Psychological Bulletin Copyright 1983 by the American Psychological Association, Inc.

1983. Vol. 93, No. 2, 203-231 0033-2909/83/9302-0203$00.75

Is Memory Schematic?
Joseph W. Alba Lynn Hasher
University of Florida Temple University

This article proposes a prototypical schema theory of memory. Such a theory

assumes the operation of four central encoding-processes: selection—a process
that chooses only some of all incoming stimuli for representation; abstraction—
a process that stores the meaning of a message without reference to the original
syntactic and lexical content; interpretation—a process by which relevant prior
knowledge is generated to aid comprehension; and integration—a process by
which a single, holistic memory representation is formed from the products of
the previous three operations. The article evaluates the supportive and critical
evidence for these processes in light of the need for any theory of memory to
account for three fundamental observations: accuracy, incompleteness, and dis-
tortion. The central retrieval process of schema theory, reconstruction, is also
discussed 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- is encoded, or stored in memory, is heavily

firms the everyday observation that although determined by a guiding schema or knowl-
one's memory for complex events is some- edge framework that selects and actively
times startlingly accurate, it is also frequently modifies experience in order to arrive at a
incomplete and occasionally even highly dis- coherent, unified, expectation-confirming and
torted. Accounting for these observations knowledge-consistent representation of an
poses the fundamental problem for any the- experience. Such theories necessarily place
ory of memory. For the past 15 years, the emphasis on the incompleteness and inac-
most popular theories (called collectively curacy of memory, although they can ac-
schema theories) have been descendents of count for accuracy as well. The concern with
a view that emphasized the incompleteness inaccuracy and distortion rather than with
and distortions characteristic of memory. accuracy has made these theories especially
This viewpoint was originally proposed by attractive in fields such as clinical, forensic,
Bartlett (1932) to account for recall of com- educational, developmental, and social psy-
plex stories. Lineal descendents of the theory chology (e.g., Ausubel, 1968; Greenwald,
largely share Bartlett's rejection of the notion 1980;Loftus, 1979a; Paris &Lindauer, 1977;
that memory representations consist of ac- Taylor & Crocker, 1981).
curate traces that are stable over long dura- Although schema theories either guide or
tions. Current proponents of this rejected are used to explain a considerable portion of
viewpoint emphasize the potential accuracy current research in human memory, it is
of memory; incompleteness in recall is at- widely agreed (e.g., Brewer & Treyens, 1981;
tributed to retrieval failure, distortions to Taylor & Crocker, 1981) that the term schema
associative encoding processes. In contrast, has no fixed definition. It is most often used
current schema theorists propose that what to refer to the general knowledge a person
possesses about a particular domain. A
We thank Robert Weisberg, Eugene Zechmeister, Mi- schema allows for the encoding, storage, and
chael McCloskey, and especially Rose Zacks for many
helpful comments; their contributions to this article were retrieval of information related to that do-
extremely important. Preparation of this manuscript was main. Beyond this, however, there are few
supported by NIMH Grant MH33140 and in part by systematic, contemporary explications of
a Biomedical Sciences Research grant to Temple Uni- schema theory.
versity.
Requests for reprints should be sent to Joseph Alba, Two exceptions to this question are nota-
Department of Marketing, 205 Matherly Hall, Univer- ble: frame theory and script theory. A frame
sity of Florida, Gainesville, Florida 32611. (Minsky, 1975) is a schema that contains
203
204 JOSEPH W. ALBA AND LYNN HASHER

knowledge about the structure of a familiar then be interpreted in such a way as to be

event, for example the knowledge a person consistent with the schema. The information
possesses about the structure of a short story. that remains will then be integrated with pre-
It specifies not the exact contents of the event viously acquired, related information that
but rather the general type of information was activated during the current encoding
expected in that situation and the order in episode. The operation of one or more of
which it should be encountered (for example, these processes is likely to result in a repre-
in a story, setting information would be fol- sentation that is less than totally accurate.
lowed by theme information and so on). A A fifth schema theory process, reconstruc-
script (Schank & Abelson, 1977) is very sim- tion, is one that operates at the time a person
ilar to a frame in that it too contains general attempts to reproduce a memory episode.
information about particular, frequently ex- This process uses whatever details were se-
perienced events (e.g., a visit to a restaurant). lected for representation and are still acces-
Scripts also contain more specific informa- sible together with general knowledge to es-
tion about the contents of the event as well sentially fabricate what might have hap-
(e.g., being seated, ordering the meal, and so pened. Several critical reviews of this process
on). Although frame and script theories have are available (e.g., Gomulicki, 1956; Spiro,
recently been used to guide some research in 1977; Zangwill, 1972). The consensus is that
memory (e.g., Bower, Black, & Turner, 1979; reconstruction is quite rare and occurs only
Graesser, Woll, Kowalski, & Smith, 1980), by under special circumstances. Because of this,
far the majority of research is not aimed at we do not undertake a full-scale review of
testing particular provisions of such well ar- reconstruction and instead focus on the en-
ticulated theories. coding processes proposed by schema theory.
Instead, the bulk of the memory literature However, because of recently renewed inter-
dealing with complex events such as prose est in the process of reconstruction (see e.g.,
learning consists of a series of studies inves- Bower et al., 1979; Graesser et al., 1980;
tigating a number of loosely connected issues Spiro, 1980a), the topic is discussed in several
whose results are either predicted by or in- sections of this article.
terpreted with reference to schema theory or In the first part of the article we explicate
its originator, Bartlett (1932). This situation the four central encoding processes of a
poses a substantial difficulty to anyone who modal schema theory and the evidence that
would evaluate the adequacy of schema theo- supports the existence of these processes. In
retics; it is difficult to evaluate a set of par- so doing, we show how schema theories ac-
tially overlapping assumptions spread across count for the three central phenomena of
15 years of research, a sizeable number of memory—accuracy, incompleteness, and
paradigms, and the research of a large num- distortion. In the second part we evaluate the
ber of investigators.1 quality of the evidence supporting the four
Our first goal then was to try to impose major schema theory encoding processes. As
some structure on the research literature by will be seen, there is a good deal of evidence
identifying the central assumptions of schema that is not as supportive as might be desired
theories. We propose that according to a as well as a fair amount of evidence that is
modal theory, schema-driven encoding of directly contradictory. In the third part of the
complex information is characterized by four article, we consider and evaluate alternative
basic processes: selection, abstraction, inter-
pretation, and integration, A schema theory 1
which asserts that all four processes occur Apparently, a similar situation confronted Taylor and
Crocker (1981) as they attempted to formulate and eval-
would state that from any environmental uate a schema theory from research in social psychology.
event, only the information that is relevant In the memory domain, there are two excellent reviews
and important to the currently activated (Gomulicki, 1956; Zangwill, 1972), but these are now
schema will be encoded. Of the information both quite out of date. In addition, there are a few scat-
tered attempts to present a cohesive schema theory, at
selected, the semantic content of the message least for limited domains, including scene perception
will be abstracted and the surface form will (Brewer & Treyens, 1981) and text comprehension and
be lost. Further, the semantic content will representation (Thorndyke & Yekovich, 1980).
 SCHEMA THEORIES 205

explanatory mechanisms. These alternatives recalled substantially more of the narrative,

offer a viable means of resolving the incon- as well as more of its essential information,
sistencies in the research literature. than did low-knowledge subjects (see also
Ausubel & Fitzgerald, 1961, 1962). Similarly,
Evidence Consistent With Schema Theory skilled chess players have a memory advan-
tage over novices when the subject of memory
The support for schema theory will be con- is the location of pieces on the chessboard of
sidered in the context of the four schema the- a real game (Chase & Simon, 1973). The ad-
ory principles mentioned above—selection, vantage of relevant prior knowledge has also
abstraction, interpretation, and integration— been found when the usual positive correla-
whose origins can be traced to Bartlett's tion between knowledge and age is reversed,
(1932) classic work. They are presented here, that is, when those with the knowledge are
not as a description of any specific current in grade school and those without are grad-
model, but rather as an outline of a proto- uate students and faculty members (Chi,
typic schema theory. 1978).
Frame theory (Minsky, 1975) also asserts
Selection the importance of prior knowledge. If the
structure of incoming information does not
Of all the concepts in a given event or match one's knowledge, memory for that
message, only some will become part of the particular type of information should be ad-
memory representation. Schema theories have versely affected. Mismatches between incom-
proposed three conditions that determine ing and existing structures can occur either
whether or not a particular piece of infor- because the structure of the incoming stim-
mation will be selected for encoding: (a) the ulus is quite deviant from the prototypical
existence of a relevant schema, (b) the acti- structure or because the person lacks ade-
vation of that schema, and (c) the importance quate knowledge about the prototypical
of the incoming information with respect to structure. In fact, when parts of a short story
the schema. The operation of each of these or fairy tale are rearranged so as to carry the
selection principles will be considered in same information but in an unexpected or-
turn. der, or when a story from a culture with dif-
Prior knowledge. A critical condition for ferent structural rules is read, comprehension
the acquisition of new knowledge is the ex- is difficult and recall of the story content is
istence of previously acquired relevant significantly reduced (Day, Stein, Trabasso,
knowledge, that is, of a well developed & Shirey, Note 1; Kintsch & Greene, 1978;
schema. In the absence of such knowledge, Kintsch, Mandel, & Kozminsky, 1977; Man-
memory is generally poor. Experimentally dler, 1978; Meyers & Boldrick, 1975; Stein
providing subjects with background infor- & Nezworski, 1978; Thorndyke, 1977).
mation or with ways to link the new infor- In agreement with schema theory, there-
mation with already stored information fore, specific domain-related prior knowledge
greatly enhances memory (Ausubel, 1960; appears to result in improved acquisition of
Ausubel & Fitzgerald, 1961, 1962; Bransford new, domain-related information. Assimila-
& Johnson, 1972, Experiment 1; Royer & tion of new information appears to be a func-
Cable, 1975, 1976; Royer & Perkins, 1977; tion of the amount of prior relevant knowl-
see also Thorndyke, 1977; Wilkes & Aired, edge, whether structural (e.g., frame) or con-
1978). tent (e.g., theme) based. Without this
Similar findings have been reported for knowledge, there is no available schema into
subjects purposely selected because their which new information can be integrated or
prior knowledge of a particular topic (here subsumed and it is therefore quickly lost. The
baseball) was known to vary substantially encoding of new information is a mapping
(Chiesi, Spilich, & Voss, 1979; Spilich, Ve- process, new onto old, which depends on a
sonder, Chiesi, & Voss, 1979). These subjects sufficiently well developed knowledge base.
were presented with narratives about typical Activation of an existing schema. The
baseball situations. High-knowledge subjects mere possession of relevant knowledge is not
206 JOSEPH W. ALBA AND LYNN HASHER

sufficient for it to play a role in encoding: The applied to only some portion of the incoming
knowledge must be activated at the time of information. Schema theory predicts that the
encoding. Research has focused on two sit- relevant information will be easily encoded
uations in which information is presented in but the remainder will either be rejected or
the absence of activated knowledge bases. In distorted so as to fit the schema. Pichert and
one case, new information is presented in Anderson (1977) presented stories that could
such a way as to avoid contact with old, rel- be viewed from either of two perspectives
evant knowledge. In the other, a schema is (e.g., a description of a house from the per-
activated but is congruent with only a portion spective of a prospective purchaser or of a
of the incoming information. burglar). Information that was relevant to
Consider the case in which prior knowl- one perspective was irrelevant to the other.
edge exists but is not activated. The classic Subjects were biased toward one perspective
studies in this area were performed by Brans- while reading the stories and later recalled
ford and Johnson (1972, Experiments 2-4). from that perspective. Results showed that
They presented subjects with short passages subjects preferentially recalled the informa-
that, without a title, were highly abstract and tion that was consistent with their perspective
extremely difficult to understand. In these at encoding (see also Kozminsky, 1977). Pre-
instances the passages were also very poorly sumably, the biasing manipulation activated
recalled. Similar memory deficits and pro- only one of the two schemata subjects pos-
cessing difficulties resulting from insufficient sessed. Information irrelevant to the acti-
contextual support have been reported fre- vated schema may never have been perma-
quently (e.g., Dooling & Lachman, 1971; nently encoded or may have been encoded
Dooling & Mullet, 1973; Johnson, Doll, but processed less elaborately than relevant
Bransford, & Lapinski, 1974; Ortony, Schal- information. Both possibilities are consistent
lert, Reynolds, & Antos, 1978; Thorndyke, with schema theory. Similar results have been
1977). Thus, when knowledge structures lie obtained using a recognition test. Given a
dormant during encoding, new knowledge recognition test containing items that were
cannot be easily assimilated; "the absence of consistent or inconsistent with respect to the
an appropriate semantic context can under two themes of a story, subjects selected items
some circumstances seriously affect the ac- that were consistent with the theme biased
quisition process" (Bransford & Johnson, during encoding (Schallert, 1976).
1973, p. 397). The congruency between a schema and the
The critical importance of schema acti- incoming information has also been studied
vation during the encoding process can also by manipulating the stimulus rather than the
be seen in studies in which schema-activating subject's frame of reference (Morris, Stein,
cues were given after a story was read or & Bransford, 1979). In simple stories, char-
heard, In such cases, recall was as poor as it acters were described as involved in an ac-
was when no cues were provided (Bransford tivity that was either consistent or inconsis-
& Johnson, 1973; Dooling & Mullet, 1973; tent with their physical attributes (e.g., The
see also Thorndyke, 1977). strong man lifted the piano; The fat man got
An interesting developmental implication stuck in a cave; or, The bald man got stuck
of these findings has been drawn by Brans- in a cave; The old man lifted the piano). After
ford and Nitsch (1978). They speculate that reading the initial passage, subjects read a
less experienced people will have greater dif- second story, which referred to the activities
ficulty than more experienced people in dis- of the characters. Recall of both the initial
covering the situational cues that can lead to and following stories was better when they
activation of a knowledge context. Hence, described or referred to character-appropri-
younger children will ordinarily show poorer ate situations than when they did not.
retention than older children and adults. On The data reviewed in this and in the pre-
those rare occasions in which children have vious section seem to indicate that both
a knowledge advantage (e.g., Chi, 1978), the structural and semantic schemata play in-
usual advantage of adults will be lost. fluential roles in the assimilation of new in-
A second line of research involves the sit- formation. Retention appears to be better
uation in which an activated schema can be when incoming information is consistent
 SCHEMA THEORIES 207

with one's possessed and activated knowledge from, the prototypical script. Economy of
and expectancies about the world. Inconsis- storage is a concept central to computer anal-
tent information may not be well represented ogies of the memory process.
in the memory code. Summary: Selection and memory theory.
The importance effect. An existing schema Considerable evidence appears to support all
that is activated during the presentation of three schema theory principles regarding
a mass of information will enable selection what will be encoded. Prior knowledge,
of only some of that information for encod- whether semantic or structural, increases the
ing. Two selection principles have been pro- likelihood of encoding new information.
posed. The first principle stems from tradi- However, the existence of prior knowledge is
tional schema theories and states that the not sufficient to guarantee encoding of in-
ideas that are most important to the theme coming information; the knowledge must be
of the information (Owens, Bower, & Black, concurrently activated. Ideas important to an
1979) and that cannot be derived from pre- activated schema are likely to have a selection
viously encoded information (Spiro, 1980b) advantage for storage.
will be given special attention and will be re- Because of the selection process, the rep-
membered best. Using stories as stimulus resentation of any event is likely to be quite
materials, several studies have demonstrated incomplete. As a result people cannot repro-
that recall of ideas is a positive function of duce from memory an exact copy of an event,
the independently rated importance of those even when they are motivated to do so.
ideas to the overall meaning of the story (e.g., Rather, according to Bartlett (1932), people
Brown & Smiley, 1977; Christie & Schu- will attempt to reconstruct the event. Stored
macher, 1975; de Villiers, 1974; Graesser, information will be recalled together with
1978a; Graesser, Robertson, Lovelace, & "probable detail" from general schematic
Swinehart, 1980; Johnson, 1970; Johnson knowledge. That people have no trouble
& Scheidt, 1977; Kintsch, Kozminsky, Streby, freely generating such details is well docu-
McKoon, & Keenan, 1975; Kozminsky, 1977; mented (Bower et al., 1979; Brockway,
Meyer, 1975, 1977; Meyer & McConkie, Chmielewski, & Cofer, 1974; Cofer, Chmie-
1973; Miller & Kintsch, 1980; Smiley, Oak- lewski, & Brockway, 1976).
ley, Worthen, Campione, & Brown, 1977; From such a perspective, accurate recall
Thorndyke, 1977; Waters, 1978). can stem from two sources: (a) aspects of the
A related selection principle comes from original event that were actually selected for
script theory (Schank & Abelson, 1977). This representation in memory and (b) chance
theory deals with situations in which people matches between the reconstruction process
process information relevant to high fre- and the original event. The likelihood of such
quency events (e.g., eating in a restaurant). matches (or, of correctly "guessing") in-
The theory predicts that the memory traces creases whenever the original event contains
representing highly typical events that oc- elements in common with other similar
curred in a particular episode will be forgot- events—the presumed source of the "prob-
ten (or omitted from the representation). able detail" from which people are thought
Typical information need not be stored, since to reconstruct.
it can always be derived from the prototypical Distortions of the original event will occur
script. One need only remember that a whenever the probable detail produced was
scripted event occurred to recall highly prob- not actually part of the original event. Much
able components. Thus, according to script of Bartlett's data consisted of recall protocols
theory, atypical information will be selected containing considerable amounts of infor-
to receive special representation in memory. mation that was not in the original stimulus.
One advantage of a representational Although Bartlett's results have been difficult
scheme like Schank and Abelson's is that it to replicate (see e.g., Gomulicki, 1956; Zang-
allows for economical storage in the space will, 1972), others have found "thematic in-
allocated for memory; highly typical individ- trusions" in retrieved information (Dooling
ual episodes that occur in real-world expe- &Christiaansen, 1977a; Kintsch etal., 1975).'
riences may be forgotten, since they are al- Similarly, people occasionally "remember"
ready represented in, and so can be recalled typical actions that never actually happened
208 JOSEPH W. ALBA AND LYNN HASHER

in the particular experience of an event (e.g., format (e.g., King & Greeno, 1974; Kintsch
Bower et al., 1979; Graesser, Gordon, & Saw- &Monk, 1972).
yer, 1979; Graesser et al., 1980; Smith & There are a considerable number of the-
Graesser, 1981). ories that assume that memory consists of
Recently, Spiro (1977, 1980a, 1980b) has sets of propositions and their relations (e.g.,
argued that the reconstructive process is most J. R. Anderson, 1976; Bransford, et al., 1972;
likely to result in distortion when one en- Brewer, 1975; Frederiksen, 1975a; Kintsch,
counters (at least under conditions in which 1974; Norman & Rumelhart, 1975; Schank,
memory is not intentional) some additional 1972, 1976). One formalized presentation of
schema-relevant knowledge that is contradic- this idea is Schank's conceptual dependency
tory to an encoded schema. Subsequent recall theory (1972). The theory asserts that all
will be based partly on both sources of in- propositions can be expressed by a small set
formation. Such a situation results in distor- of primitive concepts. All lexical expressions
tions that are a by-product of the reconstruc- that share an identical meaning will be rep-
tive process and that will serve to resolve in- resented in one way (and so stored econom-
consistencies between the sources of ically) regardless of their presentation format.
information. Again, recall will contain ad- As a result people should often incorrectly
ditional information not present in the event. recall or misrecognize synonyms of originally
The selection process is instrumental in presented words, and they do (e.g., Anderson
accounting for one of the three fundamental & Bower, 1973; R. C. Anderson, 1974; An-
memory phenomena, incompleteness of re- isfeld & Knapp, 1968; Brewer, 1975; Graes-
call. Much of an original event is simply not ser, 1978b; Sachs, 1974).
represented in memory. The other two phe- Abstraction and memory theories. Since
nomena, accuracy and distortion, are at least considerable detail is lost via the abstraction
in part the product of the reconstruction pro- process, this process can easily account for
cess that is thought to operate at retrieval. the incompleteness that is characteristic of
people's recall of complex events. In light of
Abstraction the abstraction process, the problem for
schema theories becomes one of accounting
Information that has been selected because for accurate recall. Schema theories do this
it is important and/or relevant to the schema by borrowing a finding from psycholinguistic
is further reduced during the encoding pro- research, to wit, that speakers of a language
cess by abstraction. This process codes the share preferred ways of expressing informa-
meaning but not the format of a message tion. If both the creator and perceiver of a
(e.g., Bobrow, 1970; Bransford, Barclay, & message are operating with the same prefer-
Franks, 1972). Thus, details such as the lex- ences or under the same biases, the per-
ical form of an individual word (e.g., Schank, ceiver's reproduction of the input may appear
1972, 1976) and the syntactic form of a sen- to be accurate. The accuracy, however, is the
tence (e.g., Sachs, 1967) will not be preserved product of recalling the semantic content of
in memory. Because memory for syntax ap- the message and imposing the preferred
pears to be particularly sparce as well as brief structure onto it. Thus, biases operate in a
(e.g., J. R. Anderson, 1974; Begg & Wickel- manner that is similar to the "probable de-
gren, 1974; Jarvella, 1971; Sachs, 1967, tail" reconstruction process. Biases have been
1974), the abstraction process is thought to documented for both syntactic information
operate during encoding. (J. R. Anderson, 1974; Bock, 1977; Bock &
Additional support for the notion that Brewer, 1974; Clark & Clark, 1968; James,
what is stored is an abstracted representation Thompson, & Baldwin, 1973) and lexical in-
of the original stimulus comes from studies formation (Brewer, 1975; Brewer & Lichten-
that demonstrate that after a passage is read, stein, 1974).
it takes subjects the same amount of time to Distortions may result from the abstrac-
verify information originally presented in a tion process if biases are not shared by the
complex linguistic format as it does to verify person who creates the message and the one
that same information presented in a simpler who receives it. More importantly, the ab-
 SCHEMA THEORIES 209

straction process sets the stage for distortions Given Sentence 5, people equate its meaning
because it is the necessary precondition for with Sentence 6 (Harris, 1974; Harris &
two other schema theory processes, interpre- Monaco, 1978):
tation and integration. These two are usually 5. The frightened farmer was able to raise chickens.
thought to be the major sources of dis- 6. The frightened farmer raised chickens.
tortions.
In each case, subjects are encoding infor-
Interpretation mation that goes beyond the explicit. The
distortion that results can range in degree
Thus far, distortion in the recall of com- from minor to extreme. It will be minor if
plex materials has been attributed to pro- the perceiver's interpretation corresponds to
cesses that reduce information. The encoding the meaning intended by the creator of the
deficit that results from selection and abstrac- message. Distortion will, of course, be more
tion is, in part, compensated for at recall by extreme if the perceiver misinterprets the
reconstruction and by shared speech pro- message.
duction biases. Distortions also occur be- Interpretation and comprehension. Suc-
cause those semantic propositions that are cessful comprehension heavily depends on
encoded are actually interpretations of the the interpretation process. Consider the sit-
explicitly presented information—interpre- uation in which a presented concept is gen-
tations that are based on the perceiver's ac- eral or vague. According to theory, the schema
tivated schematic knowledge. Such errors are will serve to concretize the information and
often referred to as constructive errors be- store it in its interpreted form. This process
cause they involve the addition of informa- is called instantiation. It can be demonstrated
tion to the memory representation of a com- by presenting people with sentences that con-
plex event. In contrast to reconstructive er- tain general concepts (e.g., animal, container)
rors, constructive errors are the product of but provide enough context to allow the gen-
an elaboration process occurring during or eral terms to be understood as specific ones
shortly after encoding. (e.g., dog, bottle). On a cued recall test, the
Interpretations are typically inferences (see never-presented specific terms serve as better
Harris & Monaco, 1978). Two varieties can retrieval cues than the actually presented gen-
be identified. The first, pragmatic implica- eral terms (Anderson et al., 1976). Thus, gen-
tion, involves converting explicitly stated in- eral terms are interpreted and stored as spe-
formation into its probable underlying in- cific instantiations.
tent. The second involves inferences made A second, closely related form of interpre-
during comprehension when there is a need tation-dependent comprehension occurs when
to (a) concretize vague information, (b) fill implied facts are inserted into the represen-
in missing detail, or (c) simplify complex in- tation of an incomplete stimulus. When sub-
formation. jects are given sentences that imply the pres-
Pragmatic implication.' A variety of ex- ence of an unstated object, instrument, or
amples of this aspect of interpretation can be action, the subjects construct the missing in-
found (see Harris & Monaco, 1978, for a re- formation (Paris & Lindauer, 1976; Paris,
view); a few will suffice. Given Sentence 1, Lindauer, & Cox, 1977; Paris & Upton, 1976)
people will often recall it in the form of Sen- and may then remember it as having been
tence 2 (Schweller, Brewer & Dahl, 1976): part of the original stimulus (Johnson, Brans-
1. The housewife spoke to the manager about the ford, & Solomon, 1973). Thus, subjects who
increased meat prices. hear a passage describing a person pounding
2. The housewife complained to the manager about a nail will infer the presence of a hammer as
the increased meat prices. the probable instrument and incorporate the
instrument into their representation of the
Given Sentence 3, people recall Sentence 4 passage (Johnson et al., 1973).
(Brewer, 1977): Additions made to explicit information go
3. The paratrooper leaped out of the door. beyond one-word instruments and objects.
4. The paratrooper jumped out of the plane. Frequently, speakers and writers will struc-
210 JOSEPH W. ALBA AND LYNN HASHER

ture their output in such a way as to induce Interpretation results in distortion because
listeners and readers to infer connections be- it occurs with two other processes. The first
tween ideas. Haviland and Clark (1974) and is abstraction, which eliminates the surface
Kintsch and van Dijk (1978) have argued that structure form of the stimulus and stores only
information in a text is most easily under- a meaning-based representation. The second
stood when it can be related to immediately process, discussed next, is integration.
preceding information. When this is not pos-
sible, the comprehender searches memory for Integration
some relevant information that can serve as
a connecting idea. If none is found, people That people contribute information to an
will construct an inference that bridges the incoming message is not a notion unique to
information gap. In fact, when an inference schema theory. What is unique is the predic-
must be generated in order to understand a tion made about the memory representation
recent input, people do construct the missing that is the product of interpretation. A single
information (Keenan & Kintsch, 1974; integrated memory representation is thought
McKoon & Keenan, 1974) and sometimes to be created from whatever accurate infor-
later misrecognize it as having been part of mation is selected, whatever interpretations
the text (Thorndyke, 1976). Similarly, infer- are drawn, and whatever general knowledge
ences may be made and incorporated into exists that is relevant to the stimulus. Thus,
memory when they simply increase the co- individual ideas exist only as a part of a com-
herence of otherwise understandable ideas plex semantic whole. Integration processes
(Owens et al., 1979). are thought to occur at two different stages
Frame theory predicts a related type of in- of memory: (a) when a new schema is formed
ference making. If a stimulus generally con- and (b) when an existing schema is modified.
forms to an existing frame but does not con- These will be considered in turn.
tain all of the information categories speci- Schema formation. One method of dem-
fied by that frame, the missing information onstrating integration is to show that mem-
will be provided in the form of "default val- ory for a set of related propositions consists
ues" (Minsky, 1975). Thus, if a story gram- solely of the overall intent of those proposi-
mar proposes slots for setting or theme in- tions. Consider the following sentences.
formation and a stimulus conforming to that 7. There is a tree with a box beside it and a chair
grammar possesses no information that can is on top of the box.
instantiate those slots, the missing informa- 8. The box is to the right of the tree.
tion will be inferred and attached to the rep- 9. The tree is green and extremely tall.
resentation of the stimulus (e.g., Glenn, 1978; People who read Sentences 7-9 will misrec-
Mandler & Johnson, 1977; Stein & Glenn, ognize 10 as an originally presented sentence.
Note 2). Further, if the order of ideas in a
10. The tree is to the left of the chair.
frame-like stimulus does not conform to the
ideal order, the encoding frame may be used But, they will call Sentence 11 new (Brans-
to rearrange those story ideas (Mandler, ford etal, 1972; Paris & Carter, 1973;Prawat
1978; Stein & Glenn, 1979; Stein & Nez- & Cancelli, 1976).
worski, 1978; see also Lichtenstein & Brewer, 11. The chair is to the left of the tree.
1980).
Finally, inferences will also be made if they Evidence for the integration of individual
help to simplify the prepositional structure, sentences into a holistic representation is also
especially when the information processing found when, during acquisition, subjects are
load is heavy (Frederiksen, 1975a, 1975b). presented with a set of sentences that contain
These inferences too are often incorporated different sized subsets of ideas from a more
into the recall of the text. complex, unpresented parent sentence. Sub-
The possibility of distorting an original jects' confidence that a particular sentence
event arises because the interpretation pro- was actually presented during acquisition in-
cess allows the perceiver to add to or change creases with the degree to which that sentence
the information conveyed by the stimulus. approximates the never-presented complex
 SCHEMA THEORIES 211

parent sentence regardless of whether or not mous person, misrecognized as old the items
that sentence was originally presented that were thematically related to the famous
(Bransford & Franks, 1971; Cofer, 1973; person. This did not occur for subjects in the
Griggs, 1974; Singer, 1973; Walsh & Baldwin, fictitious person condition (see also Brown,
1977). Smiley, Day, Townsend, & Lawton, 1977;
Integration during schema formation can Royer, Perkins, & Konold, 1978).
also be seen in the linear ordering paradigm. Consider next the work of Loftus and her
Here subjects are presented with a series of colleagues (e.g., Gentner & Loftus, 1979;
sentences that describe a one-dimensional Loftus, 1975; Loftus, Miller, & Burns, 1978;
ordering of objects or people. Dimensions Loftus & Palmer, 1974). In these experi-
may include weight, height, and speed. The ments, subjects are typically shown a short
sentences may simply state that a is taller film or set of slides depicting, for example,
than b, b is taller than c, d is shorter than -c, a traffic situation. Afterward, a question is
and so on. Barclay (1973) presented such sen- asked about the scene that either implies the
tences and asked subjects to determine the presence of additional information that was
relations involved in the total set by imag- never actually present or that contradicts in-
ining the entire array in its proper order. On formation that was present. On a subsequent
a subsequent recognition test, he presented memory test, subjects often misrecognize
some of the originally presented sentences new slides containing the additional or con-
along with some new sentences that either did tradictory information. Loftus and Loftus
or did not express valid relations. Subjects (1980) argue that new information intro-
were unable to discriminate actually pre- duced after an event has been witnessed may
sented sentences from new ones that were add to or replace the person's knowledge of
valid statements about the ordering (see also the original scene, resulting in a single, in-
Potts, 1973). Similar conclusions may be tegrated memory for the scene.
drawn from studies using class inclusion re- Finally, consider the phenomenon of the
lations (Potts, 1976), artificial terms (Potts, "knew-it-all-along" effect (Fischhoff, 1977).
1977), and complex nonlinear arrays (Moeser In this paradigm, subjects are asked to rate
& Tarrant, 1977). the probability of occurrence of an event
The schema theory explanation of these (Fischhoff, 1975; Fischhoff & Beyth, 1975)
findings is that the meanings inferred from or to assess the validity of factual statements
individual sentences during encoding are in- that are beyond their specific knowledge but
tegrated into a larger semantic whole. On a that pertain to topics with which they are
memory test, sentences that are consistent familiar (Fischhoff, 1977; Wood, 1978). They
with the integrated representation are judged are then told of the occurrence or nonoc-
to have been part of the original array. currence of the event or are given the truth
Schema modification. The previous stud- values of the statements. Subjects are then
ies demonstrate integration for brief and re- asked to estimate their preoutcome knowl-
lated pieces of information on topics about edge states or to recreate their original prob-
which subjects had limited prior knowledge. ability or validity judgments, uninfluenced
Another demonstration of the integration by the outcome information. The result is
process occurs when new information is pre- that subjects overestimate the degree to which
sented on a topic about which the person they were originally correct. They cannot
already has considerable information in accurately recall their judgments and the in-
memory. Consider first a study by Sulin and accuracy is systematically in the direction of
Dooling (1974). Here subjects read a short the answers supplied to them. These data sug-
passage that they believed to be about either gest that new information is immediately in-
a famous person or a fictitious person. On tegrated into the prior knowledge system that
a recognition test, subjects were given some subjects possessed about the topics, resulting
of the passage sentences along with some dis- in an inseparable combination of the new and
tractors that varied in their thematic associ- old information. When trying to recall their
ation to the famous character. Subjects who original ratings, subjects base their responses
were told that the passage described the fa- on the only representation available to them
212 JOSEPH W. ALBA AND LYNN HASHER

at test time, one which has been modified by sources: reconstruction, construction, un-
the incorporation of the new information. shared language production biases between
Integration and memory theory. Once in- the originator and receiver of the message,
tegration occurs and old knowledge has been interpretations not actually intended by an
altered or updated, accurate retrieval of ac- originator, and the integration of memory
tually presented information becomes highly episodes over time.
unlikely. New information will be integrated Incompleteness in recall is largely the
into old knowledge structures. This new in- product of two encoding operations: selection
formation can originate in the external en- and abstraction. Not all ideas are selected for
vironment (as suggested by the Loftus or representation. Not all elements are part of
Fischhoff work) or in the internal environ- the abstracted meaning. The integration pro-
ment (as suggested by the work on infer- cess further attenuates those ideas that were
ences). Regardless of the information source, selected and abstracted, thus reducing the
schema theory predicts that inaccurate re- amount of information that can be recalled.
trieval will occur because individual traces In the next section of this article we con-
of a to-be-remembered event do not exist. sider a body of research, much of it quite
recent, that is either inconsistent with or di-
Summary rectly contradictory to the central assump-
tions of schema theory. Each of the four basic
Schema theory has become an enormously encoding assumptions of schema theory will
popular framework guiding considerable be considered in turn.
amounts of research in human memory. One
source of this popularity is the apparent suc- Evidence Inconsistent With Schema Theory
cess that the major assumptions of the theory
Selection
have had in explaining experimental results.
According to schema theory what is stored The importance of the schema to the se-
is some highly selected subset of all that has lection process is based on the interpretation
been presented, and it is the schema that of five sets of findings: (a) in the absence of
guides the selection process. Memory is ab- any schema, not much can be encoded and
stractive in that a verbatim record is not left so not much will be recalled (e.g., Bransford
behind; rather, meaning appears to have high- & Johnson, 1972, 1973); (b) the greater the
est priority for storage. Memory is interpre- level of prior knowledge, the greater the
tive in that the schema serves to fill in missing amount of new, relevant information that
details and distort others so as to be schema- can be selected and subsequently recalled
consistent. Memory is integrative in that in- (e.g., Chiesi et al., 1979); (c) when an existing
coming information joins with other related schema is not activated, recall is extremely
elements in that episode, with whatever prior poor (e.g., Bransford & Johnson, 1972); (d)
knowledge is available, and with relevant, information that is congruent with an acti-
subsequent information to create a single vated schema is far more likely to be encoded
unified representation of a complex event. and so recalled than is information that is
When retrieval is required, probable detail incongruent (e.g., Pichert & Anderson, 1977);
is generated, and the representation is given (e) the more important an idea is to a schema,
a modal surface form. Thus the original the greater the likelihood that it will be re-
meaning is retained, and on occasion, the called (e.g., Johnson, 1970; Meyer & Mc-
production is a duplication of the original Conkie, 1973).
stimulus. Accuracy sterns from those ideas Consider first the idea that the encoding
that were actually selected for storage, from of new information is a function of the
culture-normal biases in expressing infor- amount of prior relevant information. A re-
mation, and from serendipitous productions view of the education literature (Barnes &
from the reconstructive process. This latter Clawson, 1975) found more research in con-
category is most likely to occur for high-prob- flict with Ausubel's (1968) version of this
ability events. principle than in support of it. Although re-
On other occasions, recall may be quite call has been shown to vary substantially with
distorted. This is attributed to a number of the amount of prior knowledge (Chiesi et al.,
 SCHEMA THEORIES 213

1979), the same research has shown that rec- Schumacher (1981), who in some conditions
ognition varies minimally (Experiments 1 did not introduce the alternative (or new)
and 2). Similarly, although an inactivated theme until 12 minutes after the original
schema may leave subjects at a considerable story had been read, long after the selection
recall disadvantage compared with subjects process should have operated. These cuing
with an activated schema (e.g., Bransford & studies suggest that more is encoded than is
Johnson, 1972), this disadvantage too dis- consistent with the operation of a schema-
appears when memory is tested using a rec- based selection principle.
ognition procedure (Alba, Alexander, Hasher, Strong evidence for the operation of re-
& Caniglia, 1981). That an activated schema trieval processes in prose recall comes from
does not benefit recognition as it does recall a study by Hasher and Griffin (1978). They
has been shown for a variety of stimulus showed that as long as 2 weeks after the en-
materials including prose (Alba et al., 1981), coding of a story the nature of recall could
individual sentences (Birnbaum, Johnson, be dramatically influenced. In that study, one
Hartley, & Taylor, 1980), lists of sentences group of subjects showed a.typical "recon-
(Hannigan, 1976), and pictures (Bower, Kar- structive" pattern, with substantial forgetting
lin, & Dueck, 1975). Finally, although im- of specific ideas along with thematically rel-
portant ideas have an advantage in recall, this evant intrusions. However, another group
advantage is not always present on a recog- showed a quite untypical pattern (termed
nition test (Britton, Meyer, Hodge, & Glynn, "reproductive" by the experimenters); these
1980, Experiment 3; Goetz, 1979; Kintsch subjects produced virtually no thematic in-
& Bates, 1977;Thorndyke&Yekovich, 1980; trusions and forgot far less of the original
Walker & Meyer, 1980; Yekovich & Thorn- story information. The manipulation that in-
dyke, 1981). duced reproductive recall was one that cre-
That recall but not recognition varies with ated doubt about the validity of the schema
the activation of a schema, with the knowl- used during encoding. Hasher and Griffin
edge content of that schema, and with the argued that subjects were then forced to re-
importance of individual ideas to that schema vise their retrieval plan (of using the encoding
poses problems for a selection process pre- schema) and so exerted special effort to gain
sumed to operate at the time of encoding. access to otherwise inaccessible information.
Information that is not part of a memory Further evidence that a schema operates at
representation should be the equivalent of retrieval is found in a study in which recall
information that was never presented; it deficits for unimportant (and so presumably
should be neither recalled nor recognized. unselected) information were greatly atten-
That there are substantial effects of a schema uated by providing subjects with both se-
on the amount that is recalled as well as on mantic (content words) and contextual (the
the content that is recalled suggests that the background color of the printed page) re-
schema influences retrieval rather than selec- trieval cues (Britton et al., 1980, Experiments
tion for storage in memory. 1 and 2).
The suggestion that a schema operates at Selection and reconstruction. Selection
retrieval to influence what is recalled is presumably allows for mnemonic distortion
strengthened considerably by the work of because the memory representation of any
Anderson and Pichert (1978). When infor- event is substantially reduced compared with
mation agrees with two different schemata, the original. According to Bartlett (1932), the
only one of which is activated during encod- subject must then rely on his or her general
ing, superior recall is seen for the information knowledge together with what few details
that is congruent with whatever schema was were actually encoded to produce a recount-
activated during acquisition (see also Pichert ing of an event. This recall procedure is called
& Anderson, 1977). However, if one tests for reconstruction. The evidence just presented,
recall a second time from the perspective of however, suggests that memory for an event
the previously inactive schema, recall of the is often quite good, though details may not
once irretrievable information is significantly always be readily accessible. In fact, the ev-
raised (Anderson & Pichert, 1978). These idence for a reconstructive process at re-
findings have been replicated by Pass and trieval is itself actually fairly weak. Under
214 JOSEPH W. ALBA AND LYNN HASHER

normal circumstances, distortion is either cially when interference is reduced (Bower et

rare or nonexistent (Zangwill, 1972). In fact, al., 1979); and (b) complete failure to dis-
a recent analysis of Bartlett's stimuli show criminate occurs only with those very few
them to be rather poorly structured and un- script actions that can be labeled very typical
representative of normal prose (Mandler & (Graesser et al., 1979)—a finding that con-
Johnson, 1977), perhaps accounting for his siderably weakens the cognitive economy ar-
apparently unique findings. Indeed, a con- gument of script theory.
temporary replication of Bartlett's proce- Possible explanations of the typicality ef-
dures showed little evidence of distortion fect that exclude the principle of selection will
(Kintsch & Greene, 1978; see also Gomu- be presented in the final portion of this ar-
licki, 1956; Paul, 1959). Recall errors that do ticle.
occur can sometimes be eliminated by ma-
nipulating recall conditions (cf. Cofer et al., Abstraction
1976), by stressing accuracy (Gauld & Ste-
phenson, 1967), or by changing the subjects' The abstraction process is thought to fur-
retrieval strategy (Hasher & Griffin, 1978). ther reduce what has been stored in memory
If reconstructive processes exist, they may by operating on what has been selected; se-
be an optional, perhaps (compared with re- mantic content is believed to be the impor-
production) even a relatively effortless, recall tant target of storage. Although it is now
strategy, rather than, as Bartlett proposed, widely held that memory is abstractive, a
and others have assumed, a necessary strat- number of recent findings cast some doubt
egy. After all, one's general knowledge on a on this process.
topic is presumably acquired by virtue of re- First, memory is not inevitably abstractive.
peated, distributed exposures to that topic. Consider memory for highly overlearned
Since both study and retrieval episodes are messages such as the Pledge of Allegiance.
implicated in recallability (see Spear, 1978), Detailed memory for such material will per-
it should be relatively easy to gain access to sist for years (Rubin, 1977; see also Smith,
general knowledge, especially when com- 1935). Even memory for information pre-
pared with the ease of gaining access to one's sented only once has been shown to contain
memory for an episode that occurred only considerable detail. College students are able
once. At the very least, reconstruction is not to discriminate actual utterances heard in the
a necessary consequence of selective encod- context of a lecture from highly similar ut-
ing; far more is encoded than the selection terances (Kintsch & Bates, 1977). Accurate
principle would suggest, and a great deal of recognition was found in that study for con-
this original information remains available tent that was largely extraneous to the topic
in memory. Finally, the circumstances at re- (e.g., jokes) as well as for low-salient, relevant
trieval play a critical role in determining what information. Memory for syntactic and lex-
can be accessed (see Tulving & Pearlstone, ical details in the lecture was quite profound;
1966, for a discussion of the distinction be- the rejected paraphrases included such subtle
tween availability and accessibility). changes from the original sentences as rear-
Yet to be considered are those reconstruc- rangement of clauses and meaning-preserv-
tion processes associated with script theory ing word substitutions.
and collectively referred to as the typicality Still more difficult paraphrases than those
effect (e.g., Bower et al., 1979; Graesser et al., used by Kintsch and Bates (1977) were in-
1979). Script theory predicts a failure on cluded in a follow-up study in which recog-
mem 017 tests to discriminate between pre- nition memory was tested for sentences from
sented and unpresented events that are typ- a television soap opera (Bates, Masling, &
ical of a given script. Although the few studies Kintsch, 1978). The paraphrases in this study
that have been performed in order to test this were constructed by substituting proper
prediction have reported decreased discrim- names for pronouns, roles for names, ellip-
inability with increased typicality, these stud- tical clauses for full clauses, and vice versa.
ies have also demonstrated that (a) people Memory for the original utterances was sur-
generally can discriminate between presented prisingly good (see also Bates, Kintsch,
and unpresented typical script actions, espe- Fletcher, & Giuliani, 1980).
 SCHEMA THEORIES 215

The ability of adults to distinguish between & Thorndyke, 1981; see also James et al.,
paraphrases and actual utterances is not eas- 1973;Kemper, 1980) and 1 week (Christiaan-
ily explained by any theory which assumes sen, 1980).
that memory is selective and abstractive. If A second version of the abstraction process
only general impressions plus a few details states that word meanings are represented
remain in memory after a lecture or a TV conceptually, and that therefore lexical in-
show, recognition performance for alterna- formation is lost (e.g., Schank, 1972). A num-
tives discriminable only on the basis of details ber of findings contradict this view. For ex-
should be at chance. A schema-theory expla- ample, people rarely make synonym substi-
nation might claim that the utterances and tution errors, are better able to identify
their paraphrases differed in terms of their sentences when they contain their original
"probable detail," or in their conformity to words rather than synonyms, and show greater
the stylistic habits of the speakers, aspects of interference in sentence verification when
messages to which listeners are presumably stimulus sentences all contain a common
sensitive. However, Keenan, MacWhinney, word than when they contain different syn-
and Mayhew (1977) demonstrated that mem- onyms of the word (Hayes-Roth & Hayes-
ory for utterances from a lecture was not Roth, 1977). Along with Anderson's (1971;
based on sensitivity to such aspects; rather, see also Hayes-Roth & Thorndyke, 1979)
subjects appear to remember the actual form finding that verbatim cues enhance recall
of the original statements. Other research more than do synonymous paraphrases of
confirms this conclusion; people are able to words, there seems to be good evidence for
discriminate between the frequency with the hypothesis that lexical information is also
which verbatim versus gist repetitions of sen- retained and forms part of the representation
tences occur (Burnett & Stevenson, 1979; of a verbal event.
Gude & Zechmeister, 1975; see also Morris, One study that can still be taken as evi-
Bransford, & Franks, 1977). dence in favor of the conceptual represen-
Thus, memory for detail has been shown tation notion did find reliable synonym sub-
for both well-learned (Rubin, 1977) and far stitution errors in recall (Brewer, 1975). How-
less well-learned information (Bates et al., ever, of the total number of possible responses,
1978; Bates et al., 1980; Keenan et al., 1977; synonym substitution occurred only 6% of
Kintsch & Bates, 1977). It should be noted the time (Experiment 1). Thus, either mem-
that in all of the experiments dealing with ory is not conceptually based, or very few
memory for once-presented messages, the words have identical synonyms. (If few words
recognition test was a surprise. Thus these do have identical synonyms, the storage ef-
findings cannot be explained away by ap- ficiency argument of computer-based mod-
pealing to unusual and "unnatural" task de- els—e.g., script theory—is not very con-
mands made under conditions of rote mem- vincing.)
orization. Memory appears to contain far more syn-
Recent research appears to stand in marked tactic and lexical detail than is consistent with
contrast to the earlier evidence often cited as the view that memory is highly abstractive.
support for the abstraction process. In fact Thus a central assumption of schema theory
it does not. What those early studies found appears to be incorrect; the representation of
was superior recognition of semantic as com- both syntactic and lexical information is
pared with syntactic information. Syntax was richer and more detailed than the abstraction
frequently remembered at levels well above process allows.
chance; that is, syntactic information was not
totally forgotten, (e.g., J. R. Anderson, 1974; Interpretation
Anderson & Paulson, 1977;Begg, 1971;Begg
& Wickelgren, 1974; Dooling & Christiaan- That people interpret and elaborate on in-
sen, 1977b; Olson &Filby, 1972; Sachs, 1967, coming information is not a notion unique
1974; Soli & Balch, 1976). Three recent stud- to schema theory. In the traditional human
ies have shown above-chance recognition of learning literature, "interpretation" pro-
lexical and syntactic detail at the surprisingly cesses were recognized and explored in re-
long retention intervals of 1 hour (Yekovich search dealing with the distinction between
216 JOSEPH W. ALBA AND LYNN HASHER

the nominal and functional stimulus (Un- have been selected, and related abstractions
derwood, 1963), with the role of mediation are joined into a single representation. The
and elaboration in the learning of simple most impressive contemporary evidence in
stimuli (Bugelski, 1968; Underwood & Schulz, support of this notion came from Bransford
1960), with subjective organization (Tulving, and Franks (1971). As is now widely ac-
1962), and more recently with the distinctionknowledged, there are many grounds on
between "fact" and "fantasy" (Johnson & which to challenge their conclusions. The
Raye, 1981). What is at issue in an evaluationmost basic is the fact that integration is not
of schema theories is the question of how in- a necessary product of encoding related
terpretive behaviors alter the memory rep- events (Moeser, 1976, 1977). In fact, the
resentation of the original stimulus. That is,Bransford and Franks (1971) findings are
are subject-generated interpretations of in- difficult to replicate unless their methods are
coming stimuli necessarily integrated with followed closely. Changes in presentation
the representation of the nominal stimulus modality (Flagg & Reynolds, 1977; see also
so as to lose tags that mark the origin of an Katz & Gruenewald, 1974), in presentation
idea? This question is considered in the next of materials (Flagg, 1976; Flagg & Reynolds,
section of the article. 1977; James & Hillinger, 1977; Katz, Atke-
There are two important points about in- son, & Lee, 1974), in instructions (James,
terpretation to be made here. Contrary to Hillinger, & Murphy, 1977), and in testing
some schema theories that predict sizable procedures (Anderson & Bower, 1973; Griggs
amounts of inference making during com- & Keen, 1977; James & Hillinger, 1977)
prehension (e.g., Schank, 1976), several re- greatly reduce or eliminate the integration
searchers have found that inference making effects found with the original Bransford-
is not an obligatory process. People may com- Franks paradigm. Finally, "integration" will
pletely fail to make even a simple inference occur for nonsemantic or arbitrary materials
(Corbett & Dosher, 1978; Hayes-Roth & (e.g., letter-digits, nonsense syllables) if the
Thorndyke, 1979; Singer, 1979, 1981); they original procedures are used (Flagg, 1976;
may generate inferences only to some infor- Katz & Gruenewald, 1974; Reitman & Bower,
mation in a message (Goetz, 1979; Walker 1973; Small, 1975). Since integration is as-
& Meyer, 1980); or they may engage in ex- sumed to be tied to similarities in the rep-
tensive inference making only when the task resentation of meaning, evidence of such a
demands require it (Frederiksen, 1975b). process operating on meaningless material is
Second, whatever retrieval errors subjects troublesome.
do make in recall may not even be indicative Results from research using the linear or-
of their underlying memory representations. dering paradigm (e.g., Barclay, 1973; Potts,
Consider a prediction of frame theory: If a 1973) were also interpreted as being the result
stimulus is structured differently from the of integration. The argument was that when
frame used to encode it, the stimulus should related propositions (e.g., a < b, b < c, c <
be reorganized to conform to the frame, and d) are presented, the individual ideas are lost,
recall studies suggest that subjects do reor- leaving a subject-constructed, holistic array
ganize information (e.g., Stein & Glenn, (a < b < c). There are now several grounds
1979; see also Mandler, 1978; Stein & Nez- on which to challenge this conclusion. Again,
worski, 1978). However, in at least certain the most basic is that the construction of a
cases, e.g., a story presented with episodes holistic order is not a necessary by-product
of encoding materials bearing ordered rela-
occurring in flashback order, there is evidence
that subjects retain input order information tions (Potts, Keller, & Rooley, 1981). In fact,
(Baker, 1978). a sizeable proportion of subjects (30%) showed
no evidence of even creating such an ordering
Integration (Potts et al., 1981). Further, recent evidence
suggests that the ordering effect may be an
Schema formation. Integration is the set . artifact of the testing procedures used. Giving
piece of schema theories. Theoretically, subjects unlimited time to make old-new
meaning is abstracted from the stimuli that decisions, as well as more information on
 SCHEMA THEORIES 217

which to base their decisions, eliminates in- told to memorize their previous knowledge
tegration effects (Lawson, 1977). Even under states, the size of the knew-it-all-along effect
less favorable conditions, an ability to make is reduced. Apparently, subjects can to some
proper discriminations seems to exist (Tzeng, extent keep old and new information sepa-
1975). Subjects appear to store individual rate. More critically, however, there is evi-
ideas, even if they are related to others in the dence that people can accurately remember
memory set. their original knowledge states even after new
Strong evidence for integration processes information has been provided, that is, after
in the formation of a schema seemed to come the integration has presumably occurred.
from both the semantic integration and linear Accurate recall of original information oc-
ordering paradigms. Integration effects ap- curs if people are led to believe that the truth
pear to be more tied to aspects of the pro- value of the new facts is unreliable (Hasher,
cedures used than to an inevitable process of Attig & Alba, 1981). Under these circum-
the human memory system. In addition, in stances, people show very little knew-it-all-
both cases there is evidence that subjects have along behavior. What is critical about this
stored in memory separate, unintegrated study is that the manipulation discrediting
units of the original stimulus complex. the new "facts" was provided after the inte-
Schema modification. A schema is a re- gration process was presumed to have oc-
pository of prior knowledge on a topic. Thus, curred, which, according to schema theory,
one might argue that the failure to demon- is an impossible finding. Thus, it appears to
strate integration across a newly acquired be possible to induce subjects to remember
knowledge base is not so critical for schema their original knowledge state, which was oth-
theory as would be the failure to demonstrate erwise thought to be lost to memory by an
integration of new facts into a preexistent integration process. These data suggest, as
knowledge base. (Such an argument creates well, that subjects can discriminate the origin
the problem of explaining where schemas of recently provided information.
come from in the first place; see Yekovich There is also a substantial volume of re-
& Thorndyke, 1981). Putting aside for the search on the integration of new with older
moment the negative evidence concerning knowledge conducted by Loftus and her col-
the abstraction process itself (on which the leagues. This research, too, appears to sup-
occurrence of integration necessarily de- port both predictions. Subjects misin-
pends), we now evaluate the evidence for in- formed—after seeing a series of slides de-
tegration of new information with well- picting an accident—that the traffic sign at
learned (or previously acquired) data bases. a critical corner was a stop sign (not a yield
Studies investigating this type of integra- sign, or vice versa) later incorrectly report
tion have attempted to show that following having seen a stop sign. Recent evidence
integration (a) people cannot remember their (Bekerian & Bowers, 1983) suggests that this
original knowledge xstate and (b) the newly dramatic and widely cited instance of inte-
acquired piece of information loses its unique, gration may be an artifact of the testing pro-
context-specifying character. Three different cedure, in which the test slides are presented
lines of research would appear to provide without regard to the original temporal se-
strong evidence for these two views. quence. The integration effect can be virtu-
The "knew-it-all-along" effect shows that ally eliminated by a testing procedure that
people tend to overestimate the quality of preserves the original presentation order.
their original knowledge about facts after Subjects can apparently use temporal-se-
they have been provided with relevant infor- quential information to distinguish the orig-
mation. The overestimation is theoretically inal slide (e.g., the yield sign) from the up-
the result of the new information's having dating slide (the stop sign). Such discrimi-
been inseparably integrated into a person's nation should not of course be possible after
general knowledge, making it impossible for the integration process has occurred; the orig-
the person to correctly remember his or her inal information should no longer exist.
original knowledge state. However, Wood Two other points about this work should
(1978) showed that if subjects are specifically be made. First, it is possible (as Loftus,
218 JOSEPH W. ALBA AND LYNN HASHER

1979a, acknowledges) that other conditions tus, 1979b) at the time of presentation—or,
necessary to demonstrate recollection of the more importantly, afterwards (Hasher et al.,
preassimilated knowledge have not yet been 1981). Subjects can, even in the face of new
discovered (for example, the Hasher et al., relevant information, continue to remember
1981, discrediting treatment might be such their original knowledge state. Discrimina-
a condition). The second point relates to the tive markers for newly provided information
generality of the data. Loftus has been suc- do not appear to be inevitably lost.
cessful in modifying people's memory of an A second set of studies more directly ad-
event only when she deals with information dressed the question of whether newly pre-
peripheral to the main focus of the event. In sented information loses its unique identity
recent studies (Alba, 1981; Loftus, 1979b), after it is integrated into a relevant, preex-
attempts at altering memory for focal infor- isting schema. Initial support for this notion
mation were shown to be almost totally in- was provided by Sulin and Dooling (1974)
effective. These findings suggest an alterna- who reported that subjects aware of the con-
tive interpretation of the original results (also nection between information in an experi-
see Shaughnessy & Mand, 1982). The relative menter-provided passage and a famous per-
ease with which nonfocal information can be son, later misrecognized information about
updated, compared with the difficulty of up- the famous person that was never actually
dating focal information, suggests that the presented. Subjects unaware of the connec-
"integration" process can work most easily tion between the story and the famous person
on memories that are either already lost or were less likely than other subjects to mis-
difficult to retrieve. In these instances the recognize the critical information. One in-
updating information will be (at least on the terpretation of these results is that the passage
basis of its recency to the memory test) more information became inextricably integrated
accessible than the original traces. It is of into the relevant knowledge schema of those
course important to account for the difficulty subjects who read it in the context of the fa-
of gaining access to nonfocal information mous person's name. Because of the integra-
(ordinarily evidence for the selection pro- tion process, subjects no longer possessed in-
cess). We will consider a possible explanation dividual traces of the experimental passage
shortly. and so on the recognition test could only re-
A further finding suggests that this differ- spond on the basis of their holistic schematic
ential retrieval hypothesis has validity: mod- knowledge.
ification of a nonfocal fact does not occur if There are several problems with these find-
updating information is presented along with ings. First, integration effects were found to
blatantly incorrect information (Loftus, be minimal at brief delay intervals (Sulin &
1979b). Similarly, updating does not occur Dooling, 1974); memory was quite accurate
if the provider of the new information is dis- immediately after the story was read. Inte-
trusted (Dodd & Bradshaw, 1980), nor if sub- gration then appears to take place over time,
jects are forewarned about the possibility of presumably (as Sulin & Dooling explain) be-
receiving misinformation (Greene, Flynn, cause memory for specific detail is only grad-
& Loftus, 1982). These three manipulations ually lost. And so with time, subject's rec-
may actually be the equivalent of the schema ollections increasingly depend on schematic
discrediting procedure (Hasher et al., 1981; knowledge. This view of integration contrasts
Hasher & Griffin, 1978), which in the latter directly with an integration mechanism that
studies apparently induced subjects to re- operates during encoding, the type of inte-
trieve ordinarily forgotten facts. gration assumed in the schema formation lit-
Thus, updating (or the integration of re- erature, as well as by Fischhoff (e.g., 1977)
lated memories) is not an inevitable conse- and by Loftus (e.g., Loftus et al., 1978).
quence of the mere provision of related, con- There are further problems with an assim-
tradictory, or supplementary information. ilation interpretation of the Sulin and Dool-
New information may be disregarded (Dodd ing data. Consider the results subsequently
& Bradshaw, 1980; Greene et al., 1982; Lof- reported by Dooling and Christiaansen
 SCHEMA THEORIES 219

(1977a).2 In this study, subjects were again knowledge content of that schema, and on
either informed or uninformed about the the importance of the incoming information
connection of the experimenter-provided pas- to the schema. Further reductions in the
sages to some famous person. Here, all sub- memory representation occur because the
jects showed very good memory for the orig- abstraction process favors the storage of
inal sentences—even over long durations. meaning rather than of individual lexical
Nonetheless, subjects in the famous condi- units or syntactic sequences. Evidence from
tion did show a reliable propensity toward a sizeable number of studies suggests that the
misrecognizing some (but not all) informa- memory trace is richer and more detailed
tion about the famous person. One possible than these information reduction processes
explanation assumes that the memory rep- would allow.
resentations for the experimental materials Schema theory also posits two processes
were equivalent for the two groups of sub- that enrich the representation of complex
jects. However, subjects who knew of the con- events: interpretation and integration. Here
nection with a famous person might easily we considered evidence that suggests that
have been operating under response biases neither of these processes is obligatory. In-
that were different from those of subjects un- deed, subjects can remember uninterpreted
informed about the connection. Informed and unintegrated information. In several
subjects might have decided that some of the places we considered studies that demon-
foils were old if they believed their memories strated that much of the original support for
imperfect, since after all, the foils were indeed schema theory encoding processes stemmed
true of the famous person. This response bias from procedural peculiarities of landmark
interpretation of these data becomes more experiments.
plausible because systematic response biases This review has revealed that memory for
have recently been discovered in research in- complex events is rather detailed, but some
vestigating the question of how people dis- aspects of information are far more easily
criminate among various sources of infor- accessed than are others. In the next section
mation (e.g., originating with themselves vs. of the article, we attempt to explain the find-
with others; see Johnson, Raye, Foley, & ings reviewed in the first two sections.
Foley, 1981).
To conclude this critique of integration, Alternative Explanations
the processes of schema formation and mod- A considerable body of research findings
ification via integration have not received
very substantial support. For both processes major that
exists is not easily accounted for by the
encoding assumptions of schema the-
there is evidence that integration is not a nec-
essary by-product of merely presenting re- ory. Unfortunately it is not currently possible
lated or orderable information. In both lit- pable of integrating all oftheory
to propose an alternative
the
that is ca-
relevant evi-
eratures there is evidence that subjects have
dence as well as generating
stored in memory unintegrated information that could not themselves be accounted for new predictions
about the original situation. That misrecog- by schema theory. What appear to exist are
nition of related information occurs is not a set of explanations, not necessarily related
deniable. The claim that integration is a ma- to one another, each of which can be tied to
jor, inevitable product of schema-based pro-
3
cessing is. 2
Dooling and Christiaansen (1977) interpret their own
data, as well as those of Sulin and Dooling (1974), with
Summary reference to the operation of reconstruction. That is, as
episodic memory for an event fades, people rely on their
Schema theory assumes that the represen- thematic knowledge to reconstruct the event. The pro-
tation of a complex event is both incomplete cess of reconstruction will be considered in the next sec-
tion.
and inaccurate. This is because what is stored 3
See Johnson and Raye (1981) for an argument that
is highly selected and depends on whether the a schema is neither a necessary nor logical source of
appropriate schema is activated, on the misrecognition errors.
220 JOSEPH W. ALBA AND LYNN HASHER

a set of findings in the schema literature as coded but will occupy a place outside of the
well as to a set of findings elsewhere in the integrated network (Kozminsky, 1977).
memory literature. We consider some of According to such models, individual
these briefly in this section of the article. propositions can have either or both of two
Again we consider in turn each major schema- sets of relations: (a) hierarchical connections
theory encoding process. descending from the theme and (b) connec-
tions that link individual propositions di-
Selection rectly to others via shared arguments.4 Com-
ponents of these two types of connections can
The representation of a prose passage ap- then serve as retrieval cues and should (as do
pears to include more details than a selection retrieval cues for simpler materials; e.g., Tulv-
process allows. Nonetheless a literature exists ing & Thomson, 1973) play a large part in
showing a predictable advantage in recall for determining how many and what particular
some ideas and not for others. There is no ideas are recalled.
doubt that sizeable differences in recall (if not Substantiating evidence exists for such a
recognition) are associated with differences network of relations. For example, when ar-
in degree of and/or activation of prior knowl- gument overlap is low, as occurs when a new
edge. Clearly, what is stored (or available) in idea cannot be related directly to a previously
memory is not always accessible (Tulving & mentioned idea (called a violation of the
Pearlstone, 1966). The question thus be- "given-new contract"; Haviland & Clark,
comes What is it that determines access- 1974), both comprehension and recall are
ibility? poor (e.g., Haviland & Clark, 1974; Hupet
Classical research in human memory has & LeBouedec, 1977; Kieras, 1978; Lesgold,
identified a number of factors that are rele- Roth, & Curtis, 1979; Yekovich, Walker, &
vant for determining what is recalled. Four Blackman, 1979). On the other hand, when
of these are important for present purposes: connections can be established (e.g., by draw-
(a) the nature of connections established dur- ing inferences to link the new and old ideas)
ing encoding, (b) differences in the number recall is good (Kieras, 1978; Yekovich &
and distribution of rehearsals, (c) the role of Manelis, 1980).
retrieval cues at recall, and (d) the order in Consider now the Bransford and Johnson
which elements or sets are recalled. We pro- (1972) materials that provided the classic
pose that these four factors are important in demonstration for the existence of the selec-
determining recall of prose materials. First, tion process. Their passages contain no ex-
however, we need to consider how textual plicit, concrete referents, and without a con-
material may be represented in memory. text to suggest exemplars for these referents,
Text processing theories such as Kintsch's none is likely to be inferred. This should in-
(1974; Kintsch & van Dijk, 1978) assume hibit construction of networks of connections
that when complex material is presented, a among the sentences at least by preventing
"text base" is formed as part of the compre- extensive use of the given-new strategy. This
hension process. Briefly, a text base is an or- would serve to reduce the possibility that one
dered hierarchy of propositions with the sentence could cue another at recall. It is not
theme at the highest level and argument-shar- surprising then that recall of these materials
ing propositions descending from it. All re- is so poor; subjects had in effect been pre-
lated propositions are interconnected in a sented with a set of unrelated sentences.
network due to argument overlap and to such It has long been known of course that units
grammatical elements as articles (de Villiers, related to one another either hierarchically
1974) and pronouns (Lesgold, 1972). Not
4
every proposition will be connected directly Such a notion is reminiscent of views about the con-
to every other one, but many should be linked nections established during the learning of arbitrary pairs
of simple verbal units; for example, connections are es-
by indirect connections. Finally, if a passage tablished between the two members of each pair as well
contains a set of propositions not related to as among all items serving as stimuli and all items serving
the theme proposition, these too will be en- as responses (Segal & Mandler, 1967).
 SCHEMA THEORIES 221

or associatively are easier to recall than are derson (1977). However, when subjects are
unrelated units (e.g., Deese, 1959). Recent given the alternative theme as a retrieval cue,
demonstrations of-this point also exist for previously irrelevant ideas become accessible
stimulus sets ranging from individual words (Anderson & Pichert, 1978). These newly rel-
(e.g., Bower & Clark-Meyers, 1980; see also evant ideas may not be particularly well re-
Bower, Clark, Lesgold, & Winzenz, 1969) to called; after all, they were not encoded in the
prose (Schustack & Anderson, 1979; Thorn- context of their theme, and so the theme's
dyke, 1977). When subjects are prevented use as a retrieval cue will likely be limited
from forming connections either by omitting (Schustack & Anderson, 1979; Thorndyke,
a cue to the existence of a hierarchy or by 1977; Tulving & Thomson, 1973).
presenting such a cue only after the story has The ability to relate sentences to one an-
been read, recall is poor for both prose other during encoding may be a determining
(Schustack & Anderson, 1979; Thorndyke, factor involved in recalling the propositions
1977) and simpler verbal materials (Bower of a text. If ideas are not or cannot be con-
et al, 1969). nected to a higher-order cue (Pichert & An-
A similar analysis applies to frame theory derson, 1977; Thorndyke, 1977) or to each
studies in which sentences constituting a other (e.g., Bransford & Johnson, 1972;
story are presented in scrambled versus story- Chiesi et al., 1979) during encoding, recall
normal order (e.g., Thorndyke, 1977).5 Ran- will be poor. Retrieval deficits ordinarily
dom sentence order could easily reduce the found with free recall may be circumvented
ability of a subject to determine the connec- by the use of either retrieval cues or recog-
tions among ideas or elements in the passage, nition tests (e.g., Alba et al,, 1981; Britton et
and the relative absence of such intersentence al., 1980; Chiesi et al., 1979).
connections should reduce recall. Similarly, Similar processes can be used to explain
recall of stories taken from other cultures the importance effect, that is, the fact that
may be poor, in part because of the difficulty ideas important to the theme are more likely
American students have in establishing ap- to be recalled than ideas that are not so im-
propriate intersentence or prepositional links portant. The majority of research on the im-
during the encoding process (Kintsch & portance variable confirms its role in recall,
Greene, 1978; Mandler & Johnson, 1977). but again, it is not necessary that the under-
On some occasions, one message may lead lying process be a differential encoding one.
to the establishment of two or more inde- A number of recent findings suggest that
pendent prepositional networks, each with items at all levels of importance receive the
its own hierarchical structure. This is the sit- same amount of initial processing (Britton,
uation that should prevail whenever a story Meyer, Simpson, Holdredge, & Curry, 1979;
can be read from two different perspectives Johnson & Scheldt, 1977; Waters, 1978).
and the ideas related to one perspective are Since low-importance information has re-
irrelevant to the other. Given one encoding cently been shown to be forgotten at the same
context, connections among related sen- rate as high-importance information (Chris-
tences should be established. All others will tiaansen, 1980), the explanation for the recall
exist outside of this network, either as un- advantage of high-importance information
related sentences or as sentences related in must lie with its advantage in accessibility.
their own network. During recall, the prop- An explanation favoring this view involves
osition at the top of the hierarchy (often the Kintsch and van Dijk's (1978) propositional
theme) is the most easily recalled proposition hierarchy algorithm for establishing a text
(for reasons to be discussed shortly) and is base. If the first idea retrieved is the highest
used to access subordinate, related ideas. If level proposition (or theme), recall will occur
the subject uses only the main theme as his
or her retrieval plan, ideas irrelevant to that 5
For recent critiques of story grammars as models of
theme (i.e., those ideas existing outside of the comprehension, representation, and recall of text ma-
text base) may never be accessed. This cor- terials see Black and Bower (1980), Brewer and Lichten-
responds to the findings of Pichert and An- stein (1981), and Thorndyke and Yekovich (1980).
222 JOSEPH W. ALBA AND LYNN HASHER

in a top-down manner with each recalled number of related subordinate actions that
proposition cuing one or more below it. The accompany them increases (Black & Bower,
model favors the recall of higher level prop- 1979). Rehearsal of connected ideas may also
ositions (Bower, 1976). For a low-level prop- account for the finding that recall of high-
osition to be recalled, the complete chain of importance ideas is good even when subjects
propositions above it must be kept intact. (in this case children) are unable to discrim-
This kind of text analysis has been quite suc- inate between ideas of high and low impor-
cessful in predicting recall performance tance (e.g., Brown & Smiley, 1977).
(Kintsch & Keenan, 1973; Kintsch et al, Thus, a number of nonschema theory
1975; Kozminsky, 1977; Meyer, 1975; memory processes appear to be involved in
Thorndyke, 1977). Others have also proposed prose recall: the existence and extent of con-
that text material is retrieved in a top-down nections among ideas; differential rehearsal
order (Britton et al., 1979; Waters, 1978; processes based on (covert) intersentence
Yekovich & Thorndyke, 1981). cuing during the comprehension process; the
That recall of prose ordinarily proceeds in availability and usefulness of retrieval cues
a top-down manner through a hierarchy at recall; and output-order produced acces-
means, of course, that items higher in the sibility problems. These are all conceptual-
hierarchy are likely to be recalled before izations with origins in nonschema theory
items lower in the hierarchy. Output order, research. Such processes, when coupled with
too, is a variable long known (see Postman contemporary associationist views of the rep-
& Underwood, 1973; Spear, 1978) to influ- resentation of prose (e.g., Anderson & Bower,
ence recall; items recalled early in a series 1973; Kintsch & van Dijk, 1978), offer the
decrease the probability that other actually possibility of accounting for differences in
stored items will be recalled. idea recallability without invoking selective
Kintsch and van Dijk's (1978) model offers encoding.7 More is stored in memory than
a reasonably objective determination of any version of a schema-based selective en-
proposition importance. Important ideas may coding mechanism can account for. Alter-
be ones that are referred to by others and so native conceptions must begin to take this
tend to be called into working memory more into consideration.
frequently and/or spend more time in work-
ing memory than others. As a consequence Abstraction
of the rehearsal processes that occur in work-
ing memory, such ideas will tend to have con- Abstraction may also be amenable to ex-
nections to a greater number of other ideas planations that emphasize retrieval rather
than will less important ideas. Components than encoding factors. Although meaning is
of each connection may then serve as re- undoubtedly the easiest aspect of a passage
trieval cues (see also Graesser, 1978a; Graes- to recall, subjects appear to store, and some-
seretal., 1980). times remember, the syntactic and lexical
Rehearsal now becomes an important de- vehicles by which meaning is conveyed.
terminant of recall of prose elements, as it Meaning may be the most persistent attribute
has long been considered to be in recall of (e.g., McKoon, 1977) in part because it is the
simpler units (e.g., Rundus, 1971). It is en-
tirely possible that what is loosely called the 6
Another nonschema explanation of the importance
"theme" is the proposition or concept that effect involves response biases. That is, there seems to
is most frequently rehearsed or referred to be a predisposition to label as the theme the initially
encountered idea in a prose passage (Kieras, 1980).Thus,
during encoding—hence its relatively easy surface structure variables may influence a person's
accessibility (Hasher & Griffin, 1978; Perfetti choice of the most important proposition.
& Goldman, 1974).6 Some evidence for the 7
We do not wish to ignore the concept of macro-
rehearsal hypothesis does exist. The more structure in the Kintsch and van Dijk model of prose
ideas that descend from a particular piece of comprehension. This aspect of the model is indeed a
schema process in that a few important propositions are
information, the greater the likelihood of its selected for representation in the macrostructure. The
recall (Meyer & McConkie, 1973). Main ac- macrostructure alone, though, cannot account for the
tions in an episode are recalled better as the levels and importance effects.
 SCHEMA THEORIES 223

information that the perceiver is most intent accounting for accurate memory of past
on remembering (Kintsch, 1977; Wanner, events; a fair number of studies have shown
1968) and/or the attribute most rehearsed, rather startling levels of accuracy even when
and/or the attribute least susceptible to in- subjects had no particular reason to expect
terference. a memory test (e.g., the work of Kintsch,
Task demands may also influence the ac- Bates, and their collaborators).
cessibility of nonsemantic detail. For exam- What is needed, in our view, is a theory
ple, instructions to process grammatical fea- of memory that can account both for accu-
tures of a sentence result in a marked in- racy and for distortion. A recent model of
crease in memory for those features (e.g., memory processes may be a useful candidate
Anderson & Paulson, 1977; Graesser & Man- (see Johnson & Raye, 1981). According to
dler, 1975; Johnson-Laird & Stevenson, 1970). this model, memory for an event will nor-
By contrast, most story materials and in- mally consist of individual traces from the
structions probably emphasize the impor- external event as well as traces from any in-
tance of meaning. ternal (subject-produced) generation done
Breakdowns in the retrieval process may during the comprehension process or during
also account for the occasional failure to re- any subsequent reprocessing of the event.
member lexical details. One theory that as- Associated with each trace are a variety of
sumes that memory contains accurate traces attributes (see Underwood, 1969), and the
even of complex events proposes that dis- relative amounts of these attributes serve as
crimination among related traces is a central a base for distinguishing between internally
source of lexical errors (e.g., Hayes-Roth & generated and externally derived traces. When
Hayes-Roth, 1977). For example, processing a trace is retrieved during a memory test, the
of a lexical unit may activate related, some- information that identifies its origin will be
times synonymous, traces in memory either used to determine whether or not the trace
through spread of activation (Hayes-Roth was part of the original stimulus. If a trace
& Hayes-Roth, 1977) or through the implicit was not part of the original event, and if ac-
generation of associative responses (e.g., Un- curacy is encouraged, the trace will not be
derwood, 1965). Errors may occur on a mem- recalled (nor will it be misrecognized), and
ory test if the associated lexical traces have accuracy will result. On some occasions,
been strongly activated and can compete with however, the critical discriminative attribute
the target information. Hence occasional information may become inaccessible, and
substitution errors may occur even though source-of-origin discriminations will not be
veridical traces are stored in memory. possible. Misrecognition of information in-
In summary, there are alternatives to ternally generated during encoding (con-
schema theory that can account for the ab- structive errors or interpretations) will result
straction errors traditionally explained by (see also Anderson & Bower, 1973; Hayes-
schema theory but that also account for the Roth & Thorndyke, 1979).
not infrequent cases of mnemonic accuracy There exists some preliminary evidence to
without appealing to "probable detail" ex- support this portion of the Johnson-Raye the-
planations. ory. Conditions that make source discrimi-
nation more difficult should increase the rate
Interpretation and Integration of confabulation. In fact, fact versus fantasy
discrimination ability for inferences de-
These are probably the most interesting creases when a delay (Owens et al., 1979) or
and widely discussed schema-theory con- an interfering task (Brockway et al., 1974) is
structs. This is so because, together with ab- inserted between acquisition and recall. Dis-
straction, they are held to be responsible for crimination increases when source-identify-
distortions of past events, and distortions are ing attributes are made more salient (Alba,
inherently interesting. However, distortions 1981; Johnson et al., 1981).
are not really all that common. Although it The reality-monitoring model has the
is the case that schema theory can account added advantage of being able to explain a
for them, schema theory is at its weakest in wide variety of mnemonic errors. The same
224 JOSEPH W. ALBA AND LYNN HASHER

explanation that is used to account for prag- Second, Johnson and Raye (1981) report
matic inference errors can also be used to that traces in the uncertain region are subject
account for logical inference misrecognition to response biases. When the source of a trace
errors (e.g., linear orderings), single word er- is uncertain, there is a tendency to attribute
rors (e.g., Underwood, 1965; Underwood & it to external sources (Johnson et al., 1981).
Freund, 1968), elaborative prose errors (e.g., Thus, when familiar ideas (e.g., thematic
Kintsch et al., 1975;Sulin&Dooling, 1974), ideas) are encountered on a recognition test,
default value errors (Minsky, 1975), and im- subjects may misrecognize them not because
portantly, alleged reconstructive errors (e.g., they are confidently recalling them from an
Spiro, 1977, 1980a, 1980b). That is, when integrated schema, but rather because they
responding on a memory test, people may have a particular response bias. Explanations
confuse internally generated memory traces that appeal to reconstruction processes such
with traces resulting from the encoding of the as the typicality effect predicted by script the-
(external) stimulus array. ory (e.g., Bower et al., 1979; Graesser et al.,
Reconstructive errors can be distinguished 1979; Graesser et al., 1980) and schema the-
from most other types of errors discussed in ory (e.g., Sulin & Dooling, 1974) may be bet-
this article in that they are believed to be the ter accommodated by the response biases
result of processes occurring at some point discussed within the framework of the reality-
after initial encoding. The most frequently monitoring model.8
cited evidence in support of the reconstruc-
tion process is the work of Spiro (1977, An Overview
1980a, 1980b). As Royer (1977) points out, As we. noted at the outset, recent research
however, the results do not necessarily imply in cognitive psychology has been greatly in-
reconstruction; reconstruction errors may be fluenced by schema theoretics. Although the
considered highly similar to other inferences present article has in large part questioned
generated during encoding and thereby per- the specifics of schema theory, there can be
fectly consistent with a reality-monitoring no denying its generally beneficial impact on
account of mnemonic error. Although it is the field of memory. Schema theory has re-
merely speculation at this point, the same vitalized the area and moved researchers to
explanation, combined with other factors consider a number of new issues about the
discussed below, may account for those re- nature of memory as well as about the pa-
constructive eiTors predicted by script theory rameters of the comprehension process (see
and known as the typicality effect. for example, Graesser, Hoffman, & Clark,
Two other aspects of the reality monitoring 1980; Kieras, 1981; Miller & Kintsch, 1980;
theory can help to explain other forms of Vipond, 1980). Although we have noted that
mnemonic inaccuracy. First, the model as- certain interpretive processes are not the sole
sumes that confidence that a particular trace domain of schema theory, it has been within
is internally or externally generated is dis-
tributed in a way similar to some models of 8
recognition memory (e.g., Atkinson & Juola, Script theory assumes that highly typical aspects of
scripted events are already stored as part of a script;
1973). That is, some traces are easily iden- when such events are actually encountered, there is no
tified as internally generated, whereas others need to restore these aspects. Subjects may, as a result,
are easily identified as externally derived. In not be able to distinguish between highly typical aspects
between, a region of uncertainty exists. How of events that did versus did not actually occur in a
these items are identified may depend on test particular instantiation of a script; this is termed the
typicality effect (e.g., Smith & Graesser, 1980). This ef-
conditions. If subjects are urged to be con- fect is the single most powerful demonstration of a re-
servative (e.g., Gauld & Stephenson, 1967) construction process. As just mentioned, response biases
or are given additional time or cues with seem a likely nonschema explanation. Some recent ev-
which to make discriminations (e.g., Hasher idence also suggests that the effect is tied to particular
confusing aspects of the "typical" procedures used in
& Griffin, 1978; Lawson, 1977), fewer inter- typicality research and is not an inevitable consequence
nally generated traces in the uncertain region of activating a script and encountering (or not encoun-
will be produced, and accuracy will increase. tering) highly typical events (Chromiak, Note 3).
 SCHEMA THEORIES 225

the context of schema theory that these pro- may impose their own interpretations on in-
cesses have been explored. coming stimuli, they appear to do so less of-
Nevertheless, we must return to the basic ten than schema theory would suggest. When
problem confronting schema theory: its def- people do generate interpretations, source
inition. Taylor and Crocker (1981) present an identifying markers are not necessarily lost.
excellent discussion of the adequacy of The final questions center on what deter-
schema theory as a theory. We will not at- mines recall. We have suggested that the
tempt to elaborate on their discussion except schema theory research can be accounted for
to once again point out the nebulous char- by assuming that a detailed representation of
acter of the theory and to restate our attempt a complex event is actually stored in mem-
to place some structure on what has trans- ory, perhaps in a format suggested by such
pired in the past 15 years. text-comprehension models as Kintsch and
We have proposed that schema theories van Dijk's (1978). What a subject shows of
share a belief in one or more of four basic what he or she has stored will hinge on a
memory encoding processes: selection, ab- variety of circumstances, some having to do
straction, interpretation, and integration. with connections formed or not formed dur-
Three of these, selection, abstraction, and in- ing encoding and with rehearsal processes
tegration, are processes that reduce the during encoding and/or the retention inter-
amount of information that will be stored in val, as well as with processes such as cuing,
memory. Selection does this by allowing only output order, response biases, task demands,
some information to receive further process- and so forth, that operate during retrieval.
ing. Abstraction does this by dropping from We have argued that there are a variety of
the representation all but the underlying nonschematic concepts available to explain
meaning of a message. Integration does this memory for complex events. Whatever the
by creating a holistic representation of what- ultimate usefulness of these latter arguments
ever remains after selection and abstraction may prove to be, however, we think it clear
with whatever relevant knowledge exists in that the stored record of any event is far more
memory. This holistic representation does detailed than prototypical schema theories
not include markers that would enable the imply. Contemporary theories of memory
person to inevitably distinguish old from new cannot disregard the richness of the stored
information. The final process, interpreta- trace.
tion, is the only one that enriches the mem-
ory representation, and it does so by adding Reference Notes
inferences, omitted details, and relevant prior 1. Day, J., Stein, N. L., Trabasso, T., & Shirey, L. A
knowledge to the incoming message. study of inferential comprehension: The use of a story
A final evaluation of schema theories schema to remember picture sequences. Paper pre-
hinges on two issues. These are (a) what is sented at the meeting of the Society for Research in
really stored in memory and (b) what causes Child Development, San Francisco, 1979.
2. Stein, N. L., & Glenn, C. G. The role of structural
distortions of factual information. Our re- variation in children's recall of simple stories. Paper
view of the literature suggests that memory presented at the meeting of the Society for Research
for complex events is far more detailed than in Child Development, New Orleans, 1977.
schematic processes would allow. At least 3. Chromiak, W. Personal Communication. May 1982.
some schematically unimportant informa-
tion is stored. At least some details are stored References
no matter what the extent of a person's prior Alba, J. W. The representational nature of inferences.
knowledge, and no matter whether that Unpublished doctoral dissertation. Temple University,
knowledge is activated at encoding, Lexical Philadelphia, 1981.
and syntactic information, along with se- Alba, J. W., Alexander, S. G., Hasher, L,, & Caniglia, K.
mantic information, are all accorded repre- The role of context in the encoding of information.
Journal of Experimental Psychology: Human Learning
sentation in memory. Memory of a topic does and Memory, 1981, 7, 283-292.
not seem to consist of a tightly integrated set Anderson, J. R. Verbatim and propositional represen-
of all relevant information. Although subjects tation of sentences in immediate and long-term mem-
226 JOSEPH W. ALBA AND LYNN HASHER

ory. Journal of Verbal Learning and Verbal Behavior, wording. Journal of Verbal Learning and Verbal Be-
1974, 13, 149-162, havior, 1971, 10, 176-181.
Anderson, J. R. Language, memory and thought. Hills- Begg, I., & Wickelgren, W. A. Retention functions for
dale, N.J.: Erlbaum, 1976. syntactic and lexical vs. semantic information in sen-
Anderson, J. R., & Bower, G. H. Human associative tence recognition memory. Memory & Cognition,
memory. Washington, D.C.: Winston, 1973. 1974, 2, 353-359.
Anderson, J. R., & Paulson, R. Representation and re- Bekerian, D. A., & Bowers, J. M. Eyewitness testimony:
tention of verbatim information. Journal of Verbal Were we misled? Journal of Experimental Psychology:
Learning and Verbal Behavior, 1977, 16, 439-451. Learning, Memory, and Cognition, 1983, 9, 139-145.
Anderson, R. C. Encoding processes in the storage and Birnbaum, I. M., Johnson, M. K., Hartley, J. T, & Tay-
retrieval of sentences. Journal of Experimental Psy- lor, T. H. Alcohol and elaborative schemas for sen-
chology, 1971, 91, 338-340. tences. Journal of Experimental Psychology: Human
Anderson, R. C. Substance recall of sentences. Quarterly Learning and Memory, 1980, 6, 293-300.
Journal of Experimental Psychology, 1974, 26, 530- Black, J. B., & Bower, G. H. Episodes as chunks in nar-
541. rative memory. Journal of Verbal Learning and Verbal
Anderson, R. C., et al. Instantiation of general terms. Behavior, 1979, 18, 309-318.
Journal of Verbal Learning and Verbal Behavior, 1976, Black, J. B., & Bower, G. H. Story understanding as prob-
/5, 667-679. lem-solving. Poetics, 1980, 9, 233-250.
Anderson, R. C., & Pichert, J. W. Recall of previously Bobrow, S. A. Memory for words in sentences. Journal
unrecallable information following a shift in perspec- of Verbal Learning and Verbal Behavior, 1970, 9, 363-
tive. Journal of Verbal Learning and Verbal Behavior, 372.
1978, 17, 1-12. Bock, J. K. The effect of a pragmatic presupposition on
Anisfeld, M., & Knapp, M. Association, synonymity, syntactic structure in question answering. Journal of
and directionality in false recognition. Journal of Ex- Verbal Learning and Verbal Behavior, 1977, 16, 723-
perimental Psychology, 1968, 77, 171-179. 734.
Atkinson, R. C., & Juola, J. F. Factors influencing speed Bock, J. K., & Brewer, W. F. Reconstructive recall in
and accuracy of word recognition. In S, Kornblum sentences with alternative surface structures. Journal
(Ed.), Attention and Performance IV. New York: Ac- of Experimental Psychology, 1974, 103, 837-843.
ademic Press, 1973. Bower, G. H. Experiments on story understanding and
Ausubel, D. P. The use of advance organizers in the learn- recall. Quarterly Journal of Experimental Psychology,
ing and retention of meaningful material. Journal of 1916,28, 511-534.
Educational Psychology, 1960, 51, 267-272. Bower, G. H., Black, J. B., & Turner, T. J. Scripts in
Ausubel, D. P. Educational psychology: A cognitive view. memory for text. Cognitive Psychology, 1979, 11,
New York: Holt, Rinehart & Winston, 1968. 177-220.
Ausubel, D. P., & Fitzgerald, D. The role of discrimi- Bower, G. H., Clark, M. C., Lesgold, A. M., & Winzenz,
nability in meaningful verbal learning and retention. D. Hierarchical retrieval schemas in recall of cate-
Journal of Educational Psychology, 1961, 52, 266- gorized word lists. Journal of Verbal Learning and
274. Verbal Behavior, 1969, 8, 323-343.
Ausubel, D. P., & Fitzgerald, D. Organizer, general back- Bower, G. H., & Clark-Meyers, G, Memories for scripts
ground, and antecedent learning variables in sequen- with organized vs. randomized presentations. British
tial verbal learning. Journal of Educational Psychol- Journal of Psychology, 1980, 71, 369-397.
ogy, 1962, 53, 243-249. Bower, G. H., Karlin, M. B., & Dueck, A. Comprehen-
Baker, L. Processing temporal relationships in simple sion and memory for pictures. Memory & Cognition,
stories: Effects of input sequence. Journal of Verbal 1975, 3, 216-220.
Learning and Verbal Behavior, 1978, 17, 559-572. Bransford, J. D., Barclay, J. R., & Franks, J. J. Sentence
Barclay, J. R. The role of comprehension in remember- memory: A constructive vs. interpretive approach.
ing sentences. Cognitive Psychology, 1973,4,229-254. Cognitive Psychology, 1972, 3, 193-209.
Barnes, B. R., & Clawson, E. U. Do advance organizers Bransford, J. D., & Franks, J. J. The abstraction of lin-
facilitate learning? Recommendations for further re- guistic ideas. Cognitive Psychology, 1971,2, 331-350.
search based on an analysis of 32 studies. Review of Bransford, J. D., & Johnson, M. K. Contextual prereq-
Educational Research, 1975, 45, 637-659. uisites for understanding: Some investigations of com-
Bartlett, F. C. Remembering: A study in experimental prehension and recall. Journal of Verbal Learning and
and social psychology. Cambridge, England: Cam- Verbal Behavior, 1972, //, 717-726.
bridge University Press, 1932. Bransford, J. D., & Johnson, M. K. Considerations of
Bates, E., Kintsch, W., Fletcher, C. R., & Giuliani, V. some problems of comprehension. In W. G. Chase
The role of pronominalization and ellipsis in texts: (Ed.), Verbal information processing. New York: Ac-
Some memory experiments. Journal of Experimental ademic Press, 1973.
Psychology: Human Learning and Memory, 1980, 6, Bransford, J. D., & Nitsch, K. E. Coming to understand
676-691. things we could not previously understand. In J. F.
Bates, E., Masling, M., & Kintsch, W. Recognition mem- Kavanagh & W. Strange (Eds.), Speech and language
ory for aspects of dialogue. Journal of Experimental in the laboratory, school, and clinic. Cambridge, Mass:
Psychology: Human Learning and Memory, 1978, 4, MIT Press, 1978.
187-197. Brewer, W. F. Memory for ideas: Synonym substitution.
Begg, I. Recognition memory for sentence meaning and Memory & Cognition, 1975, 3, 458-464.
 SCHEMA THEORIES 227

Brewer, W. F. Memory for the pragmatic implications Cofer, C. N., Chmielewski, D. L., & Brockway, J, F. Con-
of sentences. Memory & Cognition, 1977, 5, 673-678. structive processes and the structure of human mem-
Brewer, W. F., & Lichtenstein, E. H. Memory for marked ory. In C. N. Cofer (Ed.), The structure of human
semantic features versus memory for meaning. Jour- memory. San Francisco: Freeman, 1976.
nal of Verbal Learning and Verbal Behavior, 1974,13, Corbett, A. T., & Dosher, B. A. Instrument inferences
172-180. in sentence encoding. Journal of Verbal Learning and
Brewer, W. F, & Lichtenstein, E. H. Event schemas, story Verbal Behavior, 1978, 17, 479-491.
schemas, and story grammars. In J. Long & A. Bad- Deese, J. On the prediction of occurrence of particular
deley (Eds.), Attention and performance IX. Hillsdale, verbal intrusions in immediate recall. Journal of Ex-
N.J.: Erlbaum, 1981. perimental Psychology, 1959, 58, 17-22.
Brewer, W. F., & Treyens, J. C. Role of schemata in de Villiers, P. A. Imagery and theme in recall of con-
memory for places. Cognitive Psychology, 1981, 13, nected discourse. Journal of Experimental Psychology,
207-230. 1974, 103, 263-268.
Britton, B. K., Meyer, B. J., Hodge, M. H., & Glynn, Dodd, D. H., & Bradshaw, J. M. Leading questions and
S. M. Effects of the organization of text on memory: memory: Pragmatic constraints. Journal of Verbal
Tests of the retrieval and response criterion hy- Learning and Verbal Behavior, 1980, 19, 695-704.
potheses. Journal of Experimental Psychology: Human Dooling, D. J., & Christiaansen, R. E. Episodic and se-
Learning and Memory, 1980,6,620-629. mantic aspects of memory for prose. Journal of Ex-
Britton, B. K., Meyer, B. J. F., Simpson, R., Holdredge, perimental Psychology: Human Learning and Mem-
T. S., & Curry, C. Effects of the organization of text ory, 1977, 3, 428-436. (a)
on memory: Tests of two implications of'a selective Dooling, D. J., & Christiaansen, R. E. Levels of encoding
attention hypothesis. Journal of Experimental Psy- and retention of prose. In G. H. Bower (Ed.), The
chology: Human Learning and Memory, 1979, 5,496- psychology of learning and motivation (Vol. 11). New
506. York: Academic Press, 1977. (b)
Brockway, J. F., Chmielewski, D., & Cofer, C. N. Re- Dooling, D. J., & Lachman, R. Effects of comprehension
membering prose: Productivity and accuracy con- on retention of prose. Journal of Experimental Psy-
straints in recognition memory. Journal of Verbal chology, 1971, 88, 216-222.
Learning and Verbal Behavior, 1974, 13, 194-208. Dooling, D. J., & Mullet, R. L. Locus of thematic effects
Brown, A. L., & Smiley, S. S. Rating the importance of in retention of prose. Journal of Experimental Psy-
structural units of prose passages: A problem of meta- chology, 1973, 97, 404-406.
cognitive development. Child Development, 1977, 48, Pass, W., & Schumacher, G. M. Schema theory and prose
1-8. retention: Boundary conditions for encoding and re-
Brown, A. L., Smiley, S. S., Day, J. D., Townsend, trieval effects. Discourse Processes, 1981, 4, 17-26.
M. A. R., & Lawton, S. C. Intrusion of a thematic Fischhoff, B. Hindsight + foresight: The effect of out-
idea in children's comprehension and retention of sto- come knowledge on judgment under uncertainty.
ries. Child Development, 1911,48, 1454-1466. Journal of Experimental Psychology: Human Perfor-
Bugelski, B. R. Images as mediators in one-trial paired- mance and Perception, 1975, 1, 288-299.
associate learning. II: Self-timing in successive lists. Fischhoff, B. Perceived informativeness of facts. Journal
Journal of Experimental Psychology, 1968, 77, 328- of Experimental Psychology: Human Performance and
334. Perception, 1977, 3, 349-358.
Burnett, S. A., & Stevenson, J. M. Instructional effects Fischhoff, B., & Beyth, R. I knew it would happen—
on frequency judgments for sentence meanings. Amer- Remembered probabilities of once-future things. Or-
ican Journal of Psychology, 1979, 92, 711-721. ganizational Behavior and Human Performance, 1975,
Chase, W. G., & Simon, H. A. Perception in chess. Cog- 13, 1-16.
nitive Psychology, 1973, 4, 55-81. Flagg, P. W. Semantic integration in sentence memory?
Chi, M. T. H. Knowledge structures and memory de- Journal of Verbal Learning and Verbal Behavior, 1976,
velopment. In R. S. Siegler (Ed.), Children's thinking: 75,491-504.
What develops^ Hillsdale, N.J.: Erlbaum, 1978. Flagg, P. W., & Reynolds, A. G. Modality of presentation
Chiesi, H. L., Spilich, G. J., & Voss, J. F. Acquisition of and blocking in sentence recognition memory. Mem-
domain-related information in relation to high and ory & Cognition, 1977, 5, 111-115.
low domain knowledge. Journal of Verbal Learning Frederiksen, C. H. Acquisition of semantic information
and Verbal Behavior, 1979, 18, 257-274. from discourse: Effects of repeated exposures. Journal
Christiaansen, R. E. Prose memory: Forgetting rates for of Verbal Learning and Verbal Behavior, 1975, 14,
memory codes. Journal of Experimental Psychology: 158-169. (a)
Human Learning and Memory, 1980, 6, 611-619. Frederiksen, C. H. Effects of context-induced processing
Christie, D. J., & Schumacher, G. M. Developmental operations on semantic information acquired from dis-
trends in the abstraction and recall of relevant versus course. Cognitive Psychology, 1975, 7, 139-166. (b)
irrelevant thematic information from connected ver- Gauld, A., & Stephenson, G. M. Some experiments re-
bal materials. Child Development, 1975, 46, 598-602. lating to Bartlett's theory of remembering. British
Clark, H. H., & Clark, E.-V. Semantic distinctions and Journal of Psychology, 1967, 58, 39-49.
memory for complex sentences. Quarterly Journal of Centner, D., & Loftus, E. F. Integration of verbal and
Experimental Psychology, 1968, 20, 129-138. visual information as evidenced by distortions in pic-
Cofer, C. N. Constructive processes in memory. Amer- ture memory. American Journal of Psychology, 1979,
ican Scientist, 1973, 61, 537-543. 92, 366-375.
228 JOSEPH W. ALBA AND LYNN HASHER

Glenn, C. G. The role of episodic structure and of story Hasher, L., & Griffin, M. Reconstructive and reproduc-
length in children's recall of simple stories. Journal tive processes in memory. Journal of Experimental
of Verbal Learning and Verbal Behavior, 1978, 17, Psychology: Human Learning and Memory, 1978, 4,
229-247. 318-330.
Goetz, E. T. Inferring from text: Some factors influencing Haviland, S. E., & Clark, H. H. What's new? Acquiring
which inferences will be made. Discourse Processes, new information as a process in comprehension. Jour-
1979, 2, 179-195. nal of Verbal Learning and Verbal Behavior, 1974,13,
Gomulicki, B. R. Recall as an abstractive process. Ada 512-521.
Psychologies 1956, 12, 77-94. Hayes-Roth, B., & Hayes-Roth, F. The prominence of
Graesser, A. C. How to catch a fish: The memory and lexical information in memory representations of
representation of common procedures. Discourse Pro- meaning. Journal of Verbal Learning and Verbal Be-
cesses, 1978, 7, 72-89. (a) havior, 1977, 16, 119-136.
Graesser, A. C. Tests of a holistic chunking model of Hayes-Roth, B., & Thorndyke, P. W. Integration of
sentence memory through analyses of noun intru- knowledge from texts. Journal of Verbal Learning and
sions. Memory & Cognition, 1978, 6, 527-536. (b) Verbal Behavior, 1979, 18, 91-108.
Graesser, A. C., Gordon, S. E., & Sawyer, J. D. Recog- Hupet, M., & LeBouedec, B. The given-new contract
nition memory for typical and atypical actions in and the constructive aspect of memory for ideas. Jour-
scripted activities: Tests of a script pointer + tag hy- nal of Verbal Learning and Verbal Behavior, 1977,16,
pothesis. Journal of Verbal Learning and Verbal Be- 69-75.
havior, 1919,18, 319-332. James, C. T., & Hillinger, M. L. The role of confusion
Graesser, A. C., Hoffman, N. L., & Clark, L. F. Structural in the semantic integration paradigm. Journal of Ver-
components of reading time. Journal of Verbal Learn- bal Learning and Verbal Behavior, 1977, 76,711-721.
ing and Verbal Behavior, 1980, 19, 135-151. James, C. T, Hillinger, M. L., & Murphy, B. J. The effect
Graesser, A., II, & Mandler, G. Recognition memory for of complexity on confidence ratings in linguistic in-
the meaning and surface structure of sentences. Jour- tegration. Memory & Cognition, 1977, 5, 355-361.
nal of Experimental Psychology: Human Learning and James, C. T, Thompson, J. G., & Baldwin, J. M. The
Memory, 1975, 1, 238-248. reconstructive process in sentence memory. Journal
Graesser, A. C, Robertson, S. P., Lovelace, E. R., & of Verbal Learning and Verbal Behavior, 1973, 12,
Swinehart, D. M. Answers to why-questions expose 51-63.
the organization of story plot and predict recall of Jarvella, R. J. Syntactic processing of connected speech.
actions. Journal of Verbal Learning and Verbal Be- Journal of Verbal Learning and Verbal Behavior, 1971,
havior, 1980, 7 9, 110-119. 10, 409-416.
Graesser, A. C., Woll, S. B., Kowalski, D. J., & Smith, Johnson, M. K., Bransford, J. D., & Solomon, S. K.
D. A. Memory for typical and atypical actions in Memory for tacit implications of sentences. Journal
scripted activities. Journal of Experimental Psychol- of Experimental Psychology, 1973, 98, 203-205.
ogy: Human Learning and Memory, 1980,6, 503-515. Johnson, M. K., Doll, T. J., Bransford, J. D., & Lapinski,
Greene, E., Flynn, M. S., & Loftus, E. F. Inducing re- R. H. Context effects in sentence memory. Journal of
sistance to misleading information. Journal of Verbal Experimental Psychology, 1974, 103, 358-360.
Learning and Verbal Behavior, 1982, 21, 207-219. Johnson, M. K., & Raye, C. L. Reality monitoring. Psy-
Greenwald, A. G. The totalitarian ego: Fabrication and chological Revie\v, 1981, 88, 67-85.
revision of personal history. American Psychologist, Johnson, M. K., Raye, C. L., Foley, H. J., & Foley,
1980,55, 603-618. M. A. Cognitive operations and decision bias in reality
Griggs, R. A. The recall of linguistic ideas. Journal of monitoring. American Journal of Psychology, 1981,
Experimental Psychology, 1974, 103, 807-809. 94, 37-64.
Griggs, R. A., & Keen, D. M. The role of test procedure Johnson, R. E. Recall of prose as a function of the struc-
in linguistic integration studies. Memory & Cognition, tural importance of the linguistic unit. Journal oj'Ver-
1977,5, 685-689. bal Learning and Verbal Behavior, 1970, 9, 12-20.
Gude, C., & Zechmeister, E. B. Frequency judgments for Johnson, R. E., & Scheldt, B. J. Organizational encod-
the "gist" of sentences. American Journal of Psychol- ings in the serial learning of prose. Journal of Verbal
ogy, 1975, 88, 385-396. Learning and Verbal Behavior, 1977, 76, 575-588.
Hannigan, M. L. The effects of frameworks on sentence Johnson-Laird, P. N., & Stevenson, R. Memory for syn-
perception and memory. Unpublished doctoral dis- tax. Nature, 1970, 227,412.
sertation, Vanderbilt University, 1976. Katz, S., Atkeson, B., & Lee, J. The Bransford-Franks
Harris, R. J. Memory and comprehension of implica- linear effect: Integration or artifact? Memory & Cog-
tions and inferences of complex sentences. Journal of nition, 1974,2,709-713.
Verbal Learning and Verbal Behavior, 1974, 13, Katz, S., & Gruenewald, P. The abstraction of linguistic
626-637. ideas in "meaningless" sentences. Memory & Cogni-
Harris, R. J., & Monaco, G. E. Psychology of pragmatic tion, 1974,2,737-741.
implication: Information processing between the lines. Keenan, J. M., & Kintsch, W. The identification of ex-
Journal of Experimental Psychology: General, 1978, plicitly and implicitly presented information. In W.
707, 1-22. Kintsch (Ed.), The representation of meaning in mem-
Hasher, L., Attig, M. S., & Alba, J. W. I knew it all along: ory. Hillsdale, N.J.: Erlbaum, 1974.
Or did I? Journal of Verbal Learning and Verbal Be- Keenan, J. M., MacWhinney, B., & Mayhew, D. Prag-
havior, 1981, 20, 86-96. matics in memory: A study of natural conversation.
 SCHEMA THEORIES 229

Journal of Verbal Learning and Verbal Behavior, 1977, Loftus, E. F. Leading questions and the eyewitness re-
763 549-560. port. Cognitive Psychology, 1975, 7, 560-572.
Kemper, S. Memory for the form and force of declara- Loftus, E. F. Eyewitness testimony. Cambridge, Mass.:
tives and interrogatives. Memory & Cognition, 1980, Harvard University Press, 1979. (a)
5,367-371. Loftus, E. F. Reactions to blatantly contradictory infor-
Kieras, D. E. Good and bad structure in simple para- mation. Memory & Cognition, 1979, 7, 368-374. (b)
graphs: Effects on apparent theme, reading time, and Loftus, E. F., & Loftus, G. R. On the permanence of
recall. Journal of Verbal Learning and Verbal Behav- stored information in the human brain. American
ior, 1978, 17, 13-28. Psychologist, 1980, 35, 409-420.
Kieras, D. E. Initial mention as a signal to thematic Loftus, E. F., Miller, D. G., & Burns, H. J. Semantic
content in technical passages. Memory & Cognition, integration of verbal information into a visual mem-
1980, S, 345-353. ory. Journal of Experimental Psychology: Human
Kieras, D. E. Component processes in the comprehen- Learning and Memory, 1978, 4, 19-31.
sion of simple prose. Journal of Verbal Learning and Loftus, E. F., & Palmer, J. C. Reconstruction of auto-
Verbal Behavior, 1981, 20, 1-23. mobile destruction: An example of the interaction
'King, D. R. W., & Greeno, J. G. Invariance of inference between language and memory. Journal of Verbal
time when information was presented in different lin- Learning and Verbal Behavior, 1974, 13, 585-589.
guistic formats. Memory and Cognition, 1974,2, 233- Mandler, J. M. A code in the node: The use of a story
235. schema in retrieval. Discourse Processes, 1978,
Kintsch, W. The representation of meaning in memory. 7, 14-35.
Hillsdale, N.J.: Erlbaum, 1974. Mandler, J. M., & Johnson, N. S. Remembrance of
Kintsch, W. Memory and cognition. New York: Wiley, things parsed: Story structure and recall. Cognitive
1977. Psychology, 1977,9, 111-151.
Kintsch, W., & Bates, E. Recognition memory for state- McKoon, G. Organization of information in text mem-
ments from a classroom lecture. Journal of Experi- ory. Journal of Verbal Learning and Verbal Behavior,
mental Psychology: Human Learning and Memory, 1977, 16, 247-260.
1977, 3, 150-159. McKoon, G., & Keenan, J. M. Response latencies to
Kintsch, W., & Greene, E. The role of culture-specific explicit and implicit statements as a function of the
schemata in the comprehension and recall of stories. delay between reading and test. In W. Kintsch (Ed.),
Discourse Processes, 1978, 1, 1-13. The representation of meaning in memory. Hillsdale,
Kintsch, W., & Keenan, J. M. Reading rate and retention N.J.: Erlbaum, 1974.
as a function of the number of propositions in the Meyer, B. J. F. The organization of prose and its effects
base structure of sentences. Cognitive Psychology, on recall. Amsterdam: North Holland, 1975.
1973, 5, 257-274. Meyer, B. J. F. The structure of prose: Effects on learning
Kintsch, W., & van Dijk, T. A. Toward a model of text and memory and implications for educational prac-
comprehension and production. Psychological Re- tice. In R. C. Anderson, R. J. Spiro, & W. E. Montague
vim, 1978,55,363-394. (Eds.), Schooling and the acquisition of knowledge.
Kintsch, W., Kozminsky, E., Streby, W. J., McKoon, G., Hillsdale, N.J.: Erlbaum, 1977.
& Keenan, J. M. Comprehension and recall of text as Meyer, B. J. F., & McConkie, G. What is recalled after
a function of content variables. Journal of Verbal hearing a passage? Journal of Educational Psychology,
Learning and Verbal Behavior, 1975, 14, 196-214. 1973, 65, 109-117.
Kintsch, W., Mandel, T. S., & Kozminsky, E. Summa- Meyers, L. S., & Boldrick, D. Memory for meaningful
rizing scrambled stories. Memory & Cognition, 1977, connected discourse. Journal of Experimental Psy-
5, 547-522. chology: Human Learning and Memory, 1975,1, 584-
Kintsch, W., & Monk, D. Storage of complex infor- 591.
mation in memory: Some implications of the speed Miller, J. R., & Kintsch, W. Readability and recall of
with which inferences can be made. Journal of Ex- short prose passages: A theoretical analysis. Journal
perimental Psychology, 1972, 94, 25-32. of Experimental Psychology: Human Learning and
Kozminsky, E. Altering comprehension: The effect of Memory, 1980, 6, 335-354.
biasing titles on text comprehension. Memory & Cog- Minsky, M. A framework for representing knowledge.
nition, 1977, 5, 482-490. In P. H. Winston (Ed.), The psychology of computer
Lawson, R. Representation of individual sentences and vision. New York: McGraw-Hill, 1975.
holistic ideas. Journal of Experimental Psychology: Moeser, S. D. Inferential reasoning in episodic memory.
Human Learning and Memory, 1977,3, 1-9. Journal of Verbal Learning and Verbal Behavior, 1976,
Lesgold, A. M. Pronominalization: A device for unifying 15, 193-212.
sentences in memory. Journal of Verbal Learning and Moeser, S. D. Recognition processes in episodic memory.
Verbal Behavior, 1972, 77, 316-323. Canadian Journal of Psychology, 1977, 31, 41-70.
Lesgold, A. M., Roth, S. F., & Curtis, M. E. Foreground- Moeser, S. D., & Tarrant, B. L. Learning a network of
ing effects in discourse comprehension. Journal of comparisons. Journal of Experimental Psychology:
Verbal Learning and Verbal Behavior, 1979, 75, 291- Human Learning and Memory, 1977, 3, 643-659.
308. Morris, C. D., Bransford, J. D., & Franks, J. J. Levels
Lichtenstein, E. H., & Brewer, W. F. Memory for goal- of processing versus transfer appropriate processing,
directed events. Cognitive Psychology, 1980, 12, 412- Journal of Verbal Learning and Verbal Behavior, 1977,
445. 76,519-533.
230 JOSEPH W. ALBA AND LYNN HASHER

Morris, C. D., Stein, B. S., & Bransford, J, D. Prereq- Royer, J. M. Remembering: Constructive or reconstruc-
uisites for the utilization of knowledge in the recall of tive? Comments on chapter 5 by Spiro. In R. C. An-
prose passages. Journal of Experimental Psychology: derson, R. J. Spiro, & W. E. Montague (Eds.), School-
Human Learning and Memory, 1979, 5, 253-261. ing and the acquisition of knowledge. Hillsdale, N.J.:
Norman, D. A., & Rumelhart, D. E. Memory and Erlbaum, 1977.
knowledge. In D. A. Norman & D. E. Rumelhart Royer, J. M., & Cable, G. W. Facilitative transfer in prose
(Eds.), Explorations in cognition. San Francisco: Free- learning. Journal of Educational Psychology, 1975,67,
man, 1975, 116-123.
Olson, D. R., & Filby, N. On the comprehension of active Royer, J. M., & Cable, G. W. Illustrations, analogies, and
and passive sentences. Cognitive Psychology, 1972, 3, facilitative transfer in prose learning. Journal of Ed-
361-381. ucational Psychology, 1976, 68, 205-209.
Ortony, A., Schallert, D. L., Reynolds, R. E., & Antos, Royer, J. M., & Perkins, M. R. Facilitative transfer in
S. J. Interpreting metaphors and idioms: Some effects prose learning over an extended period of time. Jour-
of context on comprehension. Journal of Verbal nal of Reading Behavior, 1977, 9, 185-188.
Learning and Verbal Behavior, 1978, 17, 465-478. Royer, J. M., Perkins, M. R., & Konold, C. E. Evidence
Owens, J., Bower, G. H., & Black, J. B. The "soap-opera" for a selective storage mechanism in prose learning.
effect in story recall. Memory & Cognition, 1979, 7, Journal of Educational Psychology, 1978, 70, 457-
185-191. 462.
Paris, S. G., & Carter, A. Semantic and constructive as- Rubin, D. C. Very long-term memory for prose and
pects of sentence memory in children. Developmental verse. Journal of Verbal Learning and Verbal Behavior,
Psychology, 1973, 9, 189-197. 1977,76,611-621. .
Paris, S. G., & Lindauer, B. K. The role of inference in Rundus, D. Analysis of rehearsal processes in free recall.
children's comprehension and memory for sentences. Journal of Experimental Psychology, 1971, 89, 63-77.
Cognitive Psychology, 1976, 8, 217-227. Sachs, J. S. Recognition memory for syntactic and se-
Paris, S. G., & Lindauer, B. K. Constructive aspects of mantic aspects of connected discourse. Perception and
children's comprehension and memory. In R. V. Kail, Psychophysics, 1967, 2, 437-442.
Jr. & J. W. Hagen (Eds.), Perspectives on the devel- Sachs, J. S. Memory in reading and listening to dis-
opment of memory and cognition. Hillsdale, N.J.: Erl- course. Memory & Cognition, 1974, 2, 95-100.
baum, 1977. Schallert, D. L. Improving memory for prose: The re-
Paris, S. G., Lindauer, B. K., & Cox, G. L. The devel- lationship between depth of processing and context.
opment of inferential comprehension. Child Devel- Journal of Verbal Learning and Verbal Behavior, 1976,
opment, 1977, 48, 1728-1733. 15, 621-632.
Paris, S. G., & Upton, L. R. Children's memory for Schank, R. C. Conceptual dependency: A theory of nat-
inferential relationships in prose. Child Development, ural language understanding. Cognitive Psychology,
1976, 47,660-668. 1972,3, 552-631.
Paul, I. H. Studies in remembering: The reproduction Schank, R. C. The role of memory in language process-
of connected and extended verbal material. Psycho- ing. In C. N. Cofer (Ed.), The structure of human
logical Issues, 1959, 1(2, Whole No. 2). memory. San Francisco: Freeman, 1976.
Perfetti, C. A., & Goldman, S. R. Thematization and Schank, R. C., & Abelson, R. Scripts, plans, goals, and
sentence retrieval. Journal of Verbal Learning and understanding. Hillsdale, N.J.: Erlbaum, 1977.
Verbal Behavior, 1974, '13, 70-79. Schustack, M. W., & Anderson, J. R. Effects of analogy
Pichert, J. W., & Anderson, R. C. Taking different per- to prior knowledge on memory for new information.
spectives on a story. Journal of Educational Psychol- Journal of Verbal Learning and Verbal Behavior, 1979,
ogy, 1977, 69, 309-315. 18, 565-584.
Postman, L., & Underwood, B, J. Critical issues in in- Schweller, K. G., Brewer, W. R, & Dahl, D. A. Memory
terference theory. Memory & Cognition, 1973, 1, 19- for illocutionary forces and perlocutionary effects of
40. utterances. Journal of Verbal Learning and Verbal
Potts, G. R. Memory for redundant information. Mem- Behavior, 1976, 75, 325-337.
ory & Cognition, 1973, 7, 467-470. Segal, M. A., & Mandler, G. Directionality and organi-
Potts, G. R. Artificial logical relations and their relevance zational processes in paired-associate learning. Jour-
to semantic memory. Journal of Experimental Psy- nal of Experimental Psychology, 1967, 74, 305-312.
chology: Human Learning and Memory, 1976, 2,746- Shaughnessy, J. J., & Mand, J. L. How permanent are
758. memories for real life events? American Journal of
Potts, G. R. Integrating new and old information. Jour- Psychology, 1982, 95, 51-65.
nal of Verbal Learning and Verbal Behavior, 1977,16, Singer, M. A replication of Bransford and Franks' (1971)
305-320. "The abstraction of linguistic ideas" Bulletin of the
Potts, G. R., Keller, R. A., & Rooley, C. J. Factors af- Psychonomic Society, 1973, 7, 416-418.
fecting the use of world knowledge to complete a linear Singer, M. Processes of inference during sentence en-
ordering. Journal of Experimental Psychology: Human coding. Memory & Cognition, 1979, 7, 192-200.
Learning and Memory, 1981, 7, 254-268. Singer, M. Verifying the assertions and implications of
Prawat, R. S., & Cancelli, A. Constructive memory in language. Journal of Verbal Learning and Verbal Be-
conserving and nonconserving first graders. Develop- havior, 1981,20, 46-60.
mental Psychology, 1976, 12, 47-50. Small, D. W. The abstraction of arbitrary categories.
Reitman, J. S., & Bower, G. H. Storage and later rec- Memory & Cognition, 1975, 3, 581-585.
ognition of exemplars of concepts. Cognitive Psy- Smiley, S. S., Oakley, D. D., Worthen, D., Campione,
chology, 1973, 4, 194-206. J. C., & Brown, A. L. Recall of thematically relevant
 SCHEMA THEORIES 231

material by adolescent good and poor readers as a of Verbal Learning and Verbal Behavior, 1966, 5,
function of written versus oral presentation. Journal 381-391.
of Educational Psychology, 1977,69, 381-387. Tulving, E., & Thomson, D. M. Encoding specificity and
Smith, D. A., & Graesser, A. C. Memory for actions in retrieval processes in episodic memory. Psychological
scripted activities as a function of typicality, retention Review, 1973, SO, 352-373.
interval, and retrieval task. Memory & Cognition, Tzeng, O. J. L. Sentence memory: Recognition and in-
1981, 9, 550-559. ferences. Journal of Experimental Psychology: Human
Smith, M. E. Delayed recall of previously memorized Learning and Memory, 1975, /, 720-726.
material after twenty years. Journal of Genetic Psy- Underwood, B. J. False recognition produced by implicit
chology, 1935, 47,-477-481. verbal responses. Journal of Experimental Psychology,
Soli, S. D., & Balch, W. R. Performance biases and rec- 1965, 70, 122-129.
ognition memory for semantic and formal changes in Underwood, B. J. Stimulus selection in verbal learning.
connected discourse. Memory & Cognition, 1976, 4, In C. N. Cofer & B. S. Musgrave (Eds.), Verbal be-
673-676. havior and learning. New York: McGraw-Hill, 1963.
Spear, N. E. The processing of memories: Forgetting and Underwood, B. J. Attributes of memory. Psychological
retention. Hillsdale, N.J.: Erlbaum, 1978. Review, 1969, 76, 559-573.
Spilich, G. J., Vesonder, G. T.( Chiesi, H. L., & Voss, Underwood, B. J., & Freund, J. S. Errors in recognition
J. F. Text processing of domain-related information learning and retention. Journal of Experimental Psy-
for individuals with high and low domain knowledge. chology, 1968, 78, 55-63.
Journal of Verbal Learning and Verbal Behavior, 1979, Underwood, B. J,, & Schulz, R. W. Meaningfulness and
18, 275-290, verbal learning. Philadelphia: Lippincott, 1960.
Spiro, R. J. Remembering information from text: The Vipond, D. Micro- and macroprocesses in text compre-
"state of schema" approach. In R, C. Anderson, hension. Journal of Verbal Learning and Verbal Be-
R. J. Spiro, & W. E. Montague (Eds.), Schooling and havior, 1980,79,276-296.
the acquisition of knowledge. Hillsdale, N.J.: Erlbaum, Walker, C. H., & Meyer, B. J. F. Integrating different
1977. types of information in text. Journal of Verbal Learn-
Spiro, R. J. Accommodative reconstruction in prose re- ing and Verbal Behavior, 1980, 19, 263-275.
call. Journal of Verbal Learning and Verbal Behavior, Walsh, D. A., & Baldwin, M. Age differences in inte-
1980, 19, 84-95. (a) grated semantic memory. Developmental Psychology,
Spiro, R. J. Prior knowledge and story processing: In- 1977, 13, 509-514.
tegration, selection, and variation. Poetics, 1980, 9, Wanner, H. E. On remembering, forgetting, and under-
313-327. (b) standing sentences: A study of the deep structure hy-
Stein, N. L., & Glenn, C. G. An analysis of story com- pothesis. Unpublished doctoral dissertation, Harvard
prehension in elementary school children. In R. O. University, Cambridge, Mass., 1968.
Freedle (Ed.), New directions in discourse processing Waters, H. S. Superordinate-subordinate structure in
(Vol. 2). Hillsdale, N.J.: Erlbaum, 1979. semantic memory: The roles of comprehension and
Stein, N. L., & Nezworski, T. The effects of organization retrieval processes. Journal of Verbal Learning and
and instructional set on story memory. Discourse Pro- Verbal Behavior, 1978, 17, 587-598.
cesses, 1978, ;, 177-193. Wilkes, A. L., & Aired, G. Prose and prejudice: Some
Sulin, R. A., & Dooling, D. J. Intrusion of a thematic effects of priming context on the immediate recall of
idea in retention of prose. Journal of Experimental information. British Journal of Psychology, 1978, 69,
Psychology, 1974, 103, 255-262. 123-133.
Taylor, S. E., & Crocker, J. Schematic bases of social Wood, G. The knew-it-all-along effect. Journal of Ex-
information processing. In E. T. Higgins, C. P. Her- perimental Psychology: Human Performance and Per-
man, & M. P. Zanna, (Eds.), Social cognition: The ception, 1978, 4, 345-353.
Ontario Symposium (Vol. 1). Hillsdale, N.J.: Erlbaum, Yekovich, F. R., & Manelis, L. Accessing integrated and
1981. nonintegrated prepositional structure in memory.
Thorndyke, P. W. The role of inferences in discourse Memory & Cognition, 1980, 8, 133-144.
comprehension. Journal of Verbal Learning and Ver- Yekovich, F. R., & Thorndyke, P. W. An evaluation of
bal Behavior, 1976, 15, 437-446. alternative functional models of narrative schemata.
Thorndyke, P. W. Cognitive structures in comprehension Journal of Verbal Learning and Verbal Behavior, 1981,
and memory of narrative discourse. Cognitive Psy- 20, 454-469.
chology, 1977, 9, 77-110. Yekovich, F. R., Walker, C. H., & Blackman, H. S. The
Thorndyke, P. W., & Yekovich, F. R. A critique of, role of presupposed and focal information in inte-
schema-based theories of human story memory. Po- grating sentences. Journal of Verbal Learning and Ver-
etics, 1980, 9, 23-49. bal Behavior, 1979, 18, 535-548.
Tulving, E. Subjective organization in free recall of "un- Zangwill, O. L. Remembering revisted. Quarterly Jour-
related" words. Psychological Review, 1962, 69, nal of Experimental Psychology, 1972,24, 123-138.
344-354.
Tulving, E., & Pearlstone, Z. Availability versus acces- Received June 3, 1982
sibility of information in memory for words. Journal Revision received August 23, 1982 •