Developing Readers' Knowledge through Analogy

Developing Readers' Knowledge through Analogy Author(s): David A. Hayes and Robert J. Tierney Source: Reading Research Quarterly, Vol. 17, No. 2 (1982), pp. 256-280 Published by: Wiley on behalf of the International Reading Association Stable URL: http://www.jstor.org/stable/747486 Accessed: 19-05-2015 13:52 UTC REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/747486?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Wiley and International Reading Association are collaborating with JSTOR to digitize, preserve and extend access to Reading Research Quarterly. http://www.jstor.org This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 256 Developing readers' knowledge through analogy* DAVID A. HAYES Universityof Georgia ROBERT J. TIERNEY HarvardUniversity INFORMATION is often introduced to readers through The effect of this practice was investigated by examining analogy. three possible explanations of analogy's function: to activate specific analogous knowledge, to activate generally related knowledge, or to supply information which readers use to fashion their own comparisons. American high school students attempted to learn about the game of cricket from prose materialswhich were variously augmented with analogies drawn from the game of baseball. The students subsequently read and recalled newspaper accounts of cricket matches and made predictions and discriminations about open-ended cricket match situations. Across seven different dependent measures subjected to regression analysis, the consistent finding was that two factors-student prior knowledge about sports and baseball, and the provision of instructional texts about either baseball or cricket-explained more variance than any other factor or combinations of factors. In some instances, the more specific provision of analogies proved beneficial, especially for groups with differing levels of prior knowledge, or in conjunction with an informational text about baseball. These data were interpreted as providing strong support for a general knowledge activation hypothesis and modest support for a specific knowledge activation hypothesis, both of which were interpreted as consistent with recently emerging schema theoretic notions. UNFAMILIAR Developpement de connaissances chez les lecteurs a travers I'analogie ON INTRODUITSOUVENTaux lecteursdes informationspeufamilieresa travers l'analogie. On a etudie l'effet de cette pratique en examinant trois explications possibles de la fonction de l'analogie: activer la connaissance analogue sp6cifique, activer la connaissance habituellement rapportee, ou fournir des informations que les lecteurs utilisent pour 6laborer leurs propres comparaisons. Les lyc6ens americains essayent d'apprendre le jeu de cricket a partir de documents en prose qui ont 6t6 compl6t6s de maniere vari6e par des *The dissertation upon which this article is based was selected as IRA's Outstanding Dissertation Award winner in the 1978-79 competition. This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 257 analogies provemant du jeu de baseball. Les 6tudiants subs6quemment ont lu et se souviennent d'articles de journaux concernant les matchs de cricket et ont fait des predictions et des discriminations dans des situations de match de cricket non programme. A travers sept mesures d6pendantes diff6rentes sujettes a des analyses de r6gression, la d6couverteconsistante a eu pour r6sultatdeux facteurs - la connaissance pr6alable des 6tudiants concernant les sports et le baseball, et la fourniture de textes d'instruction de baseball ou de cricket - qui ont expliqu6 plus de variante que n'importequels autres facteurs ou combinaisonsde facteurs.Dans certainscas, la fourniture plus sp6cifique d'analogies s'est av6r6e 6tre b6n6fique, surtout pour les groupes dont les niveaux de connaissances pr6alablesdiff6raient, ou en conjonction avec un texte non formel concernant le baseball. On a interpr6t6ces donn6es comme 6tant un grand support pour une hypothese d'activation de connaissance g6n6raleet un faible support pour une hypothese d'activation de connaissance sp6cifique, ces deux ont 6t6 interpr6t6escomme 6tant consistantes avec des notions theoriques de schema se d6gageant recemment. El desarrollo del conocimiento de lectores por medio de analogia CON FRECUENCIA,lectoresson introducidosa informaci6n desconocida por medio de analogia. Se investigaron los efectos de este m6todo, examinando 3 posibles razones de la funci6n de analogia: el activar conocimientos concretos anal6gicos, el activar conocimientos generales relacionados, o contribuir informaci6n que lectores utilizan para realizar sus propias comparaciones. Alumnos estadounidenses de secundaria trataron de aprender el juego de criquet por medio de descripciones en prosa, aumentados con analogias del juego de b6isbol. Los alumnos posteriormente leyeron y repasaron descripciones de juegos de criquet en articulos de peri6dicos e hicieron predicciones y analisis de modelos de competici6n del juego de criquet en progreso. A traves de 7 diferentes medidas dependientes sometidas a analisis de regresi6n, el resultado consistente fue que 2 factores-el previo conocimiento del alumno de deportes y del b6isbol, y la inclusi6n del tema del b6isbol o criquet en los textos de instrucci6n-revelaron mas variaciones que ningun otro factor o combinaci6n de factores. En algunos casos, la inclusi6n de analogias concretas demostr6 ser de beneficio, especialmente para grupos con diferentes niveles de conocimiento previo, o en conjunci6n con un texto de referenciasobre el b6isbol. Estos datos se interpretaron de respaldar firmemente una hip6tesis de activaci6n de conocimientos generales y de dar cierto apoyo a una hip6tesis de activaci6n de conocimientos concretos, ambos interpretados de ser nociones te6ricas consistentes con el "schema"(sistema conceptual) emergiendo recientemente. This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 258 o Number2, QUARTERLY READINGRESEARCH 1982 XVII/2 Much of what is expected to be learned in school must be learned by reading and studying textual materials. Quite often, however, students lack the background knowledge necessary to profit from their reading experiences. When they do, teachers must somehow face the challenge of providing it. Many teachers believe that an effective way to meet this challenge is to provide an interpretive bridge between the unfamiliar material and the knowledge which students do have. In recent years, educators and cognitive psychologists have such a bridge can be provided through the use of analogy. that suggested to specify the influence of background knowledge on Attempting and learning, they have alluded to the role analogy might comprehension play in bringing relevant background knowledge to bear in understanding and learning new subject matter. From a theoretical perspective derived from the notion of schematism, conceived by Kant (1781/ 1966), developed by Spearman (1923), Bartlett (1932), Bruner (1960), and Ausubel (1960), and recently given prominence by Kintsch (1974), Neisser (1976), Anderson, Spiro, and Anderson (1978), Rumelhart and Ortony (1977) and others, comprehension and learning are explained essentially in terms of analogical thinking processes. Just as formal analogies isolate for comparison sets of relations held in common by different complex entities (as in traditional Pythagorean A:B::B:C or Aristotelian A:B::C:D analogies), so are thinking and learning processes thought to select from available information that which can be structured to correspond to the organization of existing ideational systems. Underlying such a notion is the assumption that intuitive organizing powers of the mind govern interactive and recursively occurring ideational systems which comprise knowledge categories held in position relative to one another in fixed patterns. The knowledge categories are roughly analogous to placeholders in formal analogical statements, and the relational patterns are analogous to the conventional ordering of such statements. The plasticity of knowledge categories allows information of to be readily assimilated into the knowledge system;the instances specific of their relational patterns provides an organizing structure to rigidity facilitate acquisition of novel sets of information which are relationally bound in similar and proportionate patterns. Mechanisms by which learning takes place have been suggested by Rumelhart and Ortony as processes of knowledge specialization and knowledge generalization. Knowledge specialization involves constraining knowledge categories so as to reduce the possible values that may be assigned to yield a representation of information; knowledge generalization involves This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 259 relaxing the constraints of knowledge categories so as to yield a representation which is more generalized. These notions suggest that comprehension and learning are transitive processes which can be accessed directly through the use of analogy. By provisionally transferringthe attributes of a familiar set of information to an unfamiliar one, new knowledge categories can be created and provided with a formula for binding them relationally. Through the reference learners make to the structure of existing knowledge, knowledge categories created for the novel content are placed into corresponding relational patterns within the existing system. Contextual differences between the existing structureand the provisional new structure constrain the knowledge categories of the latter to yield representations which ultimately render the novel material comprehensible in its own right. The experimental literature in education and psychology offers little evidence on the instructional value of an analogy, or for that matter on any attempts to increase background knowledge for learning new material. The strongest evidence has been produced by Ausubel and Fitzgerald (1961), who found that giving readersan advance expository passage on a familiar topic analogous to an unfamiliartopic to be learned from another expository passage results in superior learning. More recently Schustack and Anderson (1979) conducted memory experiments which led them to conclude that analogy promotes elaborate encoding of information and improved prose memory when there is a close relation between old and new material, but they did not discuss how that relation is to be defined. Royer and Cable (1975, 1976)studiedthe effect of advance presentation of analogous materialwhich was either concrete or abstract. but they did not directly address questions related to the instructional efficacy of analogy. Mayer (1975) found that analogies as well as illustrations and examples appear to elicit relevant knowledge structures for learning, but he drew no conclusions about the effects of analogy per se. Investigations by Dowell (1968) and Drugge (1977) found no significant effects stemming from the instructional use of analogy with high school students. Previous research on attempts to increase background knowledge through analogy is limited by its failure to address a number of issues pertinent to teaching and theory development. It is possible that the results obtained in the previous research were affected by prior knowledge or interest in the analogous materials. Of the investigations conducted to date, only the Ausubel and Fitzgerald study has attempted to assess prior knowledge of analogous materials, but the investigators This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 260 READING RESEARCH QUARTERLY * Number 2, 1982 XVII/2 did not weight its effect on subjects'performance. None of the studies has attempted to control for interest related to the analogous materials. Nor have previous studies compared the relative effects of different modes and sequences of presenting analogy in text. Furthermore, all studies have worked from a rather restricted definition of comprehension, and few studies have systematically examined the effects of analogies per se. In the study reported here, we attempted to address these concerns. The instructional effects of analogy were evaluated directly and in a context that took into account potentially important variables such as students' prior knowledge about and interest in the content from which the analogy was generated. The ultimate criterion task required American students to read and recall newspaper articles about presumably unfamiliar content, cricket matches in Australia. Prior to the criterion task, groups of students differing in their knowledge about and interest in sports and baseball were given instructional passages to read. The instructional passages provided the means through which various permutations of instruction designed to build analogies between cricket (the unfamiliartopic) and baseball (the familiar topic) were implemented. The instructional passages varied in terms of relevance to the topic and in terms of the specifity of the attempts to build analogies between the two domains, baseball and cricket; that is, some subjects read passages about irrelevanttopics, some read passages about relevant topics (baseball and cricket), and some read relevant passages in which there were attempts to draw explicit comparisons between the two domains (e.g., the wicket is like home plate). If analogies provide the kind of bridging that researchers, theorists, and philosophers suggest, then the more explicit the analogous information provided during the instructional phase, the greater should be the transfer to a task in which students are required to use knowledge about an unfamiliar topic. If on the other hand it is not the analogy per se but simply topic relevance that matters, then passages that simply discuss cricket (without any direct links between the two) ought to facilitate transfer just as well as passages which contain explicit attempts to help students see the analogy between the two. Or it may be the case that analogous instruction helps but that students can draw the analogies on their own if they are provided with the relevant information (i.e., a passage about baseball and a passage about cricket, with no attempts to compare features of the two). The present design allows for evaluation of these three competing hypotheses by comparing instructional texts consistent with each hypothesis. Furthermore, it can be argued that instructional treatment should interact with student knowledge and interest such that students knowledgeable about and interested in This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 261 baseball should benefit from the treatments to a greaterdegree than those less sports-minded. However, for purposes of instruction, a more salutory (but less sensible) finding would be that there was no interaction between treatment and sports orientation; that would imply that even students not particularly knowledgeable about or interested in the topic can benefit from the instructional interventions. To answer these concerns, the following questions were explored: 1. Would the presentation of any information about an unfamiliar domain of knowledge influence comprehension and learning from text? 2. Would attempts to increase background knowledge using alternative modes of presenting information influence comprehension and learning differentially? If so, would explicit presentation of contradistinctions between analogous domains of knowledge effect a different influence on comprehension and learning than parallel presentations which imply an analogous relation between the unfamiliar and familiar domains of knowledge? 3. Would readers with varying verbal ability and with varying amounts of background knowledge of the analogous material respond differentially to the various comprehension and learning tasks? 4. Would the comprehension and learning resulting from attempts to increase background knowledge coincide with schema-theoretic notions of comprehension and knowledge acquisition as well as other notions of reader involvement? Questions concerning the written recall of topically related text focused on the amount and type of information generated on recall tasks. It was assumed that a person's written recall performancewould be affected by the extent of knowledge related to the topic of the text and, further, that inferences could be drawn about the character of a person's cognitive processing on the basis of the amount and generality level of the information recalled. Based upon the notion (Kintsch & van Dijk, 1978; Drum, Note 1) that generality of a person's memory for text reflects extent to which its information has been assimilated, three types of information were identified in subjects' recall protocols: explicit text reproductions, text-entailed inferences, and text-evoked inferences. Other questions dealt with analogy's effect on the ability of the subjects to predict appropriate outcomes to open-ended situations related to the topic of the unfamiliar text and to discriminate between instances which This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 262 READING RESEARCH QUARTERLY o Number2, 1982 XVII/2 were and were not related to the topic of the unfamiliar text. These tasks were developed to assess the learning mechanisms of knowledge generalization and knowledge specialization. We assumed that specialized knowledge manifests itself whenever students can discriminate between instances and non-instances of a knowledge domain. Conversely, we assumed that generalized knowledge would be reflected in a student's ability to suggest plausible outcomes to open-ended situations. Method Subjects Subjects were eleventh- and twelfth-grade students of average and above-average reading ability from a rural-suburbanhigh school in Northern California. The student population represented a wide and typical range of social and economic backgrounds. The elimination of students of below-average reading proficiency was prompted by our desire to eliminate subjects who might have difficulty responding to the text. Teacherjudgment and standardizedreading test results were used as criteria for selecting students. On the basis of responses given on the surveys of interest, students from the subject pool were grouped according to three levels of baseball interest (consistently high, mixed, consistently low); a comparable number of students from each level were assigned to each treatment condition. These data were not used in the analysis. Scores on the prior knowledge test were also used to assign students evenly across treatments; however, prior knowledge was subsequently used as an independent variable in the regression analysis. The number of subjects in each treatment group actually participating in the experiment ranged from 19 to 21 subjects due to absenteeism on the day of the experiment. Experimental Materials A number of materials were developed in order to examine the effects of attempts to increase background knowledge for unfamiliar topics with passages excluding and including analogies. These materials included: a pretest survey of general interest in sports and specific interest in baseball, a pretest to assess background knowledge for baseball, an instructional text about baseball, two instructional texts on the game of cricket (one with and one without explicit analogies embedded), two control texts, two passages for recall tasks, and a discriminationprediction survey for posttest purposes. The two pretests were This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 263 administered in the first visit to the subjects' classrooms. At a later date, data on all the other experimental materials were collected in a single session in the subjects' classrooms. Pretests. General interest in sports was estimated from subjects' responses to a multiple-choice sentence item of the form, "My general feeling about sports is...." Possible responses ranged from "like sports very much" to "dislike sports very much." The data obtained suggested that all subjects involved in the study had a similar interest in sports and, therefore, general interest in sports would not be confounded with specific interest in baseball. Levels of specific interest in baseball were ascertainedthrough a rating scale and ranking procedure (Kerlinger, 1973). The first of these measures presented subjects a list of ten team sports, one of which was baseball, and asked for an indication of interest in each one independent of interest in the others. Students rated interest in each sport along a continuum for "strongly like" to "strongly dislike." Students ranked (110) the sports from most to least favorite separately for playing and for watching them. In order to assess subjects' prior knowledge of baseball, a 22item Likert-type survey was devised. The content of the baseball knowledge survey included baseball terms, rules, and situations of play that persons knowledgeable about baseball would know. Salient points of information presented in the expository text on baseball were included as well as points of baseball information to which analogical reference would later be made in the instructional text on cricket. For example, terms such as ground rule double, batter's box, and leading off were included because reference was made to them in presenting their respective cricket analogs: boundary-four, popping crease, and backing up. Items on the scale required students to select a response rangingfrom "definitely true"to "definitelyfalse;" To check on the scale's validity, the performance of assumed experts was compared with that of assumed novices. The expert population comprised students who had played interscholastic baseball; the novice population consisted of recently arrived foreign students. The difference between the expert mean (42.8) and the novice mean (1.7) provided strong support for the validity of the scale. As indicated by a split half reliability coefficient of .92, the survey reliably discrimated a variety of levels of baseball knowledge. Baseball text. The knowledge-evoking text on baseball was included in the experiment in order to examine the effect of advance presentation of information analogous to the topic of the instructional texts. In accordance with Ausubel's (1960) notion that "the most This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 264 READING RESEARCH QUARTERLY * Number 2, 1982 XVII/2 important factor in influencing learning is the quantity, clarity, and organization of the learner'spresent knowledge" (p. 50), the knowledgeevoking text devoted ample discussion to those subsuming concepts in the baseball text that would be useful for learning the analogous content of the cricket text: an overview of the game and its purpose, the playing field, scoring, the infield as the center of the game's activity, the role of the pitcher, the role of the batters and base runners, putting out batters and base runners, turns at bat, and completing the game. (Throughout the paper the knowledge-evoking text is coded B.) Cricket texts. The instructional texts were based on an article on cricket in Webster'sSports Dictionary (1976). First, an informational text strictly about the game of cricket was written to parallel the knowledge-evoking baseball text. It included an overview of the game, the playing field, scoring, the centerfield, the role of the bowler, the role of batsmen, dismissal of batsmen, turns at bat, and completing the game. Second, to create a comparable text including analogies, selected target structures were rewritten to provide direct feature comparisons with baseball. Portions of these texts appear in Figure 1. The two texts are coded C and C(A), respectively. PORTION OF TEXT WITH EMBEDDED ANALOGIES: Cricket is a bat and ball game played between two teams of 11 players each on a large grassy field. It isfrom cricket that the American game of baseball developed. In a cricket match, the teams take turns at bat. While one team bats the other team defends the field. The object of the batting team is to score runs, while the object of the fielding team is to dismiss batsmen. Unlike baseball, there are always two batsmen in play at the same time. Batsmen score runs by exchanging positions on the field. The center of activity is an area in the middle of the field called the pitch, which corresponds to the infield in baseball. At both ends of the pitch stands a wicket consisting of three vertical sticks, called stumps, with two horizontal sticks, called bails, resting across the top. Wickets are a bit like home plate in baseball. They provide a target for... CORRESPONDING PORTION OF TEXT WITHOUT ANALOGIES: Cricket is a bat and ball game played between two teams of 11 players each on a large grassy field. It is one of the mostpopular games in England andseveral other British Commonwealth countries. In a cricket match, the teams take turns at bat. While one team bats the other team defends the field. The object of the batting team is to score runs, while the object of the fielding team is to dismiss batsmen. In cricket there are always two batsmen in play at the same time. Batsmen score runs by exchanging positions on the field. The center of activity is an area in the middle of the field called the pitch, which measures 10 feet wide by 66 feet long. At both ends of the pitch stands a wicket consisting of three vertical sticks, called stumps, with two horizontal sticks, called bails, resting across the top. Wickets are 28 inches high and nine inches wide. They provide a target for... Figure1 Portions of instructionaltexts with and withoutembeddedanalogies. *It should be noted that underlining was not used in the text during the experiment. This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 265 Control texts. Two control texts were constructed. The first the structure of the knowledge-evoking text on baseball; the paralleled text on cricket. Neither contained any informational the second, about weather (Lehr, Burnett, Zin, & text was first The analogies. McNaught, 1965, pp. 10-11, 53-56); the second addressed the topic of filmmaking (Larson, 1969). Testpassages. In order to assess transfer of instructional texts to real reading situations, two test passages were created by adapting newspaper articles about cricket taken from the sports section of "The Australian" (December 1978), a daily newspaper published in Sydney, Australia. Different cricket match situations were presented within two different idea structures. This permitted identification of subjects who appeared to be more versatile in dealing with information about cricket, that is, who could use structure of the information to their advantage (Meyer, 1977; Meyer, Brandt, & Bluth, 1980). An example of one of these texts is presented in Figure 2. A hair raising century by Australian opener Graeme Wood on Friday set England back on its heels in the third test at the Melbourne Cricket Ground. Unfortunately, living dangerously eventually cost the Australians the match. Wood was caught out of his crease on the first over after lunch. Within ten more overs, the Australians were dismissed. Four were dismissed by dangerous running between creases. Two were dismissed when the English bowlers lifted the bails from the batsmen's wickets. The three remaining batsmen were caught by English fieldsmen. One was caught as he tried for a six. When the innings were complete the Australians had fallen short of the runs scored by the English. Figure2 Test Passage 1. As a check on the readabilityof the experimental materials, all texts were reviewed by classroom teachers, all of whom had taught at an eleventh-grade level, and qualified university personnel. They deemed all texts appropriate for the subjects for whom they were intended. As determined by the Fry Readability Graph (1968), all texts were assessed as being at either the sixth- or seventh-grade level. Discrimination/ Prediction task. A discrimination-prediction survey was developed to appraise the extent to which subjects learned the topic of the instructional texts. The survey presented ten cricket match This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 266 * Number2, 1982 READING RESEARCH QUARTERLY XVII/2 situations and asked subjects to choose the most likely result of that situation from a group of five results listed as optional choices for each situation. Twenty-two out of fifty choices were not related to cricket at all. In addition to choosing the most likely result for the cricket match situation, subjects were to identify those choices that were not within the scope of the game of cricket. This task produced three scores for each student: (a) the number of correct predictions; (b) the number of inappropriate predictions (predictions related to cricket but inappropriate to the situation described); and (c) the number of correct discriminations (i.e., of the 22 non-cricket choices, the number correctly identified). As a check on the appropriateness of the test, educators of Australian and British nationality who reviewed the instrument agreed the scale accurately depicted aspects of cricket. Treatments In the experiment each of the five groups of subjects read a different sequence of passages corresponding to the five treatment conditions and then responded to the test passages as well as the discrimination-prediction survey. In accordance with the sequence of passages read by each group, these treatment groups were labeled: B + C (A), X + C (A), B + C, X + C, and X + X. [B + C (A)] read a baseball passage (B), then an instructional cricket passage with analogies [C(A)]. [X + C (A)] read an unrelated passage (X), then an instructional cricket passage with analogies [C(A)]. [B + C] read a baseball passage (B), then an instructional cricket passage without analogies (C). [X + C] read an unrelated passage (X), then an instructional cricket passage without analogies (C). [X + X] read two unrelated passages. All subjects followed the same procedures in reading the and passages responding to the comprehension and learning tasks. Following the introduction to the experiment and some practice exercises, subjects were directed by their teachers to read the first two passages. The directions given by the teacher were to read and study the selection very carefully in order to learn as much as they could about the subject of the passage. They were allowed to mark the passages if they cared to do so and were told to regard the activity as a class assignment. Following the reading of the first and second passages, subjects were directed to read the first testpassage and try to remembereverythingthey This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 267 could. After reading and studying the test passage, they were instructe, to write down everything they could remember. They were given similar directions for the second test passage. Finally, they were directed to complete the discrimination-prediction survey. Dependent Measures With a system of text analysis proposed by Kintsch (1974) and codified by Turner and Green (1977), template text bases of the test passages and protocol text bases of subjects' written recalls were obtained. Protocol text bases were scored by comparing them to their corresponding template text bases according to a procedure suggested by Drum (Note 1). First, a template text base for each test passage was prepared as described above. Then, referring to the template text base, propositions of each protocol text base were identified as either repeating propositions of the template text base or representing two levels of inferences drawn from the test passages. For purposes of explaining the application of Drum's scoring procedure, a hypothetical recall protocol is given in Figure 4 and the text template in Figure 3. Propositions of the protocol text base identified as repeating propositions of the template text base were designated text reproductions. It was not necessary that a protocol proposition be recalled verbatim to be counted as a text reproduction. A protocol proposition was counted as a text reproduction if the content words used to represent its arguments and relations were synonymous with the words used in the template proposition. A protocol proposition which omitted arguments of a text base proposition was still counted as a text reproduction if the relation and at least one other argument remained intact, as in proposition 3, Figure 4. Protocol propositions representing deviations from the text were of two broad categories: propositions which could be directly linked with template propositions and propositions that could not be directly linked with template propositions. Protocol propositions that could be directly linked to the text base were designated text entailments. Three types of text-entailed propositions were identified during scoring. The first type of text entailment was a proposition which served to summarize certain propositions of the text base. This type of proposition generalized the arguments and relations of two or more specific text base propositions in such a way as to preserve their common meaning at the expense of the specific meanings represented by each in the text base (e.g., proposition This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 268 * Number2, 1982 RESEARCH READING QUARTERLY XVII/2 TEXT Darling is the proud hero of the Melbourne team. He has given spectators three centuries in as many weeks, including a score of sixes. But yesterday he was not cheered, because having scored only six runs he was caught out of his crease. TEMPLA TE TEXT BASE 1. (ISA, DARLING, HERO) 2. (QUALITY OF, HERO, PROUD) 3. (PART OF, HERO, TEAM) 4. (REFERENCE, TEAM, MELBOURNE) 5. (GIVE, A: HE, O: SPECTATORS, G: CENTURY) 6. (QUANTITY, CENTURY, THREE) 7. (INCLUDE, A: CENTURY, O: SIX) 8. (QUANTITY, SIX, TWENTY) 9. (TIME: IN, WEEKS) 10. (QUANTITY, 9, THREE) 11. (CONJUNCTION: BUT, 1, 12) 12. (CHEER, A: $, O: HE) 13. (REFERENCE, HE, DARLING) 14. (NEGATE, 12) 15. (TIME, YESTERDAY, 10) 16. (CASUALTY: BECAUSE, 17, 14) 17. (CATCH, A: $, O: HE) 18. (LOCATION: OUT OF, CREASE) 19. (REFERENCE, CREASE, HIS) 20. (SCORE: A: $, G: RUNS) 21. (QUANTITY, RUNS, SIX) 22. (QUALIFY, SIX, ONLY) Figure 3 Exampletext and templatetext base. The fans like Darling. He gets runs. He has hit three centuries. Yesterday he did six runs. A hero does better. Somebody got somewhere out of a crease. Darling plays a weird sport. CORRESPONDING PROPOSITION IN TEMPLATE TEXT BASE (Figure 3) 1. (LIKE, A: FANS, O: DARLING) 1 2. (GETS, A: HE, O: RUNS) 5,7 3. (HIT, A: HE, G: CENTURY) 5 4. (QUANTITY, CENTURIES, THREE) 6 5. (DID, A: HE, G: RUNS) 20 6. (QUANTITY, RUNS, SIX) 21 7. (TIME, YESTERDAY, 5) 15 8. (CONTRAST: BETTER, HERO, $) 9. (CATCH, A: $, O: SOMEBODY) 17 10. (LOCATION, OUT OF, CREASE) 18 11. (QUALIFY, 10, SOMEWHERE) 18 12. (PLAY, A: DARLING, 0:13) 13. (QUALITY OF, SPORT, WEIRD) Figure 4 Fictitiousrecallprotocoland protocoltext base. 2). The second type of text entailment was a proposition directly related to a single proposition of the text base. It was a text base proposition whose arguments and relations had been generalized to the extent that This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions & TIERNEY Developing reader's knowledge HAYES 269 similarity to the meaning of the text base proposition was not preserved (e.g., proposition 1). A third type of text entailment added case related information to the text base. This type of text entailment was a modifier proposition complementing a predicate proposition of the text base (e.g., proposition 9). Since these types of text entailments were not always discrete, the three types were collapsed into a single count of text entailments for purposes of analyzing the data. Protocol propositions that could not be directly linked to specific propositions of the text base were designated text-evoked propositions. Text-evoked propositions were thematically related to the content of the text base. They were generalizations drawn from the text without any connection to specific propositions of the text base (e.g., proposition 8). Drum provides a fourth category of recalls to include implausible, irrelevant, non-text-related propositions. In this study such recalls were not scored. The text-evoked category in this study included only plausible and relevant units of information. These three types of protocol propositions (text reproductions, text entailments, and text-evoked propositions) were tabulated for each subject's recall for each test passage with interrateragreement of .92 on approximately 10%of the sample. These three scores provided the raw data for the analysis of the recall protocols. Responses to the discrimination-prediction survey were scored in such a way as to yield three measures. First, an accuracy of prediction score was determined from the number of correct selections of most-likely outcomes to specific cricket match situations. A second measure was derived by counting the number of such outcomes which were within the scope of cricket regardless of their likelihood. This was similar to a context-bound discrimination. A third measure was generated by assessing whether the subjects were able to discriminate cricket from noncricket outcomes without being given a specific cricket match situation. This was similar to a less context-bound discrimination. Results The data obtained in the investigation were analyzed by hierarchical multiple regression, a multiple correlation procedure which calls for entering variables into the regression equation by a sequence consistent with the logic and purpose of the questions under examination (Cohen & Cohen, 1975). The questions of the present study were formulated to test general and specific knowledge activation hypotheses as they concern building readers' background knowledge for under- This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 270 * Number2, 1982 READING RESEARCH QUARTERLY XVII/2 standing and learning from unfamiliar text. Thus, it seemed most reasonable to account for the variance associated with the four measures of comprehension and three measures of learning by entering trait variables into the regression equation first, treatmentvariablesnext, and interactions between trait and treatment variables last. Accordingly, we entered into the regression equation first those variables which would determine the extent to which certain basic abilities accounted for the variance associated with the comprehension and learning measures. These variables included verbal ability (A), which was based on students' stanine levels of performance on the Comprehensive Test of Basic Skills (1974); prior knowledge, which was based upon their performanceon the baseball knowledge test (K); and the square of the prior knowledge test (K2)to assess any curvilinearrelations which might emerge from the data analyses. Next entered into the regression equation were the treatment which variables, represented four major contrasts of the study. Three of these contrasts provided for assessing the extent to which acquisition of new information was associated with the activation of general as well as specific relevant knowledge. To assess the general knowledge activation hypothesis, all conditions receiving at least some knowledge activation, i.e., anything about either baseball or cricket [B + C (A), X + C (A), B + C, X + C], were compared with the one treatment receiving none [X + X]. This appears as contrast (T1) in all the tables. To test the specific knowledge activation hypothesis, two other contrasts were examined. One of these contrasts (T2) measured the variance associated with membership in the two treatments providing baseball information [B + C (A), B + C] versus those treatments not providing baseball information [X + C (A), X + C]. Another contrast (T3) to examine the specific knowledge activation hypothesis involved comparing membership in any treatment condition where students were given explicit analogies [B + C (A), X + C (A)] versus those not provided with analogies [B + C, X + C]. The fourth contrast (T4) examined the interaction of the provision of information about baseball and the provision of analogies to baseball in the cricket text [X + C (A), B + C] versus [B + C (A), X + C]. The remaining variables entered into the regression equation comprised two sets of interactions. The interactions entered were between each of the treatment contrasts (Tl, T2, T3, T4) with prior knowledge (K) and verbal ability (A), respectively. Hierarchical regression analyses using these sets of variables were calculated for each dependent measure. In reporting results, we have chosen to report in the text only This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 271 those variables that added a significant (p<.05) proportion of variance explained (r2 change) to the regression. A failure to mention a variable implies, therefore, that it did not add a significant amount of explained variance to the equation. For the reader's convenience, the regression tables discriminate between p<.05 and p<.01 additions to explained variance. For ease in interpreting numbers in the text, proportions (which appear in the tables) were converted to whole percentages. Combined Recall The means and standard deviations of the recall measures for passages 1 and 2 combined are presented in Table 1. The results of the hierarchical analysis for each dependent measure are presented in Table 2. Table 1 Summary of means and standard deviations of recall measures: Passages 1 and 2 combined Total Units of Information Treatment Group 1. B 2. X 3. B 4. X 5. X + C (A) + C (A) +C +C +X Text Reproductions Text Entailments Text Evocations n M SD M SD M SD M SD 21 20 20 19 20 38.24 33.70 36.75 36.37 23.05 13.87 12.77 7.68 8.76 8.35 27.19 21.15 24.05 24.90 16.55 10.51 10.15 7.65 8.20 7.86 7.62 6.50 6.55 7.68 3.55 4.40 3.74 2.60 3.30 2.09 3.43 6.05 6.15 3.79 2.95 2.40 4.08 5.23 3.60 2.42 The results using total units of information as the dependent measure are given in the first two columns of both tables. The score for total units of information was derived by summing text reproductions, text entailments, and text evocations. According to this hierarchical regression analysis, only the prior knowledge variable (5%) and the contrast T, (20%)accounted for a significant proportion of the variance. Hence for total recall, the general knowledge activation hypothesis is supported. The results obtained from the regression analyses conducted on the other recall measures yielded similar patterns of results. For text reproductions, verbal ability (5%)as well as contrast T, (11%)accounted for a significant proportion of the variance. Since there were many more text reproductions than either entailments or evocations, the similarity between these data sets was not surprising. Wheretext-entailed recall was This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions READING RESEARCH QUARTERLY * 272 Number2, 1982 XVII/2 Table2 Partitioningof varianceandsignificanceof testsfor therecallmeasure: Passages 1 and 2 combined Total No. of Units of Information Step Source r2 1 2 3 4 5 6 7 8 A K .018 .059 .068 .264 .273 .273 .275 .275 .279 .281 .282 .383 .308 .309 .317 9 K2 T, T2 T3 T,4 T, x K T4x K T3 x K T2 x K T, x A T4 xA T3 xA T2 xA r2Change .018 .041" .009 .195** .009 .001 .001 .0004 .004 .002 .0006 .003 .025 .002 .008 Text Reproductions r2 .046 .050 .058 .164 .181 .181 .189 .190 .193 .221 .223 .223 .265 .265 .291 r2Change .046* .005 .008 .106** .016 .000 .008 .001 .003 .028 .001 .000 .041 .000 .027 Text Entailments r2 .003 .056 .057 .202 .202 .202 .219 .219 .224 .227 .235 .242 .246 .260 .260 r2Change .003 .053* .001 .146** .000 .000 .016 .000 .005 .003 .008 .007 .005 .013 .001 Text Evocations r2 r2Change .030 .085 .097 .133 .134 .137 .196 .196 .212 .262 .268 .268 .275 .276 .297 .030 .056* .012 .036* .001 .003 .058* .000 .017 .050* .005 .000 .007 .000 .021 *p<.05. **p<.01. the dependent measure, only prior knowledge (5%) and contrast T, (15%) accounted for significant proportions of variance. For text evocations as the dependent variable, the results reflected some of the same patterns but with some interestingdifferences. While prior knowledge (6%) and contrast T, (4%) accounted for significant proportions of the variance, contrast T4, the test of the interaction of baseball and analogies to baseball, also added significantly to explained variance (4%). Also, the interaction between prior knowledge (K) and contrast T3, the test for analogies, explained 5%of the variance. An examination of the means in Table 1 suggests that the locus of the T4 effect is the depressed number of evocations elicited by the [B + C (A)] group. It is interesting to note that this group had the highest number of reproductions and the second highest number of entailments. An examination of the prior knowledge by treatment means (not reported) suggested that the K X T3 interaction resulted from a tendency for more knowledgeable students who were given analogies to offer fewer evocations than did either less knowledgeable students who were given analogies or equally knowledgeable students who were not given analogies. In order to check on the reliability of these findings for recall categories summed across the two passages, separate regressions were This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge 273 HAYES& TIERNEY calculated for each passage. While there was a slight tendency for the analogy treatments to interact differently with prior knowledge or reading ability across the passages, the major findings of the combined analysis tended to hold up. That is, prior knowledge and T, tended to explain significant proportions of variance, especially on total recall and text reproductions.' Learning Measures Table 3 displays the means and standard deviations of the students' responses to the prediction task (selection of the appropriate outcomes to cricket match situations) as well as the discrimination tasks (the number of inappropriate predictions when students were asked to specify the most-likely outcome; the number of topically inconsistent choices of information specified as inconsistent by students). The results of the hierarchical regression analyses are presented in Table 4. Table3 Summaryof means and standarddeviationsof learningmeasures: Passages 1 and 2 combined Predictions Correct Treatment Group n M 1. B + C (A) 2. X + C (A) 3. B + C 4. X + C 5. X + X 21 20 20 19 20 5.81 6.10 6.25 6.05 2.80 Inappropriate Predictions Discriminations SD M SD M SD 2.36 2.00 1.52 1.18 1.58 0.76 1.35 1.85 1.21 3.45 1.04 1.57 3.05 1.08 1.67 13.76 10.08 14.65 12.79 11.80 4.66 5.85 4.63 3.54 4.43 The findings of these analyses offer support for the role of prior knowledge in addition to both a general knowledge activation and a specific knowledge activation hypothesis. Prior knowledge accounted for a significant proportion of the variance on two of the three measures (correct predictions, 5%, and inappropriate predictions, 5%). Support for the general knowledge activation hypothesis was provided by the findings for contrast T, (whether or not a student was a member of any treatment group). The contrast T, accounted for a significant proportion of the variance associated with the variables correct predictions (34%) and inappropriate predictions (18%);although contrast T, did interact with ability (4%). Some support for a specific knowledge activation hypothesis was provided by the findings for the first and third learning variables (correct predictions and discrimination of inconsistencies). In This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions READING RESEARCH QUARTERLY * 274 Number2, 1982 XVII/2 Table4 Partitioningof varianceand significancetests of learningmeasures Predictions Correct Step Source r2 1 2 3 4 5 6 7 8 A K .000 .054 .057 .401 .402 .404 .406 .407 .411 .439 .440 .477 .481 .492 .495 K2 Ti T2 T3 T4 Ti x K T4 X K T3 x K T2 x K Ti x A T4 X A T3 x A T2 x A 9 Inappropriate Predictions r2Change .000 .054* .003 .344** .000 .003 .002 .000 .004 .028* .001 .038* .004 .011 .003 r2 r2Change .000 .046 .046 .221 .221 .231 .251 .251 .254 .257 .258 .276 .276 .277 .277 .000 .046* .000 .175** .000 .010 .020 .000 .003 .002 .002 .017 .000 .001 .000 Discriminations r2 r2Change .000 .010 .020 .036 .095 .116 .117 .129 .130 .132 .134 .138 .158 .167 .167 .000 .010 .010 .015 .059* .021 .001 .012 .001 .002 .001 .004 .019 .009 .000 *p<.05 **p<.01 particular, contrast T2 (whether or not a student received any information about baseball) accounted for 6% of the variance associated with the students' discrimination performance;receiving prior information about baseball resulted in better discrimination. Also, an interaction of prior knowledge with contrast T3 (analogy versus no analogies) accounted for significant proportions of the variance (3%) associated with correct predictions. Discussion At the outset we undertook to investigate the impact, if any, of attempts to increase readers' background knowledge for understanding and learning from unfamiliar text. To guide our investigation we posed several questions. We now return to each of those questions as we discuss the findings of the investigation. Would the presentation of any information about an unfamiliar domain of knowledge influence comprehension and learning from text? Across almost all analyses, the data suggested that if students were given information about the unfamiliar topic, their performance was better on the written recall tasks as well as on prediction and discrimination tasks. On the written recall tasks students given instructional texts (treatment conditions [B + C (A)], [X + C (A)], [B + C], and [X + C]) produced significantly more information at all levels of This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES & TIERNEY 275 generality than the control subjects [X + X] who were not given such texts. In terms of performance on the prediction and discrimination tasks, subjects given background information were better able to make accurate predictions and to discriminate information consistent with a specific instance of the unfamiliar topic. In terms of the latter, subjects given information on the unfamiliar topic were better than control subjects in being able to make discriminations if the information was connected to a specific instance of the topic; that is, they made fewer inappropriate predictions. In light of the dearth of support from similar empirical endeavors, the present results can hardly be viewed as trivial. Unlike most other studies in which attempts have been made to increase background knowledge, the present data provide strong support for the efficacy of the treatment conditions over the control condition. In particular, the data supported the topic relevance or general knowledge hypothesis; the presentation of information related to the topic to be learned, regardless of the specificity of this information or mode of presentation, influenced students' comprehension and learning from text. Furthermore, the data provide some clarification of the nature of the effects of these modes intended to increase background knowledge upon comprehension and learning. Would attempts to increase background knowledge using alternative modes of presenting information influence comprehension and learning differentially? Here we observed the extent to which the variance associated with student performance could be explained by various combination comparisons (T,, T2, T3, T4). In general, the use of alternative modes of presenting information did have an impact, but one which did not consistently favor the use of analogies over the treatment conditions. The influence of anology per se manifested itself in conjunction with other factors, for example, prior knowledge of students or the provision of information about baseball. It should be pointed out that the [B + C (A)] treatment typically elicted the highest or one of the highest scores on the most of the measures except evocations. In all, it would seem that the data provided support for all three of the hypotheses investigated, but, across most of the variables under investigation, greatest support was given for the general knowledge activation hypothesis. Would readers with varying amounts of background knowledge and verbal ability respond differentially to the various comprehension and learning tasks? There was strong evidence from almost all of the analyses that a major determinant of success in comprehending and This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 276 * Number2, 1982 READING RESEARCH QUARTERLY XVII/2 learning unfamiliar material was prior knowledge. Furthermore, while verbal ability had an important impact, its importance tended to be most apparent in those recall situations measuringverbatim-typeinformation. In addition both prior knowledge and verbal ability tended to interact with the mode of presentation and the specificity with which information is presented. The findings with respect to text evocations suggested an interaction between analogy presentation and prior knowledge such that students with more verbal ability and more knowledge tended to offer more restrained recalls (fewer text evocations) than their counterparts not given analogies. The failure of the present study to find other differences across levels of analogous background knowledge may well have resulted from design limitations. On the pre-experimental baseball knowledge survey, very few of the subjects scored as low as the highest-scoring baseballnaive person who completed the survey in the validation procedures. Within the range obtained, an attempt was made to segment subjects into three levels of knowledge-high, moderate, and low. It seems doubtful, however, that the three groups representeddiscrete levels of knowledge, since there was a limited interval separatingthe scores. That several of the differences across levels of text-analogous knowledge in the present study approached significance suggests that if there had been more variation between levels of knowledge, significant differencesmight have been obtained. Would the comprehension and learning which resultedfrom to increase background knowledge coincide with schemaattempts theoretic notions of comprehension and knowledge acquisition as well as other notions of reader involvement? The data collected in the present study were examined from theoretical perspectives that explain the amount of information recalled (Kintsch, Kozminsky, Sterky, McKoon, & Keenan, 1975; Marshall, 1976) as well as the inverse relation between explicit recall and generalized recall (Kintsch & van Dijk, 1978; Spiro, 1977). According to these perspectives, the more the involvement of the reader, the more information recalled and the greater the integration of text information with reader knowledge. The data provided support for these perspectives;further,they emphasizedsome reasons for involvement or noninvolvement with text: degree of knowledge about the topic of the text together with mode of presentation. Although the measures devised to address schema specialization and generalization were neither discrete nor comparable, the performance of the subjects on the prediction and discrimination tasks suggested that knowledge gained from the instructional texts had begun This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 277 to form useful transfer mechanisms for students. Subjects given background information related to the tasks were better able to make predictions for the open-ended text-related situations. Also, they were better able to discriminate information consistent with the topic if that information was connected to a specific instance of the topic (i.e., on the inappropriate predictions measure). Subjects given background information related to the tasks were not more able than control group subjects to make discriminations between topic-consistent and topicinconsistent information when that information was not presented in some particular context (i.e., on the discrimination task). Subjects given analogy both in the advance text and embedded in the instructional text appeared to make the best context-related discriminations between instances and non-instances of the topic. In terms of the notions of schema specialization and generalization, the data from the present study did suggest knowledge appeared first to generalize to specific instances on the novel content and then to specialize within the context of those specific instances. The data indicated that independent generalization of knowledge, that is, generalization within specific applicative contexts, did not appear to be achieved by the subjects in this investigation. Implications for Research What implications can be drawn for further research which examines attempts to increase background knowledge? The present study raised more questions about increasing background knowledge than it answered. For example, the present study investigated only four ways of presenting unfamiliar information. The differential effects of other modes of presenting information need to be investigated in order to determine their instructional efficacy for meeting specific instructional goals. With respectto presentinginformationthroughanalogy,examplesof other modes which might be investigatedinclude analogical annotation, analogical questions, analogous and vicarious experiences,self-generated and selected analogies, concurrentreading on analogous topics, and analogical studyguides.Comparisonmightalso be madewith otheraids to textual instructionsuch as illustrationsand concreteexamples.Increasingbackground knowledgehas been studied in connectionwith some of these other aids to instruction, but conclusive findings as to their effectivenessare lacking. What is significant about the present study is that it confirms the worth of such endeavors and suggests some guidelines for future research studies. What guidelines for conducting similarresearchare prompted? In terms of design considerations, the present study raised several This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions 278 * Number2, READINGRESEARCH QUARTERLY 1982 XVII/2 important issues. Differences were noted across levels of treatment and prior analogous knowledge that could not have been detected by the overall measure of recall typically used in previous studies. Analysis of text recalls by levels of generality appearedto afford the detection of such differences in recalled information not obtainable by an overall measure of text recall. It suggested that differential information in the responses of readers to question types may help researchersspecify how and what other learning takes place. Alternatively, the pre-study failed to address adequately several important variables. First, the extent to which the treatment conditions as in employed the present study focused attention on the more important information of the text was not ascertainedfrom the recall data. Recalled information by level in the idea structure of the passages could not be clearly interpreted since text reproductions could not be analyzed together with text entailments at each level. For example, text entailment that summarized propositions variously located in a passage's idea structurecould not be assigned a single level in the ideational structure. Second, subjects responded to experimental texts immediately following their presentation; neither delayed posttest nor on-line processing measures were used. Third, readers' interest in the analogous material was considered in this study only to the extent of controlling its potential influence on the dependent measures. The interactive effects of analogy with subjects' interest in the analog were not investigated. From a practical standpoint such investigation appears to be warranted. Fourth, subjects involved in the present study did not represent extreme levels of background knowledge, and no attempt was made to assess individual differences in background knowledge after the introduction of the instructional texts. Given this limitation and the fact that many of the differences in measures approached conventional levels of significance, it might have been worthwhile to use subjects who represented a wide spread in knowledge about baseball. Fifth, the specific relation between the information presented, tested, and recalled was not examined. This might entail a careful examination of the relations which obtain across the information represented by the instructional text, the information students learned, and the knowledge requiredto understandthe test passages. Examples of processing, including misunderstandings, could be related to either instructionally induced or self-initiated analogizing by students. This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions Developing reader's knowledge HAYES& TIERNEY 279 Concluding Remarks In general, the findings should be considered encouraging. They support assertions by educators, philosophers, and psychologists that attempts to increase background knowledge facilitate learning unfamiliar material. Our attempts using alternative modes of presentation did promote learning from text. Of practical relevance, our findings offer support for providing readers with general information as well as explicit analogous information. The different ways in which attempts were made to increase knowledge had a general as well as differential impact on learning in accordance with individual subject differences. A need is thus indicated to move from broad notions about the utility of instructional strategies toward more refined distinctions in their application. REFERENCES ANDERSON, R.C., SPIRO, R.J., & ANDERSON, M. Schemata as scaffolding for the representation of information in connected discourse. American Educational Research Journal, 1978, 15, 433-440. AUSUBEL,D.P. The use of advanceorganizersin the learning and retention of meaningful verbal material. Journal of Educational Psychology, 1960, 51, 267-272. AUSUBEL, D.P., & FITZGERALD D. The role of discriminability in meaningful verbal learning and retention. Journal of Educational Psychology, 1961, 52, 266-274. F.C.Remembering.Cambridge:CamBARTLETT, bridge University Press, 1932. BRUNER, J.S. The process of education. Cambridge: Harvard University Press, 1960. COHEN,J., & COHEN,P. Applied multiple regression/correlation analysis for the behavioral science. Hillsdale: N.J.: Erlbaum, 1975. COMPREHENSIVE TEST OF BASIC SKILLS. Monterey, California: McGraw-Hill, 1974. R. (Ed.). Webster'ssport dictionary. COPELAND, Springfield, Mass.: Merriam, 1976. DOWELL, R.E. The relation betweenthe use of analogies and their effects on student achievement on teaching a selected concept in high school biology. Unpublished doctoral dissertation, Indiana University, 1968. DRUGGE, N.L. Thefacilitatingeffectof selected analogies on understanding of scientific explanations. Unpublished doctoral disser- tation, University of Alberta, 1977. FRY,E. A readability formula that saves time. Journal of Reading, 1968, 11(7), 513-516. KANT,I. Critique ofpure reason (F.M. Muller, trans.). Garden City N.Y.: Anchor Books, 1966. (Originally published, 1781.) KERLINGER,F.N. Foundations of behavioralresearch (2nd ed.). New York: Holt, Rinehart & Winston, 1973. KINTSCH, W. The representation of meaning in memory. Hillsdale, N.J.: Erlbaum, 1974. KINTSCH, W., KOZMINSKY, E., STERKY, W.J., MCKOON, G., & KEENAN, J.M. Comprehension and recall of text as a function of content variables. Journal of Verbal Learning and Verbal Behavior, 1975, 14, 196-214. KINTSCH,W., & VAN DIJK, T.A. Toward a model of text comprehension and production. Psychological Review, 1978, 85, 363-394. R. Youngfilmmakers. New York: Avon LARSON, Books, 1969. LEHR, P.E., BURNETT, R., ZIN, S., &MCNAUGHT, H. Weather. New York: Golden Press, 1965. N. The structure ofsemantic MARSHALL, memory for text. Unpublished doctoral dissertation, Cornell University, 1976. MAYER,R.E. Different problem-solving competencies established in learning computer programming with and without meaningful models. Journal ofEducational Psychology, 1975, 67, 725-734. B.J.F.The structure of prose: Effects on MEYER, learning and memory and implications for This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions * Number 2, 1982 READING RESEARCH QUARTERLY 280 educational practice. In R.C. Anderson, R.J. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition ofknowledge. Hillsdale, N.J.: Erlbaum, 1977. MEYER,B.J.F., BRANDT,B.M., & BLUTH,G.J. Use of top level structure in text: Key for reading comprehension of ninth-grade students. Reading Research Quarterly, 1980, 16, 72103. U. Cognition andreality: Principles and NEISSER, implications of cognitive psychology. San Francisco: Freeman, 1976. ROYER,J.M., & CABLE,G.W. Facilitated learning in connected discourse. Journal of Educational Psychology, 1975, 67, 116-123. ROYER,J.M., & CABLE,G.W. Illustrations, ana- logies and facilitative transfer of prose learning. Journal of Educational Psychology, 1976, 68, 205-209. RUMELHART, D.E., & ORTONY,A. The representa- tion of knowledge in memory. In R.C. XVII/2 Anderson, R.J. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, N.J.: Erlbaum, 1977. J.R. Effects of SCHUSTACK, M.W.,& ANDERSON, analogy to prior knowledge on memory for new information. Journal of Verbal Learning and Verbal Behavior, 1979, 18, 565-583. C. The nature of intelligence and the SPEARMAN, principles ofcognition. London: Macmillan, 1923. SPIRO,R.J.Remembering information from text: The "state of schema" approach. In R.C. Anderson, R.J. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, N.J.: Erlbaum, 1977. TURNER, A., & GREENEE. The construction and use of propositional text base (Tech. Rep. No. 63). Boulder: University of Colorado, Institute for the Study of Intellectual Behavior, April 1977. Reference Note 1. Drum, P.A. Structural and thematic variations on fourth graders types of recall. Paper presented at the meeting of the American Educational Research Association, San Francisco, April 1979. Footnote 'For the reader interested in a finer analysis, these data are available from the authors. This content downloaded from 174.6.213.164 on Tue, 19 May 2015 13:52:25 UTC All use subject to JSTOR Terms and Conditions