the CAM group. All of the 6 7-year-old children in four British Infant schools completed the Gates- MacGinitie vocabulary test. Because these children were a year younger than the older groups, a slightly lower standard on the Gates-MacGinitie was used to select the average performers to be tested on the Neale Analysis. Twenty-seven children were selected for further experimental work. The criteria for this selection were that both their accuracy and comprehension ages were within 6 months of their chronological age and that there was no more than 6 months discrepancy between the two scores. Furthermore, this group was selected so that their mean comprehension age was not significantly different from that of the less skilled comprehenders. All of the children who participated in this study spoke British English as their first language and had no known behavioral problems or learning difficulties. Table 2 presents the characteristics of the participants in this study. Materials and Procedure. The children were asked to read one practice story and four experimental stories, each of which was followed by six questions. Two of the questions asked about information that was explicitly presented in the text (literal), two required the children to combine information from different sentences in the text (text-connecting), and two required the combination of information in the text with general knowledge

5 Children s Difficulties in Text Comprehension 55 Table 3 Inference study Sample story and questions Debbie was going out for the afternoon with her friend Michael. By the time they got there, they were very thirsty. Michael got some drink out of his duffel bag and they shared that. The orange juice was very refreshing. Debbie put on her swimming costume but the water was too cold to paddle in, so they made sandcastles instead. They played all afternoon and didn t notice how late it was. Then Debbie spotted the clock on the pier. If she was late for dinner, her parents would be angry. They quickly packed up their things. Debbie changed and wrapped her swimming costume in her towel. She put the bundle in her rucksack. Then they set off for home, pedaling as fast as they could. Debbie was very tired when she got home, but she was just in time for dinner. Questions: Literal information Text-connecting inference Gap-filling inference 1. Who did Debbie spend the afternoon with? 2. Where was the clock? 3. Where did Michael get the orange juice from? 4. Where did Debbie put her towel when she packed up her things? 5. Where did Debbie and Michael spend the afternoon? 6. How did Debbie and Michael travel home? (gap-filling). An example of one of the stories and the associated questions is shown in Table 3. Results. The pattern of correct responses to the questions is shown in Table 4. An ANOVA showed that there was a main effect of Skill Group: Skilled comprehenders were better able to answer questions overall, F(2, 77) 16.33, p.0001, and a main effect of Question Type, F(2, 154) 32.33, p These effects were qualified by a significant interaction, F(4, 154) 3.24, p.02. The interaction arose because there was no difference between the three groups in their ability to answer questions that required a literal response, but there was a differential pattern of performance across groups for the questions requiring an inference. The between-group differences were tested using Bonferroni s t tests to take the number of comparisons into account (using adjusted tables in Howell, 1992). The less-skilled comprehenders did not answer significantly fewer of the literal questions correctly than the other groups, but they were poorer at answering textconnecting inference questions than both the skilled comprehenders, t (51) 5.17, p.01, and the comprehension-age match group, t (54) 4.11, p.01. The less-skilled comprehenders were also poorer than the skilled comprehenders on the gap-filling inference questions, t (51) 4.65, p.01, but the difference between the les-skilled and CAM groups on this question type was not significant. In this study, we also explored other factors that might be related to the poor comprehenders problems with inference making. We found no evidence that failure to draw inferences was caused by poor literal memory for the text and, in addition, no evidence that failure to make inferences resulted from general knowledge deficits. Indeed, when the poor comprehenders errors were pointed out to them, and they were allowed to look through the text again, they showed a substantial improvement in ability to answer the questions. Inference Making: Summary and Conclusions. The finding that the less-skilled comprehenders were not only poorer at answering inferential questions than skilled comprehenders, but were also worse than the CAM group on the text-connecting inferences, allows us to conclude that skill at drawing inferences is not simply a by-product of good reading comprehension skill; it is more likely to be a cause of good comprehension. We will later discuss the possibility that this link is not direct but may be mediated by other factors. In this study, we were able to rule out poor literal memory and lack of general knowledge as factors that were limiting the poor comprehenders inference skills (see Cain & Oakhill, in press, for details). Indeed, the improvement shown by the poor comprehenders when their errors were pointed out to them and they could search through the text again demonstrates that they were capable of making inferences, but apparently these children do not make inferences readily or spon-

6 56 Journal of Deaf Studies and Deaf Education 5:1 Winter 2000 Table 4 Mean percentages of correct responses in the inference study Question type Less-skilled group Skilled group CAM group Literal information (15.5) (16.0) (14.3) Text-connecting inferences (17.5) (15.0) (12.8) Gap-filling inferences (13.1) (14.7) (15.1) Standard deviations in parentheses. taneously when they read. It was interesting to note that the poor comprehenders still failed to make gapfilling inferences, even though all children had the requisite knowledge. It may be that strategy differences between the groups of children could underlie this finding: The less-skilled comprehenders may less able to relate general knowledge to the text in order to supply missing details. Story Structure Study With a CAM Group In another study using the CAM design (Cain & Oakhill, 1996), we explored children s understanding of story structure, by asking them to orally generate short stories in response to different types of prompts. The idea behind this experiment was twofold: First, we wanted to see whether the poor comprehenders problems in using cohesion during comprehension would also be manifest as problems in producing cohesive stories. We also wanted to explore the issue of causality, to see whether there was any evidence for use of cohesion as an underlying causal factor in children s comprehension skill. Participants. The participants in this experiment included three groups of children: 7- to 8-year-old good and poor comprehenders, and a younger CAM group. The participants for this study were selected in a manner that was exactly comparable to that described for the inference study. Materials and Procedure. The children were required to produce three stories given a topic prompt and three stories given a sequence of pictures that told a story. As well as rating their stories for overall quality of structure and cohesiveness (on a three-point scale), we also investigated the children s use of conventional devices in their storytelling. Thus, we scored their productions for the presence of traditional beginnings and endings and use of character- and scene-setting information. Examples of the children s story productions and details of the scoring procedures can be found in the original article (Cain & Oakhill, 1996). Results. We did not find any differences between the groups in their use of story conventions. Nevertheless, the good and poor comprehenders differed in their ability to produce well-structured stories: Good comprehenders produce better structured and more causally linked stories (which obtained higher ratings on our scale), F(2, 40) 8.15, p.002. In addition, the difference in story structure scores between the poor comprehenders and the CAM group was significant for the topic prompts, t(29) 2.55, p.05. The mean structure ratings for the three groups are shown in Table 5. Storytelling: Summary and Conclusions. In summary, the less-skilled comprehenders were poorer than skilled comprehenders at producing well-structured stories and were worse than the CAM group in the topic prompt condition. Therefore, skill at producing wellstructured stories is unlikely to be a by-product of good reading comprehension skill and is more likely to be a cause of good comprehension. Our feeling is that it is unlikely that there is something specific about storytelling that differentiates between good and poor comprehenders. It is more likely that an underlying factor might explain poor inference making and poor story production. This could be characterized, perhaps, as something like a drive for coherence, which has its effect on both production and comprehension.

7 Children s Difficulties in Text Comprehension 57 Table 5 Means for story structure scores Less-skilled Skilled Comprehension-age Prompt condition comprehenders comprehenders match Topic (1.77) (1.00) (1.39) Picture (1.08) (0.99) (0.94) Difference Maximum score possible for each prompt type is 6 (standard deviations in parentheses). Returning to Issues of Causality Before reaching general conclusions, we return briefly to the general issue of causality and to the link between comprehension and other skills. Our results suggest a causal link between inference making and comprehension skill, and also such a link between the ability to produce a causally coherent story and comprehension skill. However, the link between these skills and comprehension may not be direct but may be mediated by some other factor (likely candidates would be vocabulary or verbal IQ ). In fact, we have some evidence from our longitudinal data that this is not the case. In the longitudinal study, we included a number of measures that we know differentiate between groups of good and poor comprehenders (selected as above), and a number of other assessments: IQ, vocabulary, syntactic and phonological skills, and reading habits outside school. The ultimate aim of this study is to explore the stability and consequences of comprehension problems and to use the data for causal modeling. However, within each age group, we can also look at the relation between various skills and comprehension ability (as measured by the same standardized test as the one on which we matched groups). At Times 1 and 2 when the children were 7 to 8 and 8 to 9 years, respectively, we explored the predictive power of inference and other skills with verbal ability controlled. For present purposes, the data from Time 2 are the more interesting, since at that time we used the same stories to explore inference making as we did in the experiment outlined above. The data were subjected to two sets of multiple regression analyses: one with Neale Accuracy, and one with Neale Comprehension as the dependent variable. At the first four steps, a number of control variables Table 6 Summary of wave 2 regressions Step Variable R 2 change Significance level 1 Age ns 2 PIQ (t1).0171 ns 3 VIQ (t1).2376 p BPVS.1706 p Infer. total a.0750 p Literal.0159 ns 5 Story structure.232 p.061 Comprehension at time 2 as criterion (all other measures taken at time 2 unless stated); n 92. a Sum total of intersentence and contextual/gap-filling inferences. were entered. These were factors that we thought might contribute to comprehension skill: age, verbal IQ, performance IQ, and vocabulary. In order to assess the children s vocabulary, we used the British Picture Vocabulary Scale (BPVS). (This test is, essentially, a British version of the Peabody Picture Vocabulary Test) As can be seen in Table 6, both verbal IQ and BPVS vocabulary were significant predictors of comprehension skill. At the fifth step we entered, each in turn, the other variables of interest. As can be seen in Table 6, both inference making and performance on the sentence anagram task (described earlier) predicted variance in comprehension skill over and above that accounted for by verbal IQ and BPVS performance, whereas performance on the literal questions (in the inference task) was not a significant predictor. This pattern was specific to comprehension skill and did not hold when Neale Accuracy was entered as the dependent variable in the regression equation. In that case, neither the inference nor the story anagram scores predicted variance in single-word reading accuracy. Furthermore, it is interesting to note that these abilities predict variance in reading comprehension

8 58 Journal of Deaf Studies and Deaf Education 5:1 Winter 2000 even after the variance accounted for by general ability (IQ ) and more general language skills (such as vocabulary) has been entered into the regression equation. In these data, we can also see a clear dissociation between the skills that are related to comprehension and those that are related to word decoding. Of course, the real issue of interest is to explore the interrelations between the various skills and abilities across the three time points, which will only be possible once all the data are available. Conclusions In this article, we have presented evidence that comprehension and decoding skills are to some extent separable and are predicted by different underlying skills and abilities. This finding suggests that comprehension does not develop automatically once word decoding is proficient, but that it is dependent on different skills and may need specific teaching. Not only do we have evidence that we can tease apart the subskills of comprehension, but we also found that at least some components of comprehension skill (inference making and the ability to produce a causallycoherent story) are not merely by-products of being good at comprehension but, rather, may be causally implicated in comprehension development. Given the nature of this journal, it seems appropriate to make some tentative comments about the implications of our findings for deaf readers. As far as we understand, deaf children have trouble learning to read because they lack the ability to represent words in phonological memory and because their difficulties in word reading limit their comprehension. We have found that working memory capacity reliably discriminates between good and poor comprehenders who are matched for single-word reading. However, good and poor comprehenders perform comparably on tasks that assess phonological processing skills and simple short-term memory span. The difficulty with the representation of words in phonological memory that is found in deaf readers may result in a similar pattern of comprehension difficulty found in dyslexics or poor word readers (i.e., reading comprehension may be limited by the efficiency of single-word reading). Thus, the teaching of single-word decoding must be a priority for teachers of deaf children. It is not clear whether there would be such a clear dissociation between word reading and reading comprehension in deaf readers, as is found in hearing readers. It is possible that problems with phonological memory might influence both their word reading and their comprehension to a greater extent than is the case in hearing children (see Perfetti, this issue). However, as with hearing individuals, if these two main aspects of reading are dissociable, then it is important to teach both components. Thus, we would advocate that instruction in text comprehension skills occurs alongside instruction in single-word reading. In general, children who have poor word-reading skills have much better listening comprehension skills relative to their reading comprehension ability. One way to ensure that the development of text comprehension skills is not restricted by word-reading difficulties is to use listening comprehension tasks. It is important that written discourse is used in these tasks, because the vocabulary and complexity of reference differs between general conversation and written text. For deaf individuals who are fluent in signing, it would be possible to present written texts via sign language in order to teach skills such as inference making, comprehension monitoring, and the planning and structuring of stories. This instruction would enable the reader to develop the skills necessary to construct integrated and coherent representations of texts that are signed, skills that should be transferable to understanding written text. Once word-reading skills are in place, readers will not be placed at a disadvantage if they have already developed good comprehension skills. Furthermore, when text-processing strategies are fluent and automatic, the limited processing capacity of short-term memory, so vital when word-reading skills are not automatic, will not be unduly burdened with this higherlevel processing of the text. Expertise in these comprehension skills will enable deaf readers to experience the enjoyment of a text at a suitable level of complexity for their age and intelligence and may, thus, prevent the loss of motivation to read that is seen in so many children with word-reading difficulties. Finally, in contrast to the population of readers that

9 Children s Difficulties in Text Comprehension 59 we have studied, it is possible that the reading comprehension of deaf children may greatly depend on their word-reading ability, because of their phonological memory restrictions (see Gaustad, Wilbur, this issue). Thus, it is important to investigate the components of word reading, reading comprehension, and listening comprehension skill in deaf children, and their patterns of association and dissociation, in order to determine how best to aid the development of their textprocessing abilities. Received June 10, 1999; revised July 20, 1999; accepted July 23, 1999 References Baker, L., & Stein, N. (1981). The development of prose comprehension skills. In C. M. Santa & B. L. Hayes (Eds.), Children s prose comprehension: Research and practice (pp. 7 43). Newark, DE: International Reading Association. Bishop, D. (1982). Test for the Reception of Grammar. Manchester, England: Chapel Press. Cain, K., & Oakhill, J. V. (1996).The nature of the relationship between comprehension skill and the ability to tell a story. BritishJournal of Developmental Psychology, 14, Cain, K., & Oakhill, J. (1998). Comprehension skill and inference making ability: Issues of causality. In C. Hulme & R. M. Joshi (Eds.), Reading and spelling: Development and disorder. Mahwah, NJ: LEA. Cain, K., & Oakhill, J. V. (in press). Inference making and its relation to comprehension failure. Reading and Writing. Cain, K., Oakhill, J. V., & Bryant, P. E. (in press). Investigating the causes of reading comprehension failure: The comprehension-age match design. Reading and Writing. Gates, A. I., & MacGinitie, W. H. (1965). Gates-MacGinitie Reading Tests. New York: Columbia University Teacher s College Press. Howell, D. C. (1992). Statistical methods for psychology. 3rd ed. Boston: PWS-Kent. MacGinitie, W. H. & MacGinitie, R. K. (1989). Gates- MacGinitie Reading Tests. Chicago: Riverside. Neale, M. D. (1989) The Neale Analysis of Reading Ability- Revised. Windsor, Berkshire: NFER-Nelson. Oakhill, J. V. (1982) Constructive processes in skilled and lessskilled comprehenders memory for sentences, BritishJournal of Psychology, 73, Oakhill, J. V. (1984) Inferential and memory skills in children s comprehension of stories. BritishJournal of Educational Psychology, 54, Oakhill, J. V, Cain, K., & Bryant, P. E. (1999). Dissociation of single-word reading and text comprehension skills. Manuscript submitted for publication. Oakhill, J. V. Hartt, J., & Samols, D. (1996, August). Comprehension monitoring and working memory in good and poor comprehenders. Presented at the 14th Biennial ISSBD Conference, Quebec City. Oakhill, J. V., Yuill, N. M., & Parkin, A. (1986). On the nature of the difference between skilled and less-skilled comprehenders. Journal of Researchin Reading, 9, Seigneuric, A., Ehrlich, M.-F., Oakhill, J. V., & Yuill, N. M. (in press). Working memory resources and children s reading comprehension. Reading and writing. Siegel, L. S., & Ryan, E. B. (1989). The development of working memory in normally achieving and subtypes of learning disabled children. Child Development, 60, Stothard S., & Hulme, C. (1992). Reading comprehension difficulties in children: The role of language comprehension and working memory skills. Reading and Writing, 4, Yuill, N. M. & Oakhill, J. V. (1991). Children s Problems in Text Comprehension: An Experimental Investigation. Cambridge: Cambridge University Press.