Is Memory Schematic? - Home - CSU, Chico

Psychological Bulletin1983. Vol. 93, No. 2, 203-231

Copyright 1983 by the American Psychological Association, Inc.0033-2909/83/9302-0203$00.75

Is Memory Schematic?

Joseph W. AlbaUniversity of Florida

Lynn HasherTemple University

This article proposes a prototypical schema theory of memory. Such a theoryassumes the operation of four central encoding-processes: selection—a processthat chooses only some of all incoming stimuli for representation; abstraction—a process that stores the meaning of a message without reference to the originalsyntactic and lexical content; interpretation—a process by which relevant priorknowledge is generated to aid comprehension; and integration—a process bywhich a single, holistic memory representation is formed from the products ofthe previous three operations. The article evaluates the supportive and criticalevidence for these processes in light of the need for any theory of memory toaccount for three fundamental observations: accuracy, incompleteness, and dis-tortion. The central retrieval process of schema theory, reconstruction, is alsodiscussed in this context. Evidence seems to indicate that the memory represen-tation is far richer and detailed than schema theory would suggest.

The scientific literature on memory con-firms the everyday observation that althoughone's memory for complex events is some-times startlingly accurate, it is also frequentlyincomplete and occasionally even highly dis-torted. Accounting for these observationsposes the fundamental problem for any the-ory of memory. For the past 15 years, themost popular theories (called collectivelyschema theories) have been descendents ofa view that emphasized the incompletenessand distortions characteristic of memory.This viewpoint was originally proposed byBartlett (1932) to account for recall of com-plex stories. Lineal descendents of the theorylargely share Bartlett's rejection of the notionthat memory representations consist of ac-curate traces that are stable over long dura-tions. Current proponents of this rejectedviewpoint emphasize the potential accuracyof memory; incompleteness in recall is at-tributed to retrieval failure, distortions toassociative encoding processes. In contrast,current schema theorists propose that what

We thank Robert Weisberg, Eugene Zechmeister, Mi-chael McCloskey, and especially Rose Zacks for manyhelpful comments; their contributions to this article wereextremely important. Preparation of this manuscript wassupported by NIMH Grant MH33140 and in part bya Biomedical Sciences Research grant to Temple Uni-versity.

Requests for reprints should be sent to Joseph Alba,Department of Marketing, 205 Matherly Hall, Univer-sity of Florida, Gainesville, Florida 32611.

is encoded, or stored in memory, is heavilydetermined by a guiding schema or knowl-edge framework that selects and activelymodifies experience in order to arrive at acoherent, unified, expectation-confirming andknowledge-consistent representation of anexperience. Such theories necessarily placeemphasis on the incompleteness and inac-curacy of memory, although they can ac-count for accuracy as well. The concern withinaccuracy and distortion rather than withaccuracy has made these theories especiallyattractive in fields such as clinical, forensic,educational, developmental, and social psy-chology (e.g., Ausubel, 1968; Greenwald,1980;Loftus, 1979a; Paris &Lindauer, 1977;Taylor & Crocker, 1981).

Although schema theories either guide orare used to explain a considerable portion ofcurrent research in human memory, it iswidely agreed (e.g., Brewer & Treyens, 1981;Taylor & Crocker, 1981) that the term schemahas no fixed definition. It is most often usedto refer to the general knowledge a personpossesses about a particular domain. Aschema allows for the encoding, storage, andretrieval of information related to that do-main. Beyond this, however, there are fewsystematic, contemporary explications ofschema theory.

Two exceptions to this question are nota-ble: frame theory and script theory. A frame(Minsky, 1975) is a schema that contains

203

204 JOSEPH W. ALBA AND LYNN HASHER

knowledge about the structure of a familiarevent, for example the knowledge a personpossesses about the structure of a short story.It specifies not the exact contents of the eventbut rather the general type of informationexpected in that situation and the order inwhich it should be encountered (for example,in a story, setting information would be fol-lowed by theme information and so on). Ascript (Schank & Abelson, 1977) is very sim-ilar to a frame in that it too contains generalinformation about particular, frequently ex-perienced events (e.g., a visit to a restaurant).Scripts also contain more specific informa-tion about the contents of the event as well(e.g., being seated, ordering the meal, and soon). Although frame and script theories haverecently been used to guide some research inmemory (e.g., Bower, Black, & Turner, 1979;Graesser, Woll, Kowalski, & Smith, 1980), byfar the majority of research is not aimed attesting particular provisions of such well ar-ticulated theories.

Instead, the bulk of the memory literaturedealing with complex events such as proselearning consists of a series of studies inves-tigating a number of loosely connected issueswhose results are either predicted by or in-terpreted with reference to schema theory orits originator, Bartlett (1932). This situationposes a substantial difficulty to anyone whowould evaluate the adequacy of schema theo-retics; it is difficult to evaluate a set of par-tially overlapping assumptions spread across15 years of research, a sizeable number ofparadigms, and the research of a large num-ber of investigators.1

Our first goal then was to try to imposesome structure on the research literature byidentifying the central assumptions of schematheories. We propose that according to amodal theory, schema-driven encoding ofcomplex information is characterized by fourbasic processes: selection, abstraction, inter-pretation, and integration, A schema theorywhich asserts that all four processes occurwould state that from any environmentalevent, only the information that is relevantand important to the currently activatedschema will be encoded. Of the informationselected, the semantic content of the messagewill be abstracted and the surface form willbe lost. Further, the semantic content will

then be interpreted in such a way as to beconsistent with the schema. The informationthat remains will then be integrated with pre-viously acquired, related information thatwas activated during the current encodingepisode. The operation of one or more ofthese processes is likely to result in a repre-sentation that is less than totally accurate.

A fifth schema theory process, reconstruc-tion, is one that operates at the time a personattempts to reproduce a memory episode.This process uses whatever details were se-lected for representation and are still acces-sible together with general knowledge to es-sentially fabricate what might have hap-pened. Several critical reviews of this processare available (e.g., Gomulicki, 1956; Spiro,1977; Zangwill, 1972). The consensus is thatreconstruction is quite rare and occurs onlyunder special circumstances. Because of this,we do not undertake a full-scale review ofreconstruction and instead focus on the en-coding processes proposed by schema theory.However, because of recently renewed inter-est in the process of reconstruction (see e.g.,Bower et al., 1979; Graesser et al., 1980;Spiro, 1980a), the topic is discussed in severalsections of this article.

In the first part of the article we explicatethe four central encoding processes of amodal schema theory and the evidence thatsupports the existence of these processes. Inso doing, we show how schema theories ac-count for the three central phenomena ofmemory—accuracy, incompleteness, anddistortion. In the second part we evaluate thequality of the evidence supporting the fourmajor schema theory encoding processes. Aswill be seen, there is a good deal of evidencethat is not as supportive as might be desiredas well as a fair amount of evidence that isdirectly contradictory. In the third part of thearticle, we consider and evaluate alternative

1 Apparently, a similar situation confronted Taylor andCrocker (1981) as they attempted to formulate and eval-uate a schema theory from research in social psychology.In the memory domain, there are two excellent reviews(Gomulicki, 1956; Zangwill, 1972), but these are nowboth quite out of date. In addition, there are a few scat-tered attempts to present a cohesive schema theory, atleast for limited domains, including scene perception(Brewer & Treyens, 1981) and text comprehension andrepresentation (Thorndyke & Yekovich, 1980).

SCHEMA THEORIES 205

explanatory mechanisms. These alternativesoffer a viable means of resolving the incon-sistencies in the research literature.

Evidence Consistent With Schema Theory

The support for schema theory will be con-sidered in the context of the four schema the-ory principles mentioned above—selection,abstraction, interpretation, and integration—whose origins can be traced to Bartlett's(1932) classic work. They are presented here,not as a description of any specific currentmodel, but rather as an outline of a proto-typic schema theory.

Selection

Of all the concepts in a given event ormessage, only some will become part of thememory representation. Schema theories haveproposed three conditions that determinewhether or not a particular piece of infor-mation will be selected for encoding: (a) theexistence of a relevant schema, (b) the acti-vation of that schema, and (c) the importanceof the incoming information with respect tothe schema. The operation of each of theseselection principles will be considered inturn.

Prior knowledge. A critical condition forthe acquisition of new knowledge is the ex-istence of previously acquired relevantknowledge, that is, of a well developedschema. In the absence of such knowledge,memory is generally poor. Experimentallyproviding subjects with background infor-mation or with ways to link the new infor-mation with already stored informationgreatly enhances memory (Ausubel, 1960;Ausubel & Fitzgerald, 1961, 1962; Bransford& Johnson, 1972, Experiment 1; Royer &Cable, 1975, 1976; Royer & Perkins, 1977;see also Thorndyke, 1977; Wilkes & Aired,1978).

Similar findings have been reported forsubjects purposely selected because theirprior knowledge of a particular topic (herebaseball) was known to vary substantially(Chiesi, Spilich, & Voss, 1979; Spilich, Ve-sonder, Chiesi, & Voss, 1979). These subjectswere presented with narratives about typicalbaseball situations. High-knowledge subjects

recalled substantially more of the narrative,as well as more of its essential information,than did low-knowledge subjects (see alsoAusubel & Fitzgerald, 1961, 1962). Similarly,skilled chess players have a memory advan-tage over novices when the subject of memoryis the location of pieces on the chessboard ofa real game (Chase & Simon, 1973). The ad-vantage of relevant prior knowledge has alsobeen found when the usual positive correla-tion between knowledge and age is reversed,that is, when those with the knowledge arein grade school and those without are grad-uate students and faculty members (Chi,1978).

Frame theory (Minsky, 1975) also assertsthe importance of prior knowledge. If thestructure of incoming information does notmatch one's knowledge, memory for thatparticular type of information should be ad-versely affected. Mismatches between incom-ing and existing structures can occur eitherbecause the structure of the incoming stim-ulus is quite deviant from the prototypicalstructure or because the person lacks ade-quate knowledge about the prototypicalstructure. In fact, when parts of a short storyor fairy tale are rearranged so as to carry thesame information but in an unexpected or-der, or when a story from a culture with dif-ferent structural rules is read, comprehensionis difficult and recall of the story content issignificantly reduced (Day, Stein, Trabasso,& Shirey, Note 1; Kintsch & Greene, 1978;Kintsch, Mandel, & Kozminsky, 1977; Man-dler, 1978; Meyers & Boldrick, 1975; Stein& Nezworski, 1978; Thorndyke, 1977).

In agreement with schema theory, there-fore, specific domain-related prior knowledgeappears to result in improved acquisition ofnew, domain-related information. Assimila-tion of new information appears to be a func-tion of the amount of prior relevant knowl-edge, whether structural (e.g., frame) or con-tent (e.g., theme) based. Without thisknowledge, there is no available schema intowhich new information can be integrated orsubsumed and it is therefore quickly lost. Theencoding of new information is a mappingprocess, new onto old, which depends on asufficiently well developed knowledge base.

Activation of an existing schema. Themere possession of relevant knowledge is not


sufficient for it to play a role in encoding: Theknowledge must be activated at the time ofencoding. Research has focused on two sit-uations in which information is presented inthe absence of activated knowledge bases. Inone case, new information is presented insuch a way as to avoid contact with old, rel-evant knowledge. In the other, a schema isactivated but is congruent with only a portionof the incoming information.

Consider the case in which prior knowl-edge exists but is not activated. The classicstudies in this area were performed by Brans-ford and Johnson (1972, Experiments 2-4).They presented subjects with short passagesthat, without a title, were highly abstract andextremely difficult to understand. In theseinstances the passages were also very poorlyrecalled. Similar memory deficits and pro-cessing difficulties resulting from insufficientcontextual support have been reported fre-quently (e.g., Dooling & Lachman, 1971;Dooling & Mullet, 1973; Johnson, Doll,Bransford, & Lapinski, 1974; Ortony, Schal-lert, Reynolds, & Antos, 1978; Thorndyke,1977). Thus, when knowledge structures liedormant during encoding, new knowledgecannot be easily assimilated; "the absence ofan appropriate semantic context can undersome circumstances seriously affect the ac-quisition process" (Bransford & Johnson,1973, p. 397).

The critical importance of schema acti-vation during the encoding process can alsobe seen in studies in which schema-activatingcues were given after a story was read orheard, In such cases, recall was as poor as itwas when no cues were provided (Bransford& Johnson, 1973; Dooling & Mullet, 1973;see also Thorndyke, 1977).

An interesting developmental implicationof these findings has been drawn by Brans-ford and Nitsch (1978). They speculate thatless experienced people will have greater dif-ficulty than more experienced people in dis-covering the situational cues that can lead toactivation of a knowledge context. Hence,younger children will ordinarily show poorerretention than older children and adults. Onthose rare occasions in which children havea knowledge advantage (e.g., Chi, 1978), theusual advantage of adults will be lost.

A second line of research involves the sit-uation in which an activated schema can be

applied to only some portion of the incominginformation. Schema theory predicts that therelevant information will be easily encodedbut the remainder will either be rejected ordistorted so as to fit the schema. Pichert andAnderson (1977) presented stories that couldbe viewed from either of two perspectives(e.g., a description of a house from the per-spective of a prospective purchaser or of aburglar). Information that was relevant toone perspective was irrelevant to the other.Subjects were biased toward one perspectivewhile reading the stories and later recalledfrom that perspective. Results showed thatsubjects preferentially recalled the informa-tion that was consistent with their perspectiveat encoding (see also Kozminsky, 1977). Pre-sumably, the biasing manipulation activatedonly one of the two schemata subjects pos-sessed. Information irrelevant to the acti-vated schema may never have been perma-nently encoded or may have been encodedbut processed less elaborately than relevantinformation. Both possibilities are consistentwith schema theory. Similar results have beenobtained using a recognition test. Given arecognition test containing items that wereconsistent or inconsistent with respect to thetwo themes of a story, subjects selected itemsthat were consistent with the theme biasedduring encoding (Schallert, 1976).

The congruency between a schema and theincoming information has also been studiedby manipulating the stimulus rather than thesubject's frame of reference (Morris, Stein,& Bransford, 1979). In simple stories, char-acters were described as involved in an ac-tivity that was either consistent or inconsis-tent with their physical attributes (e.g., Thestrong man lifted the piano; The fat man gotstuck in a cave; or, The bald man got stuckin a cave; The old man lifted the piano). Afterreading the initial passage, subjects read asecond story, which referred to the activitiesof the characters. Recall of both the initialand following stories was better when theydescribed or referred to character-appropri-ate situations than when they did not.

The data reviewed in this and in the pre-vious section seem to indicate that bothstructural and semantic schemata play in-fluential roles in the assimilation of new in-formation. Retention appears to be betterwhen incoming information is consistent

SCHEMA THEORIES 207

with one's possessed and activated knowledgeand expectancies about the world. Inconsis-tent information may not be well representedin the memory code.

The importance effect. An existing schemathat is activated during the presentation ofa mass of information will enable selectionof only some of that information for encod-ing. Two selection principles have been pro-posed. The first principle stems from tradi-tional schema theories and states that theideas that are most important to the themeof the information (Owens, Bower, & Black,1979) and that cannot be derived from pre-viously encoded information (Spiro, 1980b)will be given special attention and will be re-membered best. Using stories as stimulusmaterials, several studies have demonstratedthat recall of ideas is a positive function ofthe independently rated importance of thoseideas to the overall meaning of the story (e.g.,Brown & Smiley, 1977; Christie & Schu-macher, 1975; de Villiers, 1974; Graesser,1978a; Graesser, Robertson, Lovelace, &Swinehart, 1980; Johnson, 1970; Johnson& Scheidt, 1977; Kintsch, Kozminsky, Streby,McKoon, & Keenan, 1975; Kozminsky, 1977;Meyer, 1975, 1977; Meyer & McConkie,1973; Miller & Kintsch, 1980; Smiley, Oak-ley, Worthen, Campione, & Brown, 1977;Thorndyke, 1977; Waters, 1978).

A related selection principle comes fromscript theory (Schank & Abelson, 1977). Thistheory deals with situations in which peopleprocess information relevant to high fre-quency events (e.g., eating in a restaurant).The theory predicts that the memory tracesrepresenting highly typical events that oc-curred in a particular episode will be forgot-ten (or omitted from the representation).Typical information need not be stored, sinceit can always be derived from the prototypicalscript. One need only remember that ascripted event occurred to recall highly prob-able components. Thus, according to scripttheory, atypical information will be selectedto receive special representation in memory.

One advantage of a representationalscheme like Schank and Abelson's is that itallows for economical storage in the spaceallocated for memory; highly typical individ-ual episodes that occur in real-world expe-riences may be forgotten, since they are al-ready represented in, and so can be recalled

from, the prototypical script. Economy ofstorage is a concept central to computer anal-ogies of the memory process.

Summary: Selection and memory theory.Considerable evidence appears to support allthree schema theory principles regardingwhat will be encoded. Prior knowledge,whether semantic or structural, increases thelikelihood of encoding new information.However, the existence of prior knowledge isnot sufficient to guarantee encoding of in-coming information; the knowledge must beconcurrently activated. Ideas important to anactivated schema are likely to have a selectionadvantage for storage.

Because of the selection process, the rep-resentation of any event is likely to be quiteincomplete. As a result people cannot repro-duce from memory an exact copy of an event,even when they are motivated to do so.Rather, according to Bartlett (1932), peoplewill attempt to reconstruct the event. Storedinformation will be recalled together with"probable detail" from general schematicknowledge. That people have no troublefreely generating such details is well docu-mented (Bower et al., 1979; Brockway,Chmielewski, & Cofer, 1974; Cofer, Chmie-lewski, & Brockway, 1976).

From such a perspective, accurate recallcan stem from two sources: (a) aspects of theoriginal event that were actually selected forrepresentation in memory and (b) chancematches between the reconstruction processand the original event. The likelihood of suchmatches (or, of correctly "guessing") in-creases whenever the original event containselements in common with other similarevents—the presumed source of the "prob-able detail" from which people are thoughtto reconstruct.

Distortions of the original event will occurwhenever the probable detail produced wasnot actually part of the original event. Muchof Bartlett's data consisted of recall protocolscontaining considerable amounts of infor-mation that was not in the original stimulus.Although Bartlett's results have been difficultto replicate (see e.g., Gomulicki, 1956; Zang-will, 1972), others have found "thematic in-trusions" in retrieved information (Dooling&Christiaansen, 1977a; Kintsch etal., 1975).'Similarly, people occasionally "remember"typical actions that never actually happened


in the particular experience of an event (e.g.,Bower et al., 1979; Graesser, Gordon, & Saw-yer, 1979; Graesser et al., 1980; Smith &Graesser, 1981).

Recently, Spiro (1977, 1980a, 1980b) hasargued that the reconstructive process is mostlikely to result in distortion when one en-counters (at least under conditions in whichmemory is not intentional) some additionalschema-relevant knowledge that is contradic-tory to an encoded schema. Subsequent recallwill be based partly on both sources of in-formation. Such a situation results in distor-tions that are a by-product of the reconstruc-tive process and that will serve to resolve in-consistencies between the sources ofinformation. Again, recall will contain ad-ditional information not present in the event.

The selection process is instrumental inaccounting for one of the three fundamentalmemory phenomena, incompleteness of re-call. Much of an original event is simply notrepresented in memory. The other two phe-nomena, accuracy and distortion, are at leastin part the product of the reconstruction pro-cess that is thought to operate at retrieval.

Abstraction

Information that has been selected becauseit is important and/or relevant to the schemais further reduced during the encoding pro-cess by abstraction. This process codes themeaning but not the format of a message(e.g., Bobrow, 1970; Bransford, Barclay, &Franks, 1972). Thus, details such as the lex-ical form of an individual word (e.g., Schank,1972, 1976) and the syntactic form of a sen-tence (e.g., Sachs, 1967) will not be preservedin memory. Because memory for syntax ap-pears to be particularly sparce as well as brief(e.g., J. R. Anderson, 1974; Begg & Wickel-gren, 1974; Jarvella, 1971; Sachs, 1967,1974), the abstraction process is thought tooperate during encoding.

Additional support for the notion thatwhat is stored is an abstracted representationof the original stimulus comes from studiesthat demonstrate that after a passage is read,it takes subjects the same amount of time toverify information originally presented in acomplex linguistic format as it does to verifythat same information presented in a simpler

format (e.g., King & Greeno, 1974; Kintsch&Monk, 1972).

There are a considerable number of the-ories that assume that memory consists ofsets of propositions and their relations (e.g.,J. R. Anderson, 1976; Bransford, et al., 1972;Brewer, 1975; Frederiksen, 1975a; Kintsch,1974; Norman & Rumelhart, 1975; Schank,1972, 1976). One formalized presentation ofthis idea is Schank's conceptual dependencytheory (1972). The theory asserts that allpropositions can be expressed by a small setof primitive concepts. All lexical expressionsthat share an identical meaning will be rep-resented in one way (and so stored econom-ically) regardless of their presentation format.As a result people should often incorrectlyrecall or misrecognize synonyms of originallypresented words, and they do (e.g., Anderson& Bower, 1973; R. C. Anderson, 1974; An-isfeld & Knapp, 1968; Brewer, 1975; Graes-ser, 1978b; Sachs, 1974).

Abstraction and memory theories. Sinceconsiderable detail is lost via the abstractionprocess, this process can easily account forthe incompleteness that is characteristic ofpeople's recall of complex events. In light ofthe abstraction process, the problem forschema theories becomes one of accountingfor accurate recall. Schema theories do thisby borrowing a finding from psycholinguisticresearch, to wit, that speakers of a languageshare preferred ways of expressing informa-tion. If both the creator and perceiver of amessage are operating with the same prefer-ences or under the same biases, the per-ceiver's reproduction of the input may appearto be accurate. The accuracy, however, is theproduct of recalling the semantic content ofthe message and imposing the preferredstructure onto it. Thus, biases operate in amanner that is similar to the "probable de-tail" reconstruction process. Biases have beendocumented for both syntactic information(J. R. Anderson, 1974; Bock, 1977; Bock &Brewer, 1974; Clark & Clark, 1968; James,Thompson, & Baldwin, 1973) and lexical in-formation (Brewer, 1975; Brewer & Lichten-stein, 1974).

Distortions may result from the abstrac-tion process if biases are not shared by theperson who creates the message and the onewho receives it. More importantly, the ab-

SCHEMA THEORIES 209

straction process sets the stage for distortionsbecause it is the necessary precondition fortwo other schema theory processes, interpre-tation and integration. These two are usuallythought to be the major sources of dis-tortions.

Interpretation

Thus far, distortion in the recall of com-plex materials has been attributed to pro-cesses that reduce information. The encodingdeficit that results from selection and abstrac-tion is, in part, compensated for at recall byreconstruction and by shared speech pro-duction biases. Distortions also occur be-cause those semantic propositions that areencoded are actually interpretations of theexplicitly presented information—interpre-tations that are based on the perceiver's ac-tivated schematic knowledge. Such errors areoften referred to as constructive errors be-cause they involve the addition of informa-tion to the memory representation of a com-plex event. In contrast to reconstructive er-rors, constructive errors are the product ofan elaboration process occurring during orshortly after encoding.

Interpretations are typically inferences (seeHarris & Monaco, 1978). Two varieties canbe identified. The first, pragmatic implica-tion, involves converting explicitly stated in-formation into its probable underlying in-tent. The second involves inferences madeduring comprehension when there is a needto (a) concretize vague information, (b) fillin missing detail, or (c) simplify complex in-formation.

Pragmatic implication.' A variety of ex-amples of this aspect of interpretation can befound (see Harris & Monaco, 1978, for a re-view); a few will suffice. Given Sentence 1,people will often recall it in the form of Sen-tence 2 (Schweller, Brewer & Dahl, 1976):

1. The housewife spoke to the manager about theincreased meat prices.

2. The housewife complained to the manager aboutthe increased meat prices.

Given Sentence 3, people recall Sentence 4(Brewer, 1977):

3. The paratrooper leaped out of the door.4. The paratrooper jumped out of the plane.

Given Sentence 5, people equate its meaningwith Sentence 6 (Harris, 1974; Harris &Monaco, 1978):

5. The frightened farmer was able to raise chickens.6. The frightened farmer raised chickens.

In each case, subjects are encoding infor-mation that goes beyond the explicit. Thedistortion that results can range in degreefrom minor to extreme. It will be minor ifthe perceiver's interpretation corresponds tothe meaning intended by the creator of themessage. Distortion will, of course, be moreextreme if the perceiver misinterprets themessage.

Interpretation and comprehension. Suc-cessful comprehension heavily depends onthe interpretation process. Consider the sit-uation in which a presented concept is gen-eral or vague. According to theory, the schemawill serve to concretize the information andstore it in its interpreted form. This processis called instantiation. It can be demonstratedby presenting people with sentences that con-tain general concepts (e.g., animal, container)but provide enough context to allow the gen-eral terms to be understood as specific ones(e.g., dog, bottle). On a cued recall test, thenever-presented specific terms serve as betterretrieval cues than the actually presented gen-eral terms (Anderson et al., 1976). Thus, gen-eral terms are interpreted and stored as spe-cific instantiations.

A second, closely related form of interpre-tation-dependent comprehension occurs whenimplied facts are inserted into the represen-tation of an incomplete stimulus. When sub-jects are given sentences that imply the pres-ence of an unstated object, instrument, oraction, the subjects construct the missing in-formation (Paris & Lindauer, 1976; Paris,Lindauer, & Cox, 1977; Paris & Upton, 1976)and may then remember it as having beenpart of the original stimulus (Johnson, Brans-ford, & Solomon, 1973). Thus, subjects whohear a passage describing a person poundinga nail will infer the presence of a hammer asthe probable instrument and incorporate theinstrument into their representation of thepassage (Johnson et al., 1973).

Additions made to explicit information gobeyond one-word instruments and objects.Frequently, speakers and writers will struc-


ture their output in such a way as to inducelisteners and readers to infer connections be-tween ideas. Haviland and Clark (1974) andKintsch and van Dijk (1978) have argued thatinformation in a text is most easily under-stood when it can be related to immediatelypreceding information. When this is not pos-sible, the comprehender searches memory forsome relevant information that can serve asa connecting idea. If none is found, peoplewill construct an inference that bridges theinformation gap. In fact, when an inferencemust be generated in order to understand arecent input, people do construct the missinginformation (Keenan & Kintsch, 1974;McKoon & Keenan, 1974) and sometimeslater misrecognize it as having been part ofthe text (Thorndyke, 1976). Similarly, infer-ences may be made and incorporated intomemory when they simply increase the co-herence of otherwise understandable ideas(Owens et al., 1979).

Frame theory predicts a related type of in-ference making. If a stimulus generally con-forms to an existing frame but does not con-tain all of the information categories speci-fied by that frame, the missing informationwill be provided in the form of "default val-ues" (Minsky, 1975). Thus, if a story gram-mar proposes slots for setting or theme in-formation and a stimulus conforming to thatgrammar possesses no information that caninstantiate those slots, the missing informa-tion will be inferred and attached to the rep-resentation of the stimulus (e.g., Glenn, 1978;Mandler & Johnson, 1977; Stein & Glenn,Note 2). Further, if the order of ideas in aframe-like stimulus does not conform to theideal order, the encoding frame may be usedto rearrange those story ideas (Mandler,1978; Stein & Glenn, 1979; Stein & Nez-worski, 1978; see also Lichtenstein & Brewer,1980).

Finally, inferences will also be made if theyhelp to simplify the prepositional structure,especially when the information processingload is heavy (Frederiksen, 1975a, 1975b).These inferences too are often incorporatedinto the recall of the text.

The possibility of distorting an originalevent arises because the interpretation pro-cess allows the perceiver to add to or changethe information conveyed by the stimulus.

Interpretation results in distortion becauseit occurs with two other processes. The firstis abstraction, which eliminates the surfacestructure form of the stimulus and stores onlya meaning-based representation. The secondprocess, discussed next, is integration.

Integration

That people contribute information to anincoming message is not a notion unique toschema theory. What is unique is the predic-tion made about the memory representationthat is the product of interpretation. A singleintegrated memory representation is thoughtto be created from whatever accurate infor-mation is selected, whatever interpretationsare drawn, and whatever general knowledgeexists that is relevant to the stimulus. Thus,individual ideas exist only as a part of a com-plex semantic whole. Integration processesare thought to occur at two different stagesof memory: (a) when a new schema is formedand (b) when an existing schema is modified.These will be considered in turn.

Schema formation. One method of dem-onstrating integration is to show that mem-ory for a set of related propositions consistssolely of the overall intent of those proposi-tions. Consider the following sentences.

7. There is a tree with a box beside it and a chairis on top of the box.

8. The box is to the right of the tree.9. The tree is green and extremely tall.

People who read Sentences 7-9 will misrec-ognize 10 as an originally presented sentence.

10. The tree is to the left of the chair.

But, they will call Sentence 11 new (Brans-ford etal, 1972; Paris & Carter, 1973;Prawat& Cancelli, 1976).

11. The chair is to the left of the tree.

Evidence for the integration of individualsentences into a holistic representation is alsofound when, during acquisition, subjects arepresented with a set of sentences that containdifferent sized subsets of ideas from a morecomplex, unpresented parent sentence. Sub-jects' confidence that a particular sentencewas actually presented during acquisition in-creases with the degree to which that sentenceapproximates the never-presented complex

SCHEMA THEORIES 211

parent sentence regardless of whether or notthat sentence was originally presented(Bransford & Franks, 1971; Cofer, 1973;Griggs, 1974; Singer, 1973; Walsh & Baldwin,1977).

Integration during schema formation canalso be seen in the linear ordering paradigm.Here subjects are presented with a series ofsentences that describe a one-dimensionalordering of objects or people. Dimensionsmay include weight, height, and speed. Thesentences may simply state that a is tallerthan b, b is taller than c, d is shorter than -c,and so on. Barclay (1973) presented such sen-tences and asked subjects to determine therelations involved in the total set by imag-ining the entire array in its proper order. Ona subsequent recognition test, he presentedsome of the originally presented sentencesalong with some new sentences that either didor did not express valid relations. Subjectswere unable to discriminate actually pre-sented sentences from new ones that werevalid statements about the ordering (see alsoPotts, 1973). Similar conclusions may bedrawn from studies using class inclusion re-lations (Potts, 1976), artificial terms (Potts,1977), and complex nonlinear arrays (Moeser& Tarrant, 1977).

The schema theory explanation of thesefindings is that the meanings inferred fromindividual sentences during encoding are in-tegrated into a larger semantic whole. On amemory test, sentences that are consistentwith the integrated representation are judgedto have been part of the original array.

Schema modification. The previous stud-ies demonstrate integration for brief and re-lated pieces of information on topics aboutwhich subjects had limited prior knowledge.Another demonstration of the integrationprocess occurs when new information is pre-sented on a topic about which the personalready has considerable information inmemory. Consider first a study by Sulin andDooling (1974). Here subjects read a shortpassage that they believed to be about eithera famous person or a fictitious person. Ona recognition test, subjects were given someof the passage sentences along with some dis-tractors that varied in their thematic associ-ation to the famous character. Subjects whowere told that the passage described the fa-

mous person, misrecognized as old the itemsthat were thematically related to the famousperson. This did not occur for subjects in thefictitious person condition (see also Brown,Smiley, Day, Townsend, & Lawton, 1977;Royer, Perkins, & Konold, 1978).

Consider next the work of Loftus and hercolleagues (e.g., Gentner & Loftus, 1979;Loftus, 1975; Loftus, Miller, & Burns, 1978;Loftus & Palmer, 1974). In these experi-ments, subjects are typically shown a shortfilm or set of slides depicting, for example,a traffic situation. Afterward, a question isasked about the scene that either implies thepresence of additional information that wasnever actually present or that contradicts in-formation that was present. On a subsequentmemory test, subjects often misrecognizenew slides containing the additional or con-tradictory information. Loftus and Loftus(1980) argue that new information intro-duced after an event has been witnessed mayadd to or replace the person's knowledge ofthe original scene, resulting in a single, in-tegrated memory for the scene.

Finally, consider the phenomenon of the"knew-it-all-along" effect (Fischhoff, 1977).In this paradigm, subjects are asked to ratethe probability of occurrence of an event(Fischhoff, 1975; Fischhoff & Beyth, 1975)or to assess the validity of factual statementsthat are beyond their specific knowledge butthat pertain to topics with which they arefamiliar (Fischhoff, 1977; Wood, 1978). Theyare then told of the occurrence or nonoc-currence of the event or are given the truthvalues of the statements. Subjects are thenasked to estimate their preoutcome knowl-edge states or to recreate their original prob-ability or validity judgments, uninfluencedby the outcome information. The result isthat subjects overestimate the degree to whichthey were originally correct. They cannotaccurately recall their judgments and the in-accuracy is systematically in the direction ofthe answers supplied to them. These data sug-gest that new information is immediately in-tegrated into the prior knowledge system thatsubjects possessed about the topics, resultingin an inseparable combination of the new andold information. When trying to recall theiroriginal ratings, subjects base their responseson the only representation available to them


at test time, one which has been modified bythe incorporation of the new information.

Integration and memory theory. Once in-tegration occurs and old knowledge has beenaltered or updated, accurate retrieval of ac-tually presented information becomes highlyunlikely. New information will be integratedinto old knowledge structures. This new in-formation can originate in the external en-vironment (as suggested by the Loftus orFischhoff work) or in the internal environ-ment (as suggested by the work on infer-ences). Regardless of the information source,schema theory predicts that inaccurate re-trieval will occur because individual tracesof a to-be-remembered event do not exist.

Summary

Schema theory has become an enormouslypopular framework guiding considerableamounts of research in human memory. Onesource of this popularity is the apparent suc-cess that the major assumptions of the theoryhave had in explaining experimental results.According to schema theory what is storedis some highly selected subset of all that hasbeen presented, and it is the schema thatguides the selection process. Memory is ab-stractive in that a verbatim record is not leftbehind; rather, meaning appears to have high-est priority for storage. Memory is interpre-tive in that the schema serves to fill in missingdetails and distort others so as to be schema-consistent. Memory is integrative in that in-coming information joins with other relatedelements in that episode, with whatever priorknowledge is available, and with relevant,subsequent information to create a singleunified representation of a complex event.

When retrieval is required, probable detailis generated, and the representation is givena modal surface form. Thus the originalmeaning is retained, and on occasion, theproduction is a duplication of the originalstimulus. Accuracy sterns from those ideasthat were actually selected for storage, fromculture-normal biases in expressing infor-mation, and from serendipitous productionsfrom the reconstructive process. This lattercategory is most likely to occur for high-prob-ability events.

On other occasions, recall may be quitedistorted. This is attributed to a number of

sources: reconstruction, construction, un-shared language production biases betweenthe originator and receiver of the message,interpretations not actually intended by anoriginator, and the integration of memoryepisodes over time.

Incompleteness in recall is largely theproduct of two encoding operations: selectionand abstraction. Not all ideas are selected forrepresentation. Not all elements are part ofthe abstracted meaning. The integration pro-cess further attenuates those ideas that wereselected and abstracted, thus reducing theamount of information that can be recalled.

In the next section of this article we con-sider a body of research, much of it quiterecent, that is either inconsistent with or di-rectly contradictory to the central assump-tions of schema theory. Each of the four basicencoding assumptions of schema theory willbe considered in turn.

Evidence Inconsistent With Schema Theory

Selection

The importance of the schema to the se-lection process is based on the interpretationof five sets of findings: (a) in the absence ofany schema, not much can be encoded andso not much will be recalled (e.g., Bransford& Johnson, 1972, 1973); (b) the greater thelevel of prior knowledge, the greater theamount of new, relevant information thatcan be selected and subsequently recalled(e.g., Chiesi et al., 1979); (c) when an existingschema is not activated, recall is extremelypoor (e.g., Bransford & Johnson, 1972); (d)information that is congruent with an acti-vated schema is far more likely to be encodedand so recalled than is information that isincongruent (e.g., Pichert & Anderson, 1977);(e) the more important an idea is to a schema,the greater the likelihood that it will be re-called (e.g., Johnson, 1970; Meyer & Mc-Conkie, 1973).

Consider first the idea that the encodingof new information is a function of theamount of prior relevant information. A re-view of the education literature (Barnes &Clawson, 1975) found more research in con-flict with Ausubel's (1968) version of thisprinciple than in support of it. Although re-call has been shown to vary substantially withthe amount of prior knowledge (Chiesi et al.,

SCHEMA THEORIES 213

1979), the same research has shown that rec-ognition varies minimally (Experiments 1and 2). Similarly, although an inactivatedschema may leave subjects at a considerablerecall disadvantage compared with subjectswith an activated schema (e.g., Bransford &Johnson, 1972), this disadvantage too dis-appears when memory is tested using a rec-ognition procedure (Alba, Alexander, Hasher,& Caniglia, 1981). That an activated schemadoes not benefit recognition as it does recallhas been shown for a variety of stimulusmaterials including prose (Alba et al., 1981),individual sentences (Birnbaum, Johnson,Hartley, & Taylor, 1980), lists of sentences(Hannigan, 1976), and pictures (Bower, Kar-lin, & Dueck, 1975). Finally, although im-portant ideas have an advantage in recall, thisadvantage is not always present on a recog-nition test (Britton, Meyer, Hodge, & Glynn,1980, Experiment 3; Goetz, 1979; Kintsch& Bates, 1977;Thorndyke&Yekovich, 1980;Walker & Meyer, 1980; Yekovich & Thorn-dyke, 1981).

That recall but not recognition varies withthe activation of a schema, with the knowl-edge content of that schema, and with theimportance of individual ideas to that schemaposes problems for a selection process pre-sumed to operate at the time of encoding.Information that is not part of a memoryrepresentation should be the equivalent ofinformation that was never presented; itshould be neither recalled nor recognized.That there are substantial effects of a schemaon the amount that is recalled as well as onthe content that is recalled suggests that theschema influences retrieval rather than selec-tion for storage in memory.

The suggestion that a schema operates atretrieval to influence what is recalled isstrengthened considerably by the work ofAnderson and Pichert (1978). When infor-mation agrees with two different schemata,only one of which is activated during encod-ing, superior recall is seen for the informationthat is congruent with whatever schema wasactivated during acquisition (see also Pichert& Anderson, 1977). However, if one tests forrecall a second time from the perspective ofthe previously inactive schema, recall of theonce irretrievable information is significantlyraised (Anderson & Pichert, 1978). Thesefindings have been replicated by Pass and

Schumacher (1981), who in some conditionsdid not introduce the alternative (or new)theme until 12 minutes after the originalstory had been read, long after the selectionprocess should have operated. These cuingstudies suggest that more is encoded than isconsistent with the operation of a schema-based selection principle.

Strong evidence for the operation of re-trieval processes in prose recall comes froma study by Hasher and Griffin (1978). Theyshowed that as long as 2 weeks after the en-coding of a story the nature of recall couldbe dramatically influenced. In that study, onegroup of subjects showed a.typical "recon-structive" pattern, with substantial forgettingof specific ideas along with thematically rel-evant intrusions. However, another groupshowed a quite untypical pattern (termed"reproductive" by the experimenters); thesesubjects produced virtually no thematic in-trusions and forgot far less of the originalstory information. The manipulation that in-duced reproductive recall was one that cre-ated doubt about the validity of the schemaused during encoding. Hasher and Griffinargued that subjects were then forced to re-vise their retrieval plan (of using the encodingschema) and so exerted special effort to gainaccess to otherwise inaccessible information.Further evidence that a schema operates atretrieval is found in a study in which recalldeficits for unimportant (and so presumablyunselected) information were greatly atten-uated by providing subjects with both se-mantic (content words) and contextual (thebackground color of the printed page) re-trieval cues (Britton et al., 1980, Experiments1 and 2).

Selection and reconstruction. Selectionpresumably allows for mnemonic distortionbecause the memory representation of anyevent is substantially reduced compared withthe original. According to Bartlett (1932), thesubject must then rely on his or her generalknowledge together with what few detailswere actually encoded to produce a recount-ing of an event. This recall procedure is calledreconstruction. The evidence just presented,however, suggests that memory for an eventis often quite good, though details may notalways be readily accessible. In fact, the ev-idence for a reconstructive process at re-trieval is itself actually fairly weak. Under


normal circumstances, distortion is eitherrare or nonexistent (Zangwill, 1972). In fact,a recent analysis of Bartlett's stimuli showthem to be rather poorly structured and un-representative of normal prose (Mandler &Johnson, 1977), perhaps accounting for hisapparently unique findings. Indeed, a con-temporary replication of Bartlett's proce-dures showed little evidence of distortion(Kintsch & Greene, 1978; see also Gomu-licki, 1956; Paul, 1959). Recall errors that dooccur can sometimes be eliminated by ma-nipulating recall conditions (cf. Cofer et al.,1976), by stressing accuracy (Gauld & Ste-phenson, 1967), or by changing the subjects'retrieval strategy (Hasher & Griffin, 1978).

If reconstructive processes exist, they maybe an optional, perhaps (compared with re-production) even a relatively effortless, recallstrategy, rather than, as Bartlett proposed,and others have assumed, a necessary strat-egy. After all, one's general knowledge on atopic is presumably acquired by virtue of re-peated, distributed exposures to that topic.Since both study and retrieval episodes areimplicated in recallability (see Spear, 1978),it should be relatively easy to gain access togeneral knowledge, especially when com-pared with the ease of gaining access to one'smemory for an episode that occurred onlyonce. At the very least, reconstruction is nota necessary consequence of selective encod-ing; far more is encoded than the selectionprinciple would suggest, and a great deal ofthis original information remains availablein memory. Finally, the circumstances at re-trieval play a critical role in determining whatcan be accessed (see Tulving & Pearlstone,1966, for a discussion of the distinction be-tween availability and accessibility).

Yet to be considered are those reconstruc-tion processes associated with script theoryand collectively referred to as the typicalityeffect (e.g., Bower et al., 1979; Graesser et al.,1979). Script theory predicts a failure onmem 017 tests to discriminate between pre-sented and unpresented events that are typ-ical of a given script. Although the few studiesthat have been performed in order to test thisprediction have reported decreased discrim-inability with increased typicality, these stud-ies have also demonstrated that (a) peoplegenerally can discriminate between presentedand unpresented typical script actions, espe-

cially when interference is reduced (Bower etal., 1979); and (b) complete failure to dis-criminate occurs only with those very fewscript actions that can be labeled very typical(Graesser et al., 1979)—a finding that con-siderably weakens the cognitive economy ar-gument of script theory.

Possible explanations of the typicality ef-fect that exclude the principle of selection willbe presented in the final portion of this ar-ticle.

Abstraction

The abstraction process is thought to fur-ther reduce what has been stored in memoryby operating on what has been selected; se-mantic content is believed to be the impor-tant target of storage. Although it is nowwidely held that memory is abstractive, anumber of recent findings cast some doubton this process.

First, memory is not inevitably abstractive.Consider memory for highly overlearnedmessages such as the Pledge of Allegiance.Detailed memory for such material will per-sist for years (Rubin, 1977; see also Smith,1935). Even memory for information pre-sented only once has been shown to containconsiderable detail. College students are ableto discriminate actual utterances heard in thecontext of a lecture from highly similar ut-terances (Kintsch & Bates, 1977). Accuraterecognition was found in that study for con-tent that was largely extraneous to the topic(e.g., jokes) as well as for low-salient, relevantinformation. Memory for syntactic and lex-ical details in the lecture was quite profound;the rejected paraphrases included such subtlechanges from the original sentences as rear-rangement of clauses and meaning-preserv-ing word substitutions.

Still more difficult paraphrases than thoseused by Kintsch and Bates (1977) were in-cluded in a follow-up study in which recog-nition memory was tested for sentences froma television soap opera (Bates, Masling, &Kintsch, 1978). The paraphrases in this studywere constructed by substituting propernames for pronouns, roles for names, ellip-tical clauses for full clauses, and vice versa.Memory for the original utterances was sur-prisingly good (see also Bates, Kintsch,Fletcher, & Giuliani, 1980).

SCHEMA THEORIES 215

The ability of adults to distinguish betweenparaphrases and actual utterances is not eas-ily explained by any theory which assumesthat memory is selective and abstractive. Ifonly general impressions plus a few detailsremain in memory after a lecture or a TVshow, recognition performance for alterna-tives discriminable only on the basis of detailsshould be at chance. A schema-theory expla-nation might claim that the utterances andtheir paraphrases differed in terms of their"probable detail," or in their conformity tothe stylistic habits of the speakers, aspects ofmessages to which listeners are presumablysensitive. However, Keenan, MacWhinney,and Mayhew (1977) demonstrated that mem-ory for utterances from a lecture was notbased on sensitivity to such aspects; rather,subjects appear to remember the actual formof the original statements. Other researchconfirms this conclusion; people are able todiscriminate between the frequency withwhich verbatim versus gist repetitions of sen-tences occur (Burnett & Stevenson, 1979;Gude & Zechmeister, 1975; see also Morris,Bransford, & Franks, 1977).

Thus, memory for detail has been shownfor both well-learned (Rubin, 1977) and farless well-learned information (Bates et al.,1978; Bates et al., 1980; Keenan et al., 1977;Kintsch & Bates, 1977). It should be notedthat in all of the experiments dealing withmemory for once-presented messages, therecognition test was a surprise. Thus thesefindings cannot be explained away by ap-pealing to unusual and "unnatural" task de-mands made under conditions of rote mem-orization.

Recent research appears to stand in markedcontrast to the earlier evidence often cited assupport for the abstraction process. In factit does not. What those early studies foundwas superior recognition of semantic as com-pared with syntactic information. Syntax wasfrequently remembered at levels well abovechance; that is, syntactic information was nottotally forgotten, (e.g., J. R. Anderson, 1974;Anderson & Paulson, 1977;Begg, 1971;Begg& Wickelgren, 1974; Dooling & Christiaan-sen, 1977b; Olson &Filby, 1972; Sachs, 1967,1974; Soli & Balch, 1976). Three recent stud-ies have shown above-chance recognition oflexical and syntactic detail at the surprisinglylong retention intervals of 1 hour (Yekovich

& Thorndyke, 1981; see also James et al.,1973;Kemper, 1980) and 1 week (Christiaan-sen, 1980).

A second version of the abstraction processstates that word meanings are representedconceptually, and that therefore lexical in-formation is lost (e.g., Schank, 1972). A num-ber of findings contradict this view. For ex-ample, people rarely make synonym substi-tution errors, are better able to identifysentences when they contain their originalwords rather than synonyms, and show greaterinterference in sentence verification whenstimulus sentences all contain a commonword than when they contain different syn-onyms of the word (Hayes-Roth & Hayes-Roth, 1977). Along with Anderson's (1971;see also Hayes-Roth & Thorndyke, 1979)finding that verbatim cues enhance recallmore than do synonymous paraphrases ofwords, there seems to be good evidence forthe hypothesis that lexical information is alsoretained and forms part of the representationof a verbal event.

One study that can still be taken as evi-dence in favor of the conceptual represen-tation notion did find reliable synonym sub-stitution errors in recall (Brewer, 1975). How-ever, of the total number of possible responses,synonym substitution occurred only 6% ofthe time (Experiment 1). Thus, either mem-ory is not conceptually based, or very fewwords have identical synonyms. (If few wordsdo have identical synonyms, the storage ef-ficiency argument of computer-based mod-els—e.g., script theory—is not very con-vincing.)

Memory appears to contain far more syn-tactic and lexical detail than is consistent withthe view that memory is highly abstractive.Thus a central assumption of schema theoryappears to be incorrect; the representation ofboth syntactic and lexical information isricher and more detailed than the abstractionprocess allows.

Interpretation

That people interpret and elaborate on in-coming information is not a notion uniqueto schema theory. In the traditional humanlearning literature, "interpretation" pro-cesses were recognized and explored in re-search dealing with the distinction between


the nominal and functional stimulus (Un-derwood, 1963), with the role of mediationand elaboration in the learning of simplestimuli (Bugelski, 1968; Underwood & Schulz,1960), with subjective organization (Tulving,1962), and more recently with the distinctionbetween "fact" and "fantasy" (Johnson &Raye, 1981). What is at issue in an evaluationof schema theories is the question of how in-terpretive behaviors alter the memory rep-resentation of the original stimulus. That is,are subject-generated interpretations of in-coming stimuli necessarily integrated withthe representation of the nominal stimulusso as to lose tags that mark the origin of anidea? This question is considered in the nextsection of the article.

There are two important points about in-terpretation to be made here. Contrary tosome schema theories that predict sizableamounts of inference making during com-prehension (e.g., Schank, 1976), several re-searchers have found that inference makingis not an obligatory process. People may com-pletely fail to make even a simple inference(Corbett & Dosher, 1978; Hayes-Roth &Thorndyke, 1979; Singer, 1979, 1981); theymay generate inferences only to some infor-mation in a message (Goetz, 1979; Walker& Meyer, 1980); or they may engage in ex-tensive inference making only when the taskdemands require it (Frederiksen, 1975b).

Second, whatever retrieval errors subjectsdo make in recall may not even be indicativeof their underlying memory representations.Consider a prediction of frame theory: If astimulus is structured differently from theframe used to encode it, the stimulus shouldbe reorganized to conform to the frame, andrecall studies suggest that subjects do reor-ganize information (e.g., Stein & Glenn,1979; see also Mandler, 1978; Stein & Nez-worski, 1978). However, in at least certaincases, e.g., a story presented with episodesoccurring in flashback order, there is evidencethat subjects retain input order information(Baker, 1978).

Integration

Schema formation. Integration is the setpiece of schema theories. Theoretically,meaning is abstracted from the stimuli that

have been selected, and related abstractionsare joined into a single representation. Themost impressive contemporary evidence insupport of this notion came from Bransfordand Franks (1971). As is now widely ac-knowledged, there are many grounds onwhich to challenge their conclusions. Themost basic is the fact that integration is nota necessary product of encoding relatedevents (Moeser, 1976, 1977). In fact, theBransford and Franks (1971) findings aredifficult to replicate unless their methods arefollowed closely. Changes in presentationmodality (Flagg & Reynolds, 1977; see alsoKatz & Gruenewald, 1974), in presentationof materials (Flagg, 1976; Flagg & Reynolds,1977; James & Hillinger, 1977; Katz, Atke-son, & Lee, 1974), in instructions (James,Hillinger, & Murphy, 1977), and in testingprocedures (Anderson & Bower, 1973; Griggs& Keen, 1977; James & Hillinger, 1977)greatly reduce or eliminate the integrationeffects found with the original Bransford-Franks paradigm. Finally, "integration" willoccur for nonsemantic or arbitrary materials(e.g., letter-digits, nonsense syllables) if theoriginal procedures are used (Flagg, 1976;Katz & Gruenewald, 1974; Reitman & Bower,1973; Small, 1975). Since integration is as-sumed to be tied to similarities in the rep-resentation of meaning, evidence of such aprocess operating on meaningless material istroublesome.

Results from research using the linear or-dering paradigm (e.g., Barclay, 1973; Potts,1973) were also interpreted as being the resultof integration. The argument was that whenrelated propositions (e.g., a < b, b < c, c <d) are presented, the individual ideas are lost,leaving a subject-constructed, holistic array(a < b < c). There are now several groundson which to challenge this conclusion. Again,the most basic is that the construction of aholistic order is not a necessary by-productof encoding materials bearing ordered rela-tions (Potts, Keller, & Rooley, 1981). In fact,a sizeable proportion of subjects (30%) showedno evidence of even creating such an ordering(Potts et al., 1981). Further, recent evidencesuggests that the ordering effect may be an

. artifact of the testing procedures used. Givingsubjects unlimited time to make old-newdecisions, as well as more information on

SCHEMA THEORIES 217

which to base their decisions, eliminates in-tegration effects (Lawson, 1977). Even underless favorable conditions, an ability to makeproper discriminations seems to exist (Tzeng,1975). Subjects appear to store individualideas, even if they are related to others in thememory set.

Strong evidence for integration processesin the formation of a schema seemed to comefrom both the semantic integration and linearordering paradigms. Integration effects ap-pear to be more tied to aspects of the pro-cedures used than to an inevitable process ofthe human memory system. In addition, inboth cases there is evidence that subjects havestored in memory separate, unintegratedunits of the original stimulus complex.

Schema modification. A schema is a re-pository of prior knowledge on a topic. Thus,one might argue that the failure to demon-strate integration across a newly acquiredknowledge base is not so critical for schematheory as would be the failure to demonstrateintegration of new facts into a preexistentknowledge base. (Such an argument createsthe problem of explaining where schemascome from in the first place; see Yekovich& Thorndyke, 1981). Putting aside for themoment the negative evidence concerningthe abstraction process itself (on which theoccurrence of integration necessarily de-pends), we now evaluate the evidence for in-tegration of new information with well-learned (or previously acquired) data bases.

Studies investigating this type of integra-tion have attempted to show that followingintegration (a) people cannot remember theiroriginal knowledge xstate and (b) the newlyacquired piece of information loses its unique,context-specifying character. Three differentlines of research would appear to providestrong evidence for these two views.

The "knew-it-all-along" effect shows thatpeople tend to overestimate the quality oftheir original knowledge about facts afterthey have been provided with relevant infor-mation. The overestimation is theoreticallythe result of the new information's havingbeen inseparably integrated into a person'sgeneral knowledge, making it impossible forthe person to correctly remember his or heroriginal knowledge state. However, Wood(1978) showed that if subjects are specifically

told to memorize their previous knowledgestates, the size of the knew-it-all-along effectis reduced. Apparently, subjects can to someextent keep old and new information sepa-rate. More critically, however, there is evi-dence that people can accurately remembertheir original knowledge states even after newinformation has been provided, that is, afterthe integration has presumably occurred.Accurate recall of original information oc-curs if people are led to believe that the truthvalue of the new facts is unreliable (Hasher,Attig & Alba, 1981). Under these circum-stances, people show very little knew-it-all-along behavior. What is critical about thisstudy is that the manipulation discreditingthe new "facts" was provided after the inte-gration process was presumed to have oc-curred, which, according to schema theory,is an impossible finding. Thus, it appears tobe possible to induce subjects to remembertheir original knowledge state, which was oth-erwise thought to be lost to memory by anintegration process. These data suggest, aswell, that subjects can discriminate the originof recently provided information.

There is also a substantial volume of re-search on the integration of new with olderknowledge conducted by Loftus and her col-leagues. This research, too, appears to sup-port both predictions. Subjects misin-formed—after seeing a series of slides de-picting an accident—that the traffic sign ata critical corner was a stop sign (not a yieldsign, or vice versa) later incorrectly reporthaving seen a stop sign. Recent evidence(Bekerian & Bowers, 1983) suggests that thisdramatic and widely cited instance of inte-gration may be an artifact of the testing pro-cedure, in which the test slides are presentedwithout regard to the original temporal se-quence. The integration effect can be virtu-ally eliminated by a testing procedure thatpreserves the original presentation order.Subjects can apparently use temporal-se-quential information to distinguish the orig-inal slide (e.g., the yield sign) from the up-dating slide (the stop sign). Such discrimi-nation should not of course be possible afterthe integration process has occurred; the orig-inal information should no longer exist.

Two other points about this work shouldbe made. First, it is possible (as Loftus,


1979a, acknowledges) that other conditionsnecessary to demonstrate recollection of thepreassimilated knowledge have not yet beendiscovered (for example, the Hasher et al.,1981, discrediting treatment might be sucha condition). The second point relates to thegenerality of the data. Loftus has been suc-cessful in modifying people's memory of anevent only when she deals with informationperipheral to the main focus of the event. Inrecent studies (Alba, 1981; Loftus, 1979b),attempts at altering memory for focal infor-mation were shown to be almost totally in-effective. These findings suggest an alterna-tive interpretation of the original results (alsosee Shaughnessy & Mand, 1982). The relativeease with which nonfocal information can beupdated, compared with the difficulty of up-dating focal information, suggests that the"integration" process can work most easilyon memories that are either already lost ordifficult to retrieve. In these instances theupdating information will be (at least on thebasis of its recency to the memory test) moreaccessible than the original traces. It is ofcourse important to account for the difficultyof gaining access to nonfocal information(ordinarily evidence for the selection pro-cess). We will consider a possible explanationshortly.

A further finding suggests that this differ-ential retrieval hypothesis has validity: mod-ification of a nonfocal fact does not occur ifupdating information is presented along withblatantly incorrect information (Loftus,1979b). Similarly, updating does not occurif the provider of the new information is dis-trusted (Dodd & Bradshaw, 1980), nor if sub-jects are forewarned about the possibility ofreceiving misinformation (Greene, Flynn,& Loftus, 1982). These three manipulationsmay actually be the equivalent of the schemadiscrediting procedure (Hasher et al., 1981;Hasher & Griffin, 1978), which in the latterstudies apparently induced subjects to re-trieve ordinarily forgotten facts.

Thus, updating (or the integration of re-lated memories) is not an inevitable conse-quence of the mere provision of related, con-tradictory, or supplementary information.New information may be disregarded (Dodd& Bradshaw, 1980; Greene et al., 1982; Lof-

tus, 1979b) at the time of presentation—or,more importantly, afterwards (Hasher et al.,1981). Subjects can, even in the face of newrelevant information, continue to remembertheir original knowledge state. Discrimina-tive markers for newly provided informationdo not appear to be inevitably lost.

A second set of studies more directly ad-dressed the question of whether newly pre-sented information loses its unique identityafter it is integrated into a relevant, preex-isting schema. Initial support for this notionwas provided by Sulin and Dooling (1974)who reported that subjects aware of the con-nection between information in an experi-menter-provided passage and a famous per-son, later misrecognized information aboutthe famous person that was never actuallypresented. Subjects unaware of the connec-tion between the story and the famous personwere less likely than other subjects to mis-recognize the critical information. One in-terpretation of these results is that the passageinformation became inextricably integratedinto the relevant knowledge schema of thosesubjects who read it in the context of the fa-mous person's name. Because of the integra-tion process, subjects no longer possessed in-dividual traces of the experimental passageand so on the recognition test could only re-spond on the basis of their holistic schematicknowledge.

There are several problems with these find-ings. First, integration effects were found tobe minimal at brief delay intervals (Sulin &Dooling, 1974); memory was quite accurateimmediately after the story was read. Inte-gration then appears to take place over time,presumably (as Sulin & Dooling explain) be-cause memory for specific detail is only grad-ually lost. And so with time, subject's rec-ollections increasingly depend on schematicknowledge. This view of integration contrastsdirectly with an integration mechanism thatoperates during encoding, the type of inte-gration assumed in the schema formation lit-erature, as well as by Fischhoff (e.g., 1977)and by Loftus (e.g., Loftus et al., 1978).

There are further problems with an assim-ilation interpretation of the Sulin and Dool-ing data. Consider the results subsequentlyreported by Dooling and Christiaansen

SCHEMA THEORIES 219

(1977a).2 In this study, subjects were againeither informed or uninformed about theconnection of the experimenter-provided pas-sages to some famous person. Here, all sub-jects showed very good memory for the orig-inal sentences—even over long durations.Nonetheless, subjects in the famous condi-tion did show a reliable propensity towardmisrecognizing some (but not all) informa-tion about the famous person. One possibleexplanation assumes that the memory rep-resentations for the experimental materialswere equivalent for the two groups of sub-jects. However, subjects who knew of the con-nection with a famous person might easilyhave been operating under response biasesthat were different from those of subjects un-informed about the connection. Informedsubjects might have decided that some of thefoils were old if they believed their memoriesimperfect, since after all, the foils were indeedtrue of the famous person. This response biasinterpretation of these data becomes moreplausible because systematic response biaseshave recently been discovered in research in-vestigating the question of how people dis-criminate among various sources of infor-mation (e.g., originating with themselves vs.with others; see Johnson, Raye, Foley, &Foley, 1981).

To conclude this critique of integration,the processes of schema formation and mod-ification via integration have not receivedvery substantial support. For both processesthere is evidence that integration is not a nec-essary by-product of merely presenting re-lated or orderable information. In both lit-eratures there is evidence that subjects havestored in memory unintegrated informationabout the original situation. That misrecog-nition of related information occurs is notdeniable. The claim that integration is a ma-jor, inevitable product of schema-based pro-cessing is.3

Summary

Schema theory assumes that the represen-tation of a complex event is both incompleteand inaccurate. This is because what is storedis highly selected and depends on whether theappropriate schema is activated, on the

knowledge content of that schema, and onthe importance of the incoming informationto the schema. Further reductions in thememory representation occur because theabstraction process favors the storage ofmeaning rather than of individual lexicalunits or syntactic sequences. Evidence froma sizeable number of studies suggests that thememory trace is richer and more detailedthan these information reduction processeswould allow.

Schema theory also posits two processesthat enrich the representation of complexevents: interpretation and integration. Herewe considered evidence that suggests thatneither of these processes is obligatory. In-deed, subjects can remember uninterpretedand unintegrated information. In severalplaces we considered studies that demon-strated that much of the original support forschema theory encoding processes stemmedfrom procedural peculiarities of landmarkexperiments.

This review has revealed that memory forcomplex events is rather detailed, but someaspects of information are far more easilyaccessed than are others. In the next sectionof the article, we attempt to explain the find-ings reviewed in the first two sections.

Alternative Explanations

A considerable body of research findingsexists that is not easily accounted for by themajor encoding assumptions of schema the-ory. Unfortunately it is not currently possibleto propose an alternative theory that is ca-pable of integrating all of the relevant evi-dence as well as generating new predictionsthat could not themselves be accounted forby schema theory. What appear to exist area set of explanations, not necessarily relatedto one another, each of which can be tied to

2 Dooling and Christiaansen (1977) interpret their owndata, as well as those of Sulin and Dooling (1974), withreference to the operation of reconstruction. That is, asepisodic memory for an event fades, people rely on theirthematic knowledge to reconstruct the event. The pro-cess of reconstruction will be considered in the next sec-tion.

3 See Johnson and Raye (1981) for an argument thata schema is neither a necessary nor logical source ofmisrecognition errors.


a set of findings in the schema literature aswell as to a set of findings elsewhere in thememory literature. We consider some ofthese briefly in this section of the article.Again we consider in turn each major schema-theory encoding process.

Selection

The representation of a prose passage ap-pears to include more details than a selectionprocess allows. Nonetheless a literature existsshowing a predictable advantage in recall forsome ideas and not for others. There is nodoubt that sizeable differences in recall (if notrecognition) are associated with differencesin degree of and/or activation of prior knowl-edge. Clearly, what is stored (or available) inmemory is not always accessible (Tulving &Pearlstone, 1966). The question thus be-comes What is it that determines access-ibility?

Classical research in human memory hasidentified a number of factors that are rele-vant for determining what is recalled. Fourof these are important for present purposes:(a) the nature of connections established dur-ing encoding, (b) differences in the numberand distribution of rehearsals, (c) the role ofretrieval cues at recall, and (d) the order inwhich elements or sets are recalled. We pro-pose that these four factors are important indetermining recall of prose materials. First,however, we need to consider how textualmaterial may be represented in memory.

Text processing theories such as Kintsch's(1974; Kintsch & van Dijk, 1978) assumethat when complex material is presented, a"text base" is formed as part of the compre-hension process. Briefly, a text base is an or-dered hierarchy of propositions with thetheme at the highest level and argument-shar-ing propositions descending from it. All re-lated propositions are interconnected in anetwork due to argument overlap and to suchgrammatical elements as articles (de Villiers,1974) and pronouns (Lesgold, 1972). Notevery proposition will be connected directlyto every other one, but many should be linkedby indirect connections. Finally, if a passagecontains a set of propositions not related tothe theme proposition, these too will be en-

coded but will occupy a place outside of theintegrated network (Kozminsky, 1977).

According to such models, individualpropositions can have either or both of twosets of relations: (a) hierarchical connectionsdescending from the theme and (b) connec-tions that link individual propositions di-rectly to others via shared arguments.4 Com-ponents of these two types of connections canthen serve as retrieval cues and should (as doretrieval cues for simpler materials; e.g., Tulv-ing & Thomson, 1973) play a large part indetermining how many and what particularideas are recalled.

Substantiating evidence exists for such anetwork of relations. For example, when ar-gument overlap is low, as occurs when a newidea cannot be related directly to a previouslymentioned idea (called a violation of the"given-new contract"; Haviland & Clark,1974), both comprehension and recall arepoor (e.g., Haviland & Clark, 1974; Hupet& LeBouedec, 1977; Kieras, 1978; Lesgold,Roth, & Curtis, 1979; Yekovich, Walker, &Blackman, 1979). On the other hand, whenconnections can be established (e.g., by draw-ing inferences to link the new and old ideas)recall is good (Kieras, 1978; Yekovich &Manelis, 1980).

Consider now the Bransford and Johnson(1972) materials that provided the classicdemonstration for the existence of the selec-tion process. Their passages contain no ex-plicit, concrete referents, and without a con-text to suggest exemplars for these referents,none is likely to be inferred. This should in-hibit construction of networks of connectionsamong the sentences at least by preventingextensive use of the given-new strategy. Thiswould serve to reduce the possibility that onesentence could cue another at recall. It is notsurprising then that recall of these materialsis so poor; subjects had in effect been pre-sented with a set of unrelated sentences.

It has long been known of course that unitsrelated to one another either hierarchically

4 Such a notion is reminiscent of views about the con-nections established during the learning of arbitrary pairsof simple verbal units; for example, connections are es-tablished between the two members of each pair as wellas among all items serving as stimuli and all items servingas responses (Segal & Mandler, 1967).

SCHEMA THEORIES 221

or associatively are easier to recall than areunrelated units (e.g., Deese, 1959). Recentdemonstrations of-this point also exist forstimulus sets ranging from individual words(e.g., Bower & Clark-Meyers, 1980; see alsoBower, Clark, Lesgold, & Winzenz, 1969) toprose (Schustack & Anderson, 1979; Thorn-dyke, 1977). When subjects are preventedfrom forming connections either by omittinga cue to the existence of a hierarchy or bypresenting such a cue only after the story hasbeen read, recall is poor for both prose(Schustack & Anderson, 1979; Thorndyke,1977) and simpler verbal materials (Boweret al, 1969).

A similar analysis applies to frame theorystudies in which sentences constituting astory are presented in scrambled versus story-normal order (e.g., Thorndyke, 1977).5 Ran-dom sentence order could easily reduce theability of a subject to determine the connec-tions among ideas or elements in the passage,and the relative absence of such intersentenceconnections should reduce recall. Similarly,recall of stories taken from other culturesmay be poor, in part because of the difficultyAmerican students have in establishing ap-propriate intersentence or prepositional linksduring the encoding process (Kintsch &Greene, 1978; Mandler & Johnson, 1977).

On some occasions, one message may leadto the establishment of two or more inde-pendent prepositional networks, each withits own hierarchical structure. This is the sit-uation that should prevail whenever a storycan be read from two different perspectivesand the ideas related to one perspective areirrelevant to the other. Given one encodingcontext, connections among related sen-tences should be established. All others willexist outside of this network, either as un-related sentences or as sentences related intheir own network. During recall, the prop-osition at the top of the hierarchy (often thetheme) is the most easily recalled proposition(for reasons to be discussed shortly) and isused to access subordinate, related ideas. Ifthe subject uses only the main theme as hisor her retrieval plan, ideas irrelevant to thattheme (i.e., those ideas existing outside of thetext base) may never be accessed. This cor-responds to the findings of Pichert and An-

derson (1977). However, when subjects aregiven the alternative theme as a retrieval cue,previously irrelevant ideas become accessible(Anderson & Pichert, 1978). These newly rel-evant ideas may not be particularly well re-called; after all, they were not encoded in thecontext of their theme, and so the theme'suse as a retrieval cue will likely be limited(Schustack & Anderson, 1979; Thorndyke,1977; Tulving & Thomson, 1973).

The ability to relate sentences to one an-other during encoding may be a determiningfactor involved in recalling the propositionsof a text. If ideas are not or cannot be con-nected to a higher-order cue (Pichert & An-derson, 1977; Thorndyke, 1977) or to eachother (e.g., Bransford & Johnson, 1972;Chiesi et al., 1979) during encoding, recallwill be poor. Retrieval deficits ordinarilyfound with free recall may be circumventedby the use of either retrieval cues or recog-nition tests (e.g., Alba et al,, 1981; Britton etal., 1980; Chiesi et al., 1979).

Similar processes can be used to explainthe importance effect, that is, the fact thatideas important to the theme are more likelyto be recalled than ideas that are not so im-portant. The majority of research on the im-portance variable confirms its role in recall,but again, it is not necessary that the under-lying process be a differential encoding one.A number of recent findings suggest thatitems at all levels of importance receive thesame amount of initial processing (Britton,Meyer, Simpson, Holdredge, & Curry, 1979;Johnson & Scheldt, 1977; Waters, 1978).Since low-importance information has re-cently been shown to be forgotten at the samerate as high-importance information (Chris-tiaansen, 1980), the explanation for the recalladvantage of high-importance informationmust lie with its advantage in accessibility.An explanation favoring this view involvesKintsch and van Dijk's (1978) propositionalhierarchy algorithm for establishing a textbase. If the first idea retrieved is the highestlevel proposition (or theme), recall will occur

5 For recent critiques of story grammars as models ofcomprehension, representation, and recall of text ma-terials see Black and Bower (1980), Brewer and Lichten-stein (1981), and Thorndyke and Yekovich (1980).


in a top-down manner with each recalledproposition cuing one or more below it. Themodel favors the recall of higher level prop-ositions (Bower, 1976). For a low-level prop-osition to be recalled, the complete chain ofpropositions above it must be kept intact.This kind of text analysis has been quite suc-cessful in predicting recall performance(Kintsch & Keenan, 1973; Kintsch et al,1975; Kozminsky, 1977; Meyer, 1975;Thorndyke, 1977). Others have also proposedthat text material is retrieved in a top-downorder (Britton et al., 1979; Waters, 1978;Yekovich & Thorndyke, 1981).

That recall of prose ordinarily proceeds ina top-down manner through a hierarchymeans, of course, that items higher in thehierarchy are likely to be recalled beforeitems lower in the hierarchy. Output order,too, is a variable long known (see Postman& Underwood, 1973; Spear, 1978) to influ-ence recall; items recalled early in a seriesdecrease the probability that other actuallystored items will be recalled.

Kintsch and van Dijk's (1978) model offersa reasonably objective determination ofproposition importance. Important ideas maybe ones that are referred to by others and sotend to be called into working memory morefrequently and/or spend more time in work-ing memory than others. As a consequenceof the rehearsal processes that occur in work-ing memory, such ideas will tend to have con-nections to a greater number of other ideasthan will less important ideas. Componentsof each connection may then serve as re-trieval cues (see also Graesser, 1978a; Graes-seretal., 1980).

Rehearsal now becomes an important de-terminant of recall of prose elements, as ithas long been considered to be in recall ofsimpler units (e.g., Rundus, 1971). It is en-tirely possible that what is loosely called the"theme" is the proposition or concept thatis most frequently rehearsed or referred toduring encoding—hence its relatively easyaccessibility (Hasher & Griffin, 1978; Perfetti& Goldman, 1974).6 Some evidence for therehearsal hypothesis does exist. The moreideas that descend from a particular piece ofinformation, the greater the likelihood of itsrecall (Meyer & McConkie, 1973). Main ac-tions in an episode are recalled better as the

number of related subordinate actions thataccompany them increases (Black & Bower,1979). Rehearsal of connected ideas may alsoaccount for the finding that recall of high-importance ideas is good even when subjects(in this case children) are unable to discrim-inate between ideas of high and low impor-tance (e.g., Brown & Smiley, 1977).

Thus, a number of nonschema theorymemory processes appear to be involved inprose recall: the existence and extent of con-nections among ideas; differential rehearsalprocesses based on (covert) intersentencecuing during the comprehension process; theavailability and usefulness of retrieval cuesat recall; and output-order produced acces-sibility problems. These are all conceptual-izations with origins in nonschema theoryresearch. Such processes, when coupled withcontemporary associationist views of the rep-resentation of prose (e.g., Anderson & Bower,1973; Kintsch & van Dijk, 1978), offer thepossibility of accounting for differences inidea recallability without invoking selectiveencoding.7 More is stored in memory thanany version of a schema-based selective en-coding mechanism can account for. Alter-native conceptions must begin to take thisinto consideration.

Abstraction

Abstraction may also be amenable to ex-planations that emphasize retrieval ratherthan encoding factors. Although meaning isundoubtedly the easiest aspect of a passageto recall, subjects appear to store, and some-times remember, the syntactic and lexicalvehicles by which meaning is conveyed.Meaning may be the most persistent attribute(e.g., McKoon, 1977) in part because it is the

6 Another nonschema explanation of the importanceeffect involves response biases. That is, there seems tobe a predisposition to label as the theme the initiallyencountered idea in a prose passage (Kieras, 1980).Thus,surface structure variables may influence a person'schoice of the most important proposition.

7 We do not wish to ignore the concept of macro-structure in the Kintsch and van Dijk model of prosecomprehension. This aspect of the model is indeed aschema process in that a few important propositions areselected for representation in the macrostructure. Themacrostructure alone, though, cannot account for thelevels and importance effects.

SCHEMA THEORIES 223

information that the perceiver is most intenton remembering (Kintsch, 1977; Wanner,1968) and/or the attribute most rehearsed,and/or the attribute least susceptible to in-terference.

Task demands may also influence the ac-cessibility of nonsemantic detail. For exam-ple, instructions to process grammatical fea-tures of a sentence result in a marked in-crease in memory for those features (e.g.,Anderson & Paulson, 1977; Graesser & Man-dler, 1975; Johnson-Laird & Stevenson, 1970).By contrast, most story materials and in-structions probably emphasize the impor-tance of meaning.

Breakdowns in the retrieval process mayalso account for the occasional failure to re-member lexical details. One theory that as-sumes that memory contains accurate traceseven of complex events proposes that dis-crimination among related traces is a centralsource of lexical errors (e.g., Hayes-Roth &Hayes-Roth, 1977). For example, processingof a lexical unit may activate related, some-times synonymous, traces in memory eitherthrough spread of activation (Hayes-Roth& Hayes-Roth, 1977) or through the implicitgeneration of associative responses (e.g., Un-derwood, 1965). Errors may occur on a mem-ory test if the associated lexical traces havebeen strongly activated and can compete withthe target information. Hence occasionalsubstitution errors may occur even thoughveridical traces are stored in memory.

In summary, there are alternatives toschema theory that can account for the ab-straction errors traditionally explained byschema theory but that also account for thenot infrequent cases of mnemonic accuracywithout appealing to "probable detail" ex-planations.

Interpretation and Integration

These are probably the most interestingand widely discussed schema-theory con-structs. This is so because, together with ab-straction, they are held to be responsible fordistortions of past events, and distortions areinherently interesting. However, distortionsare not really all that common. Although itis the case that schema theory can accountfor them, schema theory is at its weakest in

accounting for accurate memory of pastevents; a fair number of studies have shownrather startling levels of accuracy even whensubjects had no particular reason to expecta memory test (e.g., the work of Kintsch,Bates, and their collaborators).

What is needed, in our view, is a theoryof memory that can account both for accu-racy and for distortion. A recent model ofmemory processes may be a useful candidate(see Johnson & Raye, 1981). According tothis model, memory for an event will nor-mally consist of individual traces from theexternal event as well as traces from any in-ternal (subject-produced) generation doneduring the comprehension process or duringany subsequent reprocessing of the event.Associated with each trace are a variety ofattributes (see Underwood, 1969), and therelative amounts of these attributes serve asa base for distinguishing between internallygenerated and externally derived traces. Whena trace is retrieved during a memory test, theinformation that identifies its origin will beused to determine whether or not the tracewas part of the original stimulus. If a tracewas not part of the original event, and if ac-curacy is encouraged, the trace will not berecalled (nor will it be misrecognized), andaccuracy will result. On some occasions,however, the critical discriminative attributeinformation may become inaccessible, andsource-of-origin discriminations will not bepossible. Misrecognition of information in-ternally generated during encoding (con-structive errors or interpretations) will result(see also Anderson & Bower, 1973; Hayes-Roth & Thorndyke, 1979).

There exists some preliminary evidence tosupport this portion of the Johnson-Raye the-ory. Conditions that make source discrimi-nation more difficult should increase the rateof confabulation. In fact, fact versus fantasydiscrimination ability for inferences de-creases when a delay (Owens et al., 1979) oran interfering task (Brockway et al., 1974) isinserted between acquisition and recall. Dis-crimination increases when source-identify-ing attributes are made more salient (Alba,1981; Johnson et al., 1981).

The reality-monitoring model has theadded advantage of being able to explain awide variety of mnemonic errors. The same


explanation that is used to account for prag-matic inference errors can also be used toaccount for logical inference misrecognitionerrors (e.g., linear orderings), single word er-rors (e.g., Underwood, 1965; Underwood &Freund, 1968), elaborative prose errors (e.g.,Kintsch et al., 1975;Sulin&Dooling, 1974),default value errors (Minsky, 1975), and im-portantly, alleged reconstructive errors (e.g.,Spiro, 1977, 1980a, 1980b). That is, whenresponding on a memory test, people mayconfuse internally generated memory traceswith traces resulting from the encoding of the(external) stimulus array.

Reconstructive errors can be distinguishedfrom most other types of errors discussed inthis article in that they are believed to be theresult of processes occurring at some pointafter initial encoding. The most frequentlycited evidence in support of the reconstruc-tion process is the work of Spiro (1977,1980a, 1980b). As Royer (1977) points out,however, the results do not necessarily implyreconstruction; reconstruction errors may beconsidered highly similar to other inferencesgenerated during encoding and thereby per-fectly consistent with a reality-monitoringaccount of mnemonic error. Although it ismerely speculation at this point, the sameexplanation, combined with other factorsdiscussed below, may account for those re-constructive eiTors predicted by script theoryand known as the typicality effect.

Two other aspects of the reality monitoringtheory can help to explain other forms ofmnemonic inaccuracy. First, the model as-sumes that confidence that a particular traceis internally or externally generated is dis-tributed in a way similar to some models ofrecognition memory (e.g., Atkinson & Juola,1973). That is, some traces are easily iden-tified as internally generated, whereas othersare easily identified as externally derived. Inbetween, a region of uncertainty exists. Howthese items are identified may depend on testconditions. If subjects are urged to be con-servative (e.g., Gauld & Stephenson, 1967)or are given additional time or cues withwhich to make discriminations (e.g., Hasher& Griffin, 1978; Lawson, 1977), fewer inter-nally generated traces in the uncertain regionwill be produced, and accuracy will increase.

Second, Johnson and Raye (1981) reportthat traces in the uncertain region are subjectto response biases. When the source of a traceis uncertain, there is a tendency to attributeit to external sources (Johnson et al., 1981).Thus, when familiar ideas (e.g., thematicideas) are encountered on a recognition test,subjects may misrecognize them not becausethey are confidently recalling them from anintegrated schema, but rather because theyhave a particular response bias. Explanationsthat appeal to reconstruction processes suchas the typicality effect predicted by script the-ory (e.g., Bower et al., 1979; Graesser et al.,1979; Graesser et al., 1980) and schema the-ory (e.g., Sulin & Dooling, 1974) may be bet-ter accommodated by the response biasesdiscussed within the framework of the reality-monitoring model.8

An Overview

As we. noted at the outset, recent researchin cognitive psychology has been greatly in-fluenced by schema theoretics. Although thepresent article has in large part questionedthe specifics of schema theory, there can beno denying its generally beneficial impact onthe field of memory. Schema theory has re-vitalized the area and moved researchers toconsider a number of new issues about thenature of memory as well as about the pa-rameters of the comprehension process (seefor example, Graesser, Hoffman, & Clark,1980; Kieras, 1981; Miller & Kintsch, 1980;Vipond, 1980). Although we have noted thatcertain interpretive processes are not the soledomain of schema theory, it has been within

8 Script theory assumes that highly typical aspects ofscripted events are already stored as part of a script;when such events are actually encountered, there is noneed to restore these aspects. Subjects may, as a result,not be able to distinguish between highly typical aspectsof events that did versus did not actually occur in aparticular instantiation of a script; this is termed thetypicality effect (e.g., Smith & Graesser, 1980). This ef-fect is the single most powerful demonstration of a re-construction process. As just mentioned, response biasesseem a likely nonschema explanation. Some recent ev-idence also suggests that the effect is tied to particularconfusing aspects of the "typical" procedures used intypicality research and is not an inevitable consequenceof activating a script and encountering (or not encoun-tering) highly typical events (Chromiak, Note 3).

SCHEMA THEORIES 225

the context of schema theory that these pro-cesses have been explored.

Nevertheless, we must return to the basicproblem confronting schema theory: its def-inition. Taylor and Crocker (1981) present anexcellent discussion of the adequacy ofschema theory as a theory. We will not at-tempt to elaborate on their discussion exceptto once again point out the nebulous char-acter of the theory and to restate our attemptto place some structure on what has trans-pired in the past 15 years.

We have proposed that schema theoriesshare a belief in one or more of four basicmemory encoding processes: selection, ab-straction, interpretation, and integration.Three of these, selection, abstraction, and in-tegration, are processes that reduce theamount of information that will be stored inmemory. Selection does this by allowing onlysome information to receive further process-ing. Abstraction does this by dropping fromthe representation all but the underlyingmeaning of a message. Integration does thisby creating a holistic representation of what-ever remains after selection and abstractionwith whatever relevant knowledge exists inmemory. This holistic representation doesnot include markers that would enable theperson to inevitably distinguish old from newinformation. The final process, interpreta-tion, is the only one that enriches the mem-ory representation, and it does so by addinginferences, omitted details, and relevant priorknowledge to the incoming message.

A final evaluation of schema theorieshinges on two issues. These are (a) what isreally stored in memory and (b) what causesdistortions of factual information. Our re-view of the literature suggests that memoryfor complex events is far more detailed thanschematic processes would allow. At leastsome schematically unimportant informa-tion is stored. At least some details are storedno matter what the extent of a person's priorknowledge, and no matter whether thatknowledge is activated at encoding, Lexicaland syntactic information, along with se-mantic information, are all accorded repre-sentation in memory. Memory of a topic doesnot seem to consist of a tightly integrated setof all relevant information. Although subjects

may impose their own interpretations on in-coming stimuli, they appear to do so less of-ten than schema theory would suggest. Whenpeople do generate interpretations, sourceidentifying markers are not necessarily lost.

The final questions center on what deter-mines recall. We have suggested that theschema theory research can be accounted forby assuming that a detailed representation ofa complex event is actually stored in mem-ory, perhaps in a format suggested by suchtext-comprehension models as Kintsch andvan Dijk's (1978). What a subject shows ofwhat he or she has stored will hinge on avariety of circumstances, some having to dowith connections formed or not formed dur-ing encoding and with rehearsal processesduring encoding and/or the retention inter-val, as well as with processes such as cuing,output order, response biases, task demands,and so forth, that operate during retrieval.We have argued that there are a variety ofnonschematic concepts available to explainmemory for complex events. Whatever theultimate usefulness of these latter argumentsmay prove to be, however, we think it clearthat the stored record of any event is far moredetailed than prototypical schema theoriesimply. Contemporary theories of memorycannot disregard the richness of the storedtrace.

Reference Notes

1. Day, J., Stein, N. L., Trabasso, T., & Shirey, L. Astudy of inferential comprehension: The use of a storyschema to remember picture sequences. Paper pre-sented at the meeting of the Society for Research inChild Development, San Francisco, 1979.

2. Stein, N. L., & Glenn, C. G. The role of structuralvariation in children's recall of simple stories. Paperpresented at the meeting of the Society for Researchin Child Development, New Orleans, 1977.

3. Chromiak, W. Personal Communication. May 1982.

ReferencesAlba, J. W. The representational nature of inferences.

Unpublished doctoral dissertation. Temple University,Philadelphia, 1981.

Alba, J. W., Alexander, S. G., Hasher, L,, & Caniglia, K.The role of context in the encoding of information.Journal of Experimental Psychology: Human Learningand Memory, 1981, 7, 283-292.

Anderson, J. R. Verbatim and propositional represen-tation of sentences in immediate and long-term mem-


ory. Journal of Verbal Learning and Verbal Behavior,1974, 13, 149-162,

Anderson, J. R. Language, memory and thought. Hills-dale, N.J.: Erlbaum, 1976.

Anderson, J. R., & Bower, G. H. Human associativememory. Washington, D.C.: Winston, 1973.

Anderson, J. R., & Paulson, R. Representation and re-tention of verbatim information. Journal of VerbalLearning and Verbal Behavior, 1977, 16, 439-451.

Anderson, R. C. Encoding processes in the storage andretrieval of sentences. Journal of Experimental Psy-chology, 1971, 91, 338-340.

Anderson, R. C. Substance recall of sentences. QuarterlyJournal of Experimental Psychology, 1974, 26, 530-541.

Anderson, R. C., et al. Instantiation of general terms.Journal of Verbal Learning and Verbal Behavior, 1976,/5, 667-679.

Anderson, R. C., & Pichert, J. W. Recall of previouslyunrecallable information following a shift in perspec-tive. Journal of Verbal Learning and Verbal Behavior,1978, 17, 1-12.

Anisfeld, M., & Knapp, M. Association, synonymity,and directionality in false recognition. Journal of Ex-perimental Psychology, 1968, 77, 171-179.

Atkinson, R. C., & Juola, J. F. Factors influencing speedand accuracy of word recognition. In S, Kornblum(Ed.), Attention and Performance IV. New York: Ac-ademic Press, 1973.

Ausubel, D. P. The use of advance organizers in the learn-ing and retention of meaningful material. Journal ofEducational Psychology, 1960, 51, 267-272.

Ausubel, D. P. Educational psychology: A cognitive view.New York: Holt, Rinehart & Winston, 1968.

Ausubel, D. P., & Fitzgerald, D. The role of discrimi-nability in meaningful verbal learning and retention.Journal of Educational Psychology, 1961, 52, 266-274.

Ausubel, D. P., & Fitzgerald, D. Organizer, general back-ground, and antecedent learning variables in sequen-tial verbal learning. Journal of Educational Psychol-ogy, 1962, 53, 243-249.

Baker, L. Processing temporal relationships in simplestories: Effects of input sequence. Journal of VerbalLearning and Verbal Behavior, 1978, 17, 559-572.

Barclay, J. R. The role of comprehension in remember-ing sentences. Cognitive Psychology, 1973,4,229-254.

Barnes, B. R., & Clawson, E. U. Do advance organizersfacilitate learning? Recommendations for further re-search based on an analysis of 32 studies. Review ofEducational Research, 1975, 45, 637-659.

Bartlett, F. C. Remembering: A study in experimentaland social psychology. Cambridge, England: Cam-bridge University Press, 1932.

Bates, E., Kintsch, W., Fletcher, C. R., & Giuliani, V.The role of pronominalization and ellipsis in texts:Some memory experiments. Journal of ExperimentalPsychology: Human Learning and Memory, 1980, 6,676-691.

Bates, E., Masling, M., & Kintsch, W. Recognition mem-ory for aspects of dialogue. Journal of ExperimentalPsychology: Human Learning and Memory, 1978, 4,187-197.

Begg, I. Recognition memory for sentence meaning and

wording. Journal of Verbal Learning and Verbal Be-havior, 1971, 10, 176-181.

Begg, I., & Wickelgren, W. A. Retention functions forsyntactic and lexical vs. semantic information in sen-tence recognition memory. Memory & Cognition,1974, 2, 353-359.

Bekerian, D. A., & Bowers, J. M. Eyewitness testimony:Were we misled? Journal of Experimental Psychology:Learning, Memory, and Cognition, 1983, 9, 139-145.

Birnbaum, I. M., Johnson, M. K., Hartley, J. T, & Tay-lor, T. H. Alcohol and elaborative schemas for sen-tences. Journal of Experimental Psychology: HumanLearning and Memory, 1980, 6, 293-300.

Black, J. B., & Bower, G. H. Episodes as chunks in nar-rative memory. Journal of Verbal Learning and VerbalBehavior, 1979, 18, 309-318.

Black, J. B., & Bower, G. H. Story understanding as prob-lem-solving. Poetics, 1980, 9, 233-250.

Bobrow, S. A. Memory for words in sentences. Journalof Verbal Learning and Verbal Behavior, 1970, 9, 363-372.

Bock, J. K. The effect of a pragmatic presupposition onsyntactic structure in question answering. Journal ofVerbal Learning and Verbal Behavior, 1977, 16, 723-734.

Bock, J. K., & Brewer, W. F. Reconstructive recall insentences with alternative surface structures. Journalof Experimental Psychology, 1974, 103, 837-843.

Bower, G. H. Experiments on story understanding andrecall. Quarterly Journal of Experimental Psychology,1916,28, 511-534.

Bower, G. H., Black, J. B., & Turner, T. J. Scripts inmemory for text. Cognitive Psychology, 1979, 11,177-220.

Bower, G. H., Clark, M. C., Lesgold, A. M., & Winzenz,D. Hierarchical retrieval schemas in recall of cate-gorized word lists. Journal of Verbal Learning andVerbal Behavior, 1969, 8, 323-343.

Bower, G. H., & Clark-Meyers, G, Memories for scriptswith organized vs. randomized presentations. BritishJournal of Psychology, 1980, 71, 369-397.

Bower, G. H., Karlin, M. B., & Dueck, A. Comprehen-sion and memory for pictures. Memory & Cognition,1975, 3, 216-220.

Bransford, J. D., Barclay, J. R., & Franks, J. J. Sentencememory: A constructive vs. interpretive approach.Cognitive Psychology, 1972, 3, 193-209.

Bransford, J. D., & Franks, J. J. The abstraction of lin-guistic ideas. Cognitive Psychology, 1971,2, 331-350.

Bransford, J. D., & Johnson, M. K. Contextual prereq-uisites for understanding: Some investigations of com-prehension and recall. Journal of Verbal Learning andVerbal Behavior, 1972, //, 717-726.

Bransford, J. D., & Johnson, M. K. Considerations ofsome problems of comprehension. In W. G. Chase(Ed.), Verbal information processing. New York: Ac-ademic Press, 1973.

Bransford, J. D., & Nitsch, K. E. Coming to understandthings we could not previously understand. In J. F.Kavanagh & W. Strange (Eds.), Speech and languagein the laboratory, school, and clinic. Cambridge, Mass:MIT Press, 1978.

Brewer, W. F. Memory for ideas: Synonym substitution.Memory & Cognition, 1975, 3, 458-464.

SCHEMA THEORIES 227

Brewer, W. F. Memory for the pragmatic implicationsof sentences. Memory & Cognition, 1977, 5, 673-678.

Brewer, W. F., & Lichtenstein, E. H. Memory for markedsemantic features versus memory for meaning. Jour-nal of Verbal Learning and Verbal Behavior, 1974,13,172-180.

Brewer, W. F, & Lichtenstein, E. H. Event schemas, storyschemas, and story grammars. In J. Long & A. Bad-deley (Eds.), Attention and performance IX. Hillsdale,N.J.: Erlbaum, 1981.

Brewer, W. F., & Treyens, J. C. Role of schemata inmemory for places. Cognitive Psychology, 1981, 13,207-230.

Britton, B. K., Meyer, B. J., Hodge, M. H., & Glynn,S. M. Effects of the organization of text on memory:Tests of the retrieval and response criterion hy-potheses. Journal of Experimental Psychology: HumanLearning and Memory, 1980,6,620-629.

Britton, B. K., Meyer, B. J. F., Simpson, R., Holdredge,T. S., & Curry, C. Effects of the organization of texton memory: Tests of two implications of'a selectiveattention hypothesis. Journal of Experimental Psy-chology: Human Learning and Memory, 1979, 5,496-506.

Brockway, J. F., Chmielewski, D., & Cofer, C. N. Re-membering prose: Productivity and accuracy con-straints in recognition memory. Journal of VerbalLearning and Verbal Behavior, 1974, 13, 194-208.

Brown, A. L., & Smiley, S. S. Rating the importance ofstructural units of prose passages: A problem of meta-cognitive development. Child Development, 1977, 48,1-8.

Brown, A. L., Smiley, S. S., Day, J. D., Townsend,M. A. R., & Lawton, S. C. Intrusion of a thematicidea in children's comprehension and retention of sto-ries. Child Development, 1911,48, 1454-1466.

Bugelski, B. R. Images as mediators in one-trial paired-associate learning. II: Self-timing in successive lists.Journal of Experimental Psychology, 1968, 77, 328-334.

Burnett, S. A., & Stevenson, J. M. Instructional effectson frequency judgments for sentence meanings. Amer-ican Journal of Psychology, 1979, 92, 711-721.

Chase, W. G., & Simon, H. A. Perception in chess. Cog-nitive Psychology, 1973, 4, 55-81.

Chi, M. T. H. Knowledge structures and memory de-velopment. In R. S. Siegler (Ed.), Children's thinking:What develops^ Hillsdale, N.J.: Erlbaum, 1978.

Chiesi, H. L., Spilich, G. J., & Voss, J. F. Acquisition ofdomain-related information in relation to high andlow domain knowledge. Journal of Verbal Learningand Verbal Behavior, 1979, 18, 257-274.

Christiaansen, R. E. Prose memory: Forgetting rates formemory codes. Journal of Experimental Psychology:Human Learning and Memory, 1980, 6, 611-619.

Christie, D. J., & Schumacher, G. M. Developmentaltrends in the abstraction and recall of relevant versusirrelevant thematic information from connected ver-bal materials. Child Development, 1975, 46, 598-602.

Clark, H. H., & Clark, E.-V. Semantic distinctions andmemory for complex sentences. Quarterly Journal ofExperimental Psychology, 1968, 20, 129-138.

Cofer, C. N. Constructive processes in memory. Amer-ican Scientist, 1973, 61, 537-543.

Cofer, C. N., Chmielewski, D. L., & Brockway, J, F. Con-structive processes and the structure of human mem-ory. In C. N. Cofer (Ed.), The structure of humanmemory. San Francisco: Freeman, 1976.

Corbett, A. T., & Dosher, B. A. Instrument inferencesin sentence encoding. Journal of Verbal Learning andVerbal Behavior, 1978, 17, 479-491.

Deese, J. On the prediction of occurrence of particularverbal intrusions in immediate recall. Journal of Ex-perimental Psychology, 1959, 58, 17-22.

de Villiers, P. A. Imagery and theme in recall of con-nected discourse. Journal of Experimental Psychology,1974, 103, 263-268.

Dodd, D. H., & Bradshaw, J. M. Leading questions andmemory: Pragmatic constraints. Journal of VerbalLearning and Verbal Behavior, 1980, 19, 695-704.

Dooling, D. J., & Christiaansen, R. E. Episodic and se-mantic aspects of memory for prose. Journal of Ex-perimental Psychology: Human Learning and Mem-ory, 1977, 3, 428-436. (a)

Dooling, D. J., & Christiaansen, R. E. Levels of encodingand retention of prose. In G. H. Bower (Ed.), Thepsychology of learning and motivation (Vol. 11). NewYork: Academic Press, 1977. (b)

Dooling, D. J., & Lachman, R. Effects of comprehensionon retention of prose. Journal of Experimental Psy-chology, 1971, 88, 216-222.

Dooling, D. J., & Mullet, R. L. Locus of thematic effectsin retention of prose. Journal of Experimental Psy-chology, 1973, 97, 404-406.

Pass, W., & Schumacher, G. M. Schema theory and proseretention: Boundary conditions for encoding and re-trieval effects. Discourse Processes, 1981, 4, 17-26.

Fischhoff, B. Hindsight + foresight: The effect of out-come knowledge on judgment under uncertainty.Journal of Experimental Psychology: Human Perfor-mance and Perception, 1975, 1, 288-299.

Fischhoff, B. Perceived informativeness of facts. Journalof Experimental Psychology: Human Performance andPerception, 1977, 3, 349-358.

Fischhoff, B., & Beyth, R. I knew it would happen—Remembered probabilities of once-future things. Or-ganizational Behavior and Human Performance, 1975,13, 1-16.

Flagg, P. W. Semantic integration in sentence memory?Journal of Verbal Learning and Verbal Behavior, 1976,75,491-504.

Flagg, P. W., & Reynolds, A. G. Modality of presentationand blocking in sentence recognition memory. Mem-ory & Cognition, 1977, 5, 111-115.

Frederiksen, C. H. Acquisition of semantic informationfrom discourse: Effects of repeated exposures. Journalof Verbal Learning and Verbal Behavior, 1975, 14,158-169. (a)

Frederiksen, C. H. Effects of context-induced processingoperations on semantic information acquired from dis-course. Cognitive Psychology, 1975, 7, 139-166. (b)

Gauld, A., & Stephenson, G. M. Some experiments re-lating to Bartlett's theory of remembering. BritishJournal of Psychology, 1967, 58, 39-49.

Centner, D., & Loftus, E. F. Integration of verbal andvisual information as evidenced by distortions in pic-ture memory. American Journal of Psychology, 1979,92, 366-375.


Glenn, C. G. The role of episodic structure and of storylength in children's recall of simple stories. Journalof Verbal Learning and Verbal Behavior, 1978, 17,229-247.

Goetz, E. T. Inferring from text: Some factors influencingwhich inferences will be made. Discourse Processes,1979, 2, 179-195.

Gomulicki, B. R. Recall as an abstractive process. AdaPsychologies 1956, 12, 77-94.

Graesser, A. C. How to catch a fish: The memory andrepresentation of common procedures. Discourse Pro-cesses, 1978, 7, 72-89. (a)

Graesser, A. C. Tests of a holistic chunking model ofsentence memory through analyses of noun intru-sions. Memory & Cognition, 1978, 6, 527-536. (b)

Graesser, A. C., Gordon, S. E., & Sawyer, J. D. Recog-nition memory for typical and atypical actions inscripted activities: Tests of a script pointer + tag hy-pothesis. Journal of Verbal Learning and Verbal Be-havior, 1919,18, 319-332.

Graesser, A. C., Hoffman, N. L., & Clark, L. F. Structuralcomponents of reading time. Journal of Verbal Learn-ing and Verbal Behavior, 1980, 19, 135-151.

Graesser, A., II, & Mandler, G. Recognition memory forthe meaning and surface structure of sentences. Jour-nal of Experimental Psychology: Human Learning andMemory, 1975, 1, 238-248.

Graesser, A. C, Robertson, S. P., Lovelace, E. R., &Swinehart, D. M. Answers to why-questions exposethe organization of story plot and predict recall ofactions. Journal of Verbal Learning and Verbal Be-havior, 1980, 7 9, 110-119.

Graesser, A. C., Woll, S. B., Kowalski, D. J., & Smith,D. A. Memory for typical and atypical actions inscripted activities. Journal of Experimental Psychol-ogy: Human Learning and Memory, 1980,6, 503-515.

Greene, E., Flynn, M. S., & Loftus, E. F. Inducing re-sistance to misleading information. Journal of VerbalLearning and Verbal Behavior, 1982, 21, 207-219.

Greenwald, A. G. The totalitarian ego: Fabrication andrevision of personal history. American Psychologist,1980,55, 603-618.

Griggs, R. A. The recall of linguistic ideas. Journal ofExperimental Psychology, 1974, 103, 807-809.

Griggs, R. A., & Keen, D. M. The role of test procedurein linguistic integration studies. Memory & Cognition,1977,5, 685-689.

Gude, C., & Zechmeister, E. B. Frequency judgments forthe "gist" of sentences. American Journal of Psychol-ogy, 1975, 88, 385-396.

Hannigan, M. L. The effects of frameworks on sentenceperception and memory. Unpublished doctoral dis-sertation, Vanderbilt University, 1976.

Harris, R. J. Memory and comprehension of implica-tions and inferences of complex sentences. Journal ofVerbal Learning and Verbal Behavior, 1974, 13,626-637.

Harris, R. J., & Monaco, G. E. Psychology of pragmaticimplication: Information processing between the lines.Journal of Experimental Psychology: General, 1978,707, 1-22.

Hasher, L., Attig, M. S., & Alba, J. W. I knew it all along:Or did I? Journal of Verbal Learning and Verbal Be-havior, 1981, 20, 86-96.

Hasher, L., & Griffin, M. Reconstructive and reproduc-tive processes in memory. Journal of ExperimentalPsychology: Human Learning and Memory, 1978, 4,318-330.

Haviland, S. E., & Clark, H. H. What's new? Acquiringnew information as a process in comprehension. Jour-nal of Verbal Learning and Verbal Behavior, 1974,13,512-521.

Hayes-Roth, B., & Hayes-Roth, F. The prominence oflexical information in memory representations ofmeaning. Journal of Verbal Learning and Verbal Be-havior, 1977, 16, 119-136.

Hayes-Roth, B., & Thorndyke, P. W. Integration ofknowledge from texts. Journal of Verbal Learning andVerbal Behavior, 1979, 18, 91-108.

Hupet, M., & LeBouedec, B. The given-new contractand the constructive aspect of memory for ideas. Jour-nal of Verbal Learning and Verbal Behavior, 1977,16,69-75.

James, C. T., & Hillinger, M. L. The role of confusionin the semantic integration paradigm. Journal of Ver-bal Learning and Verbal Behavior, 1977, 76,711-721.

James, C. T, Hillinger, M. L., & Murphy, B. J. The effectof complexity on confidence ratings in linguistic in-tegration. Memory & Cognition, 1977, 5, 355-361.

James, C. T, Thompson, J. G., & Baldwin, J. M. Thereconstructive process in sentence memory. Journalof Verbal Learning and Verbal Behavior, 1973, 12,51-63.

Jarvella, R. J. Syntactic processing of connected speech.Journal of Verbal Learning and Verbal Behavior, 1971,10, 409-416.

Johnson, M. K., Bransford, J. D., & Solomon, S. K.Memory for tacit implications of sentences. Journalof Experimental Psychology, 1973, 98, 203-205.

Johnson, M. K., Doll, T. J., Bransford, J. D., & Lapinski,R. H. Context effects in sentence memory. Journal ofExperimental Psychology, 1974, 103, 358-360.

Johnson, M. K., & Raye, C. L. Reality monitoring. Psy-chological Revie\v, 1981, 88, 67-85.

Johnson, M. K., Raye, C. L., Foley, H. J., & Foley,M. A. Cognitive operations and decision bias in realitymonitoring. American Journal of Psychology, 1981,94, 37-64.

Johnson, R. E. Recall of prose as a function of the struc-tural importance of the linguistic unit. Journal oj'Ver-bal Learning and Verbal Behavior, 1970, 9, 12-20.

Johnson, R. E., & Scheldt, B. J. Organizational encod-ings in the serial learning of prose. Journal of VerbalLearning and Verbal Behavior, 1977, 76, 575-588.

Johnson-Laird, P. N., & Stevenson, R. Memory for syn-tax. Nature, 1970, 227,412.

Katz, S., Atkeson, B., & Lee, J. The Bransford-Frankslinear effect: Integration or artifact? Memory & Cog-nition, 1974,2,709-713.

Katz, S., & Gruenewald, P. The abstraction of linguisticideas in "meaningless" sentences. Memory & Cogni-tion, 1974,2,737-741.

Keenan, J. M., & Kintsch, W. The identification of ex-plicitly and implicitly presented information. In W.Kintsch (Ed.), The representation of meaning in mem-ory. Hillsdale, N.J.: Erlbaum, 1974.

Keenan, J. M., MacWhinney, B., & Mayhew, D. Prag-matics in memory: A study of natural conversation.

SCHEMA THEORIES 229

Journal of Verbal Learning and Verbal Behavior, 1977,763 549-560.

Kemper, S. Memory for the form and force of declara-tives and interrogatives. Memory & Cognition, 1980,5,367-371.

Kieras, D. E. Good and bad structure in simple para-graphs: Effects on apparent theme, reading time, andrecall. Journal of Verbal Learning and Verbal Behav-ior, 1978, 17, 13-28.

Kieras, D. E. Initial mention as a signal to thematiccontent in technical passages. Memory & Cognition,1980, S, 345-353.

Kieras, D. E. Component processes in the comprehen-sion of simple prose. Journal of Verbal Learning andVerbal Behavior, 1981, 20, 1-23.

'King, D. R. W., & Greeno, J. G. Invariance of inferencetime when information was presented in different lin-guistic formats. Memory and Cognition, 1974,2, 233-235.

Kintsch, W. The representation of meaning in memory.Hillsdale, N.J.: Erlbaum, 1974.

Kintsch, W. Memory and cognition. New York: Wiley,1977.

Kintsch, W., & Bates, E. Recognition memory for state-ments from a classroom lecture. Journal of Experi-mental Psychology: Human Learning and Memory,1977, 3, 150-159.

Kintsch, W., & Greene, E. The role of culture-specificschemata in the comprehension and recall of stories.Discourse Processes, 1978, 1, 1-13.

Kintsch, W., & Keenan, J. M. Reading rate and retentionas a function of the number of propositions in thebase structure of sentences. Cognitive Psychology,1973, 5, 257-274.

Kintsch, W., & van Dijk, T. A. Toward a model of textcomprehension and production. Psychological Re-vim, 1978,55,363-394.

Kintsch, W., Kozminsky, E., Streby, W. J., McKoon, G.,& Keenan, J. M. Comprehension and recall of text asa function of content variables. Journal of VerbalLearning and Verbal Behavior, 1975, 14, 196-214.

Kintsch, W., Mandel, T. S., & Kozminsky, E. Summa-rizing scrambled stories. Memory & Cognition, 1977,5, 547-522.

Kintsch, W., & Monk, D. Storage of complex infor-mation in memory: Some implications of the speedwith which inferences can be made. Journal of Ex-perimental Psychology, 1972, 94, 25-32.

Kozminsky, E. Altering comprehension: The effect ofbiasing titles on text comprehension. Memory & Cog-nition, 1977, 5, 482-490.

Lawson, R. Representation of individual sentences andholistic ideas. Journal of Experimental Psychology:Human Learning and Memory, 1977,3, 1-9.

Lesgold, A. M. Pronominalization: A device for unifyingsentences in memory. Journal of Verbal Learning andVerbal Behavior, 1972, 77, 316-323.

Lesgold, A. M., Roth, S. F., & Curtis, M. E. Foreground-ing effects in discourse comprehension. Journal ofVerbal Learning and Verbal Behavior, 1979, 75, 291-308.

Lichtenstein, E. H., & Brewer, W. F. Memory for goal-directed events. Cognitive Psychology, 1980, 12, 412-445.

Loftus, E. F. Leading questions and the eyewitness re-port. Cognitive Psychology, 1975, 7, 560-572.

Loftus, E. F. Eyewitness testimony. Cambridge, Mass.:Harvard University Press, 1979. (a)

Loftus, E. F. Reactions to blatantly contradictory infor-mation. Memory & Cognition, 1979, 7, 368-374. (b)

Loftus, E. F., & Loftus, G. R. On the permanence ofstored information in the human brain. AmericanPsychologist, 1980, 35, 409-420.

Loftus, E. F., Miller, D. G., & Burns, H. J. Semanticintegration of verbal information into a visual mem-ory. Journal of Experimental Psychology: HumanLearning and Memory, 1978, 4, 19-31.

Loftus, E. F., & Palmer, J. C. Reconstruction of auto-mobile destruction: An example of the interactionbetween language and memory. Journal of VerbalLearning and Verbal Behavior, 1974, 13, 585-589.

Mandler, J. M. A code in the node: The use of a storyschema in retrieval. Discourse Processes, 1978,7, 14-35.

Mandler, J. M., & Johnson, N. S. Remembrance ofthings parsed: Story structure and recall. CognitivePsychology, 1977,9, 111-151.

McKoon, G. Organization of information in text mem-ory. Journal of Verbal Learning and Verbal Behavior,1977, 16, 247-260.

McKoon, G., & Keenan, J. M. Response latencies toexplicit and implicit statements as a function of thedelay between reading and test. In W. Kintsch (Ed.),The representation of meaning in memory. Hillsdale,N.J.: Erlbaum, 1974.

Meyer, B. J. F. The organization of prose and its effectson recall. Amsterdam: North Holland, 1975.

Meyer, B. J. F. The structure of prose: Effects on learningand memory and implications for educational prac-tice. In R. C. Anderson, R. J. Spiro, & W. E. Montague(Eds.), Schooling and the acquisition of knowledge.Hillsdale, N.J.: Erlbaum, 1977.

Meyer, B. J. F., & McConkie, G. What is recalled afterhearing a passage? Journal of Educational Psychology,1973, 65, 109-117.

Meyers, L. S., & Boldrick, D. Memory for meaningfulconnected discourse. Journal of Experimental Psy-chology: Human Learning and Memory, 1975,1, 584-591.

Miller, J. R., & Kintsch, W. Readability and recall ofshort prose passages: A theoretical analysis. Journalof Experimental Psychology: Human Learning andMemory, 1980, 6, 335-354.

Minsky, M. A framework for representing knowledge.In P. H. Winston (Ed.), The psychology of computervision. New York: McGraw-Hill, 1975.

Moeser, S. D. Inferential reasoning in episodic memory.Journal of Verbal Learning and Verbal Behavior, 1976,15, 193-212.

Moeser, S. D. Recognition processes in episodic memory.Canadian Journal of Psychology, 1977, 31, 41-70.

Moeser, S. D., & Tarrant, B. L. Learning a network ofcomparisons. Journal of Experimental Psychology:Human Learning and Memory, 1977, 3, 643-659.

Morris, C. D., Bransford, J. D., & Franks, J. J. Levelsof processing versus transfer appropriate processing,Journal of Verbal Learning and Verbal Behavior, 1977,76,519-533.


Morris, C. D., Stein, B. S., & Bransford, J, D. Prereq-uisites for the utilization of knowledge in the recall ofprose passages. Journal of Experimental Psychology:Human Learning and Memory, 1979, 5, 253-261.

Norman, D. A., & Rumelhart, D. E. Memory andknowledge. In D. A. Norman & D. E. Rumelhart(Eds.), Explorations in cognition. San Francisco: Free-man, 1975,

Olson, D. R., & Filby, N. On the comprehension of activeand passive sentences. Cognitive Psychology, 1972, 3,361-381.

Ortony, A., Schallert, D. L., Reynolds, R. E., & Antos,S. J. Interpreting metaphors and idioms: Some effectsof context on comprehension. Journal of VerbalLearning and Verbal Behavior, 1978, 17, 465-478.

Owens, J., Bower, G. H., & Black, J. B. The "soap-opera"effect in story recall. Memory & Cognition, 1979, 7,185-191.

Paris, S. G., & Carter, A. Semantic and constructive as-pects of sentence memory in children. DevelopmentalPsychology, 1973, 9, 189-197.

Paris, S. G., & Lindauer, B. K. The role of inference inchildren's comprehension and memory for sentences.Cognitive Psychology, 1976, 8, 217-227.

Paris, S. G., & Lindauer, B. K. Constructive aspects ofchildren's comprehension and memory. In R. V. Kail,Jr. & J. W. Hagen (Eds.), Perspectives on the devel-opment of memory and cognition. Hillsdale, N.J.: Erl-baum, 1977.

Paris, S. G., Lindauer, B. K., & Cox, G. L. The devel-opment of inferential comprehension. Child Devel-opment, 1977, 48, 1728-1733.

Paris, S. G., & Upton, L. R. Children's memory forinferential relationships in prose. Child Development,1976, 47,660-668.

Paul, I. H. Studies in remembering: The reproductionof connected and extended verbal material. Psycho-logical Issues, 1959, 1(2, Whole No. 2).

Perfetti, C. A., & Goldman, S. R. Thematization andsentence retrieval. Journal of Verbal Learning andVerbal Behavior, 1974, '13, 70-79.

Pichert, J. W., & Anderson, R. C. Taking different per-spectives on a story. Journal of Educational Psychol-ogy, 1977, 69, 309-315.

Postman, L., & Underwood, B, J. Critical issues in in-terference theory. Memory & Cognition, 1973, 1, 19-40.

Potts, G. R. Memory for redundant information. Mem-ory & Cognition, 1973, 7, 467-470.

Potts, G. R. Artificial logical relations and their relevanceto semantic memory. Journal of Experimental Psy-chology: Human Learning and Memory, 1976, 2,746-758.

Potts, G. R. Integrating new and old information. Jour-nal of Verbal Learning and Verbal Behavior, 1977,16,305-320.

Potts, G. R., Keller, R. A., & Rooley, C. J. Factors af-fecting the use of world knowledge to complete a linearordering. Journal of Experimental Psychology: HumanLearning and Memory, 1981, 7, 254-268.

Prawat, R. S., & Cancelli, A. Constructive memory inconserving and nonconserving first graders. Develop-mental Psychology, 1976, 12, 47-50.

Reitman, J. S., & Bower, G. H. Storage and later rec-ognition of exemplars of concepts. Cognitive Psy-chology, 1973, 4, 194-206.

Royer, J. M. Remembering: Constructive or reconstruc-tive? Comments on chapter 5 by Spiro. In R. C. An-derson, R. J. Spiro, & W. E. Montague (Eds.), School-ing and the acquisition of knowledge. Hillsdale, N.J.:Erlbaum, 1977.

Royer, J. M., & Cable, G. W. Facilitative transfer in proselearning. Journal of Educational Psychology, 1975,67,116-123.

Royer, J. M., & Cable, G. W. Illustrations, analogies, andfacilitative transfer in prose learning. Journal of Ed-ucational Psychology, 1976, 68, 205-209.

Royer, J. M., & Perkins, M. R. Facilitative transfer inprose learning over an extended period of time. Jour-nal of Reading Behavior, 1977, 9, 185-188.

Royer, J. M., Perkins, M. R., & Konold, C. E. Evidencefor a selective storage mechanism in prose learning.Journal of Educational Psychology, 1978, 70, 457-462.

Rubin, D. C. Very long-term memory for prose andverse. Journal of Verbal Learning and Verbal Behavior,1977,76,611-621. .

Rundus, D. Analysis of rehearsal processes in free recall.Journal of Experimental Psychology, 1971, 89, 63-77.

Sachs, J. S. Recognition memory for syntactic and se-mantic aspects of connected discourse. Perception andPsychophysics, 1967, 2, 437-442.

Sachs, J. S. Memory in reading and listening to dis-course. Memory & Cognition, 1974, 2, 95-100.

Schallert, D. L. Improving memory for prose: The re-lationship between depth of processing and context.Journal of Verbal Learning and Verbal Behavior, 1976,15, 621-632.

Schank, R. C. Conceptual dependency: A theory of nat-ural language understanding. Cognitive Psychology,1972,3, 552-631.

Schank, R. C. The role of memory in language process-ing. In C. N. Cofer (Ed.), The structure of humanmemory. San Francisco: Freeman, 1976.

Schank, R. C., & Abelson, R. Scripts, plans, goals, andunderstanding. Hillsdale, N.J.: Erlbaum, 1977.

Schustack, M. W., & Anderson, J. R. Effects of analogyto prior knowledge on memory for new information.Journal of Verbal Learning and Verbal Behavior, 1979,18, 565-584.

Schweller, K. G., Brewer, W. R, & Dahl, D. A. Memoryfor illocutionary forces and perlocutionary effects ofutterances. Journal of Verbal Learning and VerbalBehavior, 1976, 75, 325-337.

Segal, M. A., & Mandler, G. Directionality and organi-zational processes in paired-associate learning. Jour-nal of Experimental Psychology, 1967, 74, 305-312.

Shaughnessy, J. J., & Mand, J. L. How permanent arememories for real life events? American Journal ofPsychology, 1982, 95, 51-65.

Singer, M. A replication of Bransford and Franks' (1971)"The abstraction of linguistic ideas" Bulletin of thePsychonomic Society, 1973, 7, 416-418.

Singer, M. Processes of inference during sentence en-coding. Memory & Cognition, 1979, 7, 192-200.

Singer, M. Verifying the assertions and implications oflanguage. Journal of Verbal Learning and Verbal Be-havior, 1981,20, 46-60.

Small, D. W. The abstraction of arbitrary categories.Memory & Cognition, 1975, 3, 581-585.

Smiley, S. S., Oakley, D. D., Worthen, D., Campione,J. C., & Brown, A. L. Recall of thematically relevant

SCHEMA THEORIES 231

material by adolescent good and poor readers as afunction of written versus oral presentation. Journalof Educational Psychology, 1977,69, 381-387.

Smith, D. A., & Graesser, A. C. Memory for actions inscripted activities as a function of typicality, retentioninterval, and retrieval task. Memory & Cognition,1981, 9, 550-559.

Smith, M. E. Delayed recall of previously memorizedmaterial after twenty years. Journal of Genetic Psy-chology, 1935, 47,-477-481.

Soli, S. D., & Balch, W. R. Performance biases and rec-ognition memory for semantic and formal changes inconnected discourse. Memory & Cognition, 1976, 4,673-676.

Spear, N. E. The processing of memories: Forgetting andretention. Hillsdale, N.J.: Erlbaum, 1978.

Spilich, G. J., Vesonder, G. T.( Chiesi, H. L., & Voss,J. F. Text processing of domain-related informationfor individuals with high and low domain knowledge.Journal of Verbal Learning and Verbal Behavior, 1979,18, 275-290,

Spiro, R. J. Remembering information from text: The"state of schema" approach. In R, C. Anderson,R. J. Spiro, & W. E. Montague (Eds.), Schooling andthe acquisition of knowledge. Hillsdale, N.J.: Erlbaum,1977.

Spiro, R. J. Accommodative reconstruction in prose re-call. Journal of Verbal Learning and Verbal Behavior,1980, 19, 84-95. (a)

Spiro, R. J. Prior knowledge and story processing: In-tegration, selection, and variation. Poetics, 1980, 9,313-327. (b)

Stein, N. L., & Glenn, C. G. An analysis of story com-prehension in elementary school children. In R. O.Freedle (Ed.), New directions in discourse processing(Vol. 2). Hillsdale, N.J.: Erlbaum, 1979.

Stein, N. L., & Nezworski, T. The effects of organizationand instructional set on story memory. Discourse Pro-cesses, 1978, ;, 177-193.

Sulin, R. A., & Dooling, D. J. Intrusion of a thematicidea in retention of prose. Journal of ExperimentalPsychology, 1974, 103, 255-262.

Taylor, S. E., & Crocker, J. Schematic bases of socialinformation processing. In E. T. Higgins, C. P. Her-man, & M. P. Zanna, (Eds.), Social cognition: TheOntario Symposium (Vol. 1). Hillsdale, N.J.: Erlbaum,1981.

Thorndyke, P. W. The role of inferences in discoursecomprehension. Journal of Verbal Learning and Ver-bal Behavior, 1976, 15, 437-446.

Thorndyke, P. W. Cognitive structures in comprehensionand memory of narrative discourse. Cognitive Psy-chology, 1977, 9, 77-110.

Thorndyke, P. W., & Yekovich, F. R. A critique of,schema-based theories of human story memory. Po-etics, 1980, 9, 23-49.

Tulving, E. Subjective organization in free recall of "un-related" words. Psychological Review, 1962, 69,344-354.

Tulving, E., & Pearlstone, Z. Availability versus acces-sibility of information in memory for words. Journal

of Verbal Learning and Verbal Behavior, 1966, 5,381-391.

Tulving, E., & Thomson, D. M. Encoding specificity andretrieval processes in episodic memory. PsychologicalReview, 1973, SO, 352-373.

Tzeng, O. J. L. Sentence memory: Recognition and in-ferences. Journal of Experimental Psychology: HumanLearning and Memory, 1975, /, 720-726.

Underwood, B. J. False recognition produced by implicitverbal responses. Journal of Experimental Psychology,1965, 70, 122-129.

Underwood, B. J. Stimulus selection in verbal learning.In C. N. Cofer & B. S. Musgrave (Eds.), Verbal be-havior and learning. New York: McGraw-Hill, 1963.

Underwood, B. J. Attributes of memory. PsychologicalReview, 1969, 76, 559-573.

Underwood, B. J., & Freund, J. S. Errors in recognitionlearning and retention. Journal of Experimental Psy-chology, 1968, 78, 55-63.

Underwood, B. J,, & Schulz, R. W. Meaningfulness andverbal learning. Philadelphia: Lippincott, 1960.

Vipond, D. Micro- and macroprocesses in text compre-hension. Journal of Verbal Learning and Verbal Be-havior, 1980,79,276-296.

Walker, C. H., & Meyer, B. J. F. Integrating differenttypes of information in text. Journal of Verbal Learn-ing and Verbal Behavior, 1980, 19, 263-275.

Walsh, D. A., & Baldwin, M. Age differences in inte-grated semantic memory. Developmental Psychology,1977, 13, 509-514.

Wanner, H. E. On remembering, forgetting, and under-standing sentences: A study of the deep structure hy-pothesis. Unpublished doctoral dissertation, HarvardUniversity, Cambridge, Mass., 1968.

Waters, H. S. Superordinate-subordinate structure insemantic memory: The roles of comprehension andretrieval processes. Journal of Verbal Learning andVerbal Behavior, 1978, 17, 587-598.

Wilkes, A. L., & Aired, G. Prose and prejudice: Someeffects of priming context on the immediate recall ofinformation. British Journal of Psychology, 1978, 69,123-133.

Wood, G. The knew-it-all-along effect. Journal of Ex-perimental Psychology: Human Performance and Per-ception, 1978, 4, 345-353.

Yekovich, F. R., & Manelis, L. Accessing integrated andnonintegrated prepositional structure in memory.Memory & Cognition, 1980, 8, 133-144.

Yekovich, F. R., & Thorndyke, P. W. An evaluation ofalternative functional models of narrative schemata.Journal of Verbal Learning and Verbal Behavior, 1981,20, 454-469.

Yekovich, F. R., Walker, C. H., & Blackman, H. S. Therole of presupposed and focal information in inte-grating sentences. Journal of Verbal Learning and Ver-bal Behavior, 1979, 18, 535-548.

Zangwill, O. L. Remembering revisted. Quarterly Jour-nal of Experimental Psychology, 1972,24, 123-138.

Received June 3, 1982Revision received August 23, 1982 •

Is Memory Schematic? - Home - CSU, Chico

Documents

Transcript of Is Memory Schematic? - Home - CSU, Chico