University of Illinois at Urbana - Champaign Library Large - scale Digitization Project , 2007 .

A study was conducted to identify poor readers and to characterize weaknesses in their knowledge and use of story structure in comprehension and recall. Subjects were 80 year-3 children, 20 good readers and 60 poor readers. The poor readers were then divided into relatively homogeneous subgroups, using measures of language-reading comprehension, according to a numerical classification procedure. This procedure helped identify specific weaknesses in their language-reading comprehension. All children listened to 3 stories and retold the stories under freeand probe-recall conditions. Comparison of recalls between the good readers and each of the subgroups of poor readers showed that poor readers in 2 subgroups evidenced reduced sensitivity to story structure. The children in these subgroups recalled less of the stories overall, recalled less information from story grammar categories to varying extents, and showed patterns of category recall which differed from those of normal readers. Children in one of the subgroups also displayed poor perception of causal relations across story episode boundaries. These results provide evidence of marked heterogeneity in poor readers' story comprehension and recall. Certain subgroups of poor readers may have qualitatively different problems processing stories, relative to other poor readers, which may require a more concerted approach to instruction in story structure. Wilkinson, Elkins, & Bain INDIVIDUAL DIFFERENCES IN STORY COMPREHENSION AND RECALL OF POOR READERS A common approach to research on children with reading difficulties is to compare the status of groups of readers of different ability. The research designs define "good" and "poor" readers according to some criterion and then compare the status of these readers on measures of cognitive performance. The goal of some of these studies is to identify the underlying cause(s) of children's reading difficulties. Such status studies have compared groups of readers on a wide variety of cognitive measures, including isolated word recognition, oral reading, vocabulary knowledge, and memory (for reviews, see Aulls, 1981; Kleiman, 1985; Lipson & Wixson, 1986). However, to date, such studies have failed to identify consistently an area of cognitive performance that is responsible for reading difficulty. This inconsistency in findings has been particularly true of studies in the area of story comprehension and recall. In these studies, children listen to stories and then perform a comprehension or recall task. By comparing the good and poor readers' comprehension and recall, researchers are able to make inferences about children's relative sensitivity to aspects of story structure. The rationale for these studies is that if good and poor readers differ in sensitivity to story structure, independent of their decoding abilities, then deficiency in story schema, or failure to use story schema, may be responsible for some of the difficulties experienced by poor readers. Results have shown quantitative differences between readers of different ability--poor readers comprehend and recall less of a story than do good readers (although even here there is some inconsistency)--but results have been inconclusive as to whether there are qualitative differences in comprehension and recall. Thus, it is not yet clear whether good and poor readers differ in sensitivity to story structure. Studies have employed a variety of methods to assess sensitivity to story structure. One group of studies has investigated students' awareness of relative importance of idea units in stories. Levels of importance of idea units have been empirically defined using Johnson's (1970) technique. This approach is largely atheoretical with respect to the role of idea units in story comprehension and recall. As such, the studies do not necessarily implicate story-specific knowledge. Results have been mixed. On the one hand, Smiley, Oakley, Worthen, Campione, and Brown (1977) and Wong (1979) found that poor readers recalled fewer idea units than did good readers and that they were sensitive to fewer gradations of structural importance than were good readers. On the other hand, Luftig and Greeson (1983) found no differences in sensitivity to gradations of importance between educable mentally retarded and normal children, and Worden and Nakamura (1982) found no such differences between learning-disabled and normal college students. Worden and Nakamura found no differences even in overall recall of the 2 groups (though these results may have been due to students' repeated exposure to the stories). Another group of studies has investigated students' sensitivity to story structure using more theoretically motivated propositional analyses. Hansen (1978), using Kintsch's (1974) propositional model, found that learning-disabled children recalled fewer propositions overall and recalled fewer superordinate propositions than did normal children. The 2 groups did not differ in recall of subordinate propositions (see also Weisberg, 1979). By contrast, Feagans and Short (1984) parsed their stories in terms of "action units" and failed to find any major differences between reading-disabled and normal children in comprehension and recall of these units. Wolman (1991), using Trabasso, Secco, and van den Broek's (1984) causal network analysis, found that children with mild disabilities recalled less than did children without disabilities but, again, failed to find differences between the groups in sensitivity to causal connections in stories. By far the largest group of studies has investigated students' sensitivity to story grammar categories. Story grammars are analytical tools that describe the structural components of narrative text (Mandler Individual Differences 2 Wilkinson, Elkins, & Bain & Johnson, 1977; Rumelhart, 1975; Stein & Glenn, 1979; Thorndyke, 1977). Although story grammars differ in detail, the categories of information are essentially the same. Stein and Glenn's (1979) wellknown grammar proposes that a story consists of a major setting (main character), minor setting (time and place) followed by one or more episodes. The episodes comprise 6 categories: initiating events, internal responses, internal plans, attempts, direct consequences, and reactions. Story grammars are assumed to approximate the readers' (or listeners') cognitive schema that guide the encoding and retrieval of story information. As such, studies of readers' sensitivity to grammatical categories more closely implicate story-specific knowledge. Results of studies comparing good and poor readers have again been equivocal. A classic study by Weaver and Dickinson (1982) (see also Weaver, 1978; Weaver & Dickinson, 1979) examined the story recall of 10and 13-year-old "dyslexic" boys and compared their results with those from Stein and Glenn's (1979) normal grade-5 readers using the Stein and Glenn grammar. They found no differences between the younger disabled and normal readers in overall recall (comparison with the older disabled readers was not made because of the age difference) and only 2 differences in recall of information within categories--the disabled readers recalled somewhat less of the character's thoughts or feelings about the outcome (the reaction category) and of the story context (minor setting). Moreover, they found only minor differences in the pattern of recall of story grammar categories (the rank order of recall of the categories). A number of other studies comparing good readers with various categories of poor readers have reported similar results, finding little evidence of differential sensitivity to story grammar categories (Backman, Lundberg, Nilsson, & Ohlsson, 1984; McConaughy, 1985; Summers, 1980; Worden, Malmgren, & Gabourie, 1982; see also Gold, 1983). These results stand in marked contrast to those of Fitzgerald (1984), Hinchley and Levy (1988) and Rahman and Bisanz (1986) who reported having identified poor readers who demonstrated reduced sensitivity to story grammar categories (see also Barnhart, 1990). A fundamental problem with the above studies, and with status studies generally, that may account for the inconsistency in findings, is heterogeneity in the samples of poor readers. This was suggested by Wiener and Cromer (1967) and elaborated by Applebee (1971), Elkins (1978), Kleiman (1985), and Singer (1982). Most studies examining the performance of poor readers, relative to that of good readers, have ignored individual differences and assumed that the poor readers constitute a homogeneous group. However, Applebee and others pointed out that there may be considerable heterogeneity of reading difficulties in samples of poor readers. They argued that if there were relatively homogeneous subgroups within the poor reader sample, and these subgroups were ignored by averaging across the subgroups, then differences between the good and poor readers may be obscured. Comparison of the poor readers' performance with that of good readers may reveal no differences or differences that were unstable (i.e., sample specific) and that did not apply to any one subgroup (see also Backman, Mamen, & Ferguson, 1984; Harris, 1978-1979; Lipson & Wixson, 1986). Note that this argument posits the existence of systematic individual differences. Individuals naturally differ from each other in a variety of ways. The existence of systematic differences suggests that there are similarities as well as differences in the ways students perform and that it is possible to distinguish subgroups reflecting the systematic rather than random component of variation between students. Students in any given subgroup should share a pattern of performance on variables that defines that subgroup and distinguishes its members from those of other subgroups (Applebee, 1971; Kareev, 1982). There are indications that the problem of heterogeneity in samples of poor readers is implicated in story comprehension research. Weaver and Dickinson noted large variation in performance within their group of dyslexic students (see especially Weaver & Dickinson, 1979). Indeed, when they divided their dyslexics into subgroups based on verbal-performance IQ discrepancy scores, the few significant Individual Differences 3 Wilkinson, Elkins, & Bain differences obtained in their dyslexic-normal comparison seem to have been due to only one subgroup of disabled readers (the less verbally proficient) and the non-significant finding for total recall obtained only for comparison of the normal readers with another subgroup (the more verbally proficient). Hinchley and Levy (1988) also obtained results that suggest that there may be a large individual difference component in story comprehension and recall and that only some poor readers have deficits in story-structure knowledge. Another problem that may account for the inconsistency in findings is that the measures employed may not have assessed relevant aspects of students' knowledge or use of story structure. Story grammars specify not only certain categories of story information but also the relations among the categories. Stein (1982) has argued that differences between good and poor readers may be found only if students are required to perform tasks that deal with the relational properties of stories. Indeed, studies that have used tasks requiring students to deal with the relational properties of stories, by having them anticipate upcoming story information or recall stories that deviate from the canonical form prescribed by a grammar, have found significant differences between good and poor readers in sensitivity to story structure (Fitzgerald, 1984; Hinchley & Levy, 1988; Rahman & Bisanz, 1986). The purpose of the present study was to identify poor readers who show weaknesses in their knowledge and use of story structure. The study sought to address the 2 problems described above. First, to take into account individual differences among poor readers, we identified homogeneous subgroups within the poor reader sample by numerically classifying the children on the basis of the component structure of reading comprehension ability. There is ample evidence that poor readers can be grouped into distinguishable subgroups (e.g., Carr, Brown, Vavrus, & Evans, 1990; Doehring, Trites, Patel, & Fiedorowicz, 1981; Lovett, 1984; Torgesen, 1982), although it is uncertain whether knowledge of story structure relates to any of the groupings. We reasoned that if any deficits in story-specific knowledge could be found, they would obtain for only some subgroup(s) of poor readers. We used the numerical classification procedure as a device to address the problem of heterogeneity in our sample of poor readers. Using this approach, we hoped to be able to separate systematic individual differences from differences due to random error and, thereby, detect ability-group differences in story comprehension and recall that heretofore may have gone unnoticed. Our point is that previous story grammar studies may have not only overlooked some interesting findings about differences within the group of poor readers, they may have also lumped the variance associated with these differences together with error variance, thus decreasing the power of tests of differences between groups of good and poor readers. Second, to assess children's perception of relations among story information, in addition to assessing children's free recall of stories, we examined their probe recall of causal relations. The probes were designed to provide a more structure-dependent measure of recall. We decided to focus on causal relations as these have been found to be an important determinant of reproduction probability in summarization tasks (Graesser, 1981; Lehnert, Black, & Resier, 1981) and have received attention in recent systems of text analysis (Trabasso & van den Broek, 1985). Probe questions were constructed for each story and targeted at either interor intra-episodic causal relations. Higher order probes were also used so that the extent of a causal relation perceived by a child could be assessed.