Reality Monitoring - Yale University · Johnson, Department of Psychology, State University of New York, Stony Brook, New York 11794. something may make it seem as though it was perceived

Psychological Review198!, Vol. 88, No. 1,67-85

Copyright 1981 by the American Psychological Association, Inc.0033-295X/81 /8801 -0067S00.75

Reality Monitoring

Marcia K. JohnsonState University of New York at Stony Brook

Carol L. RayeBarnard College

People remember information from two basic sources: that derived from externalsources (obtained through perceptual processes) and that generated by internalprocesses such as reasoning, imagination, and thought. Of particular interest tous are the processes people use in deciding whether information initially had anexternal or an internal source, which we call "reality monitoring." We proposea working model of reality monitoring to account for both discrimination andconfusion between memories for thoughts and memories for perceptions. Ex-amples of questions the model addresses are, What types of information are morelikely to be represented in memories of external events than in memories ofinternal events? What cues allow people to decide the origin of a memory? Whatis the nature of the decision processes involved? Which processes, and underwhat conditions, are likely to break down and lead to unreliable memory? Whatassumptions do individuals have about their memory for their thoughts com-pared to their memory for their perceptions? How accurate are these assump-tions? We summarize some research that is encouraging as far as the tractabilityof some of these problems is concerned and that demonstrates the usefulness ofthe particular working model proposed here.

Both perception of external stimuli andthought produce memories. We have re-ferred to the processes by which a personattributes a memory to an external or aninternal source as reality monitoring (John-son, 1977; Johnson & Raye, Note 1). Ofcourse, in one sense, the memories createdby thoughts are no less "real" than thosecreated by perceptual experiences, and theformer can be shown to have important con-sequences; for example, thinking about

This research was supported by Grant BNS-7813054from the National Science Foundation. We would liketo thank Mark Aronoff, Ed Casey, Don Fry, Julie Hoch-berg, Emil Menzel, and Tom Taylor for their partici-pation in discussions that clarified some of the ideas inthis article and for their helpful comments on an earlierdraft of the manuscript.

The general issue addressed here was raised at theannual meeting of the American Psychological Asso-ciation, Chicago, 1975. The proposed model and someof the present data were initially presented at the annualmeeting of the Psychonomic Society, San Antonio,Texas, 1978, and at the annual meeting of the RockyMountain Psychological Association, Las Vegas, Ne-vada, 1979.

Carol Raye is now at Bell Telephone Laboratories,Piscataway, New Jersey 08854.

Requests for reprints should be sent to Marcia K.Johnson, Department of Psychology, State Universityof New York, Stony Brook, New York 11794.

something may make it seem as though itwas perceived more often than it actuallywas (e.g., Johnson, Taylor, & Raye, 1977).However, real commonly refers to thingsexisting outside of oneself, and it is this sensethat is implied by reality monitoring. Theterm is intended to suggest similarities withthe concepts of reality testing and memorymonitoring. Reality testing generally refersto the process of distinguishing a presentperception from a present act of imaginationor act of remembering (Cameron, 1963;Freud, 1895/1966, p. 325, 1925/1961, p.237; Horowitz, 1978).

Reality testing has received some experi-mental attention from investigators inter-ested in perception (Perky, 1910; Segal &Fusella, 1970) and from those interested inclinical problems, especially because of thecentral role confusion between the real andthe imagined plays in diagnoses of schizo-phrenia (McGuigan, 1966; Mellor, 1970;Mintz & Alpert, 1972). In contrast, realitymonitoring refers to the process of distin-guishing a past perception from a past actof imagination, both of which resulted inmemories. Reality monitoring thus alsoshares with Hart's (1967) memory monitor-ing concept the idea of making judgments

67

68 MARCIA K. JOHNSON AND CAROL L. RAYE

about information in memory. Reality mon-itoring is particularly interesting in thatfairly extreme errors about the origin ofmemories are probably more common thanreality testing failures such as hallucina-tions. Reality monitoring failures may alsogenerally be more difficult to correct thanmisperceptions. You cannot, for example,move your head to look for changes in oc-clusion or reach out and try to touch an er-roneous memory.

One way to regard the events we call fail-ures in reality monitoring is as the extremeconsequences of elaborative processes thathave consistently interested cognitive psy-chologists. Cognitive approaches to memoryemphasize memories as a joint product ofexternal and internal events (e.g., Bartlett,1932; Bruner, 1957; James, 1890/1950;Krech, 1949; Kohler, 1947/1959; Lashley,1967; Postman, 1951; Tolman, 1948; and,more recently, Bower, 1972a; Bransford &Johnson, 1973; Cofer, 1973; Craik & Lock-hart, 1972; Dooling & Lachman, 1971;Hochberg, 1979; Kintsch, 1976; Mandler,1967; Neisser, 1967; Paivio, 1969; Tulving,1968). With rare exceptions (e.g., Gibson,1966, 1979), few modern investigators pro-pose that there is a straightforward rela-tionship between physical stimuli and themind's impression of those stimuli, much lessour memory for them. Rather, the more gen-eral view is that most, if not all, stimuli aresubject to alternative interpretations andelaboration and that what is remembered isbased on the way that events are interpreted,organized, and thought about. Given thistheoretical framework, is there any reasonto make a distinction between memories ofperceptions and memories of thoughts?

The idea that there can be no memoriesof pure sensory experience without some de-gree of cognitive elaboration is certainlycompelling. Yet equally compelling is theidea that memories do vary in the extent towhich they are tied to external events (i.e.,publicly available stimulus conditions). Inordinary usage, memory is by definition dis-tinguished from imagination by virtue of itsveridicality. Furthermore, the importance ofthe distinction between externally and in-ternally generated information is not simplydefinitional. For example, memories of child-

hood experiences such as sexual encountersmay suggest different psychological con-structs depending on whether they representearly imaginative processes or actual per-ceptual experiences. Freud's decision to treatsuch memories as largely the product of fan-tasy greatly influenced the development ofpsychoanalytic theory, especially of the ideaof infantile sexuality (Heidbreder, 1933, pp.385-386). In short, the fact that the criteriaof veridicality may not be entirely clear tomemory theorists does not argue against thecentral role veridicality plays in the conceptof memory.

Our position is not really in conflict witha general cognitive view emphasizing thecontribution of thought to comprehensionand memory (e.g., Bransford & Johnson,1972, 1973). However, in attempting todemonstrate that people do think while per-ceiving and remembering, there has been atendency to overemphasize the hopeless en-tanglement of memories derived from per-ception and thought. Many investigatorshave assumed not only that a memory is ajoint product of the external and internalevents but also that this joint memory existsin place of what might be called the externaland the internal components. For example,"We have strong evidence that the list-as-presented is not the list-as-encoded, just asa CVC presented may not lead to a CVCstored, but to an NLM [natural languagemediator] instead" (Klatzky, 1980, p. 222,emphasis added). The notions that exter-nally generated information is discardedonce it is used to generate a construction(Sachs, 1967) or that externally and inter-nally generated traces are equivalent in ev-ery respect have led to a number of expec-tations—for example, that ideas and theirparaphrases should be indistinguishable(Bransford & Franks, 1971; Jenkins, 1974),that both perceived and inferred componentsof a mental map should be equally available(e.g., Levine & Jankovic, Note 2), and thatif inferences do not intrude in recall, theymust not have been made previously (e.g.,Corbett & Dosher, 1978).

Thus, cognitive theorizing has sometimesassumed that the inability to disentangle in-ternally and externally generated informa-tion is the defining characteristic of a "cog-

REALITY MONITORING 69

nitive" system. However, it is not logicallynecessary to prove that specific memory (i.e.,memory faithful to perceptual events) doesnot exist in order to show that organizedmental representations do exist. Yet it isprecisely the tendency to equate evidence forconstructive or imaginative processes withevidence against accurate memory (and viceversa) that limits the questions cognitive psy-chologists have posed. In addition, in em-phasizing the many transformations a stim-ulus might undergo, we run the danger offorgetting the dire functional implicationsof a memory system that is assumed to beso loosely tied to external events (althoughthe problem is beginning to be recognized;Neisser, 1976).

If thought is assumed to produce memoryrepresentations but not necessarily to sup-plant those yielded by perception, a differentset of questions for research is suggested.What are the characteristics of both mem-ories derived primarily from perceptual ex-perience and memories more remotely re-lated to perceptual experience? How are therepresentations of externally and internallygenerated events distinguished, and whenand why are they confused? Understandingthe differences and similarities in the natureof internally and externally generated mem-ories is important for any theory of the wayinformation is represented in memory, andan adequate model must be able to accountfor both confusion and discrimination be-tween past imaginations and past percep-tions (Johnson, Note 3). J. S. Mill (1869/1967) suggested that "the only difficultyabout Memory, when once the laws of As-sociation are understood, is the differencebetween it and imagination" (p. 339). Toparaphrase this notion to fit the presentproblem, once we have acknowledged therole of imagination in memory, we still havethe problem of accounting for the condi-tions under which we know the differencebetween memories of external events andmemories of imaginations. What are our"cues to reality?"

For the most part, philosophers and psy-chologists have not been concerned with theproblem of reality monitoring, althoughJames Mill (1829/1967) did assert that be-cause sensations and ideas "are distinguish-

able in the having, it is likely that the copyof the sensation should be distinguishablefrom the revival of the idea" (p. 334). Al-though little attention has been given to theissue of discriminating memories producedby perceptions from those produced bythought, a number of philosophers have ad-dressed the issue of how sensations and ideasare distinguishable initially ("in the hav-ing"). Two general lines of thought havebeen that perception and imagination pri-marily differ in that percepts are strongeror more vivid (e.g., Hume, 1739/1978;James, 1890/1950) and the opposing ideathat "sensation . . . and imagination, evenwhere they have the same object, are oper-ations of a quite different nature, and per-fectly distinguishable by those who aresound and sober" (Reid, 1764/1975, chap.2, section 5, p. 18). The model presentedbelow incorporates aspects of both of thesehypotheses and further suggests that someof the same cues that allow judgments aboutthe nature (present perception or presentimagination) of conscious experience arepreserved and become the bases for judg-ments about the nature of past experience.

The Present Approach

Some Preliminary Assumptions

Although we assume that perceptual eventsproduce persistent memory traces (perhapsin a kind of continuous record of experience;Landauer, 1975), we also assume that in-ternally generated events produce persistentmemory traces. Not all, but many of theerrors we see in memory may be the con-sequence of a failure to discriminate the or-igin of a memory trace. The ideas that mem-ory traces for perceptual experiences areaccurate, in that they reflect the character-istics of stimulus-determined processes andthat they exist side by side or intertwinedwith memories produced by imaginative pro-cesses have particular implications for the-ories of representation that are based on er-ror data such as false recognitions. Insofaras false recognitions reflect a failure to dis-criminate the origin of a trace, their appro-priateness as an indication of the nature ofthe memory representation of external events


is somewhat questionable. That is, errorsmay reflect coexistence of information frommultiple sources as well as integration, andthis distinction is beginning to be investi-gated (e.g., Cole & Loftus, 1979; Hasher& Griffin, 1978). Although this argumentmay at first seem to be quibbling over ter-minology, these two metaphors certainly im-ply different directions for research. Theymight also lead to different expectationsabout the potential success of introspectiveeffort (and more extreme measures such ashypnosis) in separating the real from theimagined.

Types of Self-Generated Information

There are many types of thoughts andimaginal events that would be relevant forreality monitoring. These types can tenta-tively be separated into three major cate-gories: (a) re-representation, (b) cotemporalthoughts, and (c) fantasy. The first categoryrefers to the re-representation of perceptualexperience or remembering something pre-viously experienced. In re-representation, in-formation that has dropped out of conscious-ness or working memory is reactivated at alater time in the absence of the original ex-ternal stimulus (in contrast to, for example,rehearsing recently perceived informationseveral times in succession). Cotemporalthought refers to the sort of elaborative andassociative processes that augment, bridge,or embellish ongoing perceptual experiencebut that are not neccessarily part of the ve-ridical representation of perceptual experi-ence. Most work on imaginal processes inmemory has been directed at this category(e.g., Bartlett, 1932; Bower, 1972b; Neisser,1967). Fantasy involves novel combinationsof information that produce imaginary eventsthat take place only in our imagination.Making up a story or dreaming would fallprimarily in this category. While recognizingthat there are some problems with this clas-sification scheme (where does "re-represen-tation" end and "fantasy" begin?), we havefound it useful as a starting point. To date,our research has primarily consisted of lab-oratory investigations of the first two cate-gories.

As stated above, we use the term reality

monitoring as a way of referring to the ac-tivity of discriminating between memoriesprimarily derived from external events andthose primarily derived from internal events.However, this term is not meant to implythat the processes involved in discriminatingthe origin (internal or external) of infor-mation in memory are always the same.Rather, the details of the processing (acti-vating information, applying decision strat-egies) might reasonably be expected to varydepending on the nature of the rememberedinformation, the conditions under which themonitoring occurs, the cost of mistakes, andso forth. Finally, reality monitoring can bethought of as one of the general class ofmetamemory processes (e.g., see Brown,1975; Kreutzer, Leonard, & Flavell, 1975)by which people reflect varying degrees ofunderstanding about the nature of their ownmemories.

A Working Model

As a vehicle for interpreting available dataand as a stimulus for further research, wehave developed the following working modelof reality monitoring. Basically, we proposethat reality monitoring is affected by twomajor factors: the nature of the traces beingevaluated and the types of decision processesapplied. Either of these should influence theamount of confusion a person displays aboutthe origin of information. Our first assump-tion is that any memory potentially consistsof many types of information or attributes(e.g., Bower, 1967; Craik & Lockhart, 1972;Posner & Warren, 1972; Underwood, 1969;Wickens, 1970). Thus, for example, if aword is presented or imagined, informationmay be stored with respect to its physicalproperties (e.g., pitch) and its semantic prop-erties. Types of potential attributes thatshould be particularly important for realitymonitoring include information about thesensory characteristics of the stimulus pre-sentation (e.g., auditory information), thetype of cognitive processing engaged (e.g.,imagery), semantic content, and contextualinformation (e.g., spatial and temporal in-formation). Various subcategories could bedefined within each of these four. For ex-ample, semantic content for a spoken word,


as used here, would include physical aspectsof its referent such as size and color, func-tional characteristics of the referent, emo-tional connotation, and so on. An issue thatcould plague us, as well as most recent the-ories of memory based on "attributes," "lev-els," or "codes" (e.g., Craik & Lockhart,1972; Paivio, 1971; Underwood, 1969), is,How is it possible to separate sensory fromsemantic dimensions of memory represen-tations? The fact that something is seen andnot heard (or said by a low frequency voicerather than a voice higher in frequency)might have significance in the semanticsense, as might the spatial or temporal po-sition of an event. These categories are notnecessarily proposed as hard and fast divi-sions (any more than one can make a cleardistinction between internal and external)but represent sensible and useful analyticcategories for characterizing and under-standing the nature of information repre-sented in memory.

We propose that, as a class, internallygenerated memories may differ from theclass of externally generated memories alongspecific dimensions. First, externally gener-ated memories in general may have morespatial and temporal contextual attributescoded in the representation of the event thaninternally generated memories do. Second,they should also have more sensory attri-butes, although imaginal processes presum-ably also generate some sensory information(e.g., see Atwood, 1971; Brooks, 1968; Finke,1979; Kosslyn, 1976). Third, we propose thatexternally generated representations aremore semantically detailed—that is, containmore information or more specific infor-mation—than internally generated represen-tations. Thought tends to be more schematic.Finally, we propose that internally generatedmemories may typically have more opera-tional attributes associated with them (codedin the trace). This latter notion is basedpartly on the assumption that perception isusually somewhat more "automatic" (Brown,1975; Hasher & Zacks, 1979; Posner & Sny-der, 1975) than imaginal processes and thatattention (e.g., voluntarily engaging in cre-ating images) increases the chances thatcoded information about operations will beavailable later (see Posner & Warren, 1972).

Although for some cases these character-istics might not hold, if the two classes ofrepresentations generally differ in these ways,then these dimensions could be used to de-cide the origin of a trace regardless of theparticular content of the memory involved.Assuming that the populations of externaland internal memories form distributions oneach of these four dimensions, a reasonablemodel of the process involved in decisionsabout the origin (internal or external) of amemory might be similar to some models ofold-new recognition decisions (e.g., Atkin-son & Juola, 1973; Mandler, 1972; Raye,1976). That is, criteria may be set on eachdimension, one for identifying probable in-ternal events and one for probable externalevents, and any event registering betweenthe two cutoffs would be of uncertain originand require further processing. Figure 1 il-lustrates such a decision structure for oneof the proposed dimensions.

Information about these four dimensionscould of course be combined in differentways. Decisions could be based on a strategythat always or usually weighted the four ina constant fashion, for example, sensory at-tributes might always be weighted moreheavily than amount of detail, and amountof detail more heavily than contextual at-tributes, and so on. On the other hand, dif-ferent dimensions might be given moreweight in varying instances depending onwhat other information the individual hasabout the situation. In general, then, itshould be possible to produce confusion

PROBABLEINTERNAL

0o

OAO

UNCERTAIN ORIGIN

AAO

AOA A

PROBABLEEXTERNAL

AAOAAAA

STRENGTH OR AMOUNT OF CONTEXTUAL INFORMATION

Figure I. Representation of a set of decision rules forjudging the origin of a memory on the basis of theamount of contextual (time and place) information itincludes. (In this sample, externally derived memoriesare represented by triangles and internally derived onesby circles.)


about the origin of an event by manipula-tions that cause it to have abnormally highor low values on one or more dimensions,compared to typical memories of its class;an interesting question is whether some di-mensions are more important in this regardthan others.

In addition to atypical class characteris-tics, a major vehicle for confusion should bespecific similarity (with respect to any at-tribute) between particular memories fromthe two classes. We have discussed "a mem-ory" as if it could be isolated, but this is justfor convenience. Clearly, when a memory isactivated, so are other memories, and thecharacteristics of these other memories mayinfluence or dominate the decision process.Thus in considering the origin of a memorythat was actually generated internally, onemight be influenced by extensive sensorycharacteristics of semantically similar tracesthat are also activated while the target isbeing evaluated. The extreme case would beone in which a retrieval cue activates a sim-ilar trace from the opposite class but not thetarget, causing an apparent error in identi-fication of origin when, in fact, the targettrace is not being considered at all.

As mentioned earlier, decisions about theorigin of a memory based on the proposedtypical class differences may be inconclusive,and the person may access further infor-mation. The person then may use this furtherinformation to reason toward a decisionabout the origin of the memory. This furtherinformation may consist of qualitative char-acteristics of the trace being evaluated incombination with the person's prior knowl-edge. For example, a person may rememberan extremely vivid dream about a money treebut reason that it could not be an externallygenerated memory because money does notgrow on trees. Or this further informationmay consist of characteristics of relatedtraces. For example, suppose an architectwas not sure whether a design she had inmind was initially hers or her associate's.She then remembers not only the idea butalso her associate's starting to draw a roughsketch and breaking his pencil. Therefore,she reasons, the idea must have been his.

Finally, this type of reasoning also prob-ably uses strategies based on certain meta-

memory assumptions or beliefs held by theindividual. For example, subjects might havecertain "biases" about memories derivedfrom their own thoughts (e.g., "I'd remem-ber if it had been my idea"). These latterprocesses need not always wait for the out-come of the more general decision based onclass characteristics discussed above butmay be initiated simultaneously. However,since they are assumed often to involve morespecific information, which may require ad-ditional retrieval from memory, they arepresumed normally to take longer and forthis reason might be considered a secondstage. Which processes play the predomi-nant role in a particular decision should de-pend on such factors as the amount of timea person has, availability of additional in-formation in memory, and the cost of mis-takes. Thus, if the consequence of assertingthat something was perceived when in factit might only have been inferred is serious,a person should be particularly likely to at-tempt to retrieve additional information thatwould tend to support or confirm one or theother decision.

In general, both perception and thoughtresult in memories that persist. Confusionof the two does not necessarily happen viadegradation of traces but rather can happenvia other mechanisms.

Empirical Evidence

The model developed above (see Table 1)represents a set of interrelated hypotheses,and we have tried in a number of ways totranslate some of these hypotheses into ex-periments.

Evidence That the Classes of Externallyand Internally Generated Memories Differ

A fundamental assumption of the modelis that the classes of externally and internallyderived memories differ. Two lines of evi-dence suggest this is a reasonable startingpoint.

Superiority effects. Slamecka and Graf(1978) reported the results of several exper-iments specifically directed at determiningwhether there is a memory advantage forself-generated information. Their general


Table 1Summary of the Model

Types of attributes potentially comprising memoriesContextualSensorySemanticCognitive operations

Dimensions on which the classes of externally generated and internally generated memories typically differExternal have more contextual attributesExternal have more sensory attributesExternal have more semantic detailInternal include more information about cognitive operations

Processing characteristics of reality monitoringDecisions about the origin of a specific trace may be based on a weighted combination of the results of

comparing the target trace attributes indicated above to criteria defining the general classes of external andinternal representations

Decisions may be based on a reasoning process involvingQualitative characteristics of the target traceCharacteristics of related tracesMetamemory assumptions

Which processes take place should depend on such factors as time, availability of different types ofinformation, cost of mistakes, and so forth

Sources of errors in reality monitoringTarget trace not typical of its classCharacteristics of similar incorrect tracesFailure in a reasoning process (e.g., failure to retrieve additional information, incorrect metamemory

assumptions)

procedure was to present subjects with astimulus and either to present a response(E items) or to require the subject to gen-erate a response (S items) according to arule and a specified first letter. The rule andfirst letter constrained the generated re-sponses (e.g., synonym-rapid-f almost al-ways produced the response fast), allowingSlamecka and Graf to counterbalance theassignment of items to conditions such thatE and S items were the same across subjects.Across a number of types of items and in-structional conditions, recognition and recallwere better for the subject-generated items(the "generation effect").

Using a similar procedure, we have alsofound a generation effect after retention in-tervals as long as 10 days (Johnson, Raye,Foley, & Foley, in press, Experiment 2). Inaddition, in one study we manipulatedwhether subjects overtly responded by writ-ing down E and S words as they occurredor whether they were simply required to re-spond covertly by listening to E words andthinking about S words. The generation ef-fect was as great in both free and cued recallwhen the subjects thought about but did notovertly express their responses (Johnson,

Raye, Foley, & Foley, in press, Experiment1). This outcome supports the idea that thelocus of the advantage is in the generationprocess itself and does not, for example, de-pend on overt expression of the generateditem and consequent sensory components(e.g., hearing one's own voice or seeing one'sown handwriting) or on some combinationof generation and sensory components.

In several studies, we compared frequencyjudgments of externally generated eventswith frequency judgments of self-generatedevents (Johnson, Taylor, & Raye, 1977;Raye, Johnson, & Taylor, 1980; Taylor,Johnson, Birnbaum, & Raye, Note 4). Thebasic paradigm involves two kinds of trials,presentation and generation. On presenta-tion trials, subjects see a number of cues andto-be-remembered words; on generationtrials, the cues are shown, and subjects areasked to generate the items. The cues arepreselected so that they will elicit the ap-propriate response with almost 100% cer-tainty (e.g., color-blue). Over the course ofthe procedure, each item is seen either two,five, or eight times and is generated by sub-jects either two, five, or eight times. Finally,subjects are given a surprise frequency judg-


ment task. For example, they are asked toestimate the number of times they actuallysaw each item, and we explain to them thatwe are not interested in the number of timesthey generated it but only the number oftimes we presented it. Or, they are asked toindicate the number of times they generatedan item, ignoring the number of times theyperceived it.

These studies had two major outcomes.The first was that subjects were more sen-sitive to the relative frequency of self-gen-erated events than to the relative frequencyof externally generated events. The secondmajor finding concerned confusion betweenthe two types of events. If there were no con-fusion between presentations and genera-tions, it would not matter, for example, howoften an item had been generated when sub-jects judged how often it was presented.However, generating events systematicallyincreased the frequency with which thoseevents seemed to have been perceived. Sim-ilarly, perceiving items systematically in-creased the frequency with which thoseitems seemed to have been generated. How-ever, the two types of events did not interferewith each other equally. The extent to whichsubjects were confused by internal eventswhen they judged external events was greaterthan vice versa. Thus both the steeper fre-quency discrimination function for self-gen-erated information and the greater confusionin judging external events indicated that in-ternally generated memories were superiorto or had some advantage over externallygenerated memories in this situation, as wellas in recall and recognition tests. A numberof related findings also suggest there is someadvantage associated with self-generated in-formation (e.g., Bobrow & Bower, 1969;Davies, Milnee, & Glennie, 1973; Erdelyi,Buschke, & Finkelstein, 1977; Jacoby, 1978;Jarvella & Collas, 1974; Ross & Sicoly,1979).

Another indication that external and in-ternal memories differ comes from Raye etal. (1980, Experiment 2). Half the peoplejudging each type of event were given a limitfor their estimates—they were asked to usenumbers between 0 and 10. The other halfof the subjects were not given any limit orrange for their estimates. We expected this

manipulation to affect the criteria subjectsused for including memories in an "eventcount." Of special interest was whether thetwo types of judgments would be affectedsimilarly. If they were, it would emphasizethe similarity between externally and inter-nally derived memories. However, if the twowere affected differently, the results wouldpoint to differences in the memories or judg-ment processes in the two cases. For ex-ample, if memories for one type of infor-mation are more stable and faithful, theymight be less susceptible to variations injudgment criteria. In fact, the limit did notaffect the magnitude of confusion when peo-ple were judging the frequency of internallygenerated events. This finding suggests thateven without the limit, people were using afairly stringent criterion of trace appropri-ateness. (Perhaps specific would be moreaccurate than stringent, since whatever theyused allowed for good discrimination of rel-ative frequency of self-generated events.)When external event frequency was judged,however, the limit reduced confusion, ap-parently because subjects selectively ex-cluded more inappropriate, self-generatedevent memories. Thus, these two types oftraces must differ in some way that allowsthe subject to discriminate between them.Raye et al. argued that these results wereconsistent with the present working modelif it were assumed that when judging thefrequency of internally generated events,subjects were relying heavily on the presenceof information about cognitive operations inmaking decisions about whether to includememories in an event count (see Raye et al.,1980, for an expanded discussion).

Comparing between- and within-classdiscriminations. Another line of researchis based on the logic that if two classes ofevents are different, discriminating betweeninstances from within a class should be moredifficult than discriminating between in-stances drawn from separate classes. Forexample, is reality monitoring subject tomore or less error than monitoring theorigins of information derived from two ex-ternal sources? Our model proposes that,along with more specific information, thereare a number of general dimensions on whichexternally and internally derived memories


may characteristically differ and which maybe used in the process of reality monitoring.That is, decisions could be based simply onthe amount of certain types of information(e.g., if it has a high sensory component,classify it as external). In contrast, twomemories derived from external sourcescould not be distinguished along these gen-eral dimensions. Thus memories from exter-nal sources would always have to be discrim-inated on the basis of specific content (specificsensory information rather than amount,etc.). It follows that, given similar situations,it should be easier to identify the origin ofinternally versus externally derived memo-ries than to identify the sources of memoriesderived from two external sources.

In one experiment testing this prediction,subjects participated in small groups duringthe first phase of the experiment (Raye &Johnson, 1980). Within a group they wererandomly assigned to play different roles:speakers, recorders, or listeners. All subjectsthought that the speakers and recorders wereserving as "experimenters," that the listenerswere the "subjects," and that the listenerswould later be tested for their memory ofthe "conversation." The group was given atopic of conversation to start them off, forexample, snow. Speaker A then gave ahighly related word such as white. SpeakerB gave a word related to Speaker A's word,for example, black, then Speaker A replied,for example, coal, and so on, alternating.One recorder wrote down all of the wordsspoken by Speaker A; the other recorderwrote down all the words given by SpeakerB. The listeners were instructed to listencarefully to both speakers in preparation fora memory test later. After a 1-hour, filled,retention interval, all subjects received a rec-ognition test which had the standard re-quirement that subjects discriminate be-tween old and new items and the additionalrequirement that, for old items, they identifywhich speaker said each.

The measure of primary interest was thepercent correct identification of origin: thetotal number of words correctly attributedto each of the two speakers divided by thetotal number of words correctly identified asfrom the conversation (multiplied by 100).As predicted by the model, the speakers, who

were discriminating between internal andexternal sources, were better at identifica-tion of origin than the listeners and re-corders, who were discriminating betweentwo external sources of information.

This result was taken as support for theidea that external events as a class differ insome ways from the class of internal events.These differences potentially provide addi-tional cues that allow the origin of memoriesfor internally generated events and externalevents to be distinguished more easily thanthe memories for two comparable externalevents. Such cues would operate in the firststage of the model. For example, greater sa-lience of the cognitive operations that pro-duced internally generated traces may serveas a discriminative cue available in external-internal discriminations that is not availablein external-external discriminations. An-other possibility is that the semantic contentsof the memories differ in some way. Intu-itively, internally generated responses shouldbe more idiosyncratically determined thanexternal events. Internally generated eventsare not necessarily more meaningful thanexternal events, but they are more likely torefer to biographical information or infor-mation relevant to the self. Internal eventsreflect idiosyncratic probabilities, whereasexternal events reflect normative probabili-ties. The presence of a difference in averageidiosyncracy value for external and internalitems could provide a discriminative cue:Would I have said this? Is this a high fre-quency response for me? Does this responsehave some special meaning for me? This fac-tor might operate in the reasoning stage ofreality monitoring.

In a second conversation study (Raye &Johnson, 1980, Experiment 2), the content(as well as the idiosyncracies) of the con-versation was controlled. A subject, the"director," asked the speakers questionssuch as, "What is a word that is the oppositeof fast, beginning with s and ending withwl" The director indicated which of the twospeakers was to answer, and the speakers'responses were essentially determined by thequestions, which, obviously, left little roomfor idiosyncratic contributions to the contentof the conversations. Again, there were re-corders and listeners, and only the listeners


thought they would get a memory test. Allsubjects, including the directors, were latergiven a recognition and identification-of-or-igin test.

Again, the speakers were significantly bet-ter at identifying the origin of an item theyrecognized as old than were the recordersand listeners, and the recorders and listenersdid not differ from one another. The direc-tors had the lowest discrimination scores.This finding rules out, we think, the possi-bility that the entire difference in the pre-vious experiment could be attributed to dif-ferences across conditions in the idiosyncraticvalue of the content of the items. In fact, themagnitude of the difference was almost ex-actly the same as it was in the previous ex-periment, suggesting that idiosyncratic cuesdid not add much. Of course, we would notwant to offer this as a general conclusion,since we would expect idiosyncratic valuetypically to be one of the cues used in thereasoning processes involved in reality mon-itoring. However, the present results are con-sistent with the possibility that, indepen-dently of the personal significance of whatan individual says, the cognitive operationsthat go into generating information persistin memory and become potential cues as tothe origin of that information.

A parallel logic suggests that it should beeasier to discriminate between memories ofexternal and internal events than betweenmemories of two internally generated eventsbecause, for example, the latter would bothinclude large amounts of cognitive opera-tions information, minimizing cognitive op-erations information as a discriminative cue.In two developmental studies, Foley, John-son, and Raye (Note 5) had subjects engagein two types of self-generated acts, eithersaying items aloud or only thinking aboutthem. Subjects then attempted to identifywhich items they said and which theythought. Consistent with expectations fromour model, this condition was more difficultthan one in which the subjects discriminatedbetween items they had said and items some-one else had said.

Compared to older subjects, 6-year-oldchildren were at no disadvantage in discrim-inating what they said from what someoneelse said (see also Johnson, Raye, Hasher,

& Chromiak, 1979). In contrast, comparedto older subjects, 6-year-olds did have a par-ticularly difficult time discriminating be-tween words they had said aloud and wordsthey had only thought. Within the presentframework, these data suggest two interest-ing possibilities. First, children as young as6 years may have learned to take advantageof cues in memory that differentiate the selffrom others but not those that differentiatevarious self-initiated acts such as speech andthought. Thus the special confusion of youngchildren could be a consequence of a lessdeveloped reasoning process and not a con-sequence of trace characteristics. Alterna-tively, 6-year-olds, compared to older sub-jects, may produce memories when thinkingthat are, in fact, more like the memoriesproduced from talking. Similarity betweenthought and speech would be increased, forexample, by subvocalizing during thought,and there is some evidence that children ofthis age do show relatively high levels of sub-vocalization (Garrity, 1977).

Specific Dimensions Proposed asDiscriminators

Cognitive operators. According to thereality monitoring model, increasing auto-maticity in responding ought to reduce in-formation regarding cognitive operations.The stronger or more specific the cue for aresponse, the more automatically the re-sponse should be generated, reducing searchand decision processes that later could beused as cues to the origin of information.Conversely, if a response requires somewhatmore effort or search, stored informationabout these operations should become a po-tential cue for identifying the source of amemory. In order to manipulate this searchprocess, we used a procedure in which sub-jects heard category cues followed by cate-gory instances on half of the trials and gen-erated their own category instances for theother half of the trials. Half of the subjectswere cued with the first letter of the re-sponse, and half were not cued with the firstletter. Our reasoning was that providing thefirst-letter cue should more completely de-termine the response, making it more au-tomatic, than not providing the first letter.


Later, subjects were asked to discriminatebetween category instances that were pre-sented and instances they generated. Gen-erating instances of categories to fit first-let-ter cues resulted later in lower correctidentification of origin than generating in-stances of categories without first-letter cues(Johnson, Raye, Foley, & Foley, in press,Experiment 1). In Johnson et al.'s Experi-ment 2, increasing the difficulty of gener-ating a response and thus presumably in-creasing cognitive operations information byrequiring a less typical response (e.g., ani-mal-p-, rather than animal-d-) increasedcorrect identification of origin. These find-ings support the notion that cognitive oper-ations information is important in realitymonitoring decisions.

The cue value of remembered cognitiveoperations may explain why mediators gen-erated by subjects in list learning situationsare rarely intruded. That is, there is ampleevidence that in paired-associate learning,for example, subjects generate verbal or pic-torial elaborators (e.g., Martin, Boersma,& Cox, 1965; Yuille, 1973) and that specificqualitative characteristics of these elabora-tors influence recall (Hasher, Griffin, &Johnson, 1977; Hasher & Johnson, 1975).In addition, self-generated information hasa recall advantage (Slamecka & Graf, 1978),even when the self-generated informationoccurs only covertly (Johnson, Raye, Foley,& Foley, in press, Experiment 1). Why isit then that intrusions of the mediator or self-generated components of an image or othertypes of elaborators are so rare? From thepresent perspective, it is important that whensubjects create verbal or imaginal elabora-tors, the strategy is usually purposefully em-ployed and not the result of automatic orincidental processes (Brown, 1975; Hasher& Zacks, 1979; Posner & Snyder, 1975;Schneider & Shiffrin, 1977). The best ar-gument for the strategic as opposed to au-tomatic nature of elaborators is that it isquite easy to improve people's performanceby instructing them to make elaborators(e.g., Bower, 1972b; Paivio, 1971).

On the other hand, those elaborations thatmore automatically arise on the basis of pastexperience, such as those involved in pro-cessing meaningful prose, should be more

likely to be confused with external events.Intrusions seem to occur more often withprose. It is not that more thought necessarilygoes into processing prose compared to wordlists; it is perhaps that more conscious (inthe sense of purposeful) thought may be ex-pended on the word lists.

Similarly, daydreams often seem less realthan dreams and rarely lead to the momen-tary confusion that dreams produce. Day-dreams are typically more controlled by con-sciously selected schemata than are dreamsduring sleep. In general, those self-generatedmemories that do not have a voluntary qual-ity but rather have been elicited by externalevents or generated when voluntary pro-cesses are suspended should be susceptibleto greater confusion. Creating a thought un-der voluntary control (or bringing a previ-ously involuntary idea under voluntary con-trol) should provide a powerful cue forlocalizing it as self-generated rather than asexisting externally. Consistent with thisspeculation is at least one finding: Whereasrepeated re-representations of perceiveditems increase confusion (e.g., Johnson, Raye,Wang, & Taylor, 1979; Johnson, Taylor,& Raye, 1977), generating an initially self-generated item for a second time increasedcorrect identification of origin (Johnson,Raye, & Durso, 1980).

Contextual information. In one experi-ment (Johnson, Raye, Foley, & Kim, Note6), we compared the spatial location infor-mation associated with perceived pictureswith the spatial location information asso-ciated with imagined pictures. Subjects sawa label followed by an appropriate line draw-ing projected on a screen to the left or ascreen to the right, or they saw a label andthen themselves created an appropriate im-age projected to the left or to the right. Asa cover task, subjects rated the pictures andimages with respect to how easy they wouldbe to draw. This task produced equal rec-ognition of items that had been presented aspictures and items that had been imagined(one indication of the effectiveness of thiscover task). Of primary interest was thefinding that, as predicted by the model, sub-jects were significantly more accurate in des-ignating the location for perceived as com-pared to imagined items. In addition, for


those items whose location was correctly des-ignated, subjects were more confident aboutthe location of the perceived pictures.

We adapted a similar procedure to ad-dress the question of whether temporal in-formation is better for internally or exter-nally generated information. Subjects heardlabels and then either saw an appropriateline drawing or imagined one. The cover taskwas the rating task just described. Followingthis, each subject received a booklet, and oneach page, in random order, were labels foreight items, each drawn from a differenteighth of the list. The eight items either allreferred to pictures or all referred to images.The subject's task was to indicate the tem-poral order of occurrence of the eight items.We used this procedure to eliminate differ-ential response bias across positions of thelist (Toglia & Kimble, 1976). This taskturned out to be quite difficult, but the sub-jects were able to identify the temporal po-sition of very early and very late items. Foritems presented early, performance did notdiffer on imagined as compared to perceiveditems, whereas the subjects were signifi-cantly less accurate in identifying temporalorder of imagined items for the most recentportion of the list. We are currently follow-ing this up with an easier task; however, thisinitial finding is particularly interesting be-cause some investigators (Bjork & Whitten,1974; Crowder, 1976, pp. 461-464) haveproposed that the recency portion of the se-rial position curve especially reflects the con-tribution of temporal information.

Sensory information. The model alsopredicts that sensory information should begreater for perceived compared to imagineditems. In one experiment (Johnson, Raye,Foley, & Foley, Note 7), subjects saw a wordthat was then pronounced by a male or fe-male experimenter or saw a word and thenimagined it as pronounced by the male orfemale experimenter. Craik and Kirsner(1974) had previously shown significantmemory for voice of perceived items, andGeiselman and Glenny (1977) had shownmemory for the "voice" of an imagined item.Our interest was in comparing the infor-mation about voice for perceived and imag-ined items. As a cover task, subjects rated

the harshness of the perceived or imaginedsound.

We asked a similar question with a pro-cedure in which subjects saw or heard itemson some trials or imagined seeing or imag-ined hearing them on other trials. The covertask required subjects to rate the angularityof the visual display of a perceived or imag-ined word or the harshness of the sound ofan auditorially perceived or imagined item.

Across four experiments, two requiring adiscrimination between male and femalevoices and two between visual and auditoryinformation, we have found a consistent, butnot always significant, advantage for per-ceived information. We think these covertasks very closely equated the informationproduced during perception and thought,primarily by restricting it to a single type ofsensory information, and were thus a con-servative test of the hypothesis. The subject-report data described below suggest thatwhen semantic processing is manipulatedand sensory information is left free to vary,sensory information is more frequently re-ported for perceived than for subject-gen-erated items.

Of course, the model does not demandthat sensory information always be better formemories created by perception than forthose created by thought; it only indicatesthat this is the more typical case. The verycapacity for some thoughts to have moresensory detail than some perceptions createsone of the conditions for confusion. Fur-thermore, with the right orienting task, spa-tial location and temporal information mightbe equated for imagined and perceived items.One problem for future work would be toattempt to characterize those situations inwhich contextual, semantic, or sensory in-formation from imagination is likely to beunusually high, or from perception, unusu-ally low.

For example, the amount of sensory in-formation in an imaginary experience shouldvary among individuals. Good imagers, com-pared to poor imagers, should be more ableto regenerate accurately the details of theirexperiences. Although it may be difficult tospecify exactly the differences in the natureof the representations of good and poor im-


agers (Pylyshyn, 1973), a likely differenceis in the amount of sensory detail that isrepresented in the images they create. Wemight then expect that good imagers wouldmore readily confuse re-representations ofexperiences with similar perceptions. In oneexperiment (Johnson, Raye, Wang, & Tay-lor, 1979, Experiment 1; Johnson & Raye,Note 8), we separated subjects into groupsof good and poor imagers on the basis of animagery test. Using our re-representationparadigm, subjects saw pictures of commonobjects taken from children's books on pre-sentation trials and were asked to generateimages of the pictures on generation trials.On a later test of presentation frequency, thejudgments of good imagers were more af-fected by the number of times they hadimagined a picture than were the judgmentsof poor imagers. (Good and poor imagers didnot differ in their ability to recall the set ofpictures).

Although it simplifies the discussion tolook at one type of information at a time,we should keep in mind that all memoriesinclude several types of information and thatany reality monitoring decision is the resultof processes that weight and integrate in-formation across these dimensions. The pre-viously discussed Raye et al. study (1980,Experiment 2), which compared frequencyjudgments with and without a limit, dem-onstrated that subjects can change theweighting given to various dimensions. Whichcomponents of memories will weigh moreheavily in reality monitoring judgments maydepend on the nature of the events beingjudged. For example, cognitive operationsmay be particularly important in decisionsabout language, especially words and/orsentences given in isolation. Sensory infor-mation may prove to be the more dominantcomponent in decisions about the origin ofvisual information such as pictures.

Furthermore, that decisions are not simplymade on the basis of one type of information,without regard for other attributes, is illus-trated by some results reported by Durso andJohnson (1980). Subjects saw a concept rep-resented by a word or a line drawing, andon a later recognition test (where both tar-gets and distractors were given auditorially),

the subjects were asked to indicate the orig-inal mode of presentation of the items theyidentified as old. The orienting task at ac-quisition was manipulated for differentgroups of subjects. Some subjects were askedto perform tasks that required that they at-tend to the formal properties of the wordsymbol (what is the last letter of the nameof this concept?). Other subjects were askedto perform tasks that explicitly required thatthey create an image of the concept (Howlong would it take for an artist to make aline drawing of this concept?). And somesubjects were asked to perform tasks thatrequired them to consult semantic infor-mation regarding the referent of the concept(e.g., What is this thing used for?). The mostinteresting data for the present purposes arethe confusions between the two modes ofrepresentations, that is, the mistaken judg-ments that words were presented as picturesor that pictures were presented as words.The relative frequency of these two types ofconfusions did not differ for either the ori-enting tasks involving the formal propertiesof words or those involving explicit imagery.However, for the semantic conditions, sub-jects were significantly more likely to say aword had been presented as a picture thanto say a picture had been presented as aword. This finding is all the more strikingbecause there was clearly a significant biason false positives (calling a completely newitem old) to say they were presented as wordsrather than as pictures. The high rate ofmisidentifications of words as pictures bythose subjects who did semantic processingsuggests they had been incidentally creatingimages or consulting sensory informationduring the orienting tasks and that they latermistook this self-generated pictorial infor-mation for externally presented pictorial in-formation.

Memories for words created in the explicitimagery conditions should also have hadconsiderable sensory information; however,the memories also very likely included cuesindicating that the image was purposefullycreated. On the other hand, the incidentallycreated images, which had less salient cuesthat they were generated by cognitive op-erations yet which had high sensory infer-


mation, should be more likely to be mistakenfor externally presented pictures. Thus, hereis a case where it makes sense to assume thatdecisions about the origin of memories tookinto account both sensory and cognitive op-erations information.

Metamemory Assumptions

An important point incorporated in theworking model of reality monitoring thatappears to be born out by our experimentsis the idea that subjects' assumptions abouthow their memories work will play a criticalrole in decision strategies and biases oper-ating during reality monitoring.

What subjects report. The good-poorimager study above attempted to manipulatethe similarity between perceptions and re-representations of those perceptions. Wehave also attempted to vary the relationshipbetween perceptions and cotemporal, self-generated information (Johnson, Raye,Foley, & Foley, in press, Experiment 3). Anexperimenter read a list of common, unre-lated words. Subjects in the related condi-tion were told to generate a word for eachstimulus word that had some meaningfulrelationship to the stimulus word. Subjectsin the unrelated condition were told to givea word that was not in any way related tothe stimulus item. Subjects in the first-lettercondition were told to give a word thatstarted with the same first letter as the wordthey heard. Thus in the related and unre-lated conditions subjects gave some attentionto the meaning of each externally presenteditem in order to perform the task, but theinternally generated product should havebeen different in the two cases. The relatedand the first-letter conditions should bothproduce related items, but these relation-ships should primarily be semantic in therelated group and orthographic in the first-letter condition. All subjects wrote downboth presented and generated words.

A week later, subjects received an iden-tification-of-origin test. For those words cor-rectly recognized as old, the mean percentscorrect identification of the source of an itemwere 62, 68, and 74, for related, first-letter,and unrelated, respectively, indicating thatas cotemporal thoughts become more related

to perceptual events, the likelihood of laterconfusing their memory representations in-creases. However, even though the relatedsubjects showed the least ability to identifythe origin of an item, they were still signif-icantly above a chance score of 50%.

Following the identification-of-origin test,subjects in this study were asked to describehow they were able to identify the source ofvarious items. Subjects tended to mentionsensory cues more often with respect to ex-perimenter-generated items ("I differen-tiated words which you said by rememberingyour pronunciation"; "I could visualize yoursaying it"; or "The words which the exper-imenter stated were remembered in hervoice"). Cognitive processing, additional in-formation, and a consideration of specificsemantic content were mentioned more inconjunction with subject-generated items("When I was very sure [about my words]I could remember I had a very specific rea-son for making the association. If the word[only] seemed familiar, I would say that itwas the experimenter's word"; "I made thedecision by knowing what my train of thoughtwas during the exercise"; "Sometimes thewords I chose went together with a certainscene, i.e., pond, cloud, tree.'And when I sawthe words again I tried to remember if theyfit in any of the images I had").

Biases in judgments. If you ask them,many subjects expect self-generated infor-mation to have an advantage in memory(Johnson, Raye, Foley, & Foley, in press,Experiment 2). This expectation is consis-tent with a notable bias we found: Whensubjects felt that a completely new item wasfamiliar (a false positive), they displayed amuch greater willingness to attribute it toan external source than to say they gener-ated it. We found this "it-had-to-be-you"("I'd remember if it were me") effect withthe three different instructions for generat-ing responses described above (related, un-related, first letter), for high and low fre-quency category instances and opposites(Johnson, Raye, Foley, & Foley, in pressExperiment 2) and for identification of ori-gin of words from sentences (Johnson, Raye,& Durso, 1980). The pervasiveness of thisbias suggests that it is probably an importantcomponent of performance in memory tasks,


contributing, perhaps, to false positives incertain recognition paradigms (e.g., Cramer,1965; Johnson, Bransford, & Solomon, 1973;Kimble, 1968; Sulin & Dooling, 1974; Un-derwood, 1965) and unedited intrusions inrecall (Brewer, 1977; Deese, 1959). Undersome circumstances, subjects can be inducedwith a stricter criterion to perform in a waythat more closely matches real externalevents (Raye et al., 1980). An investigationof the circumstances that might affect thisbias could prove interesting. For example,we have recently found that alcohol intoxi-cation reduces (but does not eliminate) thisbias, reflecting perhaps people's belief thatone consequence of drinking is that it di-minishes one's ability to remember what onedid (Taylor et al., Note 4). Of course, thebias might be completely reversed if the so-cial desirability of having been the sourceof an idea were markedly increased ("I don'tremember generating it, but that idea is sogood it has to be mine").

Remembering Occurrence VersusRemembering Origin

Do we need a model of reality monitoringthat is distinct from the theories of memorythat have been proposed to account for rec-ognition and recall in general? For example,could we simply say that "deep" processingof information leads to better memory, in-cluding better identification of origin? Astrength theorist might propose that identi-fication of origin simply requires more of thesame sort of information that recall and rec-ognition depend on. However, attempts tocharacterize memory functioning in termsof a single dimension have always run intodifficulties. For example, in the present con-text, both strength and levels theories wouldexpect a consistent correlation betweenmemory for occurrence measures (recallor recognition) and identification-of-originmeasures. Memory for occurrence and mem-ory for origin sometimes do respond to thesame variables. For example, both recogni-tion and identification of origin were betterfor low as compared to high frequency cat-egory instances (Johnson, Raye, Foley, &Foley, in press, Experiment 2). However,these two measures are by no means always

affected similarly. For example, directors,who asked the questions, were as good asspeakers in later identifying answers thathad been given but far worse at identifyingwho said what (Raye & Johnson, 1980, Ex-periment 2). And 6-year-olds who said somewords and thought others were as good atidentifying which words had occurred aswere 6-year-olds who had listened to twospeakers say the words; however, the chil-dren found it much more difficult to dis-criminate what they had said from what theyhad thought than to discriminate betweenwhat the two speakers had said (Foley et al.,Note 5).

In some cases, memory for occurrence andidentification of origin appear actually to beaffected in opposite ways by the same vari-able. For example, recall and recognitionwere both better when subjects generatedsemantically related responses than whenthey generated responses sharing the samefirst letter as the stimuli. However, discrim-inating between items heard and items gen-erated was more difficult in the related case(Johnson, Raye, Foley, & Foley, in press).

In the experiments in which Durso andJohnson (1980) compared memory for pic-tures and words under different orientingtasks, attending to the referent for a wordrather than properties of the verbal name ofthe concept resulted in higher recognition,but it also resulted in more words that werelater identified as having been pictures. Withthe explicit imagery tasks, there was also anincrease in recognition relative to verbaltasks, but this was not accompanied by anincrease in identifying words as pictures. Insummary, it does not appear to us that theinformation and processes that are impor-tant in discriminating the origin of infor-mation can be equated with those that pro-duce voluntary recall or a feeling offamiliarity.

The factor that probably most distin-guishes the process of voluntary recall fromrecognition is the dependence of recall onorganizational information (Mandler, 1967;Tulving, 1968). In contrast, recognition isnotable for the variety of types of informa-tion that can produce a "familiarity" re-sponse (e.g., rapid activation of semantic orsensory information, cognitive operations in-


formation, context cues). Reality monitor-ing, as a process, has its own properties, al-though to a certain extent it of course drawson information that is operative in recall andrecognition. (For example, the potential use-fulness in recognition of cognitive opera-tions, especially, has received recent atten-tion, e.g., Kolers, 1975; Russo & Wisher,1976.) Although a model of reality moni-toring does not offer a theory of recall orrecognition, neither do theories of recall orrecognition constitute theories of realitymonitoring. Furthermore, understandingreality monitoring can contribute to our un-derstanding of recall and recognition pro-cesses, for example, by providing mecha-nisms for editing out inappropriate responsesand by accounting for failures in editing.

Summary and Conclusions

We have intended to focus attention onthe processes involved in distinguishing be-tween the perceived and the self-generatedin memory (reality monitoring). We startedwith the assumption that perceiving differsfrom self-generated events such as thinkingand imagining. That is, they are either dif-ferent operations of the mind or operationsof the mind resulting in characteristicallydifferent amounts of various types of infor-mation. Furthermore, we believe that theprocesses involved in reality monitoring,when they are functioning optimally, capi-talize on these characteristic differences. Wehave outlined a working model of realitymonitoring and have investigated some ofthe questions it raises. The present modelproposes that reality monitoring can bebased on some general dimensions—ratherindependent of the particular content of amemory—that serve as distinguishers be-tween the two classes of memories. It is pro-posed that dimensions or attributes partic-ularly central to reality monitoring includeamount of contextual information (time andplace), semantic detail, sensory information,and cognitive operations. The model addi-tionally assumes that reality monitoring mayinvolve more extended reasoning processesthat take into account the content of partic-ular memories and their relation to other

knowledge available to the person, as wellas metamemory assumptions of the person.

Overall, our results support these ideas:Memories from external and internal sourcesappear to differ in class-characteristic ways;confusion is increased by semantic and sen-sory similarity between memories from thetwo sources; confusion is reduced with in-creases in the information about cognitiveoperations associated with internally gen-erated memories; memories based in percep-tion have better spatial, temporal, andsensory information; and people's tacitassumptions about these characteristic dif-ferences are reflected in metamemory as-sumptions that influence reality monitoringjudgments.

For the most part, the materials and pro-cedures we have used seem to be extremelyconducive to confusing external and internalsources of memories. The words, sentences,and pictures were not, after all, embeddedin social or emotional contexts that mightprovide cues from related information storedin memory about the origin of a particulartrace. Given the seemingly meager cues, itis perhaps remarkable that significant realitymonitoring was obtained in some conditionswith retention intervals as long as 7 and 10days. This sensitivity to the origin of infor-mation, when a discrimination is called for,further recommends the usefulness of mak-ing a distinction between external and in-ternal sources of information and the use-fulness of attempting to incorporate such adistinction into models of memory.

The overall picture suggested by the evi-dence reviewed here is one of a memory sys-tem that preserves information about theorigin of information remarkably well; thedecision criteria through which this infor-mation is filtered, however, allow for someerror in attributing memories to sources. Wefeel that the working model proposed hereis a viable framework for generating ques-tions and integrating results of studies di-rected at the processes involved in realitymonitoring.

Reference Notes

1. Johnson, M. K., & Raye, C. L. A working model ofreality monitoring. Paper presented at the annual


meeting of the Psychonomic Society, San Antonio,Texas, November 1978.

2. Levine, M., & Jankovic, I. N. An introduction tospatial problem solving. Manuscript submitted forpublication.

3. Johnson, M. K. Constructive aspects of memory:Historical antecedents. Paper presented at the sym-posium on constructive processes in memory, annualmeeting of the American Psychological Association,Chicago, September 1975.

4. Taylor, T. H., Johnson, M. K., Birnbaum, I. M.,& Raye, C. L. Alcohol and reality monitoring. Man-uscript submitted for publication.

5. Foley, M. A., Johnson, M. K., & Raye, C. L. Adevelopmental study of reality monitoring, compar-ing two types of judgments about memories. Paperpresented at the annual meeting of the Eastern Psy-chological Association, Hartford, Connecticut, April1980.

6. Johnson, M. K., Raye, C. L., Foley, M. A., & Kim,J. Pictures and images: Spatial and temporal infor-mation compared. Manuscript in preparation, 1980.

7. Johnson, M. K., Raye, C. L., Foley, M. A., & Foley,H. J. Perceptions and imaginations: sensory infor-mation compared. Manuscript in preparation, 1980.

8. Johnson, M. K., & Raye, C. L. Fact and fantasy:Confusion between perceptual experience and imag-inations. Paper presented at the annual meeting ofthe Psychonomic Society, St. Louis, Missouri, No-vember 1976.

References

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

Atwood, G. An experimental study of visual imaginationand memory. Cognitive Psychology, 1971, 2, 290-299.

Bartlett, F. C. Remembering: A study in experimentaland social psychology. Cambridge, Eng.: CambridgeUniversity Press, 1932.

Bjork, R. A., & Whitten, W. B. Recency-sensitive re-trieval processes in long-term free recall. CognitivePsychology, 1974, 6, 173-189.

Bobrow, S. A., & Bower, G. H. Comprehension andrecall of sentences. Journal of Experimental Psy-chology, 1969, 80, 455-461.

Bower, G. H. A multi-component theory of the memorytrace. In K. W. Spence & J. T. Spence (Eds.), Thepsychology of learning and motivation (Vol. 1). NewYork: Academic Press, 1967.

Bower, G. H. A selective review of organizational factorsin memory. In E. Tulving & W. Donaldson, (Eds.),Organization of memory. New York: AcademicPress, 1972. (a)

Bower, G. H. Mental imagery and associative learning.In L. W. Gregg (Ed.), Cognition in learning andmemory. New York: Wiley, 1972. (b)

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

Bransford, J. D., & Johnson, M. K. Contextual prereq-

uisites for understanding: Some investigations of com-prehension and recall. Journal of Verbal Learningand Verbal Behavior, 1972, 11, 717-726.

Bransford, J. D., & Johnson, M. K. Considerations ofsome problems of comprehension. In W. Chase (Ed.),Visual information processing. New York: AcademicPress, 1973.

Brewer, W. F. Memory for pragmatic implications ofsentences. Memory & Cognition, 1977, 5, 673-678.

Brooks, L. R. Spatial and verbal components of the actof recall. Canadian Journal of Psychology, 1968, 22,349-368.

Brown, A. L. The development of memory: Knowingabout knowing, and knowing how to know. In H. W.Reese (Ed.), Advances in child development and be-havior (Vol. 10). New York: Academic Press, 1975.

Bruner, J. S. Going beyond the information given. InContemporary approaches to cognition. Cambridge,Mass.: Harvard University Press, 1957.

Cameron, N. Personality development and psychopa-thology: A dynamic approach. Boston: HoughtonMifflin, 1963.

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

Cole, W. G., & Loftus, E. F. Incorporating new infor-mation into memory. American Journal of Psychol-ogy, 1979, 92, 413-425.

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

Craik, F. I. M., & Kirsner, K. The effect of speaker'svoice on word recognition. Quarterly Journal of Ex-perimental Psychology, 1974, 26, 274-284.

Craik, F. I. M., & Lockhart, R. S. Levels of processing:A framework for memory research. Journal of VerbalLearning and Verbal Behavior, 1972, / / , 671-684.

Cramer, P. Recovery of a discrete memory. Journal ofPersonality and Social Psychology, 1965, /, 326-332.

Crowder, R. G. Principles of Learning and memory.Hillsdale, N.J.: Erlbaum, 1976.

Davies, G. M., Milnee, J. E., & Glennie, B. S. On thesignificance of "double encoding" for the superior re-call of pictures to names. Quarterly Journal of Ex-perimental Psychology, 1973, 25, 413-423.

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

Dooling, D. J., & Lachman, R. Effects of comprehen-sion on retention of prose. Journal of ExperimentalPsychology, 1971, 88, 216-222.

Durso, F. T., & Johnson, M. K. The effects of orientingtasks on recognition, recall, and modality confusionof pictures and words. Journal of Verbal Learningand Verbal Behavior, 1980, 19, 416-429.

Erdelyi, M., Buschke, H., & Finkelstein, S. Hyper-amnesia for socratic stimuli: The growth of recall forinternally generated memory list abstracted from aseries of riddles. Memory & Cognition, 1977, 5, 283-286.

Finke, R. A. The functional equivalence of mental im-ages and errors of movement. Cognitive Psychology,1979, //, 235-264.

Freud, S. Negation (Standard Ed., Vol. 19. James


Strachey, Ed.). Toronto, Canada: Clarke, Irwin,1961. (Originally published, 1925.)

Freud, S. Project for a scientific psychology (StandardEd., Vol. 1. James Strachey, Ed.). Toronto, Canada:Clarke, Irwin, 1966. (Originally published, 1895.)

Garrity, L. I. Electromyography: A review of the currentstatus of subvocal speech research. Memory & Cog-nition, 1977, 5, 615-622.

Geiselman, R. E., & Glenny, J. Effects of imaginingspeakers' voices on the retention of words presentedvisually. Memory & Cognition, 1977, 5, 499-504.

Gibson, J. J. The senses considered as perceptual sys-tems. Boston: Houghton Mifflin, 1966.

Gibson, J. J. The ecological approach to visual percep-tion. Boston: Houghton Mifflin, 1979.

Hart, J. T. Memory and the memory-monitoring pro-cess. Journal of Verbal Learning and Verbal Behav-ior, 1967, 6, 685-691.

Hasher, L., & Griffin, M. Reconstructive and repro-ductive processes in memory. Journal of Experimen-tal Psychology, 1978, 4, 318-330.

Hasher, L., Griffin, M., & Johnson, M. K. More oninterpretive factors in forgetting. Memory & Cogni-tion, 1977, 5, 41-45.

Hasher, L., & Johnson, M. K. Interpretive factors inforgetting. Journal of Experimental Psychology:Human Learning and Memory, 1975, /, 567-575.

Hasher, L., & Zacks, R. T. Automatic and effortfulprocesses in memory. Journal of Experimental Psy-chology: General, 1979, 108, 356-388.

Heidbreder, E. Seven psychologies. New York: Apple-ton-Century-Crofts, 1933.

Hochberg, J. Perception. Englewood Cliffs, N.J.: Pren-tice-Hall, 1979.

Horowitz, M. J. Image formation and cognition. NewYork: Appleton-Century-Crofts, 1978.

Hume, D. A treatise of human nature. Oxford, Eng.:Clarendon Press, 1978. (Originally published, 1739.)

Jacoby, L. L. On interpreting the effects of repetition:Solving a problem versus remembering a solution.Journal of Verbal Learning and Verbal Behavior,1978, 17, 649-667.

James, W. The principles of psychology. New York:Dover, 1950. (Originally published, 1890.)

Jarvella, R. J., & Collas, J. G. Memory for the inten-tions of sentences. Memory & Cognition, 1974, 2,185-188.

Jenkins, J. J. Remember that old theory of memory?Well, forget it! American Psychologist, November1974, pp. 785-795.

Johnson, M. K. What is being counted none the less?In I. M. Birnbaum & E. S. Parker (Eds.), Alcoholand human memory. Hillsdale, N.J.: Erlbaum, 1977.

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

Johnson, M. K., Raye, C. L., & Durso, F. T. Realitymonitoring: Second perceptions and thoughts. Bul-letin of the Psychonomic Society, 1980, 15, 402-404.

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

Johnson, M. K., Raye, C. L., Hasher, L., & Chromiak,

W. Are there developmental differences in realitymonitoring? Journal of Experimental Child Psy-chology, 1979, 27, 120-128.

Johnson, M. K., Raye, C. L., Wang, A. Y., & Taylor,T. H. Fact and fantasy: The roles of accuracy andvariability in confusing imaginations with perceptualexperiences. Journal of Experimental Psychology:Human Learning and Memory, 1979, 5, 229-240.

Johnson, M. K., Taylor, T. H., & Raye, C. L. Fact andfantasy: The effects of internally generated events onthe apparent frequency of externally generated events.Memory & Cognition, 1977, 5, 116-122.

Kimble, G. A. Mediating associations. Journal of Ex-perimental Psychology, 1968, 76, 263-266.

Kintsch, W. Memory for prose. In C. N. Cofer (Ed.),The structure of human memory. San Francisco:Freeman, 1976.

Klatzky, R. L. Human memory: Structure and pro-cesses. San Francisco: Freeman, 1980.

Kohler, W. Gestalt psychology. New York: MentorBooks, 1959. (Originally published, 1947.)

Kolers, P. A. Specificity of operations in sentence rec-ognition. Cognitive Psychology, 1975, 7, 289-306.

Kosslyn, S. M. Can imagery be distinguished from otherforms of internal representation? Evidence from stud-ies of information retrieval times. Memory & Cog-nition, 1976, 4, 291-297.

Krech, D. Notes toward a psychological theory. Journalof Personality, 1949, 18, 66-87.

Kreutzer, M. A., Leonard, C., & Flavell, J. H. An in-terview study of children's knowledge about memory.Monographs of the Society for Research in ChildDevelopment, 1975, 40 (1, Serial No. 159).

Landauer, T. K. Memory without organization: Prop-erties of a model with random storage and undirectedretrieval. Cognitive Psychology, 1975, 7, 495-531.

Lashley, K. S. The problem of serial order in behavior.In N. J. Slamecka (Ed.), Human learning and mem-ory. New York: Oxford University Press, 1967.

Mandler, G. Organization and memory. In K. W.Spence & J. T. Spence (Eds.), The psychology oflearning and motivation (Vol. 1). New York: Aca-demic Press, 1967.

Mandler, G. Organization and recognition. In E. Tulv-ing & W. Donaldson (Eds.), Organization of memory.New York: Academic Press, 1972.

Martin, C. J., Boersma, F. J., & Cox, D. L. A classi-fication of associative strategies in paired-associatelearning. Psychonomic Science, 1965, 3, 455-456.

McGuigan, F. J. Covert oral behavior and auditory hal-lucinations. Psychophysiology, 1966, 3(1), 73-80.

Mellor, C. S. First rank symptoms of schizophrenia. I:The frequency in schizophrenics on admission to hos-pital. II: Differences between individual first ranksymptoms. British Journal of Psychiatry, 1970, 117,15-23.

Mill, J. Analysis of the phenomena of the human mind,by James Mill (edited with additional notes by J. S.Mill). New York: Kelley, 1967. (Originally published,1829.)

Mill, J. S. Analysis of the phenomena of the humanmind, by James Mill (edited with additional notes byJ. S. Mill). New York: Kelley, 1967. (Originally pub-lished, 1869.)


Mintz, S., & Alpert, M. Imagery vividness, reality test-ing, and schizophrenic hallucinations. Journal of Ab-normal Psychology, 1972, 79, 310-316.

Neisser, U. Cognitive Psychology. New York: Apple-ton-Century-Crofts, 1967.

Neisser, U. Cognition and reality. San Francisco: Free-man, 1976.

Paivio, A. Mental imagery in associative learning andmemory. Psychological Review, 1969, 76, 241-263.

Paivio, A. Imagery and verbal processes. New York:Holt, Rinehart & Winston, 1971.

Perky, C. W. An experimental study of imagination.American Journal of Psychology, 1910, 21, 422-452.

Posner, M. I., & Snyder, C. R. R. Attention and cog-nitive control. In R. L. Solso (Ed.), Information pro-cessing and cognition: The Loyola Symposium. Hills-dale, N.J.: Erlbaum, 1975.

Posner, M. I., & Warren, R. E. Traces, concepts andconscious construction. In A. W. Melton & E. Martin(Eds.), Coding processes in human memory. Wash-ington, D.C.: Winston, 1972.

Postman, L. Toward a general theory of cognition. InJ. H. Rohrer & M. Sherif (Eds.), Social psychologyat the crossroads. New York: Harper, 1951.

Pylyshyn, Z. W. What the mind's eye tells the mind'sbrain: A critique of mental imagery. PsychologicalBulletin, 1973, 80, 1-23.

Raye, C. L. Recognition: Frequency or organization?American Journal of Psychology, 1976, 89, 645-658.

Raye, C. L., & Johnson, M. K. Reality monitoring vs.discrimination between external sources. Bulletin ofthe Psychonomic Society, 1980, 15, 405-408.

Raye, C. L., Johnson, M. K., & Taylor, T. H. Is theresomething special about memory for internally-gen-erated information? Memory & Cognition, 1980, 8,141-148.

Reid, T. An inquiry into the human mind, on the prin-ciples of common sense. In K. Lehrer & R. E. Bean-blossom (Eds.), Thomas Reid's inquiry and essays.Indianapolis, Ind.: Bobbs-Merrill, 1975. (Originallypublished, 1764.)

Ross, M., & Sicoly, F. Egocentric biases in availabilityand attribution. Journal of Personality and SocialPsychology, 1979, 37, 322-336.

Russo, J. E., & Wisher, R. A. Reprocessing as a rec-ognition cue. Memory & Cognition, 1976, 4, 683-689.

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

Schneider, W., & Shiffrin, R. M. Controlled and au-tomatic human information processing: I. Detection,search, and attention. Psychological Review, 1977,84, 1-66.

Segal, S. J., & Fusella, V. Influence of imaged picturesand sounds on detection of visual and auditory signals.Journal of Experimental Psychology, 1970, 83, 458-464.

Slamecka, N. J., & Graf, P. The generation effect:Delineation of a phenomenon. Journal of Experi-mental Psychology: Human Learning and Memory,1978,4,592-604.

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

Toglia, M. P., & Kimble, G. A. Recall and use of serialposition information. Journal of Experimental Psy-chology: Human Learning and Memory, 1976, 2,431-445.

Tolman, E. C. Cognitive maps in rats and men. Psy-chological Review, 1948, 55, 189-208.

Tulving, E. Theoretical issues in free recall. In T. R.Dixon & D. L. Horton (Eds.), Verbal behavior andgeneral behavior theory. Englewood Cliffs, N.J.:Prentice-Hall, 1968.

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

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

Wickens, D. D. Encoding categories of words: An em-pirical approach to meaning. Psychological Review,1970, 77, 1-15.

Yuille, J. C. A detailed examination of mediation inpaired-associate learning. Memory & Cognition, 1973,/, 333-342.

Received February 29, 1980 •

Documents

Reality Monitoring - Yale University · Johnson, Department of Psychology, State University of New York, Stony Brook, New York 11794. something may make it seem as though it was perceived