TheFour-DimensionalModel: Interaction ofSchematain the Process ofFL Reading Comprehension AdinaLevineand Thea Reves This paper reports on a studywhich ofeach ofthe three typesofschemata? investigated the interactiveeffectofdif What is the influence of schemata ferent types of schemata on successful inducing activities on FL reading text comprehension in a Foreign Lan comprehension? guage. The tentative model of the The results ofthe studyconfirm that study is four-dimensional: the di all the three types of schemata mension of overall reading compre contribute to overall text compre hensionwasassumedtobetheresult of hension, both separately and cumula an interaction oflinguistic, content and tively, although to differing extents. It formalschemata. can also be assumed that schemata The following research questions inducing activities may indirectly affect were specifically asked: What is the overall text comprehension by arousing combined effect of the three types of thereader'sappropriateschemata. schemata? Whatisthe relative weight The introduction of the schema-theoretical approach to the analysis of the reading process has had a profound impact on the study ofreading comprehension. According to the advocates ofthis theory, reading is an interactive process in which the writer's perspective, ideas, development of arguments, intentions and conclusions are all interpreted through the reader's experiences, cultural background and biases. Well-organized background knowledge, i.e., a set of schemata, has a facilitating effect: the reader develops appropriate expectations integrating this knowledge with information derived from the text. (Adams & Collins, 1979; Anderson & Pearson, 1984); Rummelhart, 1980; Stanovich, 1980). Three areas of schema have basically been acknowledged as playing a part in the act of reading: linguistic schemata (The skills ofdecoding and discourse processing), content schemata (knowledge of the content area of the text) and formal schemata (recognizing the rhetorical structure of the text) (Carell, 1983, 1984, 1985, 1991; Carell & Eisterhold, 1983). The question arises then, to what extent the schema-theory of reading is appropriate also for Second Language (L2) and Foreign Language (FL) reading research or, in other words, to what extent the unique problems inherent in L2 and FL reading should be specifically addressed. Some research studies indeed suggest that reading in L2 and FL involves the same processes as reading Ll. TESLCANADAJOURNAI)REVUE TESLDUCANADA VOL. 11,NO.2, SPRING1994 71 Thus, Hudson (1982) examined Clarke's "short circuit" theory (Clarke, 1980), according to which "a language ceiling in L2 effectively prohibits the complete transfer of L1 reading skills to L2 reading". Hudson's research suggested that the readers' initial schemata override, to a great extent, their linguistic limitations. Carrell and Eisterhold (1983) drew similar conclusions, i.e., that only a few reading problems are related to the deficiencies in language skills. They claimed that what is crucial for reading comprehension is the reader's ability to apply the already existing background knowledge to the information derived from the text. A number of scholars, on the other hand, have maintained that, while for a fluent L1 reader the possession of appropriate language skills can be taken for granted, an L2 or FL reader's inadequate language proficiency may prevent the application of the reader's schemata to the reading of the text. Thus, Eskey emphasized the role of identification skills in L2 reading, i.e., the importance of "the straightforward recognition of the lexical units and the grammatical signals required for the simple decoding of the text" (1983, p. 36). A similarview on the role oflanguage proficiency in L2 reading was expressed by Grabe, who stated that L2 readers must reach a stage of automatic text processing in order to possess, together with relevant content and formal schemata, "the critical mass of knowledge which would enable them to stop learning to read and only read to learn" (1988, p. 36). So far L2 and FL reading researchers have examined either the effect of only one type of schemata or the combined effect of two types (Barnitz, 1986; Carrell, 1984; 1987; Johnson, 1982). Carrell (1984) examined the effect of rhetorical organization on reading, while holding the effect of content schemata constant. Two such studies in L2 reading were conducted: one with narrative texts and one with expository texts. The studies enabled the author to make a number of suggestions regarding text organization and the effect of the rhetorical pattern on text processing. It was found that tightly organized comparison, causation and problem/solution types of organization tend to aid the recall of text ideas. In her 1987 study, Carrell addressed the combined effect of both content and formal schemata in L2 reading and showed the "each component-content and form-plays a significant but different role in the comprehension of a text". The study was the first of its kind and the author concluded that "further research of the combined effects ofcontent and form in ESL reading is clearly needed". It should, however, be emphasized that although L2 reading has shown that language proficiency has a significant impact on 72 ADINALEVINEflHEAREVES comprehension (Alderson, 1984; Clarke, 1980; Cziko, 1980; Devine, 1987; Mcleod & McLaughlin, 1986), the schematic effect of linguistic skills in their combination with content and formal schemata on the process of FL reading has not been investigated. Thus the central issue in our present research is the interactive effect of all the three types of schemata (linguistic, content and formal) on successful text comprehension in the FL. Another aspect examined in the present study was the extent to which FL reading is affected when certain types of schemata are induced at the pre-reading stage. Empirical studies of Ll and FL reading have both supported and disclaimed the facilitating effect of induced schemata on reading comprehension. While Hudson (1982) and carrell (1984) found that the pre-teaChing of relevant knowledge and the activating of appropriate schemata were beneficial, Johnson (1981) reported that the pre-teaChing of target vocabulary, for example, did not significantly affect successful text processing. In fact, research on the effectiveness of pre-reading activities has shown that not all such activities are equally helpful. Thus, Hudson (1982) found that the pre-teaching ofvocabulary was more effective for advanced students than for those of the intermediate or beginners' level. Lee (1986) reported that providing a title and a picture page enhanced comprehension only when the text was familiar. Since the purpose ofour studywas to examine the role played by each ofthe three schemata, it was important to investigate ifand to what extent readers would utilize different types of induced schemata and how the inducing ofspecific types of schemata would contribute to text comprehension. The tentative model on which the study is based is four-dimensional: the dimension of overall reading comprehension was assumed to be the result of an interaction to the three dimensions of linguistic schemata (skills of decoding and discourse processing); content schemata (knowledge related to the content domain of the text); and formal schemata (familiarity with the rhetorical patterns in which information is presented). FIGURE 1 I -H I Linguistic OverallReading Content Schemata Comprehension Schemata '1 Formal Schemata TESLCANADAJOURNAUREVUE TESLDUCANADA VOL. 11,NO.2,SPRING1994 73 The following three research questions were specifically asked: (1) What is the combined effect of the three types of schemata on FL reading comprehension of academic texts of different rhetorical structures? (2) What is the relative weight of each of the three schemata in the interactive process of FL reading? and (3) What is the influence of induced schemata related activities on FL reading comprehension? PROCEDURE Ninety-five first-year university students enrolled in advanced EFL courses were the subjects of the study. During the course of instruction the subjects were tested in reading comprehension at regular intervals. Altogether nine tests were administered; the text organization in each test was descriptive (illustrative), analytical or argumentative-presented in random order to avoid any biasing effects. The texts were all evaluated by independent raters in order to ensure the same level of text difficulty. The raters based their text-evaluation on the criteria established by the previous research on the development of the criterion-referenced test (Reves & Levine, 1992), namely, subject-matter, cultural reference, various aspects of vocabulary, syntactic complexity, and implications requiring complex inferencing. In order to examine the effect to the three types ofschemata and the relative contribution of each of the three schemata to text comprehension, the test questions were specifically geared to the reader's application of each of the schemata in the process of reading comprehension. Thus, two questions referred to the structural features of the text (they were meant to tap the reader's awareness of text organization and its effect on comprehension, i.e., formal schemata). Two questions referred to the linguistic aspects of the text (they were meant to tap the reader's application of decoding and discourse skills, i.e., linguistic SChemata). Two questions were geared to the content domain of the text (they were meant to tap the reader's text-oriented and background knowledge, i.e., content SChemata). Each of the six questions was followed by the task of justifying the given answers. These justifications were coded by independent raters according to the specific schemata type. The scores on the two questions on each of the schemata were combined into a single score (Cronbach-alpha formula was used to check their internal consistency). The seventh item on the test was the task of writing a summary of the whole text with reference to the main idea(s), supporting idea(s), author's intention and 74 ADINALEVINEfIHEAREVES conclusions. The validity of this type of summary as a reliable measure of the reader's overall mastery understanding of the text was suggested by the findings of the previous research (Reves & Levine, 1992). The score on this item was considered the score of overall reading comprehension (OS). To capture full comprehension of the text and to avoid the stumbling block of the FL written expression, all the answers to the questions, including the summary, were given in the reader's L1. All the tests were piloted on a group ofseven high-school teachers ofEFL. In order to test to what extent schemata induced in the pre reading stage could influence successful text comprehension (see Research Question 3), the Treatment- and Control-group design was used. In an attempt to arouse schemata, the Treatment-group was given pre-reading, schemata-activating materials, while the Control-group answered the test-questions without any preliminary sChemata-inducing activities. The schemata-inducing activities were of three types: For linguistic schemata, we used pre-test questions related to decoding, discourse, and vocabulary items in minimal context; for content schemata, we used independent association questions referring to the content area of the text to be read; and for formal schemata, we used the pre-reading of a text in L1, the rhetorical structure (text organization) of which was the same as that ofthe FL (English) text to be read. The schema-inducing activities, as well as the texts of different text-organization, were presented in random order, to avoid a biasing effect of the same rhetorical structure always receiving the same induced schema-treatment. Information on the reader's text processing activities as well as on individual variations in readers' schemata was gathered from two sources: (1) Readers' immediate retrospection reports, referring to each question on all the tests (A sample group of subjects were asked to justify their answers, referring either to the text or to any source they Chose.); and (2) Interviews conducted with randomly chosen subjects immediately after the completion of the test; they were meant to capture information on whether and how the three types of schemata were applied to the reading ofthe text. These two sources were intended to provide specific information on each reader's individual implementation of schemata as well as to shed some light on the issue ofvariation in readers' schemata. lESLCANADAJOURNAUREVUETESLDUCANADA VOL. 11,NO.2,SPRING1994 75 DATAANALYSIS The questions on all the tests as well as the questions which the Treatment-group had to answer as part of the pre-reading schemata activating treatment were rated by two sets of raters (inter-rater reliability: .91). The transcripts of the interviews were categorized and analysed by independent raters. To find out to what extent the three schemata contributed to overall text comprehension, regression analysis on the scores obtained on questions related to specific schemata was preformed. To examine the interaction among the three types of schemata, regression and correlational analyses among the combined scores on each type ofschemata were done. In order to find out if the reader's application of schemata was similar or different in the processing of texts of different rhetorical structures, correlational and regression analyses were carried out. To see to what extent induced schemata in the pre reading phase could or could not render a more effective text comprehension, the test results of the Treatment and of the Control-group were correlated. To identify the readers' actual application of the schemata and their metacognition of them, the retrospection reports and interview questions related to specific schemata were analysed. All the results presented below are statisticallysignificant, at a .0001 or .001 level. RESULTSAND DISCUSSION Text ComprehensionandSchemata Regarding the question of the contribution of each of the three types of schemata to successful text comprehension the following findings can be reported. The Stepwise Regression Analysis, carried out on tests on all texts, ignoring the subjects' justification of their answers, showed that overall text comprehension was best predicted by content schemata (45%). The second best predictor was the linguistic schemata (27%), while the formal schemata predicted overall text comprehension onlyvery weakly (2%). These findings are partly supported by correlational analysis: the highest correlation observed is between the mean scores on questions geared to content schemata and the mean overall score (OS) (r=.67). The correlation between the mean scores on linguistic schemata geared questions and the mean of the overall score is slightly lower: 7=.54. The correlation between mean scores on formal schemata geared questions and the mean of the overall score is even lower (.26). 76 ADINALEVINE{TI-lEAREVES TABLE 1 StepwiseRegression Analysis ofPredictionofOverallText ComprehensionbySchemata (thedatawereanalysedwith noreferencetospecifictextorganization) Predictor Partial ProblF Variables B-value R**2 significant Content schemata 1.32 0.45 0.0001 Linguistic schemata 0.51 0.27 0.001 Formal schemata 0.39 0.02 0.001 TABLE2 Correlations BetweenScores on Schemata-geared Questions and Overall Score Content Linguistic Formal schemata schemata schemata OverallScore .67 .54 .26 p=O.OOI p=0.003 p=0.03 Regression analysis indicates that the scores on the questions related to specific schemata mutually predict one another (all around R**2=.36). This interaction among the three types of schemata is also supported by high correlations among the scores on the three types of specific schemata-geared questions (all around r=.52). RhetoricalStructureandSchemata To find out whether readers' application of schemata is affected by the rhetorical structure of the text, separate analysis was carried out with reference to each of the three types of text organization. In this analysis both the answers to the questions and their justificationwere considered. The analysis shows that in tests ondescriptive texts the overall text comprehension is best predicted by the linguistic schemata (50%). The content schemata is the second best predictor (19%), while the prediction of text comprehension by formal schemata is not of statistical significance. Likewise, in tests on analytical texts both TESLCANADAJOURNAUREVUE rESLDUCANADA VOL 11,NO.2,SPRING1994 77 linguistic and content schemata predict overall text comprehension, with linguistic schemata being here, too, a slightly better predictor (Li: 24%; Co: 18%). The prediction by formal schemata is not statistically significant in this case either. When the rhetorical structure of the texts was argumentative, however, the prediction of text comprehension by the formal schemata is the strongest (52%). In this case, the second best predictor is the linguistic schemata (21%), while the content schemata did not significantly predict overall text comprehension. TABLE3 StepwiseRegressionAnalysis ofPredictionofScores on OverallText ComprehensionbySchemata (thedatawereanalysedwithreferencetospecifictextorganization) Predictor Partial ProblF Text Variables B-value R"2 significant Linguis-tic schemata 1.00 0.50 0.0001 Descriptive Content schemata 0.97 0.19 0.0001 Content schemata 0.55 0.18 0.0001 Analytical Linguistic schemata 0.59 0.24 0.001 Formal schemata 0.98 0.52 0.0001 Argumentative Linguistic schemata 1.00 0.21 0.001 The difference between the results of data-analysis, when specific text organization was not considered (Tables 1 and 2) and when it was (Tables 3 and 4), can be accounted for in three ways: only in the second case was the subjects' justification of the answers included in the evaluation; the subjects made extensive references to linguistic markers in justifying their answers, thus adding weight to the contribution of linguistic schemata; and the raters coded some of the justifications as being of linguistic character, even when the questions themselves weregeared to the content area ofthe text. 78 ADINALEVINE/fHEAREVES TABLE4 ContributionofSchemata Related to DifferentText-organization Descriptive Analytical Argumentative texts texts texts 8 8 a content formal linguistic schemata schemata schemata descriptive linguistic formal schemata 50 schemata schemata 52 One ofthe surprising findings is the increase in the weight of the linguistic schemata and the decrease in that of the content schemata, when the data were analysed with regard to texts of different rhetorical structure. A tentative explanation which can be offered here is that the content schemata-geared questions on the tests may have been too closely text-bound, rather than being geared to a more general content domain. Answering these questions probably required close text-decoding, which by itself is of linguistic character. Correlations between the mean of the overall scores on each of the three sets of tests (i.e., on descriptive, analytical and argumentative texts) on the one hand, and scores on schemata geared questions on the other, show that in tests on descriptive texts the scores on linguistic schemata-geared questions correlate significantly with the overall score (r=.34). In tests on analytical and argumentative texts, scores on questions geared to all three types ofschemata yield high correlations with the score on overall reading comprehension (all ranging from .43 to .51). InducedSchemataand Text Comprehension Referring to the overall effect of induced schemata on FL text comprehension, the comparison of the mean of the overall reading comprehension score of the Treatment-group with that of the Control-group show that the treatment produced only a mild effect on overall comprehension (Treatment-group: X=19.27, SD=2.49; Control-Group: X=17.98, SD=2.59). The Treatment-group also scored only slightly higher on specific schemata-geared questions; on linguistic schemata-geared questions, the difference in the mean score was 0.8, on content schemata-geared questions 0.6 and on formal schemata-geared questions the difference was 0.2. TESLCANADAJOURNAUREVlJE TESLDUCANADA VOL.11,NO.2,SPRING1994 79 Correlations, on the other hand, suggest that the relationship between the schemata-inducing activities and overall text comprehension is rather impressive in the case of the Treatment group: correlations between scores on specific schemata-geared questions and the overall score are higher in the case of the Treatment-group than in that of the Control-group. The difference is biggest in the correlations between questions geared to the linguistic schemata and the overall reading comprehension score (Treatment-group: r=.79; Control-group: r=.30). Correlations between questions geared to content schemata and the overall reading comprehension score are weaker (Treatment-group: r=.76; Control group: r=.63) and no difference is found between the Treatment-group and the Control-group in the relationship of formal schemata-geared questions and the overall reading comprehension score. Correlations between the scores on sChemata-inducing activities and the scores on questions geared to specific schemata indicate a clear relationship between the inducing activities of linguistic nature, on the one hand, and the questions geared to linguistic schemata (r=.53) and to content schemata (r=.51), on the other. Regression analysis indicates that induced linguistic schematawell predicted the score on the linguistic SChemata-geared question (R**2=.28), while induced content and formal schemata predicted the score on the SChemata-geared questions only to a very small degree (Fo: R**2=.15; Co: R**2=.06). These findings are confirmed by Stepwise Regression analysis: the overall reading comprehension score is predicted only by the induced linguistic schemata (43%). Regarding the texts of different rhetorical structures, it seems that only the induced linguistic schemata affect overall text comprehension in tests on all the three rhetorical structures: the correlations between the scores on induced linguistic schemata and the overall reading comprehension score are high, ranging between .42 and .70. The relationship between induced content and formal schemata and overall text-comprehension, on the other hand, vary according to the rhetorical organization of the text. Thus, inducing content schemata shows a rather strong relationship with the overall reading comprehension score ofboth descriptive and analytical types of texts (r=.63), while the inducing of formal schemata seems to have no relationship with the reading of the same types of text (no significant correlations were found). There is, however, a clear relationship between the reading of a tightly organized argumentative text and induced formal schemata (r=.54). The 80 ADINALEVINE/IHEAREVES