Meaning construction in early oral reading.
Keywords: early reading, reading comprehension, miscues, evaluation
The general aim of this study is to report a classroom-friendly method for exploring students' reading comprehension/meaning construction consistent with such longitudinal investigations focused on emergent writing and reading skill development as those developed by Cameron, Hunt, and Linton (1996), Cameron and Wang (1999), Cameron and Hutchison (2009), Dyson (2003), and Wells (2009). Theoretical accounts of reading comprehension, operationally defined here as the "interactive construction of text meaning" (Brandt, 2009; Farstrup & Samuels, 2002; K. S. Goodman, 1985; Olson & Torrance, 2009; Samuels, 1987; van Dijk & Kintsch, 1983), emphasize the importance of text and reader contributions to a successful reading experience. Making sense of written communication involves lower, automatic processing levels related to the visual encoding of small linguistic units of text (Perfetti, Landi, & Oakhill, 2005) in interaction with higher, memory-dependent processing levels related to knowledge a reader can draw on to build a context of meaning (Conway, 2002).
The Role of Meaning Construction in Emergent Reading
Perfetti et al.'s (2005) model of reading development is based on the multiple processing levels instrumental in developing reading fluency. Basic processes include letter recognition, lexical and semantic access, and word-decoding skills that develop under general constraints of memory systems. Second, fluent reading involves syntactic processing, the integration and linkage of individual word meanings derived from decoding into sentence and text meanings, and develops with experience, as well as being constrained by memory. According to Perfetti et el., a goal of beginning readers should be to make use of context, as they construct text meanings. Yet for reading fluency to develop, beginning readers need skill in context-free decoding so that attention and memory are available for higher levels of syntactic processing. Stanovich (1980, 2003) has similarly posited an interactive-compensatory reading mechanism: Inefficient processing at one level leads to compensation at another level; when decoding skills are inefficient, more reliance on context and syntactic processing is necessary. Beginning readers are thus more dependent on context than fluent readers, and skilled readers show less reliance on context in word recognition than poor readers (Cameron et al., 1996). Thus, such theorists as K. S. Goodman (1985) and Smith (2005) have asserted the critical importance of context usage in emergent reading, citing observational evidence that good beginning readers make more use of context to recognize words. From a contrasting point of view, Perfetti et el. (2005) and Stanovich (1980, 2003) argue that fluent reading develops from decoding skill mastery: decoding processes may be facilitated by linguistic processes based on context use, but development of higher levels of text processing skill is limited by decoding skill attainment. Less-skilled readers' syntactic processing is directed at word recognition, whereas more-skilled readers may build text meanings pursuant to syntactic processing. Pads (2005) describes an "uneasy compromise" in the debate on the developmental order and the relative importance of decoding and comprehension skills in emergent reading, whereby both are "components to be acquired and assembled," and that "a balance of both kinds of skills" should be achieved in early instruction (Paris, 2005, p. 187).
Advancing an Ecologically Valid Index of Meaning Construction
Bronfenbrenner (1979) urged researchers to strive for ecological validity in developmental investigations, and Sameroff (2010) focused attention on the transactional nature of developmental processes. There might be nothing more important to aim for when indexing meaning construction than observing transactions with meaning making in ecological context. In evaluating meaning construction processes in reading, most traditional tests merely assess memory for text content and do not afford such validity. Recall testing, with questions inserted at the end of sentences or paragraphs in a text, provides an indication of the current state of meaning construction, but such insertions can influence subsequent processing of text meaning, as readers' goals change in response to the task, thus interfering with naturally developing meaning construction. "Schematic techniques," "conceptual grids," "story grammars," and "propositional analyses" have been useful in research on the structure and content of recall. However, without significant modifications, these techniques are not amenable to the measurement of concurrent processing during emergent reading, making them less appropriate as online evidence of meaning construction processing more useful for classroom formative evaluations.
The Reading Miscue Inventory (RMI; Y. Goodman, Watson, & Burke, 2005; Y. M. Goodman & Burke, 1972) provides such oral-reading error (or "miscue") probes, sensitive indices of semantic and/or syntactic processing in early readers. The meaning change component of the RMI assesses the discrepancy between a written word's meaning and the oral reading error or miscued meaning of that word. The RMI philosophy assumes reading to be a "psycholinguistic guessing game," and that reading need not necessarily be accurate to be meaningful. There are miscues that do little damage to meaning construction, just as there are miscues that significantly disrupt meaning. Our interest in the RMI lies in its potential to provide quantitative and qualitative, formative, noninvasively collected indices of meaning construction. However, scoring reliability of miscue meaning changes is difficult to achieve when miscues are not under educator/researcher control, so the RMI is most suited to psycholinguistic descriptions and formative rather than summative evaluations. We furthered this goal by developing a process measure of meaning construction by manipulating meaning change within a miscue analysis protocol, inserting textual disruptions such as might be found in editing a child's own textual productions and thus placing those errors under experimental control.
Additionally, the concern for more objective scoring of meaning change in this study led us to apply Faigley and Witte's classic (1981) distinctions between surface changes and text-based meaning changes in writing revision. Surface changes are either formal (superficial; e.g., spelling or punctuation), or meaning preserving (paraphrases of text information) changes. Disruptive meaning changes are divided into microstructural and macrostructural changes; the latter category alters the summary of a text. Our modification of this meaning change taxonomy indexed each child's reading of intact text to allow for more detailed evaluation of miscues and increased meaning change score reliability. In the latter part of the text, we inserted disruptions that resembled these 2nd-graders' oral-reading miscues. Participants' adaptations to these disruptions were divergences between what was written and what was read aloud, allowing examination of responses to different levels of meaning change.
The Research Question and Hypotheses
The specific goal of this study was to devise an efficient, objective, and ecologically valid procedure for indexing young readers' transactions in making a text meaningful during oral classroom reading. Our research question was "What is the relationship between adaptations made during the reading of disrupted text and the production of miscues during the reading of undisrupted text?" It was hypothesized that an inverse relationship would be found between text disruption adaptation scores and reading miscue scores. It was more specifically hypothesized that children who do not produce macrostructure miscues would be better equipped to adapt extensive text disruptions. In sum, a relationship between meaning change ratings and disrupted-text adaptation scores was expected to provide convergent evidence in support of using either meaning change scores, or strategic text disruption adaptation scores on their own, to index meaning construction during reading, thus accounting for the constructive, interactive, contextually sensitive nature of text processing levels under development in young readers (Cameron & Hutchison, 2009; Conway, 2002; Paris, 2005).
Participants were 51 2nd-graders attending school in a small, predominantly middle-class, provincial capital city in eastern Canada. Although participants came from two classrooms located in separate schools, there were no apparent differences between schools on the basis of parents' socioeconomic status. There were 24 girls and 27 boys, ranging in age from 7 years 5 months to 9 years 4 months, with a mean age of 8 years 1 month. English, the language of instruction, was the first language of most children. All of the children were participants in a longitudinal literacy study, for whom parental permission for participation had been obtained, and procedures were in accord with ethical guidelines set down by the Society for Research in Child Development (2007). There were 22 girls and 27 boys in the final sample, as two children were omitted from analysis--one because she did not finish reading the disrupted text and the other because she adapted the last half of the disrupted text by picking out occasional text items and weaving a story around them, without "fixing mistakes" as the other children did.
"The Bradleys Move In" (in Clymer & Martin, 1978) is a 618-word story with 95 lines of text and a Spache readability index of 2.6 (Spache, 1966). In it, a boy and a girl meet and make friends with a family that moves in next door. The story was transcribed in bold type, and the original illustrations were traced and colored. Text disruptions were inserted in the last 150 words (on the last two pages of the eight-page story). The drawings in the disrupted text were left uncolored as an additional cue that this part of the story was different. This point of division between texts was chosen because it was estimated from pilot data that readers would have produced a large enough body of miscues to create a valid RMI rating by this point in the story, having already read up to 468 words. Pilot 2nd-graders had produced an average of one miscue for every eight words read. Within one standard deviation of that figure, some readers produced as many as one miscue for every three words read, whereas others produced as few as one miscue for every 20 words. These figures were used as guidelines in determining the number of disruptions to insert. As few as eight disruptions or as many as 50 could have been inserted in the text while staying within the range of these guidelines. A total of 27 disruptions representing three instances of each level (structural, minimal, and extensive) and type of disruption (substitution, insertion, omission) were thus inserted into the text.
Structural substitutions were created by replacing nouns with nonwords similar to the original noun, except for a change in a vowel sound represented by the spelling. Insertions of this level of disruption were obtained by adding an s to a singular noun, and omissions at this level involved the deletion of the past tense ed verb ending.
Minimal substitutions were obtained by replacing nouns with nonwords similar to the original noun except for a reversal of two of the phonemes. Word shape remained the same. We created minimal insertions by placing the before a character's name, and omissions at this level of disruption involved deleting part of a compound verb.
We replaced characters' names with nonwords that kept all of the original text items' letters to create extensive substitutions, although consonants were printed first and then vowels. Insertions at this level were of whole nouns placed before nouns already in the text. Omissions were of whole one-word verbs. Appendix A contains a copy of the disrupted text in which the disruptions are underlined and numbered.
The researcher was introduced to each classroom on a first visit. The teacher told the children that they would be invited to read a story with the researcher and that each of them would get a turn. The reading took place in a corner of the classroom sectioned off by a low movable partition. The children were read with and recorded individually. For reading the intact text, children were told that they could sound out hard words, guess them, or skip them and go on. Whether or not they finished reading the intact text, all participants were congratulated on their performance. The intact text narrative was summarized by the researcher and the participants were told the next part of the story would be different: it would contain mistakes that could be changed if they didn't seem right, they could add words if any were missing, and they could skip words and go on. The children were told to make sense of the story as best they could, and to read it so that it sounded right to them. There was no time limit for reading the disrupted text. After finishing this task, all children were again congratulated on their performance and thanked for their help.
Meaning-changing miscues. The reader's most appropriate or least meaning-damaging miscue of a word was scored for meaning change, rather than the first miscued attempt. The reader's first 25 miscues formed the sample to be scored for meaning change. Each miscue was scored with regard to the accurately read words and the miscued words that surrounded it. The crucial factor in determining meaning change here (using Faigley & Witte's 1981 taxonomy) is the extent to which miscues affect the subject and/or verb of main clauses. Because oral reading miscues sometimes result in semantically and/or syntactically unacceptable structures, we differentiate between the two. Classifications are further based on the type of linguistic units (i.e., nouns vs. verbs vs. adjectives and adverbs) being changed, the operations by which the changes are effected (omissions, insertions, substitutions), the interconnection of changes, and whether the changes are stated explicitly in contradiction to text information or implied by a lack of specification. Appendix B provides exemplars for scoring meaning change.
Meaning-preserving miscues (score = 1) are the most superficial changes in this taxonomy. They do not change the number, person, or gender
of any subjects or verbs; they do not affect subject-verb agreement; and they do not alter any of the object, spatial, causal, or temporal relations expressed in the text. The insertion of that in "He hoped ^that there would be a boy next door," and the omission of Up in "(Up) she went, higher--still higher" are good examples of this category. Formal miscues (score = 2) do not alter the essence of subjects and verbs found in main clauses either, although they could implicitly change subject-verb agreement by generalizing the subject referent or by pluralizing the verb. Generally, they do not create explicit contradictions in text information. None of these miscues explicitly alter object, spatial, causal, or temporal relations. Omitting the ed in "climbed" in the sentence "And the cat that climbed up in your oak tree was MY CAT ..." is a formal miscue, as is reading they for he, high for higher, etc. Microstructure miscues (score = 3) can result in explicit contradictions in the subject-verb agreement of main clauses by singularizing plural subjects or verbs and include changes in function words that result in the temporal, causal, spatial, or object relations being explicitly contradicted. Reading they as he in the sentence, "They waited by the front door till the woman looked at them and smiled" is a miscue of this category. So is reading waiting as sitting in the sentence, "A woman was waiting by the front door" (the woman could still be waiting). Macrostructure miscues (score = 4) alter basic meaning of subjects and/or verbs of main clauses. They do so by explicitly resulting in semantically unacceptable structures. Omissions of whole subjects or verbs from main clauses, or their substitution for words that are not at all synonyms, both fit into this category. Reading till as tell in "William and Lucy waited by the front door till the woman looked at them and smiled," or substituting she for he, are examples of macrostructure miscues. Nonwords also fit into this category.
If a direct, obvious, and meaningful link was noted between a miscue and one or more of the other miscues, that miscue was scored one point lower than it would have been had there been no such link. If a miscue was corrected, it also received a score one point lower than it would have without correction. For example, for reading Stilly on a first attempt at Stanley and then correcting, an instance of correcting a formal miscue (score = 2) was scored 1. This took into account that a miscue was produced, but it was not considered so damaging to meaning as it might have been had it been left uncorrected. Corrected meaning-preserving miscues were included in the sample scored, although they received a score of 0. When these guidelines were followed, the sum of meaning change scores provided an index of the degree of meaning change contained in a reader's sample of miscues.
Text-disruption adaptations. We identified five different levels of adaptation to disruptions. Generally, adaptations were characterized as being: less disruptive and contextually appropriate (evidence of the highest meaning making in this context, score = 4), less disruptive but contextually inappropriate (score = 3), and so forth for equally disruptive and more disruptive. Examples of each level of adaptation are demonstrated with the nonword disruption frinz as: friendly (less disruptive, contextually appropriate); freed (less disruptive, contextually inappropriate); freenz (equally disruptive, score = 2); fr--(more disruptive, score = 1). Adaptation was defined with regard to disruption level and disruption type, so that what might have been a successful adaptation (or a less disruptive utterance) at one level or on one type might not have been successful at another level, or on another disruption type. Accurate reading of a disruption (e.g., pronouncing frinz so that it rhymed with pins) received the lowest score for adaptation, because lack of adaptation in this context was operationalized as a lack of evidence of adaptation. This definition was not meant to imply that children who read disruptions accurately were not gleaning meaning from their reading, although it did mean that no evidence of sensitivity to problems in the text was apparent in the recorded oral reading sample. Also, adaptation did not necessarily mean "correction," because a response to a disruption might include productions other than the original text meaning, or unacceptable productions.
Reliability Estimates for Scoring
The children's oral reading performance was scored, although not all psychometric properties of the scoring systems are yet known. A trained research assistant independently scored one quarter of the data to calculate inter-rater agreement. The weighted kappa (Kw) for ratings of meaning change was .8203, p < .01 (99% confidence intervals [CI] [.8198, .8208]). For ratings of adaptations, Kw was .8750, p < .01 (99% CI [.8749, .8752]).
Miscue scores were compiled and subjected to descriptive statistical analyses for reporting mean frequencies and then inferential ([tau] correlational) analyses were applied to the mean scores. The inferential Wilcoxon matched-pairs signed-ranks test (z) was applied to median reading speed scores on undisrupted and disrupted texts. Finally, a Friedman two-way ANOVA by ranks for related samples ([chi square]) was calculated to test the overall effect of disruption type (substitutions, insertions, omissions) on adaptation scores.
The children produced an average of 39.04 miscues while reading intact text at a mean rate of 11.73 lines per min, with a mean total meaning change score of 52.94. The average number of lines of intact text read was 60.69 out of a possible 73. Fifteen (31%) of the children did not finish reading the intact text in the time allotted and 11 (23%) of the children produced fewer than 25 miscues.
Miscue ratings for the group were as follows: on average 3.14 were rated as macrostructural miscues, 6.74 were rated as microstructural, 6.31 were rated as formal miscues, 7.55 miscues were rated as meaning preserving, and 1.27 miscues were rated as having no meaning change. Please see Table 1.
The average adaptation score on the disrupted text was 79.51 and the text was read at 6.17 lines per minute on average (see Table 2).
Inferential Statistical Analyses
Total meaning change scores and total adaptation scores were highly negatively correlated, [tau] = -.55, p < .001, confirming our main hypothesis that children's adaptation scores to disrupted texts would be inversely related to meaning change miscue scores based upon reading intact text. Thus, children earning higher meaning change scores obtained lower adaptation scores. Furthermore, miscue meaning change scores were related to a simple accuracy measure--in this case, total number of miscues, [tau] = .44, p < .001--and were also negatively related to number of lines of the intact text read, [tau] = -.39, p < .001. Meaning changes and adaptations were also negatively related as follows: number of macrostructural miscues related negatively to adaptations on extensive text disruptions, [tau] = -.34, p < .01, and number of microstructural miscues related to adaptations on minimal disruptions, [tau] = -.22, p < .05.
Number of formal miscues were not related to adaptation scores on structural disruptions, [tau] = .07, p > .05. Therefore, children who produced few macrostructure miscues produced more meaning-restoring adaptations on extensive disruptions, indirectly confirming our second hypothesis that children who did not produce macrostructure miscues would make more successful adaptations to extensive text disruptions. Children who produced few microstructure miscues produced more meaningful adaptations on minimal disruptions, although this association was not so strong. Indeed, average adaptation score on each of the three levels of text disruption were 30.00 for structural disruptions, 26.43 for minimal, 23.08 for the extensive level of text disruptions, and the order of relations expected was based in part on the assumption that structural disruptions would be most readily adapted, followed by minimal and that extensive disruptions would be the most difficult to adapt. This was tested post-hoc using Page's (1963) test for ordered alternatives: Page's L statistic = 656.5 (for three treatments and N = 49), p < .001. In order, median adaptation score totals for structural, minimal, and extensive disruptions were 32, 29, and 22, respectively, confirming expectations.
Intact text reading rate (11.73 lines per minute--Table 1) was not just negatively related to meaning change (Table 1) [tau] = -.43, p < .001, but it was also positively associated with disrupted-text adaptation (Table 2) [tau] = .47, p < .001. These relationships are important because they respond to questions about how disruptive were the latter portions of text, which were read at a rate of 6.17 lines per minute on average overall (Table 2). A comparison of individual children's reading rates on each type of text was carried out using a Wilcoxon matched-pairs signed-rank test, which showed that the children read disrupted text significantly more slowly (Md = 6.17 lines per minute) than undisrupted text (Md = 11.61 lines per minute), as N = 49, z = -5.74, p < .001.
A median split at a meaning change score of 54 was used to differentiate children classified as low (N = 23, median score = 46), or high meaning change readers (N = 26, median score = 58.5). The intact text reading rates of low meaning change readers (median = 14.50) were greater than high meaning change readers' reading rates on either text (median rate for intact = 8.55, median rate for disrupted = 5.18). No difference between readers was found on disrupted text reading rates: 5.74 lines per minute for low meaning change readers and 5.18 lines per minute for high meaning change readers. This reader by text differential is graphically depicted in Figure 1.
To test the overall effect of disruption type (substitutions, insertions, omissions) on adaptation scores, a Friedman two-way ANOVA by ranks for related samples indicated ([chi square] (2, N = 49) = 57.44, p < .001) that adaptation scores of at least one type of disruption were different from scores on at least one other type of disruption. Post hoc tests using Daniel's (1978) multiple comparison procedure showed that scores on substitutions were different from scores on insertions (rank total difference = 56.84, p < .01) and omissions (rank total difference = 71.05, p < .01). No differences were found between scores on insertions and scores on omissions (Table 3).
Mean adaptation scores for individual disruptions were compared to the mean adaptation score for all disruptions (M = 2.95, SD = .65) and scores beyond the 2.3 to 3.6 range (range = M + or - 1 SD) were considered noteworthy. Adaptation scores to substitutions: Disruptions 2, 14, 19, and 27 (see Appendix A), and adaptations to structural insertions: disruptions 16 and 22 appear high. Adaptations to extensive insertions: Disruptions 8, 17, 24 and extensive omissions: Disruptions 13 and 21 appear low. Disruptions 8, 17, and 24 constituted adjectivally ambiguous insertions with little cue to their being mistakes to fix. As for Disruptions 13 and 21, both omissions of a verb, it may also be the case that there was not enough sentence or text-level context to provide cues for better adaptations.
When mean adaptation for each disruption level is used as a guideline, a different picture emerges. The children's average adaptation score for structural disruptions ranged from 2.68 to 3.98, minimal disruption adaptations ranged from 2.08 to 3.89, and adaptations for extensive disruptions ranged from 1.85 to 3.27. Any disruption for which the mean adaptation score was beyond the range for that disruption's level was considered noteworthy. In this manner, adaptations on Disruptions 4, 19, and 27 (all character name substitutions that each contain a jumble of the letters contained in the name they replace, see Appendix A) appear high for extensive disruptions; they may have been readily identifiable as a specific character name because of the contextual and grapho-phonemic cues they retain in the story structure. Adaptations on Disruptions 8 and 21 appear low, indicating a general lack of adaptation evident in the readers' performance. Post-hoc examination of these two disruptions and the participants' performance on them points to Disruption 8 not being a salient enough mistake to fix, while Disruption 21 may simply have been too difficult or puzzling to resolve. No other mean adaptation scores were noted using these guidelines. All individual disruption adaptation scores are related to the total adaptation score except for scores on Disruptions 14, 18 and 22, which are only marginally significant. These three disruptions and perhaps Disruptions 4 and 19 do not appear to discriminate between high and low total adaptations because participants' adaptations to them were generally consistent across readers.
Overall, the children demonstrated enthusiasm and interest in performing these ecologically valid reading tasks, a significant bonus to classroom engagement in reading comprehension formative evaluation. The range of reading performances studied here in miscues and adaptations shows that young readers are not only sensitive to text characteristics, they also engage in a wide range of levels of processing aimed at making use of a variety of textual features, as evident in the ranges of scores reported in Tables 1 and 2, and as might be expected by many reading researchers, such as Smith (2005) and Y. Goodman et al. (2005). Furthermore, as our inferential analyses showed, less skilled readers' syntactic processing was directed at word recognition, whereas the more skilled readers built text meaning based on syntactic processing, much as Perfetti et al. (2005) and Stanovich (2003) hold. Further, participants provided evidence of variations in flexibility in dealing with the changing demands between reading tasks reported here, suggesting that these manipulations could effectively serve useful formative diagnostic classroom functions. Inverse relationships were established between miscues and adaptations, as was predicted in Hypotheses 1, and, as stated in Hypothesis 2, higher quality adaptations were made by children who produced the least invasive macrostructural miscues.
Contributions of Miscue Scores
Readers' miscue data were rated in terms of relative meaning change. When more objective scoring of meaning change was achieved in the present study by applying Faigley and Witte's (1981) distinctions between surface changes and text-based changes in writing revision, high scorer agreement was reached. The achievement not only adds to the utility of RMI meaning change classifications, modifying the meaning change taxonomy, but also indexes each child's reading of intact text and allows for a more detailed evaluation of miscues while increasing the reliability of their meaning change scores. Appendix B provides examples of the types of miscues for each category scored.
Distinguishing between surface changes that do not add or remove information contained in a summary of a text and text-based changes that do modify a given text summary was refined here for reading miscues that were merely formal (miscues that reflected little meaning change) versus those that were meaning preserving (e.g., paraphrases). Rating a miscue as microstructural was meant to reflect a change in meaning that was close to altering the text summary at a more local or sentence level, whereas a macrostructural miscue was considered a greater change to the overall story. Thirteen percent of the sampled miscues were rated as macrostructural on average, 27% were rated as microstructural, 25% were rated as formal miscues, 30% were meaning-preserving, and 5% of the miscues were rated as involving no change in meaning, on average. These results show that, overall, this study's beginning readers were generally able to maintain the story meaning despite sometimes diverging from what was written.
Contributions of Adaptation Scores
As for adaptations, we found that scores on substitutions were different from scores on insertions and omissions, and this may be because substitutions more or less retained some visual and phonetic similarity to the original text item they replaced (see Appendix A). It could have been assumed that they would provide more cues for higher degrees of adaptation than insertions, which themselves contained all of the original text item information accompanied by superfluous linguistic elements. Omissions retained the least of the original text information and provided none of the visual cues of disruption usually associated with cloze tests, for example. Cloze protocols cue readers to missing words with blank spaces to be filled; these blank spaces are usually underlined as another cue to some text information missing. Such spacing and underlining was not available in the disrupted text our participants read. Although no aspect of the three-way relation described above was predicted, the foregoing description of the features of disruption types offers an interpretation of the difference found between substitution disruptions and both of the other types in exploratory analyses. Adapting these disruptions required context-level and syntactic processing to make up for the meaning disruptions in the second part of the story, accommodating Stanovich's (1980, 2003) view of interactive-compensatory reading mechanisms.
Further exploration of adaptations to individual disruptions found high adaptations on most substitutions and structural disruptions, and low adaptations on extensive insertions and omissions. Interestingly, high adaptations on extensive substitutions involving story character names suggest that these disruptions may not have been as representative of that level of disruption as extensive insertions and omissions were. For example, substituting Gordon, Harold, and Betty for William, Stanley, and Lucy might have been more appropriate.
Overall, lower adaptation scores were obtained on omissions. The combined effect of a larger number of omissions in one text portion may have been greater for less-skilled readers than it was for more-skilled readers, as Perfetti et al.'s (2005) model of reading development would hold. Also, it may be that both of the texts used were too short to demonstrate other between reader patterns.
Relations Between Intact and Disrupted Text Performances
Although meaning construction is a multidimensional phenomenon, the performances displayed by these participants indicate that the extent to which they stayed "on track" with the intact text' s meaning is strongly related to the degree to which they brought the meaning of the disrupted text back on track. The strong negative relationship between these young readers' meaning-change-miscue and disruption-adaptation scores implicates the validity of both as indices of performance in meaning construction in text comprehension. The underlying psychological processes in reading performance as reflected in meaning change scores and the processing of disrupted text as reflected in adaptation scores are presumably somewhat general ones, as the data suggest that they reflect meaning construction in reading comprehension.
The negative relationship between meaning change and adaptation to disruptions supports the use of the former classification by providing evidence that the overall degree of meaning change evidenced in a child's oral reading is linked to the overall degree of adaptation on disruptions. This, in turn, is based on the child's responses to textual features that appear to represent degrees of meaning disruption. Post-hoc analyses confirmed that, overall, adaptations differed following the class of disruptions intended to represent levels of change in text meaning. Adaptation data lent credence to the psychological relevance of meaning change scores, for it is evident in the performance of these children that disruptions labeled "extensive," representing more severe meaning disruption, elicited less adaptation than "minimal" disruptions, which in turn elicited less adaptation than "structural" disruptions intended to represent the most superficial level of damage to text meaning. This supports use of the Faigley and Witte (1981) inspired meaning change classification in longitudinal investigations focused on emergent writing and reading skill development where a whole language approach is favored (as in Cameron & Hutchison, 2009; Dyson, 2003; Wells, 2009).
Because number of macrostructural miscues related negatively to adaptations on extensive disruptions, we can say that children who produced few macrostructure miscues were highly adaptive of extensive disruptions. Also, children who produced few microstructure miscues adapted effectively to minimal disruptions, as number of microstructural miscues related to adaptations on minimal disruptions. Clarifying these relations requires further investigation of the meaning change classification's scaling. This could also help to elucidate why the number of formal miscues produced was not related to adaptation on structural disruptions, raising the possibility that they may both be relatively constant across readers.
The absence of detectable intact text trends for high and low meaning change readers might be due to the small miscue sample size. On the other hand, the large variation in amount of text covered by miscue samples between readers may have reduced the likelihood of establishing differences. The absence of difference trends in intact text processing is inconsistent with the notion that high meaning change participants' reading would be characterized by impeded text meaning construction, which in turn would result in an increasingly error-full and meaning-disrupted reading. Although it may be that even high meaning change readers were able to maintain a level of comprehension construction that was consistent throughout their reading of intact text, this hypothesis may need to be revisited in further research on the meaning change metric.
The children's adaptations to textual disruptions, representing as they do their responses to explicit instructions to "fix mistakes" and to "make sense" of a disrupted text "as best they could," can be viewed as instances in which various levels of text processing would have been called into play for adaptations to be evident. These two indices may actually represent different component processes that vary across task demands and between readers, but within readers there was a strong correspondence between the scored oral reading performances. These two indicators of meaning construction may represent degrees of phonetic, syntactic, and semantic processing, as well as levels of process monitoring. The meaning construction classification system used here appears reliable and useful. Having repeated measures of a child's meaning change scores from different oral reading samples and relating them to the formative data acquired by encouraging them to fix mistakes in the representatively disrupted texts of another could provide insight into the pattern of evolving comprehension for that child. This would allow patterns of strengths and weaknesses in comprehension to be recorded and used for remedial work with that child. This study's two indices of meaning construction in reading also could be used in literacy studies in the manner described in Cameron, Hunt, and Linton (1988, 1996).
The two reading tasks diverged in what they required of readers. At no time during their performance was any mention made to the children of a reading rate to be achieved in either task. Overall, intact text reading rates were higher than disrupted text rates. And, overall, the children did spend more time on disrupted text lines than they did on intact lines. What this actually means is made clearer by the differential reading rates by texts. For children who scored low on meaning change, the disrupted text had a more marked effect than it did for children who scored high on meaning change, another source of evidence in support of meaning change as an index of sensitivity to text meaning, concern for text meaning or ability to make sense of text. As they read the disrupted text, low meaning change readers (i.e., more skilled readers) adjusted their rate of text processing to accommodate adaptation challenges and obtained higher adaptation scores for doing so. Making the text meaningful, through the execution of phonological, semantic, and syntactic processing skills, allowed low meaning change readers effectively to shift strategies in response to the somewhat different task, a finding consistent with reports of Cameron et al. (1996), Perfetti et al. (2005), and Stanovich (2003). All children produced miscues while reading the intact text and none of them adapted the disrupted text in such a manner as to reflect completely and accurately the last quarter of the original version of the story. Finally, all readers showed some evidence of adaptation.
Contributions to Classroom Practice
It can be seen from the performance of these children that a broad range of reading skills and reading strategies were put into effect. Using such relatively uninvasive and straightforward techniques in classroom practices as miscue and adaptation coding of individual students' oral reading could help teachers identify challenges each student is facing in constructing meaning from any classroom text. Information from these assessments could assist in targeting instructional alternatives and this, in turn, could help young readers of varying abilities better understand how to monitor their reading by focusing on seeking meaning, self-correcting miscues, and restoring meaning to texts that contain errors, whether in their own writing or someone else's. Discussions about reading and/or writing errors could add to these formative experiences, facilitated by teachers and researchers adopting a dialogic stance in the classroom (Dyson, 2003; Wells, 2009), with individual students and with a class as a whole.
The significance of meaning change for adaptation lies in the implication that children's concern for meaning comprehension in reading, their ability to stay with the sense of the story's meaning, and their ability to monitor their reading all may be instrumental in the extent of understanding they attain. Because these processes are assumed to be involved in maintaining a text's meaning despite the production of miscues, they also could contribute to restoring meaning to texts when meaning is unclear or disrupted.
More needs to be known about the internal structure of this classification in order that a model of the scoring system be developed and tested. Therefore, further investigation of the psychometric properties of the meaning change scoring system is the next logical step in the exploration of this tool's usefulness. Once these properties are known, reasonable predictions could be made between it and other classifications of meaning construction. An investigation of the patterns of association between meaning change scores, adaptation scores, and writing and editing performance is of special interest and a fruitful line of research in emergent literacy.
APPENDIX A DISRUPTIONS AND ADAPTATIONS
The disruptions inserted in the last quarter of the text used William and Stanley and the (1) Lucy went down the stilt (2).
"This -- (3) where OUR family lives, stnlaey (4)" Lucy said. "We live here with our mother and thafer (5). And the cat that climb -- (6)
up in your oak tree was MY CATS (7).
You and the van (8) cat have to be frinz (9)."
"Stanley -- (10) not know about talk," William say- (11).
"How do you know?" Lucy asked.
William didn't -- (12) how to answer.
He -- (13) about this for a long temi (14).
The (15) Stanley was having a good times (16).
Every now and then he ran away, looking in door (17) people's gardens. But each time, he came fight back and walk -- (18) beside wllmiia (19).
William -- (20) feeling better now, but he still -- (21) that a boy had moved next doors (22).
"This is a nice dog," the (23) William said. "It's almost as good as having a basket (24) boy next door."
"It's not atmosl (25) as good as having a boe (26) next door, lcuy (27) said, "It's better."
Note: None of the disruptions were underlined or numbered in the text the children read, nor were there any spaces allotted to cue omissions.
Adaptation exemplars for each type and level of disruption, with scores in parentheses Disruption Adaptation exemplars "the William William, Wilm, Will, Willy (all scored 4); "the woman (3); a William (2); sounds out "th-e Wi-ll-i-am," then omits William (1); the William (0) stnlaey dog, Stanley, Stan, doggie (4); Staniel, Stanel, William, Willy (3); Steenle-ay, st-n-le (2); s-t- n-a-y, stn, omit whole word (1); stnlaey (0) Thafer dad, father, papa (4); fathers, feather (3); tafey, thafer, after, thafey, thafa (2); thaf, aftha, th- a-f-er, omit word (1); thafer, thayfer (0) climb(ed) was climbing, clambered, went, ran, climbed (4); climbs, will be climbing (3); run, go, claim (2); clim, cl-, c-1-i-m-b, omit whole word (1); climb (0) CAT"S cat, kitten, kitty (4); animal, pet (3); kittens, kitties, pets, animals (2); cates, cuts, ca-, c- a-t, omit whole word (1); Cats (0) "van cat fun cat, vain cat, cat (4); cat fan, vandal cat (3); truck cat, van's cat (2); van, vank, v-van or c-a-t alone (1); van cat (0) frinz friendly, friends, nice to each other (4); freed, French, fringe (3); freenz, frines, (2); fra, fr-, f-r-i-n-z (1); frinz (0) (does) did not know, will not know, knows not know not (4); has not know, don't know, knows, knew, is knowing, never know (3); Know not, never know, nothing know (2); no know, know, kanow, not kanow, not now, n-o-t, kn-o-w, omit both (1); not know (0) (thought) was thinking, thinked (4); felt, considered, dreamed, knew (3); Aaah, hmmm, pause (2); linking "he" and "about" as "hebout" for `He (thought) about ..." (1); no pause, no insertion (0)
APPENDIX B EXEMPLARS FOR EACH MEANING CHANGE CATEGORY
1. Meaning-preserving miscues
--omitting or inserting a semantically and syntactically correct word in a sentence (e.g., "He hoped ^that there would be a boy in the family. (Up) she went, higher-still higher.")
--substituting a word with an acceptable synonym (e.g., reading auto for car, home for house)
--reversing the order of said and a character name or referent
--syllabication followed by a whole rendering of a word
--pauses between words if such pauses do not precede or follow another miscue, do not occur at the ends of lines or sentences, and do not contain the intonation of questions or exclamations (e.g., "He hoped there / would be ...")
--substituting a character name with the same name (or nonword) produced on the previous encounter with the character name (e.g., always reading Lucy as Lucky)
--simple repetition of an accurately read text item
--omitting Woof Woof
2. Formal miscues:
--slight grammatical changes like omitting or inserting verb endings, or part of a compound verb in a subordinate clause (e.g., reading climb for climbed, move for moves, hope for hoped, etc.)
--inflectional changes (adding or omitting s, es, ed, ing, er, est) to adjectives, adverbs, and object nouns; also, omitting possessive inflections
--pluralizing (or generalizing) a singular form (e.g., they for he, "The van were (was) coming")
--omitting an article, conjunction or preposition
--omitting or inserting punctuation (e.g., reading Stanley is a DOG? without interrogative intonation)
--omitting adjective and adverb modifiers (e.g., "A (great) big dog," "Stanley was (very) happy")
--substituting a noun with another which is a "loose synonym" (e.g., bag for basket)
--syllabication of a word if it is not followed by a rendering of the whole word
3. Microstructure miscues:
--extensive grammatical changes that reduce main clause subjects and verbs to their semantic root (e.g., reading the van come for the van was coming)
--substituting article, conjunction, or preposition with contradictory function words
--a first miscue of a character name, substituting it with an existing word that does not usually stand as a character name (Still for Stanley), or a second miscue of a character name, substituting it with a real or potential (nonword) name different fro m the first miscue. Neither of the miscues are another story character's name
--omitting a surname, but reading the title of address (e.g., "Mrs. (Bradley)")
--singularizing a plural form
--omitting an adjective, an adverb, or an object noun
--changes to subordinate clauses which involve the omission of whole subjects and or verbs
--the, this, that, for a
--misarticulations (no more than one phoneme reversal, substitution, or insertion) that result in nonwords (e.g., woo-man for woman, pleople for people)
--intonation shifts internal to the word that result in other known word(s) (e.g., high-her for higher, eve-wreathing for everything, fall-low for follow)
4. Macrostructure miscues:
--major content changes, nonwords, and explicit contradictions to subjects and verbs of main clauses
--generally any change that is not restricted to the sentence level, or a change that is contextually appropriate at the sentence, but changes the text summary (e.g., reading Help, she said ... for Hello, she said ... reading Lucy's cat and William were all happy for Lucy's cat and William were not happy at all)
--a first or a second miscue of a character name, by substituting it with another of the story's character names or, a first or a second miscue of a character name which is a complete omission of the name
--substituting or inserting non synonym adjectives, adverbs and object nouns which explicitly contradict text information
--substituting a specific referent with another specific referent referring to someone or something else (e.g., she for he, woman for William, Stanley for Bradley)
--omitting a whole verb, a pronoun or a noun
--any nonword substitution (on any text item) which does not retain the basic sound of the text item, and thus cannot count as a misarticulation (more than one phoneme reversal, substitution, or insertion) (e.g., fine-t-sed for finished, wullda for would" famy for family
--omitting the text title
Note. The following symbols, () and ^, represent omission of the word in parentheses and insertion of the word following the arrow respectively.
We thank the participants for their enthusiasm for and commitment to demonstrating their skills at making meaning from their early reading, and their dedicated teacher, Anne Kathryn Hunt, without whom this research would not have been conducted.
Brandt, D. (2009). Literacy and learning: Reflections on writing, reading, and society. San Francisco, CA: Jossey-Bass.
Bronfenbrenner, U. (1979). The ecology of human development." Experiments by nature and design. Cambridge, MA: Harvard University Press.
Cameron, C. A., Hunt, A. K., & Linton, M. J. (1988). Medium effects on children's story rewriting and story retelling. First Language, 8, 3-18.
Cameron, C. A., Hunt, A. K., & Linton, M. J. (1996). Written expression as recontextualization: Children write in social time. Educational Psychology Review, 8, 125-150.
Cameron, C. A., & Hutchison, J. (2009). Telephone-mediated communication effects on young children's oral and written narratives. First Language, 29(4), 343-367.
Cameron, C. A., & Wang, M. (1999). Frog, where are you? Children's narrative expression over the telephone. Discourse Processes, 28, 217-236.
Clymer, T., & Martin, P. M. (1978). The dog next door and other stories. Toronto, Canada: Ginn & Company.
Conway, M. A. (Ed.). (2002). Levels of processing 30 years on. Brighton, England: Psychological Press.
Daniel, W. W. (1978). Applied nonparametric statistics. Boston, MA: Houghton Mifflin.
Dyson, A. H. (2003). The brothers and sisters learn to write: Popular literacies in childhood and school cultures. New York, NY: Teachers College Press.
Faigley, L., & Witte, S. (1981). Analyzing revision. College Composition and Communication, 32, 400-414.
Farstrup, A. E., & Samuels, S. J. (2002). What research has to say about reading instruction (3rd ed.) Newark, DE: International Reading Association.
Goodman, K. S. (1985). Reading: A psycholinguistic guessing game. In H. Singer & R. B. Ruddell (Eds.), Theoretical models and processes of reading (3rd ed., pp. 813-840). Newark, DE: International Reading Association. (Original work published 1984)
Goodman, Y., Watson, D., & Burke, C. L. (2005). Reading miscue inventory: From evaluation to instruction (2nd ed.). Katonah, NY: Richard C. Owen Publishers.
Goodman, Y. M., & Burke, C. L. (1972). Reading miscue inventory. New York, NY: Macmillan.
Olson, D. R., & Torrance, N. (2009). The Cambridge handbook of literacy. New York, NY: Cambridge University Press.
Page, E. B. (1963). Ordered hypotheses for multiple treatments: A significance test for linear ranks. Journal of the American Statistical Association, 58, 216-230. Retrieved from www.jstor.org/stable/2282965
Paris, S. G. (2005). Reinterpreting the development of reading skills. Reading Research Quarterly, 40(2), 184-202. doi: 10.1598/RRQ.40.2.3
Perfetti, C. A., Landi, N., & Oakhill, J. (2005). The acquisition of reading comprehension skill. In M. J. Snowling & C. Hulme (Eds.), The science of reading: A handbook (pp. 227-253). Oxford, England: Blackwell.
Sameroff, A. (2010). A unified theory of development: A dialectic integration of nature and nurture. Child Development, 81(1), 6-22. doi:0009-3920/2010/8101-0002
Samuels, S. J. (1987). Factors that influence listening and reading comprehension. In R. Horowitz & S. J. Samuels (Eds.), Comprehending oral and written language (pp. 295-325). San Diego, CA: Academic Press.
Smith, E (2005). Reading without nonsense (4th ed.). New York, NY: Teachers' College Press.
Society for Research in Child Development. (2007). Ethical standards for research with children. Retrieved from www. srcd.org/index2.php?option=com_content&task=view&id=68&pop=1&page=0&Itemid=499
Spache, G. D. (1966). Good reading for poor readers. Champaign, IL: Gerrard.
Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading fluency. Reading Research Quarterly, 16, 32-71. Retrieved from http://www.jstor.org/stable/747348
Stanovich, K. E. (2003). Understanding the styles of science in the study of reading. Scientific Studies of Reading, 7(2), 105-126. doi: 105-126.10.1207/S1532799XSSR0702_1
van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. New York, NY: Academic Press.
Wells, G. (2009). The meaning makers: Learning to talk and talking to learn (2nd ed.). Bristol, England: Multilingual Matters.
Paul David Leger
University of New Brunswick, Fredericton, Canada
Catherine Ann Cameron
University of British Columbia, Vancouver, Canada
Submitted December 21, 2011; accepted May 31, 2012.
This research was supported by a grant from the Social Sciences and Humanities Research Council of Canada (410-2001-95).
Address correspondence to Catherine Ann Cameron, UBC Psychology Department, 2136 West Mall, Vancouver, BC V6T 1Z4, Canada. E-mail: firstname.lastname@example.org
TABLE 1 Descriptive Statistics for Reading Scores on Intact Text Variable M SD Median Minimum Maximum Meaning change 52.94 11.12 54.00 15.00 75.00 Number of miscues 39.04 17.69 37.00 11.00 99.00 Lines read 60.69 18.99 72.00 8.00 73.00 Reading time (a) 6.13 2.17 5.87 2.88 12.02 Reading rate (b) 11.73 6.12 11.61 1.03 24.97 Number of miscues rated as: Macrostructure 3.14 2.84 3.00 0.00 11.00 Microstructure 6.74 2.64 7.00 1.00 12.00 Formal 6.31 2.22 6.00 1.00 13.00 Meaning-preserving 7.55 2.77 7.00 2.00 13.00 No change 1.27 2.56 0.00 0.00 13.00 (a) ln minutes. (b) ln lines per minute. TABLE 2 Descriptive Statistics for Reading Scores on Disrupted Text Variable M SD Median Minimum Maximum Adaptation 79.51 17.43 80.00 33.00 103.00 Reading time (a) 4.46 2.42 4.08 1.22 14.50 Reading rate (b) 6.17 3.04 6.17 1.52 18.08 Adaptation by disruption levels: Structural 30.00 5.89 32.00 9.00 36.00 Minimal 26.43 7.70 29.00 9.00 36.00 Extensive 23.08 6.39 22.00 12.00 36.00 Adaptation by disruption types: Substitutions 33.30 5.06 33.00 9.00 36.00 Insertions 26.00 6.75 26.00 11.00 36.00 Omissions 23.00 7.56 23.00 8.00 35.00 (a) ln minutes. (b) In lines per minute. TABLE 3 Mean Adaptation Scores for Each Disruption Disruption type Substitution Insertion Omission Disruption Level No. M No. M No. M Structural 2 3.63 7 3.35 6 3.27 9 3.31 16 3.69 11 2.98 26 3.14 22 3.76 18 2.88 Minimal 5 3.37 1 3.04 10 2.33 14 3.63 15 2.61 12 2.80 25 3.35 23 3.00 20 2.31 Extensive 4 3.51 24 2.00 3 2.47 19 3.63 8 1.76 13 1.94 27 3.90 17 2.08 21 1.80 Note. No. = disruption number.
|Printer friendly Cite/link Email Feedback|
|Author:||Leger, Paul David; Cameron, Catherine Ann|
|Publication:||Journal of Research in Childhood Education|
|Date:||Oct 1, 2013|
|Previous Article:||Early language and literacy achievement of early reading first students in kindergarten and 1st grade in the United States.|
|Next Article:||Requirements of high-quality kindergarten programs according to Jordanian parents: an exploratory study.|