Deep-level Comprehension of Scientific Text

download Deep-level Comprehension of Scientific Text

of 20

Transcript of Deep-level Comprehension of Scientific Text

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    1/20

    Top Lang DisordersVol. 25, No. 1, pp. 6 5-8 3 2005 Uppinc ott Williams & Wilkins, Inc.

    Deep Level Comprehension

    of Science TextsThe Role of the Reader and the TextRachel M. Best PhD; Michael Rowe, MS;Yasuhiro Ozuru, P hD; Danielle S. McNam ara, PhD

    Many students from elementary school through college encounter difficulty understanding theirscience textbooks, regardless of whether they have language disorders. This article discusses someof the particular difficulties associated with science tex t comprehension and possible remedies forfacilitating and enhancing comprehension of challenging expository text materials. Specifically,we focus on the difficulties associated with generating inferences needed to comprehend sciencetexts. The successful generation of inferences is affected by factors such as students' prior knowl-edge and reading strategies, and the manner in which science texts are written. Many studentslack the necessary prior knowledge and reading strategies to generate inferences and thus com-prehen d science texts only poorly. Further, science texts are typically low

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    2/20

    66 TOPICS IN LANGUAGE DISORDERS/JANUARY-MARCH 2005

    populations when reading science texts, in-cluding problems resulting from one text-related factor (text cohesion) and tw o reader-

    related factors (domain-relevant knowledgeand reading strategies). Finally, we proposetext-based and reader-based remediation tech-niques to facilitate and enhance science textcomprehension and thus overcome compre-hension problem s associated w ith reading dif-ficult texts.

    DNFERENCING AS THE KEY TO DEEP

    COM PREHENSION OF SCIENCE TEXTS

    Reading comprehension can be definedas the ability to obtain meaning from w rit-ten text for some purpose (Vellutino, 2003,p. 51). To comprehend successfully, thereader must identify a series of letters as aword, access the meaning of the word fromthe lexicon or mental dictionary (Perfetti,1985), and integrate individual word mean-

    ings into a coherent sentence-level represen-tation (Vellutino, Scanlon, Small, Tanzman,1994). Deep comprehension, however, re-quires more than the mere interpretation ofindividual sentences; the reader must also beable to integrate individual sentence mean-ings into a coherent text-level representa-tion (Kintsch, 1988, 1998). In other words,to achieve deep comprehension, the readermust construct a global meaning that inte-

    grates multiple sentences. The primary focusin this article is on inference generation, a pro-cess that leads to text-based and knowledge-based connections to, and across, sentences.The process of creating connections whilereading by generating inferences underliesthe successful and deep-level comprehensionof science texts.

    Making inferences is a critical feature ofunderstanding the overall meaning of texts(Kintsch, 1988) because inferences combinethe individual sentence meanings distributed

    the information relevant to the situations orevents. Therefore, to successfully compre-hend a text, the reader must generate infer-

    ences to fill in missing information and builda coherent mental model that incorporates allthe information in the text (Zwaan & Singer2003). For example, consider the followingsentence pair:

    1. Plants lack a nervous system.2. They carmot make quick responses to

    stimuli.The first inference that must be made afterthe first sentence is that the pronominal ref-

    erent for they is plants. Comprehensionalso depends on inferring tha t the ability tomake quick responses is somehow related tothe nervous system. Building this inferentialconnection requires the use of logic, syntacticknowledge, and the ability to access knowl-edge from semantic memory, or to recall rel-evant information cited in earlier parts of thetext. This form of inference is called a back-ward causal inference because it involves at-

    tributing the cause of a phenom enon or eventdescribed in a given sentence to a thing orevent described in a previous section of thetext. In the cu rrent example, successful, deep-level understanding of this passage requiresthe inference that the nervous system is re-sponsible for quick responses. It illustrates thepoint that successful comprehension of textsrequires more than word-decoding, vocabu-lary, and syntactic skills. Deep-level compre-

    hension also requires the ability to make in-ferential links across individual sentences inorder to construct a global picture of underly-ing concepts.

    According to Kintsch's (1988, 1998) CImodel, to accomplish this, readers must firstuse their language-related know^ledge (e.g.,morphology, syntax, semantics) to construct arepresen tation on the basis of the informationexplicitly stated in the text. This initial levelof representation is often far removed froma coherent model regarding the situations or

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    3/20

    Deep-Level Comprehension of Science Texts 67

    To go beyond a purely text-level compre-hension, the CI m odel then posits that readersmust draw on additional information, such asprior domain know ledge, or information citedin previous parts ofthe text, and integrate thisinformation into a more complete m ental rep-resentation of the events that the text aimsto describe. In this sense, the representationthat readers must construct to com prehend atext's deeper meanings does not necessarilycorrespond w^ith the representation that theycan construct directly from information pro-vided explicitly in the text. The related impli-cation of the CI model is that th e quality of thetext determines the amount and type of infer-ences that the reader needs to generate in or-der to com prehend the events or situations ina coherent manner.

    TEXT COHESION A ND INFERENCING

    The CI model suggests that comprehension

    can be improved when text cohesion is im-proved, thus reducing the need to make in-ferences. This improvement would be par-ticularly true for readers who are unable togenerate inferences from difficult texts. Textcohesion refers to properties of the text thatdetermine the degree to which readers needto generate inferences to construct a coher-ent mental representation from the informa-tion explicitly stated in the text. Texts are

    considered to be low cohesion w^hen con-structing a coherent representation from thetext requires many inferences. Texts are con-sidered high cohesion w hen elements withinthe text provide more explicit clues to rela-tions within and across sentences.

    Cohesion is important at both the globaland local levels of text. The term global co-hesion refers to the overall cohesion of thetext. Signals for global cohesion include intro-ductory paragraphs, headers, summary para-graphs, and the semantic overlap between

    and explanations for difficulty terms or con-cepts. Texts are considered to be locally co-hesive to the extent that the relationship be-tween adjacent sen tences (or clauses) is madeexplicit through linguistic cues. For exam-ple, argument overlap can be increased byrepeating a referent in a sentence and usingfewer pronominal referents or synonyms. Be-low are examples of sentence pairs that doand do not repeat the target referent sexualreproduction :

    1. We tend to take the existence of sex-ual reproduction for granted. From an

    evolutionary standpoint, this is a seriouspuzzle.2. We tend to take the existence of sex-

    ual reproduction for granted. However,from an evolutionary standpoint, sexualreproduction is a serious puzzle.

    In this example, the second sentence pair ispresumed easier to process because the au-thor has explicitly stated that sexual repro-duction is the puzzle. Also, by adding the

    connective, however (other examples includetherefore, because, etc.), the author has pro-vided more explicit cues for readers on the na-ture ofthe link between the sentences, gener-ating the expectation among readers that thenature of the puzzle is about to be explained.

    A second aspect of cohesion is termed ex-planatory cohesion. This aspect concernsthe degree to w^hich the background informa-tion necessary to make connections among

    relevant sentences is provided explicitly bythe text. Specifically, the presence of domain-specific background information necessaryfor the coherent linking of sentences has alarge impact on explanatory cohesion. Re-turning to the previous example about plants,readers would be more likely to recognize therelationship between the two sentences, andto construct a representation that integratedtheir meanings if the sentences Plants lack

    a nervous system and They cannot makequick response to stimuli were precededb th l t t Th

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    4/20

    68 TOPICS IN LANGUAGE DISORDERS/JANUARY-MARCH 2005

    helps the reader construct inferences w hich,in turn, facilitate deeper comprehension.

    A common problem of many sciencetextbooks, however, is that authors oftenleave out information they assume to beprior knowledge for target readers. Textbookanalyses have revealed that such omissionseven occur for information that is critical tothe construction of situations or events de-scribed by the text (Beck, McKeown, Sinatra, Loxterman, 1991). Studies also have show nthat improving text quality by increasing localand/or global cohesion facilitates comprehen-

    sion as assessed with a variety of comprehen-sion measures, including recall, open-endedquestions, multiple-choice questions, andcard-sorting tasks. For example, the positiveinfluence of increasing text cohesion on com-prehension is observed with both narrativetexts (e.g.. Beck, McKeown, Omanson, &Pople, 1984) and expository texts (e.g.. Becket al., 1991; Britton & Gulgoz, 1991; Britton,Gulgoz, & Glynn, 1993; Linderholm et al.,

    2000), especially when readers do not havesufficient domain knowledge to generateinferences necessary for comprehension(McNamara, 2001; McNamara, Kintsch,Songer, & Kintsch, 1996; McNamara &Kitsch, 1996).

    READER ABBLITY AND INFERENCING

    Revising a text to increase cohesion can re-duce the need for readers to make inferencesabout relationships among ideas in texts, butthere w ill always be a need for students to gen-erate inferences, if they are to build a men-tal model of the global concepts conveyedby the text. Several theoretical accounts havebeen posited for why some readers are ableto generate inferences better than others (seeMcNamara & O'Reilly, 2004, for a review).Some researchers have proposed that read-ers' ability to generate inferences is a func-tion of their working mem ory capacity Gust

    are better able to make inferences, becausethey can hold and process more of the texat the same time (Engle Marshall, 1983). Inthis article, we focus on two other importanfactors that influence readers' abilities to generate inferences. One factor concerns readers' preexisting domain-specific knowledgethe other concerns their degree of competency in using reading strategies. We emphasize roles of domain knowledge and readingstrategies because, unlike working memorycapacity, they are subject to intervention.

    Domain-specific knowledge refers to the

    degree to which readers possess knowledgethat is specifically related to the text content. Reading strategies refer to a generaset of abilities that facilitate active processing of a text's content, and are consideredto be closely related to metacognitive abilities, such as knowledge of cognition andthe ability to monitor and regulate ongoingcognitive processes (Hacker, 1998). Domain-specific knowledge and the ability to use

    reading strategies are not completely inde-pend ent. Com petent readers are able to integrate them and use them in concert. On theother hand, our research suggests that thetwo factors are distinguishable because theytend to be related to different aspects ofthe reading comprehension process. Whereasdomain-specific knowledge is closely relatedto how easily one can comprehend materialat a given level by accessing relevant knowl-

    edge, the ability to use reading strategies ismore closely related to the effort and use ofactive processing techniques, such as elabo-rative and bridging inferences (Best, Ozuru, &McNamara, 2004; Ozuru, Best, McNamara,2004).

    Contributions of domain-specificknowledge to inferencing

    With respect to the benefits of domain-specific knowledge, readers with rich and or-ganized topic-relevant knowledge structures

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    5/20

    Deep-Level Comprehension of Science Texts 69

    For example, readers are known to gener-ate backward causal inferences (discussedearlier) routinely when the topic of the textis generally familiar, as they are in narra-tive texts that tend to regard familiar top-ics such as relationships between individ-uals and common problems encounteredin everyday life (e.g., Graesser, Singer, &Trabasso, 1994). On the other hand, readersare less likely to draw backward causal infer-ences when reading unfamiliar expositorytexts (Noordman, Vonk, & Kempff, 1992).These findings suggest that readers typicallydo not generate backward causal inferenceswhen they have less knowledge about thetext topic. Also, a wealth of empirical ev-idence demonstrates that a reader's back-ground knowledge facilitates and enhancescomprehension and learning from exposi-tory materials (e.g., Afflerbach, 1986; Chi,Feltovich, & Glaser, 1981; Chiesi, Spilich, &Voss, 1979; Lundeberg, 1987; McNamara &Kintsch, 1996; Means & Voss, 1985; O'Reilly

    McNamara, 2002).Having rich domain-specific knowledgeseems to be an important factor supporting in-ference generation, but it is not an essential.Studies have show^n that some aspects of in-ference generation are observed even w^henknowledge about a topic is limited. For exam-ple, Noordman et al. (1992) conducted twoexperiments in w^hich they showed that read-ers are able to draw^ backward causal infer-

    ences even when reading expository texts onunfamiliar topics. In a similar vein. Singer,Harkness, and Stewart (1997) found evidenceof inference generation for unfamiliar expos-itory materials, that is, as long as the read-ers were not rushed in their examinationof the texts. Together, these studies showthat readers sometimes generate inferenceseven when they possess little domain-specificknowledge, indicating that inference genera-

    tion depends not only on the read er's knowl-edge level, but also on other factors. Likely

    did i l d di h i l

    Rich domain-specific knowledge alsointeracts in interesting ways with text-basedfactors. McNamara and colleagues (e.g.,McNamara, 2001; McNamara et al., 1996;McNamara & Kintsch, 1996) investigatedthe interactions of text cohesion and priordomain-specific knowledge in text compre-hension by manipulating text cohesion andmeasuring the effects on comprehensionby low- and high-knowledge readers atthe middle-school and college levels. Inter-estingly, whereas low-knowledge readersdemonstrated better comprehension whenreading high-cohesion texts, high-knowledgereaders experienced better comprehensionwhen reading low-cohesion texts. One inter-pretation of this finding is that the redun-dancy between information in high-cohesiontexts and background knowledge led high-knowledge readers to engage in passiveprocessing, thus keeping their comprehen-sion at a surface level (see also Gilabert,Martinez, & Vidal-Abarca, 2005). According

    to this explanation, increased text-basedcohesion may interfere with high-know^ledgereaders' ability to actively process texts bydecreasing the need for spontaneously gen-erated inferences. In contrast, low-cohesiontexts require readers to generate manyinferences. Thus, high-knowledge readersare induced by low-cohesion text to workmore actively to integrate information in thetext with their prior know^ledge. This active

    processing allows high-know^ledge readersto achieve deeper levels of comprehension.However, low-knowledge readers do notpossess the knowledge necessary to generatethe inferences required by low-cohesion text.As a consequence, low-know^ledge readersare disadvantaged by a combination oflow text-cohesion and low domain-specificknowledge.

    There is also evidence that readers withhigh reading ability (as measured with theNelson-Denny test; Brown, Fishco, & Hanna,

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    6/20

    70 TO PIC S IN LANGUAGE DisoRDERS,(fANUARY-MARCH 2005

    that only high-knowledge readers with lowscores on the Nelson-Denny Reading Testbenefited from reading low cohesion texts. Incontrast, high-knowledge readers with higherscores on the Nelson-Denny Reading Testbenefited from reading high-cohesion texts.This pattern of results indicates that whereashigh-knowledge readers with lower readingabilities tend to process texts actively onlywhen the low-cohesion version of the textdemands the active generation of inferences,high-know^ledge readers with good readingabilities tend to actively process texts regard-less of the text cohesion level or task de-mands. This research suggests that how ac-tively a reader processes text by generatinginferences may be influenced not only by thereader's domain-specific knowledge but alsoby an ability to use knowledge actively andstrategically when reading science texts. Inthe ne xt section, we conside r the role of read-ing strategy use in generation of inferencesthat facilitate deep-level comprehension ofscience texts.

    Con tributions o f reading strategiesto inferencing

    Solely having domain-specific knowledgeis often not enough; to be used effectively,read ers mu st deliberately activate their knowl-edge or it may rem ain somew^hat inert andfail to contribute to deeper understanding.That is, students may have some domain and

    general knowledge that can be used to gen-erate inferences and thus understand a textat a deeper level, but they may not real-ize that their knowledge is applicable to aparticular reading situation. Indeed, severalresearchers have found that readers some-times encounter difficulty invoking preex-isting knowledge in novel situations (e.g.,Bransford, 1979; Ha sher & Zacks, 1979;Nitsch, 1977).

    This situation is w he re reading strategy con-tributions come into play. Competency in

    di id ddi i l

    erat ion (and hence, deep comprehension) isbolstered by several sources of evidence. Ac-cording to McNamara and Scott (2001), theuse of reading strategies, such as the care-ful monitoring of text contents at the time ofreading, is an important component of read-ing comprehension ability. Good comprehen-ders (e.g., as assessed using the Nelson-DennyReading Co mp rehe nsion Test) are more likelyto generate inferences that repair concep-tual gaps between clauses, sentences, andparaphrases (Cain & Oakhill, 1999; Long,Op py, & Seely, 1994 ; Magliano & Millis, 2003;Magliano, Wiem er-Hastings, M illis, Munoz , &McNamara, 2002; OakhiU, 1982,1984; Oakhill& Yuill, 1996). Furthermore, good compre-hend ers have more metacognit ive knowledge(Baker, 1982; Wong, 1985) and a re mo re likelyto use reading strategies to repair gaps intheir understanding than poor comprehen-ders (e.g.. Garner, 1987; Long & Golding,1993; Oakhill, 1982, 1983). Finally, interven-tions that promote the active and strategicuse of knowledg e has be en show^n to improv ereading com pre hen sion (e.g., Bereiter & Bird,1985; Bulgren, Deshler, Schum aker, & Lenz,2000; Chi, de Leeuw, Chiu, & La Vancher,1994; Cornoldi & Oakhill, 1996; Kucan &Beck, 1997; McNamara, 2004, McNamara& Scott, 1999; O'Reilly, Best, & McNamara,2004; O'Reilly, Sinclair, & McNamara, 2004a,2004b). Collectively, these studies indicatethat reading strategies are important to suc-cessful comprehension, and that readerscan be taught strategies to improve theircomprehension.

    CHALLENGES OF COMPREHENDINGSCIENCE TEXTS

    Students of all ages have been foundto experience difficulty comprehending andlearning from science texts (Brand-Gruwel,

    Aarn outse, & Van den Bos, 1998; Nichols,Rupley, & Willson, 199 7). The p rob lem s w ith

    i b k h i h

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    7/20

    Deep-Level Com prehension of Science Texts 71

    Science texts contain difficult vocabulary andsyntax, and also place greater emphasis oninferential thinking and the use of priorknowledge (Allington, 2002). The increasedexposure to challenging expository materi-als, including science textbooks, at a timewhen domain knowledge is still developingplaces greater cognitive demands on youngreaders, and may account for some of the read-ing comprehension difficulties experiencedby children in the third to fifth grades, whichis sometimes called the fourth-grade slump(McNamara, Floyd, Best, & Louwerse, 2004;Meichenbaum & Biemiller, 1998; Sweet &Snow, 2003).

    In this section of the article, we discuss fac-tors that account for why the comprehensionof science text is particularly difficult for manystudents. We focus on three factors of read-ing comprehension identified earlier in thediscussion (text quality, background knowl-edge, and reading skill). Our goal is to providea detailed picture of the nature of problems

    commonly experienced by students in thisparticular curricular area to assist clinicians indevising possible rem edies.

    KNOWLEDGE DEFICITS AN DMISCONCEPTIONS

    As noted previously, the CI model of textcomprehension emphasizes domain-specificknowledge as an important factor driving text

    comprehension. Further, a number of studiesindicate that knowledge is, arguably, the mostimportant factor determining expository textcom prehension (e.g., Afflerbach, 1986; Chi,Feltovich, & Glaser, 1981; Chiesi, Spilich, &Voss, 1979; Lundeberg, 1987; Means Voss,1985). According to this view, it is inevitablethat many students experience significant dif-ficulty comprehending expository texts and,in particular, texts dealing with scientific con-

    cepts because their existing know ledge is lim-ited. Science textbooks, along with other ex-

    i d i h l f

    no background know ledge about the domain-specific expository text content when theyfirst encounter it.

    Students' knowledge deficits may take, atleast, two different forms. First, many studen tsmay lack the knowledge of specific conceptsoutlined by the text (e.g., osmosis, gravity,etc.). As our earlier plant example indicated,w^hen students do no t have a sufficient under-standing of a particular concept, they oftenhave a problem generating inferences to linkconcepts w ithin or across sentences. As a re-sult, their understanding of the text remainsfragmented and isolated, causing a failure toform a coherent mental representation of theoverall text content.

    Second, students' knowledge deficits maytake the form of preexisting misconcep-tions based on common knowledge or per-sonal experience, rather than scientific con-cepts. Understanding scientific phenomenaoften requires adopting a completely differentperspective from that acquired from every-

    day perceptual experiences. Indeed, Vygotsky(1978) pointed out the problem of integrat-ing spontaneously developed knowledge with scientific or theoretical concep ts, whichare abstract in nature and cannot be acquiredfrom direct perceptual experience. For exam-ple, children's direct perceptual experienceof the m ovement of the sun in the sky can cre-ate the misunderstanding that the sun movesacross the sky. This perceptual experience

    can make it difficult for children to under-stand that the sun moves as a function of theearth rotating and revolving around the sun,and not the sun's movement around the earth.

    Children with misconceptions about themovement of the sun may fail to compre-hend a text about the solar system accuratelybecause they cannot reconcile informationstated in the text with preexisting backgroundknowledge gained from everyday perceptualexperiences. According to the CI model,the utilization of preexisting misconceptions( b k d k l d )

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    8/20

    72 TOPICS IN LANGLIAGE DISORDERS/JANUARY-MARCH 2005

    inferences), which contrasts with the scien-tific model of the situation intended by theauthors of the text. Such misconceptions maygive rise to what Piaget (1985) describedas the assimilation of the incoming infor-mation to the preexisting knowledge struc-ture, instead of accommodation of the stu-dent's current knowledge structure to thetext. Thus, for students with misconceptions,deep, or accurate, text comprehension re-quires not only accessing and using priorbackground knowledge, but also recognizingmisconceptions, or contradictions betweentheir prior knowledge and the text content.Perhaps m ost importantly, successful compre-hension requ ires that readers subsequently re-pair any erroneous aspects of their mentalmodels (Chi et al., 1994).

    TEXT COHE SION INFLUENCESREVISITED

    As discussed earlier, several studies haveshown text with high cohesion to be partic-

    ularly beneficial to readers with less knowl-edge about the text domain (Beck et al., 1991;Britton et al., 1993; Britton & Gulgoz, 1991;McNamara, 2001; McNamara et al., 1996;Vidal-Abarca, Martinez, & Gilabert, 2000). Itis important to bear in mind that there is notan optimal level of explicitness or cohesionfor all readers. Rather, comprehension suc-cess also depends on the knowledge level andreading strategies that the reader brings into

    the reading situation. As mentioned earlier,high-cohesion text is not necessarily helpfulto high-know^ledge readers w ith a low level ofcom petency in reading strategy use. It is theo-rized that this reverse cohesion effect occursbecause high-knowledge readers are morelikely to actively engage with and compre-hend texts that contain fewer cohesion cues.However, many, if not most, students lack thenecessary domain-specific knowledge to gen-

    erate inferences when reading their sciencetextbooks, particularly if the books are w^rit-ten with low cohesion

    Moreover, there is reason to believe that theselection of science texts may be carriedout by textbook selection committees with-out sufficient attention to all of the textualfeatures that influence the understandabUityof the texts. First, texts are often classifiedand selected on the basis of traditional read-ability formulas that rely on simple indicessuch as word frequency, word length, andsentence length (Beck et al., 1991; Brittonet al., 1993). According to these formulas,texts that comprise short words and sen-tences are considered less difficult, and thusmore suitable for novice readers. Short wordstend to be m ore frequently used and encoun-tered, and thus are considered to be de factomore familiar (Zipf, 1932). Short sentencesplace fewer processing dem ands pertaining tolower level cognitive process (e.g., syntacticprocessing), but short sentences do not nec-essarily facilitate deep-level comprehensionbecause they offer fewer cohesion supportsfor low-knowledge readers. That is, texts char-acterized by short w ords and short sentencesare likely to lack adequate cohesion becausethe sentences are not likely to contain con-nectives. In addition, a focus on readability interms of short words and sentences is likelyto diminish the focus on elements of cohe-sion such as referential and explanatory cohe-sion. The lack of cohesion supp orts built intothe text places other demands on the reader,such as the need to make inferences about

    the nature of the links within and betweensentences, which will tend to interfere withhigher level processing. In sum, a reliance onreadability formulas is likely to result in a lackof attention to other textual properties thathave an important bearing on deep-level textcomprehension.

    A second complicating factor is that text-books are written by experts on the topic.Experts' high-level knowledge of a subject

    matter can interfere w^ith writing for lessknowledgeable readers such as school stu-d t (B itt t l 1993) A l b d f

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    9/20

    Deep-Level Comprehension of Science Texts 73

    1999). In other words, vast domain knowl-edge, which is obvious and natural to domainexperts, can lead them to assume an unrealis-tic amount of prior knowledge on the part ofreaders (Beck et al., 1991). As a consequence,textbooks often lack sufficient background in-formation for non-domain-expert readers tounderstand the con tent, impeding the forma-tion of a coherent situation model in thesereaders' minds.

    READING STRATEGY EVELUENCESREVISITED

    As reported earlier, the use of strategic read-ing practices, such as the active generation ofinferences using backward causal inferencing,plays a critical role in th e deep-level compre-hension of science texts. However, there arereasons to believe that it is difficult to attainsufficient com petency in reading strategy useby the time children begin to read challeng-ing science tex ts. We discuss two possible rea-sons why students struggle to acquire and ap-

    ply reading strategies.First, students, particularly those at theelementary school level, may not have mas-tered efficient and accurate lower level com-prehension skills (e.g., efficient word decod-ing), which are necessary for the execution ofhigher level comprehension processes (e.g.,strategic reading practices, metacomprehen-sion, backward causal inferences). Slow orinaccurate word recognition skills may af-

    fect comprehension by consuming too muchworking memory capacity, thus constrainingresources that could be used for deep compre-hension processes involving inference gener-ation (Cain, Oakhill, & Bryant, 2004; Curtis,1980; Hannon & Daneman, 2001; Perfetti,1985). Many studies have indicated that chil-dren may not have mastered word decodingskills by the time they are introduced to ex-pository materials (Brand- Gruwel, Aarnoutse,

    Van den Bos, 1988; Mommers, 1987; Roth,Speece, Cooper, De La Paz, 1996; Taschow,1969 V ll ti 2003) d th tUl t g

    Second, even if students have acquired suf-ficient word decoding ability, they still maynot have attained sufficient com petency in us-ing reading strategies necessary for success-ful comprehension and learning from sciencetexts. For many students, strategic readingpractices needed to support inference gen-eration are unlikely to develop automaticallyjust by virtue of reading narrative texts inclassroom reading activities. Instead, knowl-edge of reading strategies and effective use ofthese strategies in comprehending challeng-ing expository texts may need to be explicitlytaught.

    A report published by Educational TestingService (ETS, 2004) on current fourth-gradereading instruction indicated that teachers doprovide students with instructions on tech-niques such as predicting con tents of thematerial they are reading and making gen-eralizations and inferences about reading con-tents (p. 24). However, there is little evidencethat deep-level reading strategies are explic-

    itly taught as a part of the normal curricu-lum. It seems tha t the majority of reading com-prehension training administered at schoolto children focuses on lower level readingskills such as efficient word decoding skills(Mommers, 1987; Nichols et al., 1997; Press-ley & Wharton-McDonald, 1997; Roth et al.,1996; Wilson & Rupley, 1997). It is a con-cern if such instruction occurs in the absence,or at the expense, of the explicit teaching of

    higher level processes, such as strategic read-ing practices. Investigations regarding the useof metacognitive strategies, in fact, have in-dicated that teachers spend little time givingmetacognitive and strategy-oriented instruc-tion pertaining to reading comprehension(Baker, 1996; Graesser, Person, & Magliano,1995; Kurtz, Schneider, Carr, Borkowski, &Rellinger, 1990; Moley et al, 1992). Further-more, reading problems for children who

    have learned higher level reading skills suchas inference generation strategies in the con-t t f t t ith f ili ti t t

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    10/20

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    11/20

    Deep-Level Com prehension of Science Texts 75

    adjacent and distant sentences using La-tent Semantic Analysis (Landauer & Dumais,1997), which was developed by researchers atthe University of Colorado. As discussed ear-lier, the number of connectives contained ina text is an important measure of cohesion.Connectives provide explicit cues for readerswhen generating inferences. Similarly, argu-ment overlap, which is measured by the de-gree to which adjacent sentences share con-tent words, is known to be related to localcohesion because having lexical overlap helpsreaders relate sentences and construct highertext level representation. However, simpleword matching may not provide a compre-hensive measure of overlap.

    As described by Landauer and Dumais(1997), Latent Semantic Analysis (LSA) bol-sters argument overlap analysis by comput-ing the co-occurrence of two or more Un-guistic units (w^ords, phrases, sentence , etc.)with respect to a given reference point w ithinthe LSA space. The LSA space is a mathemat-

    ical, dimensional space for plotting vectorsderived from the linguistic units (e.g., sen-tences) being compared. Landauer and Du-mais showed that LSA betw^een two linguisticunits correlates w ith th e conceptual similarityamong linguistic units, not just the identicalwords each unit possesses. The LSA is there-fore useful for measuring th e approxim ate co-hesion between two sentences in terms ofthe overlap of conceptually related w^ords, not

    just the words themselves. Such a concep tualoverlap between words would not be cap-tured in a simple analysis of argument overlap(i.e., word matching).

    Ultimately, the Coh-Metrix tool is designedto be used by researchers and educationalpractitioners to measure aspects of texts thatcontribute to reading comprehension diffi-culties. Although the tool does not yet in-clude established criteria for distinguishing

    low-cohesion texts from high-cohesion texts,it does offer information about differences inl l f h i R h d

    ers with more or less domain know ledge. ACoh-Metrix analysis of texts from publishedstudies of text cohesion (McNamara, Ozuru,Louwerse, & Graesser, 2005) indicates thatthe tool reliably captures critical aspects oftext manipulations (argument overlap, in-crease of explanatory cohesion) such as thoseused in experimental research of text revision(e.g.. Beck et al., 1991; McNamara et al., 1996;Voss Silfies, 1996).

    Overall, we believe that research investi-gating text features will provide a more sys-tematic and reliable method of measuring andclassifying text difficulty. Furthermore, new^measures of text difficulty should offer animportant alternative to the simplistic mea-sures currently used (e.g., reading grade levelformulas). Consequently, tools such as Coh-Metrix are anticipated to positively contributeto the process of assigning more appropri-ate texts to students who may lack sufficientbackground knowledge.

    Reader focused interventions: Teachingreading strategies

    If problems that students encounter under-standing their science textbooks stem largelyfrom deficits in, or erroneous, backgroundknowledge, it is important to find reme-dies that compensate for poor knowledgeby teaching them comprehension strategies.Even if a tool such as Coh-Metrix (Graesseret al., 2004) becom es capable of providing in-

    formation that helps improve text cohesion,it is impractical or even impossible to pro-vide all the cohesion cues and background in-formation necessary for low-knowledge read-ers. In addition, such an overly cohesive textmight interfere with the active processing thatis necessary for deep-level com prehension byhigher ability readers (e.g., McNamara et al.,1996). Thus, reading comprehension difficul-ties relating to the comprehension of science

    texts need to be approached not only from theperspective of the text, but also from the per-

    ti f th d

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    12/20

    76 TO PIC S IN LANGUAGE DisoRDERS/jANtJARY-MARCH 2005

    learning new information. Thus, most read-ers appro ach th e task of reading science textswith low levels of topic-relevant knowledge.The circularity of this phenomenon is suchthat readers cannot comp rehe nd th e text con-tents at a deep level without learning newconcepts or information from the text. Thiscontrasts w^ith the reading process of narra-t ive texts, w her e com prehen sion is supportedby sufficient knowledge about the content sothat inferences necessary for deep-level com-prehension may be generated relatively auto-matically. The comprehension processes nec-essary for science texts, therefore, requiregreater effort because the information rele-vant to understand ing a given senten ce, or re-lations between sentences, is not often read-ily accessible in long-term memory. Readersoften need to search for relevant informa-tion through various forms of linking (e.g.,association, analogy). Effortful generation ofinfer enc es is psyc hologic ally different fromthe inferences that readers generate to under-stand familiar narrative materials (see Kintsch,1993).

    For this reason, not only readers, but alsoinstructors, need to be aware of the nature ofthe difficulties associated with reading com-prehension and learning from science texts.They also need to actively deal with thesepro ble m s by (1) explicitly discussing w ith stu-dents the nature of comprehension problemspertaining to science texts and (2) teaching

    and using techniques that help students withlimited prior topic knowledge deal effectivelyand strategically with the challenges of sci-ence text comprehension.

    Numerous reading strategy interventionshave been d evelop ed and tested. There is con-verging evidence to show that the provisionof explicit reading strategy training. In whichreade rs are taught and trained to actively pro-cess a text using specific re ading strategies, is

    effective for facilitating deeper level compre-hensio n of texts. These interven tion pro gram s

    f f f

    adolescents with comprehension difficultiesand language disorders (Bulgren et al., 2000Deshler, & De nton , 1984; Fisher, Schumak er,& Deshler, 200 2).

    Our aim here is not to provide an exhaus-tive list of interventions, but to show the waysin which reading strategies can help studentsovercome knowledge deficits. We focus pre-dominantly on strategy training developed inour laboratory at the University of Memphisfor typically developing students from themiddle-school level and be yond (i.e., stude ntsmost frequently exposed to science texts).Generally, our efforts have concentrated onteaching reading strategies to help typical stu-dents who lack topic-relevant knowledge forcomprehending science texts. In the remain-der of this section, we first describe four spe-cific reading strategies that are known to con-tribute to active processing of science texts,and then describe Self-Explanation ReadingTraining, a reading strategy training interven-tion program developed in our lab that pro-

    vides instruction and practice on the use ofthese specific strategies.

    Comprehension monitoring

    Students' ability to mo nitor com pre hen sioncritically and accurately is the foundation ofstrategic and active reading (Hacker, 1998).That is, readers need to be aware of notonly w he ther they are having com prehen sionproblems, but also the nature of the problem

    they encounter (e.g. , word meaning, syntax,or relat ions between sentences). However,there is evidence that most readers are ratherpoor at monitoring their own comprehen-sion (e.g., Glenberg, Wilkinson, & Epstein,1982). Thus, providing training to improvecomprehension monitoring is important topro m ote th e active and strategic processing ofexpository texts. Specific methods of teach-ing comprehension monitoring include self-

    questioning and checking texts for contentconsistency. Research conducted with ele-

    h l d h i di d h

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    13/20

    Deep-Level Com prehension of Science Texts 11

    about the material and to check the text forcontent consistency (Baker, 1985).

    Paraphrasing

    Paraphrasing is an important techniquefor facilitating the active processing of texts(Rosenshine & Meisler, 1994). Paraphrasingrequires transforming the surface characteris-tics of the sentence by replacing the contentwords or syntactic structure of the sentencewith equivalent forms, hence forcing readersto process the information actively by access-ing related but different lexical items. Para-

    phrasing also externalizes one's understatingof the information in the text, which , in turn,helps readers monitor comprehension moreclosely. Difficulty when paraphrasing is a clearsign of a comprehension problem.

    Elaboration

    Generating elaborative inferences based onthe reader's personal experience or common-sense knowledge relating to information de-scribed by the text is assumed to be use-ful for facilitating deep-level comprehension(McNamara, 2004). For one thing, elabora-tive inferences help readers overcome gapsin domain-specific knowledge. As our discus-sion has indicated, understanding texts at adeep level requires the reader to generate in-ferences to fill n gaps present in text-based in-formation, such that information distributedacross text can be encoded as coherent andin an integrated m anner. Also, the use of priorknowledge is important for helping the readerretain novel information learned from text inmemory; new^ information canno t be learnedand retained without being integrated withprior know ledge (Pressley et al., 1992). Thus,when readers do not have sufficient domainknowledge, they need to forcefully integratethe new information in the text with exist-ing knowledge by forming a link be tween thenew information and indirectly related gen-eral knowledge or personal experience. As

    problem can be minimized if readers com bineelaborations based on general knowledge andpersonal experience with inferences using in-

    formation cited in the prior sections of thetext.

    Bridging

    The process of generating inferences usinginformation stated in previous sections of thetext plays an integral role in helping readersbuild a global represen tation o fthe text. Morecohesive texts provide a stifficient amount ofbackground information for readers to build

    a global understanding of the overall text bycontinuously adding and integrating newlyintroduced information with previouslycited information (Clark & Haviland, 1977;Gernsbacher & Hargreaves, 1988). Althoughmany school texts may not contain sufficientbackground information and the cohesioncues that help readers link informationpresented in different sentences in an ap-propriate manner (e.g.. Beck et al., 1991),

    students should be taught to maximize theuse of information provided within a textto understand the meaning of individualsentences as well as the overall meaning ofthe text. When struggling to disambiguateand/or narrow down the meaning of a givensentence, readers need to learn that informa-tion stated in the previous section can be apowerful and reliable source of informationthat can aid their comprehension.

    Self-Explanation ReadingTraining SERT)

    In summary, we postulate that the combi-nation of the above-mentioned reading strate-gies constitute a powerful combined tool forhelping readers comprehend and learn fromexpository texts. The essence ofthe approachis to teach readers to compensate for bothknowledge and text deficits by learning todraw deeper inferences based on the active,constructive processing of connections that

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    14/20

    78 TOPICS IN LANGUAGE DISORDERS/JANUARY-MARCH 2005

    active and strategic processing of sciencetexts by combining the above specific read-ing strategies through a method called self-

    explanation . Self-explanation refers to the pro-cess of explaining text contents to oneselfwhile reading (Chi, Bassok, Lewis, Reimann,& Glazer, 1989). There are a number of ben-efits to using the self-explanation method, in-cluding facilitating inference generation andrepairing erroneous mental models by in-creasing the extent to which the reader mon-itors the ongoing comprehension process(Chi, de Leeuw, Chiu, LaVancher, 1994).

    Our program, called Self-ExplanationReading Training (SERT; McNamara, 2004;McNamara Scott, 1999) uses the self-explanation technique to teach effective read-ing comprehension strategies (monitoringcomprehension, paraphrasing, elaborative in-ferences, and generating bridging inferences).SERT is divided into three components: (1)Introduction, in which students are taughtthe reading comprehension strategies; (2)

    Demonstration, in which the strategies aredemonstrated to students; and (3) Practice,in which students practice self-explainingscience texts using the reading strategies. Inthis way, our program aims to promote bothexplicit understanding of reading strategies(declarative knowledge) and the skills of ap-plying the strategies w^hUe reading challeng-ing science texts (procedural knowledge).

    Thus far, research conducted with studentsin middle-school, high-school, and college lev-els has indicated that SERT increases self-explanation quality and subsequent compre-hension of science texts. Eor example, a studyconducted with 38 middle-school students(O'Reilly McNamara, 2004) demonstratedthat students trained with SERT performedbetter on a comprehension assessment thanuntrained students. In a study conducted with42 college undergraduates, McNamara (2004)found that SERT improved the quality ofself-explanations (as reflected in students' in-

    O'Reilly, Best, and McNamara (2004) foundthat SERT improved science text compre-hension among low-know^ledge high-school

    readers. In this study, low-knowledge readerstrained by SERT performed better on a com-prehension task than low-knowledge readerstrained using a different form of reading strat-egy training (Previewing) or low-knowledgereaders assigned to a control condition.

    Current research in our lab has begun to fo-cus more closely on the effects of individualdifferences in reading strategy intervention,with the aim being to build a student-

    adaptive reading strategy tutor called Interac-tive Strategy Training for Active Reading andThinking (iSTART; McNamara, Levinstein, Boonthum, 2004). In its prototype, iSTARTis an automated, interactive tutor, which cur-rently incorporates SERT but is being ex-panded to include other reading strategies.The current iSTART system continuously eval-uates what students know about the strategiesand their ability to use the strategies so as to

    tailor scaffolding and feedback to the level ofthe student. Eor example, students strugglingto produce elaborations are encouraged to add more details to their self-explanations.Evaluations of iSTART indicate that the sys-tem improves the students' use of readingstrategies and comprehension of science textsat both the middle-school and college level(Magliano et al., 2005; O'Reilly et al., 2004a,2004b). Our future research will continue to

    explore the relations between individual dif-ferences and the benefits of reading strategytraining to develop reading strategy interven-tions that are increasingly adaptive to each stu-dent's needs.

    Preliminary findings suggest that auto-mated versions of reading strategy trainers areuseful not only for individual intervention butalso for assisting teachers to implement strat-egy training in educational settings such asthe classroom. One of the factors impedingstudents' ability to acquire and practice effec-

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    15/20

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    16/20

    8 0 T O P I C S I N LANGUAGE DisoRDERS/jANtjARY-MARCH 2 0 0 5

    Enyedy, A. S. Nixon, & E Herre ra (Eds.), Proceedings ofthe sixth intemationai conference ofthe learning sci-ences: Embracing diversity in the learning sciences(pp, 89-96), Mahwah, NJ: Eribaum.

    Bowen, B. A, ( t9 99 ). Four puzzles in adult literacy: Re-flections on the national adult literacy survey. Journalof Adolescent and Adult Literacy, 42, 314-323 .

    Brand-Gruwel, S., Aarno utse, C, A, J,, & Van den Bos, K, P,(1998). Improving text comprehension strategies inreading and listening settings. Learning and Lnstruc-tion, 8, 6 3 - 8 1 .

    Bransford, J, (1979). Hum an cognition: Learning,understanding, and remembering. Belmont, CA:Wadsworth,

    Bransford, J, D., Vye, N. J,, Adams, L. T, Perfetto,G, A, (1989), Learning skills and the acquisition ofknowledge. In R, Glaser & A. Lesgold (Eds,), Founda-tions for a psychology of education (pp, 199-249) .Hillsdale, NJ: Eribaum .

    Britton, B, K., Gulgoz, S, (1991). Using Kintsch's com-putational m odel to improv e instructional text: Effectsof repairing inference calls on recall and cognitivestructures. Journal of Educational Psychology, 83,329-345 ,

    Britton, B, K,, Gulgoz, S,, Glynn, S, (1993). Im pact ofgood and poo r writing on learners: Research and the-ory. In B. K, Britton, A, Woodward, M, R. Binkley(Eds,), Leamingfrom textbooks: Theory and practice(pp, 1-46), Hillsdale, NJ: Eribaum,

    Brown, J,, Fishco, V, Hanna, G, (1993). Manual forscoring and interpretation. Forms G H. Chicago:Riverside Press,

    Bulgren, J, A., Deshler, D, D,, Schumaker, J, B,, Lenz,B, K, (2000), The use and effectiveness of analogicalinstruction in diverse secondary content classrooms.Journal of Educationai Psychology, 92, 426-441 ,

    Cain, K,, OakhiU, J, V (1999). Inference making abU-ity and its relation to compreh ension failure in youngchildren, Reading and Writing, 11, 489-503 ,

    Cain, K,, OakhiU, J,, Bryant, P, (2004), ChUdren's read-ing com prehen sion ability: Conc urrent pred iction byworkin g mem ory, verbal ability, and co mp on en t skills,Journai of Educationai Psychology, 96, 31-42,

    Chall, J,, Squire, J, R, (1991). The publishing industryand textbooks. In R, Barr, M, L, Kamil, P B, Mosen-thal, & P D, Pearson (Eds,), Handbook of Reading Re-search (Vol, 2, pp , 12 0-146 ), New York: Longman,

    Chi, M, T, H,, Bassok, M,, Lewis, M, W,, Reimann, P,& Glaser, R, (1989). Self-explanations: How s tudentsstudy and use examples in learning to solve problem s.Cognitive Science, 13, 145-182,

    Chi, M, T, H,, De Leeuw, N,, Chiu, M,, LaVancher,C, (1994), Eliciting self explanations improves under-standing. Cognitive Science, 18, 439-477,

    Chiesi, H, I,, Spilich, G, J,, & Voss, J, E (197 9). A cquisitionof domain related information in relation to high andlow domain knowledge , /oMm a/ of Verbal Learningand Verbai Behavior, 18, 275-290,

    Clark, H,, & Haviland, S, (1977), Com prehension and th egiven new contr act. In R, Freedle (ed,), Discoursepro-duction and comprehension (pp , 1-40), H illsdale, NJ:Erlbatun,

    Cornoldi, C , Oakhill, J, (1996). In C, Cornoldi J,Oakhill (Eds,), Reading comprehension difficulties.HUlsdale, NJ: Eribaum ,

    Cote, N,, Goldman, S, R,, Saul, E, U, (1998), Studentsmaking sense of informational text: Relations betweenprocessing and representation. Discourse Processes,25, 1-53.

    Curtis, M, E, (1980 ), Dev elopm ent of com po ne nts of read-

    ing skill ,/oMm / of Educational Psychology, 90,294-311.

    Educational Testing Service, (2004), The fourth-gradereading classroom. Princeton, NJ: Policy Evaluationand Research Center,

    Engle, R,, Marshall, K, (1983), Do developmentalchanges in digit span result from acquisition strate-gies?/oMr/ of Experimental Child Psychology, 36,429-436,

    Fisher, J, B,, Schum aker, J, B ,, & Deshler, D, D, (200 2), Im-proving th e reading comprehension of at-risk adoles-cents . In C, C, Block & M, Pressley (Eds,), Comprehen-sion instruction: Research-based best practices, (pp,351-364) New York: Guildford Press.

    Garner, R. (1987), Metacognition and reading compre-hension. Norw ood, NJ: Ablex,

    Gernsbacher, M, A, (1997), Coherence cues mappingduring comprehension. In J , Co stermans M, Fayol(Eds,), Processing interclausal relationships in theproduction and comprehension of text (pp, 3-21) ,Hillsdale, NJ: Eribaum,

    Gernsbacher, M, A,, Hargreaves, D, (1988), Accessingsentenc e participants: The advantage of first men tion.Journal of Mem ory and Language, 27, 699-717 ,

    Gilabert, R,, Martinez, G,, Vidal-Abarca, E, (2005).Some texts are aiways better: Text revision tofoster inferences of readers with high and lowbackground knowledge. Manuscript submitted forpublication,

    Glenb erg, A, M,, Wilkinson, A, C, Epstein, W (1982),The illusion of knowin g: Failure in the self assessm entoicotaptthension. Memory and Cognition, 10, 597-602,

    Graesser, A, C, McNamara, D, S,, Louwerse, M, M,, & Cai,Z, (2004), Coh-Metrix: Analysis of text on cohesion

    and language, Behaviorai Research Methods, Instru-ments and Computers, 36, 193-202,Graesser, A. C, Person, N, K,, MagUano, J, P (1995),

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    17/20

    Deep-Level Com prehension of Science Texts 81

    Constructing inferences during narrative text compre-hension./'yvc*o/ogfcfl/J?ef

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    18/20

    82 TOPICS IN LANGUAGE DISORDERS/JANUARY-M ARCH 2005

    structures. Journal of Memory and Language, 24,

    746-757.

    Meichenbaum, D., Biemiller, A. (1998). Nurturing

    independent learners: Helping students take charge

    of their learning. Cambridge, MA: Brookline Books.Moiey, B. E., Hart, S. S., Leal, L., SantuUi, K., Rao, N.,

    Johnson, T, et al. (1992). The teacher's role in facili-

    tating memory and study strategy development in the

    elementary school classroom. Child Development, 63,

    653-672.

    Mommers, M. J. (1987). An investigation into the relationbetween word recognition skills, reading comprehen -

    sion and spelling skills in the first two years of primary

    school. Journal of Research in Reading, 10, 122-

    143.Nelson, N. W. (1998). Childhood language disorders

    in context: Infancy through adolescence (2nd ed.).

    Boston: Allyn and Bacon.

    Nichols, W. D., Rupley, W; H., Willson, V L. (1997).A cross sectional comparison of the relationship

    between decoding components, conceptual knowl-

    edge, and metacognitive knowledge to reading com-

    prehension for readers in grades four and five

    (ERIC Document Reproduction Service No. 407 670).

    Williamsburg, KY.

    Nickerson, R. S. (1999). How we knowand sometimes

    misjudgewhat others know: Imputing one's own

    knowledge to others. Psychological Bulletin, 125,

    Nitsch, K. E. (1977). Structuring decontextualizedforms

    of knowledge. Unpublished doctoral dissertation, Van-

    derbilt University.

    Noordman, L. G., Vonk, W., Kempff, H. J. (1992). Causalinferences during the reading of expository texts.

    Journal of Memory and Language, 31, 573-590.

    Oakhill, J. (1982). Constructive processes in skilled and

    less skilled comprehenders' memory for sentences.

    Britishjoumai of Psychoiogy, 73, 13-20.

    Oakhill, J. (1983). Instantiation in skilled and less skilled

    comprehenders. Quarterly Journal of ExperimentalPsychology, 35, 441-450.

    Oakhill, J. (1984). Inferential and memory skills in chil-

    dren's comprehension of stories. British Journal of

    Educationai Psychology, 54, 31-39.

    OakhiU, J., Patel, S. (1991). Can imagery training help

    children who have comprehension problems?/oM>7ia/

    of Research in Reading, 14, 106-115.

    Oakhill,J., YuilI, N. (1996). Higher orde r factors in com-

    prehension disability: Processes and remediation. In

    C. Cornaldi J. Oakhill (Eds.), Reading comprehen-

    sion difficulties: Processes and Intervention (pp. 69-

    72). Mahwah, NJ: Eribaum.O'Reilly, T, Best, R., McNamara, D. S. (2004) . Self ex-

    planation reading training: Effects for low knowledge

    O'Reilly, T, McNamara, D. S. (2002). What 's a science

    student to do? In Proceedings of the Twenty-fourth

    Annual Meeting ofthe Cognitive Science Society (pp.

    726-731). Mahwah, NJ: Eribaum.

    O'Reilly, T, McNamara, D. S. (2004). Good texts can bebetter for skilled comprehenders./oMma/ of Educa-

    tional Psychology.

    O'Reilly, T. P, Sinclair, G. P, McNamara, D. S. (2004a).

    iSTART: A web-based reading strategy inte rvention

    that improves students' science comprehension. In

    Kinshuk, D. G. Sampson, P Isaias (Eds ), Proceed-

    ings of the IADIS Intemationai Conference Cog-

    nition and Exploratory Learning in Digital Age:

    CELDA 2004 (p p. 173-180) . Lisbon, Portugal: IADIS

    Press.

    O'Reilly, T. P, Sinclair, G. P, McNamara, D. S. (2004b).

    Reading strategy training: Automated versus live. In K.Eorbus, D. Gentner, T. Regier (Eds.), Proceed ings ofthe 26th Annual Meeting of the Cognitive Science So-

    ciety (pp . 1059-1064) . Mahwah, NJ: Eribaum.

    Ozuru, Y, Best, R., McNamara, D. S. (2004). Contribu-

    tion of reading skill to learning from expository text s.

    In K. Forbus, D. Gentner, T. Regier (Eds.), Proceed-

    ings ofthe 26th Annuai Meeting ofthe Cognitive Sci-

    ence Society (pp. 1071-1076) . Mahwah, NJ: Eribaum.

    Perfetti, C. (1985). Reading ability. New York: Oxford

    University Press.

    Piaget, J. (1985). The equilibrium of cognitive struc-

    tures. Chicago: Chicago University Press.

    Pressely, G. M. (1976). Mental imagery helps eight year

    olds remember what they read. Journal of Educa-

    tional Psychology, 68, 355-359.

    Pressley, M., Wharton-McDonald, R. (1997). Skilled

    comprehension and its development through instruc-

    tion. School Psychology Review, 26, 448-466.

    Pressley, M., Wood, W. W, Woloshyn, V E., Martin, V,

    King, A., Menke, D. (1992). Encouraging mind-

    ful use of prior knowledge: Attempting to construct

    explanatory answers facilitates learning. Educationai

    Psychoiogist, 27, 91-100.

    Rosen, V, Engle, R. (1997). The role of work ing mem-

    ory capacity in retrieval./oMma/ of ExperimentaiPsy-

    chology: General, 126, 211-227.

    Rosensh ine, B., Meister, C. (1994). Cognitive strategy

    instruc tion in reading. In D. A. Hayes S. A. Stahl

    (Eds.), Instructional models in reading (pp. 85-107).

    Hillsdale, NJ: Eribaum.

    Roth, E P, Speece, D. L., Cooper, D. H., De La Paz, S.

    (1996). Unresolved mysteries: How do metalinguistic

    and narrative skills connect with early reading?/owr-

    nal of Special Education, 30, 257-277.

    Schumaker, J. B., Deshler, D. D., Demon, P. H. (1984).An integrated system for providing content to learn-

    ing disabled adolescen ts using an audio-taped format.

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    19/20

    Deep-Level Com prehension of Science Texts 83

    Singer, M., Harkness, D., & Stewart, S. T. (1997). Con-structing inferences in expository text comprehen-sion. Discourse Processes, 24, 199-228.

    Snow, C. (2002). Reading for understanding: Toward

    an R D program in reading com prehension. SantaMon ica, CA: RAND.

    Sweet, A. P., & Snow, C. E. (Eds.). (2003). Rethinkingreading comprehension. Ne w York: Guilford Press.

    Taschow, H. G. (1969). Reading improvement in mathe-matics. Reading Improvem ent, 6, 62 -67 .

    Veliutino, E R. (2003). Individual differences as sourcesof variability in reading comprehension in elementaryschool children. In A. P Sweet C. E. Sno w (Eds.), Re-thinking reading comprehension (pp. 51-81) . NewYork: Guilford Press.

    Vellutino, E R., Scanlon, D. M., & Tanzman, M. S. (1994).

    Components of reading ability: Issues and problemsin operationalizing word identification, phonologicalcoding, and orthographic coding. In G. R. Lyon (Ed.),Frames of reference for the assessment of learningdisabilities: New views on measuremen t issues ( p p .279-329). Baltimore: Paul H. Brookes.

    Vidal-Abarca, E., Martinez, G., & Gilabert, R. (2000). Twoprocedures to improve instructional text: Effects on

    memory and learning. Joumal of Educational Psy-chology, 92, 107-116.

    Voss, J., & Silfies, L. (19 96). Learning from histo ry text:The interaction of knowledge and comprehension

    skill with text structure. Cognition and Instruction,14 , 4 5 - 6 8 .

    Vygotsky, L. il97S). Mind in society. Cambridge, MA: MrTPress.

    Wilson, V L., & Rupley, W. H. (1997). A structural equa-tion model for reading compre hension based on back-ground, pho nem ic, and strategy knowledge. ScientificStudies of Reading, 1, 45 -63 .

    Wong, B. (1985). Metacognition and learning disabilities.In D. Forrest Pressley, G. MacKinnon, & T Waller(Eds.), Metacognition, cognition, and human perfor-mance (Vol. 2, pp . 13 7-18 0). New York: Academ icPress.

    Zipf, G. K. (1932). Selected studies o f the principle ofrelative frequency in language. Cam bridge , MA: Har-vard University Press.

    Zwaan, R. A., & Singer, M. (2003). Text comprehension.In A. C. Graesser, M.A. Gernsbaeher, & S. R. Gold-man (Eds.). Handbo ok of discourse processes (pp .83-1 21). Mahwah, NJ: Erlbaum.

  • 8/11/2019 Deep-level Comprehension of Scientific Text

    20/20