Contexts of Cryptomnesia: May the Source Be with You

25
Social Cognition, Vol. 17, No. 3, 1999, pp. 273-297 CONTEXTS OF CRYPTOMNESIA: MAY THE SOURCE BE WITH YOU C. NEIL MACRAE University of Bristol GALEN V. BODENHAUSEN Northwestern University GUGLIELMOCALVINI University of Bristol Cryptomnesia represents an intriguing type of mental illusion in which people mis takenly believe that they have produced a new idea when in fact they have simply unwittingly retrieved an old, previously encountered idea from memory. Drawing on recent research on this form of inadvertent plagiarism, we hypothesized that perceivers' susceptibility to the illusion that other people's responses were actually self-generated would be influenced by contextual variables that impact upon the efficiency of source monitoring. The results of three studies, examining different contextual factors (i.e., Experiment 1, perceptual similarity; Experiment 2, cogni tive distraction; Experiment 3, retrieval context), confirmed this prediction. We consider how difficulties in source monitoring may inform our understanding of the process and consequences of cryptomnesia in everyday life. One of the most disheartening experiences of old age is discovering that a point you have just made so significant, so beautifully expressed was made by you in something you published a long time ago. Skmncr (1983, p. 242) Information is only as reliable as the source from which it springs, so it is perhaps not too surprising that social impressions, judgments, and eval uations are calibrated in terms of the credibility of their original source. When a given source's credibility is low, information issuing from that source tends to have little impact on perceivers' judgments (e.g., The authors thank John Skowronski, Rich Marsh, and an anonymous reviewer for his or her helpful comments on this work. Address correspondence to C. Neil Macrae, Department of Experimental Psychology, University of Bristol, 8 Woodland Road, Bristol BS8 1TN, UK. E-mail: c.n.macrae@bris- tol.ac.uk. 273

Transcript of Contexts of Cryptomnesia: May the Source Be with You

Social Cognition, Vol. 17, No. 3, 1999, pp. 273-297

CONTEXTS OF CRYPTOMNESIA:

MAY THE SOURCE BE WITH YOU

C. NEIL MACRAE

University of Bristol

GALEN V. BODENHAUSEN

Northwestern University

GUGLIELMOCALVINI

University of Bristol

Cryptomnesia represents an intriguing type of mental illusion in which people mis

takenly believe that they have produced a new idea when in fact they have simply

unwittingly retrieved an old, previously encountered idea from memory. Drawingon recent research on this form of inadvertent plagiarism, we hypothesized that

perceivers' susceptibility to the illusion that other people's responses were actually

self-generated would be influenced by contextual variables that impact upon the

efficiency of source monitoring. The results of three studies, examining different

contextual factors (i.e., Experiment 1, perceptual similarity; Experiment 2, cognitive distraction; Experiment 3, retrieval context), confirmed this prediction. We

consider how difficulties in source monitoring may inform our understanding of the

process and consequences of cryptomnesia in everyday life.

One of the most disheartening experiences of old age is discovering that a

point you have just made so significant, so beautifully expressed was

made by you in something you published a long time ago.Skmncr (1983, p. 242)

Information is only as reliable as the source from which it springs, so it is

perhaps not too surprising that social impressions, judgments, and eval

uations are calibrated in terms of the credibility of their original source.

When a given source's credibility is low, information issuing from that

source tends to have little impact on perceivers' judgments (e.g.,

The authors thank John Skowronski, Rich Marsh, and an anonymous reviewer for his or

her helpful comments on this work.

Address correspondence to C. Neil Macrae, Department of Experimental Psychology,

University of Bristol, 8 Woodland Road, Bristol BS8 1TN, UK. E-mail: c.n.macrae@bris-

tol.ac.uk.

273

274 MACRAE, BODENHAUSEN, AND CALVINI

Hovland & Weiss, 1951; Macrae, Shepherd, & Milne, 1992; Pratkanis &

Aronson, 1992). Of course, our own ideas, perceptions, and evaluations

are likely to be seen as highly credible. So, on average, items that we generate ourselves are likely to exert greater influence on our judgmentsand behavior than information we receive from sources whose reliabil

ity and credibility is less certain (e.g., Pyszczynski & Greenberg, 1987).

Given, therefore, their privileged status within the processes of social

perception and judgment, it is obviously important to have a memory

system capable of distinguishing these self-generated products from in

formation we acquire from external, potentially questionable sources.

Fortunately for the social perceiver, our memory system is

well-equipped to keep track of the sources of stored information (John

son, Hashtroudi, & Lindsay, 1993; Raye, Johnson, & Taylor, 1980). Nev

ertheless, on occasion, memory can play some mischievous tricks on

unsuspecting social perceivers, prompting a variety of troublesome ef

fects. For example, we may tell a friend an hysterical anecdote, failing to

appreciate that either we told her the same story on a previous occasion,

or worse still that she was actually the originator of the tale. Of all our

memory lapses, this latter oversight is particularly troubling, as it im

plies that we may periodically take credit for other people's creative out

puts (e.g., research ideas, recipes, jokes), erroneously believing that theyare in fact our own. Such instances of unintentional plagiarism, or

cryptomnesia as it is termed, occur when consciousness is populated bymemories, but perceivers do not explicitly recognize the items as such

(Taylor, 1965). Instead, they are taken to be the original products of cur

rent cognitive processing. As Brown and Murphy (1989) put it;

"Cryptomnesia refers to generating a word, an idea, a song, or a solutionto a problem, with the belief that it is either totally original, or at least

original within the present context. In actuality, the item is not original,but one which has been produced by someone else (or even oneself) atsome earlier time" (p. 432).As it turns out, cryptomnesia extends beyond the theft of witty anec

dotes from their owners indeed, science, literature, and the music in

dustry are littered with cases of inadvertent plagiarism. From Freud to

GeorgeHarrison to the makers of the recentmovie The FullMonty (1997),

many have been castigated for allegedly stealing other people's creative

products. Even for we lesser mortals, however, cryptomnesic episodescan shape our behavioral outputs. In discussion with a friend, for exam

ple, we may unwittingly repeat a phrase (e.g., "over the hill") or word

(e.g., "decrepit") she has just uttered, yet be quite unaware that this du

plication has occurred (Jacoby & Kelley, 1987). As these mundane in

stances of inadvertent plagiarism are rarely detected, there are typicallyno costs associated with the theft of other people's behavioral products.

CONTEXTS OF CRYPTOMNESIA 275

In other contexts (e.g., literature, music), however, cryptomnesia can

have decidedly worrisome implications for perceivers, ranging from

private embarrassment to public censure. Plagiarize unintentionally, it

would appear, at one's potential peril.If, then, perceivers have a propensity to steal other people's behavioral

products, it is obviously useful to know exactly why and when this theft

occurs. For example, is it possible to identify processing conditions under

which perceivers are particularly likely to plagiarize others? If indeed it

is, then what are these conditions and how frequently do they occur in

daily life? Simply stated, is cryptomnesia something we should worry

about in our everyday interactions with others? In a related vein, when

duplication occurs, are perceivers more likely to plagiarize Self or others?

While both types of plagiarism are undeniably problematic, clearly the

latter variety has more troubling implications for perceivers. Finally, can

cryptomnesia be understood in terms of existingmodels ofmemory func

tion or must new theoretical frameworks be advanced to accommodate

the emerging empirical evidence? In the research reported here, we de

scribe three experiments that attempt to widen our understanding of the

process and consequences of unintentional plagiarism.

SOURCE MONITORING AND CRYPTOMNESIA

Inadvertent plagiarism is believed to derive from the failure of some basic

memory operations specifically, perceivers' capacity to enlist the source

monitoring processes that enable them to identify the origin of their recol

lections (Brown & Halliday, 1991; Brown & Murphy, 1989; Johnson et al.,

1993; Johnson & Raye, 1981; Landau &Marsh, 1997;Marsh & Bower, 1993;

Marsh & Landau, 1995; Marsh, Landau, & Hicks, 1997). Put simply, unin

tentional plagiarism occurs when perceivers misconstrue their memories

as original thoughts. Given this attributional failing, cryptomnesia is

clearly a variant of source forgetting, a phenomenon whereby perceiversdemonstrate an impaired ability to identify the origin (i.e., when, where) of

their recollections (Johnson et al., 1993; Schacter, Harbluk, & McLachlan,

1984). This memory lapse has some all too familiar consequences. For ex

ample, perceivers may be unable to discern whether an event was experienced or imagined (e.g., visiting the zoo as a small child), to statewhich of

multiple targets produced a particular behavior (e.g., was it Uncle John or

Aunt Jean who broke the bathroom window?), or to report the setting or

time frame inwhich a specific event occurred (e.g., eating anchovies for the

first time). In each of these cases, the problem is the same; although

perceivers are capable of retrieving and scrutinizing fragments of a particular memory, they are singularly unable to recollect the context in which

the memory was acquired. As a result, errant source attributions occur

276 MACRAE, BODENHAUSEN, AND CALVINI

(Jacoby & Kelley, 1987; Johnson et al., 1993; Johnson & Raye, 1981). In

cryptomnesia, of course, the attribution process is complicated still further

as perceivers have no idea that the contents of consciousness are memo

ries; instead, they are taken to be original thoughts. Little wonder, there

fore, that perceivers periodically purloin the research ideas, anecdotes,

and melodies of others. So why then does source forgetting (hence

cryptomnesia) occur?

According to the influential framework proposed by Johnson and her

colleagues (Johnson, 1988; Johnson et al., 1993; Johnson & Raye, 1981),

successful source monitoring depends upon two factors: (1) the charac

teristics of stored memorial representations; and (2) the decision pro

cesses that are employed when perceivers attempt to establish the originof their recollections. Memories differ enormously in content, varyingboth in terms of the amount and quality of information they convey.

Generally speaking, whenever the characteristics of stored representations are plentiful and highly differentiated, perceivers are unlikely to

make source-monitoring errors. Instead, these errors tend to occurwhen

memories have overlapping or common properties (Johnson et al., 1993;

Johnson & Raye, 1981). The accuracy of source monitoring is also af

fected by the decision processes that perceivers recruit when they at

tempt to resolve problems of source identification. In this respect, two

general information-processing strategies are available (Chaiken,

Lieberman, & Eagly, 1989; Johnson et al., 1993). First, the determination

of source can rely on heuristic processes that inspect only the quality (i.e.,amount of information, clarity of detail) of stored representations. These

processes are relatively automatic in character (i.e., effortless, precon-scious, fast), prompting the rapid, although fallible, computation of

source judgments (Chaiken et al., 1989; Johnson et al., 1993). Second,

source can also be determined by systematic processes that are generallyslower, more accurate, and more attentionally demanding than their

heuristic counterparts (Johnson et al., 1993). Responsive to perceivers'goal states, these processes are susceptible to disruption and interfer

ence from competing mental activities. Operating in tandem, heuristic

and systematic processes combine to resolve problems of source identi

fication, with the relative contribution of each process determined byperceivers' goal states and the characteristics of the task at hand (Johnson et al., 1993).

Accepting the proposition that cryptomnesia is a variant of source for

getting, it is possible to advance the following empirical prediction.Under contexts in which source monitoring is problematic (Johnson et

al., 1993; Johnson & Raye, 1981 ), onemight expect levels of unintentional

plagiarism to be elevated. The reasoning here is quite straightforward: If

perceivers find it difficult to establish the identity of an item in con-

CONTEXTS OF CRYPTOMNESIA 277

sciousness, then they may be particularly inclined to conclude that the

item is an original product of their current thought processes. As such, in

contexts that reliably promote source confusion, perceivers may displaya pronounced tendency to plagiarize others, or perhaps even them

selves. In the present article, we investigate this intriguing possibility(see also Landau & Marsh, 1997). Prior to describing our experiments,however, we first detail how cryptomnesia can be elicited in the labora

tory; we then speculate on why at least certain forms of inadvertent pla

giarism may be moderated by difficulties in source monitoring.

ELICITING CRYPTOMNESIA

Despite its potentially troublesome implications for perceivers, it is onlywithin the last decade or so that cryptomnesia has attracted serious em

pirical attention. Previous attempts to investigate this topic were ham

pered by the daunting task of eliciting inadvertent plagiarism under

controlled conditions. Thanks to the development of two experimental

analogues of cryptomnesia, however, it is now possible to investigatethis fascinating phenomenon in the laboratory. In the original experimental procedure, Brown and Murphy (1989) had groups of partici

pants take turns in orally generating exemplars from various semantic

or orthographic categories (e.g., sports, musical instruments, words be

ginning with the letters TH). Crucially, prior to the commencement of

this task (i.e.,generate items), participantswere admonished not to repeat

any of the items. As such, instances of unintentional plagiarism occurred

whenever participants duplicated their own or another person's re

sponses. Quite remarkably, despite the instruction not to repeat any of

the items, participants displayedmodest levels of plagiarism during this

task. More alarmingly, however, on other indices of cryptomnesia, con

siderably higher levels of duplication were detected. Following the

item-generation task, Brown and Murphy (1989) required participantsto perform two additional activities: first, to reproduce their earlier re

sponses (i.e., recall own); and second, to generate additional exemplars

(i.e., generate new) that had not been offered (by either Self or others) in

the generation task. In the recall-own task, plagiarism occurred when

ever participants duplicated another person's response. In the gener

ate-new task, it was apparent whenever participants produced an item

that had been offered before. The results were revealing: Elevated levels

of cryptomnesia were observed on both measures, although partici

pants generally preferred to steal the products of others than to plagiarize their own previous efforts. Summarizing their findings, Brown and

Murphy (1989) concluded that "recently experienced information can be

inadvertently plagiarized shortly after the initial experience" (p. 439).

278 MACRAE, BODENHAUSEN, AND CALVINI

In the second experimental analog of cryptomnesia,Marsh and Bower

(1993) investigated inadvertent plagiarism using a word puzzle task

called Boggle. Adapting the essential features of the Brown and Murphy

(1989) procedure, Marsh and Bower required participants to form English words by stringing together adjacent items in a 4 x 4 matrix of let

ters. Participants performed this task against a computer partner and

were instructed not to reproduce either their own or the computer's so

lutions to the puzzle. As in the original Brown andMurphy (1989) procedure, in the course of the experiment, participants performed three

tasks: generate-items; recall-own; and generate-new. Importantly, on all

threemeasures of cryptomnesia,Marsh and Bower replicated the effects

observed by Brown and Murphy (1989). Indeed, if anything, the inci

dence of inadvertent plagiarism was greater on the Boggle puzzle than

Brown and Murphy's category-exemplar generation task. Taken to

gether, then, these two ingenious procedures provide converging evi

dence for the existence of inadvertent plagiarism that is amenable to

experimental investigation.But is there any reason to believe that cryptomnesia is moderated by

difficulties in source monitoring (Johnson et al., 1993)? Four years ago,Marsh and Landau (1995) were skeptical about this relationship and

concluded that, "There appears to be very little evidence that the factors

shown to influence source monitoring alter the incidence of

cryptomnesia" (p. 1580). In a recent publication, however, these authors

have modified their position somewhat, arguing instead that source

confusion can exacerbate cryptomnesia, but only certain types of inad

vertent plagiarism (Landau &Marsh, 1997). So why this change of view

point?Motivating Landau and Marsh's (1997) revised position was their

observation that quite different decision processes are associated with

the tasks used to measure cryptomnesia in the laboratory. Specifically,in the two generative tasks (i.e., generate-items; generate-new)

perceivers need only implement an elementary decision process that enables them to identify whether an item in consciousness is old or new.

For example, the familiarity of an item can serve just such a purpose

(Jacoby & Kelley, 1987; Johnson et al., 1993). Given that recognition judgments are generally not affected by source confusion (Johnson et al,

1993), it is therefore unlikely that difficulties in source monitoringshould increase the frequency of generative cryptomnesic errors. In therecall-own task, however, matters are considerably more complicated,as perceivers must supplement a basic recognition (i.e., old-new) judgment with a secondary assessment of whether the item was

self-generated or provided by a partner (Marsh et al., 1997). Critically,this additional decision process (i.e., was the item yours or mine?) de-

CONTEXTS OF CRYPTOMNESIA 279

mands an extended reasoning typical of source-monitoring tasks,where

perceivers must distinguish between classes of formerly encountered

items Qohnson et al., 1993; Johnson & Raye, 1981). Thus, of the three

tasks used to measure cryptomnesia, only the recall-own task demands

the cognitive operations needed in source monitoring. As such, only

performance on this task should be affected by standard manipulationsof source confusion. But is this really the case?

Inspection of the available literature provides only limited support for

the contention that source confusion can increase the incidence of at least

certain forms of cryptomnesia (Landau & Marsh, 1997; Marsh et al.,

1997). Perhaps the most compelling evidence comes from a recent study

by Landau and Marsh (1997). Modifying the basic Boggle procedure,Landau and Marsh created task conditions where Self- and

Other-generated solutions to the word puzzle were highly confusing.This confusionwas achieved by requiring some participants to guess the

computer's solutions during the item-generation task, an activity be

lieved to increase the cognitive similarity between the computer's re

sponses and self-generated solutions. Based on an application of

Johnson et al.'s (1993) source-monitoring framework, it was anticipatedthat when the two sources (i.e., Self and computer) were highly similar,

levels of inadvertent plagiarism would increase, but only in the re

call-own task. Interestingly, this is exactly what Landau and Marsh

found. Thus, at least where the theft of puzzle solutions from a computer

is concerned, source confusion appears to increase the incidence of

cryptomnesia. But just how widespread are these cryptomnesic effects?

Under conditions of source confusion, do perceivers reliably steal other

people's behavioral products? In the present research, we consider this

important empirical question.

THE PRESENT RESEARCH

Given the noted perils of inadvertent plagiarism, in the presentwork we

reconsider the contention that source confusion increases the incidence

of cryptomnesia, but only certain types of unintentional duplication (i.e.,

recall-own errors). As this prediction gains only limited support in the

available literature (Landau & Marsh, 1997), we deemed it appropriateto undertake a detailed investigation of the effects of source confusion

on cryptomnesia. In so doing, we made a number of modifications to

previous research on this topic. First, we employed a different

item-generation task. Second, and more important, to establish the gen

erality of cryptomnesic effects across task contexts, we considered

whether different manipulations of source confusion would prompt

280 MACRAE, BODENHAUSEN, AND CALVINI

comparable outbreaks of inadvertent plagiarism. This latter issue is

noteworthy, of course, as it is only by explicating the task conditions

under which inadvertent plagiarism reliably occurs that one can beginto anticipate social settings in which perceivers may be particularly sus

ceptible to this type of memorial lapse.To elicit cryptomnesia in the laboratory, we utilized Brown and

Murphy's (1989) category-exemplar generation task. Participantswere required to offer, in a round-robin fashion, six exemplars from

each of four orthographic categories, while under the general admoni

tion not to repeat any of the previous responses. Having generated the

items, participants were then requested to reproduce their earlier ef

forts (i.e., recall-own) and to provide a number of additional exem

plars (i.e., generate-new) from each orthographic category. In

adopting this paradigm, we made one important modification to the

original procedure. Although Brown and Murphy (1989) had partici

pants perform the item-generation task in groups of four, it turned out

that the most likely source of a plagiarized item was the person speak

ing immediately before each participant. As such, recognizing that

this individual is the most likely source of a plagiarized response, in

the present procedure participants were required to perform the

item-generation task in pairs.Our assumption, following Landau and Marsh (1997), was that source

confusion would increase the incidence of cryptomnesia, but only in the

recall-own task. We did not anticipate that standard manipulations of

source confusion would increase the incidence of generative

cryptomnesic errors. To provide a comprehensive investigation of the

effects of source confusion on unintentional plagiarism, in each of the re

ported experiments the difficulty of source monitoring was manipulated in a different manner (i.e., Experiment 1, perceptual similarity;

Experiment 2, cognitive distraction; Experiment 3, retrieval context).We

varied the basis of source confusion for three reasons: First, we wanted

to establish the generality of the finding that source confusion increases

the incidence of inadvertent plagiarism, at least with respect to

perceivers' tendency to take credit for other people's efforts. Second, we

hoped to identify the type of processing conditions that may exacerbate

cryptomnesia outside the laboratory. Third, we wanted to establish the

extent towhich Johnson et al.'s (1993) source-monitoring framework can

inform our theoretical understanding of cryptomnesia. To realize this fi

nal objective, we therefore employed experimental manipulations that

obstructed the encoding (i.e., Experiments 1 & 2) and post-encoding (i.e.,

Experiment 2) operations believed to underlie successful source moni

toring (see Johnson et al., 1993).

CONTEXTS OF CRYPTOMNESIA 281

EXPERIMENT 1

PERCEPTUAL SIMILARITY AND CRYPTOMNESIA

When the perceptual or semantic similarity of two sources is increased,

perceivers typically experience difficulty recollecting exactly who saidwhat (Ferguson, Hashtroudi, & Johnson, 1992; Lindsay, 1990; Lindsay,Johnson, & Kwon, 1991). For example, perceivers are more likely to

misattribute statements made by one speaker to another when the two

sources describe the same event than when they describe different

events (Lindsay et al., 1991). Similarly, one is more likely to confuse the

source of a recollection when the potential candidates are two women

than when they are a man and a woman (Johnson, Nolde, & De

Leonardis, 1996; Lindsay et al., 1991). This latter effect is corroborated byan extensive literature in social psychology, which demonstrates that

perceivers make more within-group (e.g., female-female) than be-

tween-group (e.g., female-male) confusions when recollecting the be

havior of others (see Taylor, Fiske, Etcoff, & Ruderman, 1978).

Extrapolating these findings to the investigation of cryptomnesia, an in

teresting empirical prediction emerges. Specifically, when participants

generate items in same-sex (e.g., female-female) rather than mixed-sex

dyads (e.g., female-male), difficulties in source monitoring may in

crease the incidence of cryptomnesia, at least with respect to partici

pants' tendency to believe that their partner's responses were in fact

self-generated. In our first experiment, we investigated this prediction.

METHOD

Participants and Design. Twenty-eight undergraduates (21 women and

7men) were paid2 ($3.20) for their participation in the experiment. The

experiment had a single factor (group: same-sex or mixed-sex) be-

tween-participants design.Stimulus Material and Procedure. Participants arrived at the laboratory

in pairs (i.e., female-female or female-male) and were greeted by a fe

male experimenter who explained that the experimentwas an investigation of people's word-production skills. Following Brown and Murphy(1989), the experimental procedure was divided into three phases: item

generation; written recall of own responses; and written retrieval of new

exemplars. In the item-generation task, participants were told that theywould take turns in orally generating exemplars from four different or

thographic categories (i.e., words beginning with the letter pairs BE, FO,

MA, and TH), and that each categorywould be completed in turn. Partic

ipants were instructed to produce a new exemplar each time and not to

282 MACRAE, BODENHAUSEN, AND CALVINI

TABLE 1. Percentage of Incorrect Responses as a Function of Task, Error Type, and

Group Composition (Experiment 1)

Task

Genera te-Items Recall-Own Generate-New

Self-

Plagiarism(%)

Partner-

Plagiarism

(%)

Partner-

Plagiarism(%)

New-

Errors

(%)

Self-

Plagiarism(%)

Partner-

Plagiarism(%)

GroupSame-Sex

Mixed-Sex

0

0

0

0.8

24.4

14.5

6.7

8.2

2.6

2.6

4.6

6.2

duplicate any of the previous responses. The order of presentation of the

orthographic categories was counterbalanced and participants pro

vided six exemplars for each category. The experimenter wrote down

items as they were produced and also tape recorded the session as a

backup. Following the item-generation phase and a 3-minLite distractor

task (i.e., counting backwards in 3s from 2000), participantswere given a

recall sheet listing the four orthographic categories with six blank spacesbeneath each letter pair. The experimenter instructed participants to

write down the six examples they had provided during the generationtask (i.e., recall-own).When this taskwas completed, the recall sheetwas

removed and replaced with a second sheet that listed the four ortho

graphic categories with six blank spaces beneath each letter pair. On this

sheet, participants were instructed to provide six completely new exam

ples for each orthographic category (i.e., generate-new). It was emphasized that all the blank spaces on the sheets must be completed. Upon

completion of the experiment, participants were debriefed, paid,thanked, and dismissed.

RESULTS AND DISCUSSION

As is conventional in research of this kind (Brown & Murphy, 1989), the

proportion of errors produced on each task were analyzed separately in aseries of 2 (group) x 2 (error type) analyses of variance (ANOVAs), the results of which are summarized below (see Table 1 for treatment means).Generate-items. During the item-generation task, plagiarized re

sponses can come from two sources: Specifically, participants can duplicate either their own responses (i.e., self-plagiarism) or the responses oftheir partner (i.e., partner-plagiarism). As can be seen in Table 1, how

ever, errors on this taskwere extremely uncommon and did not vary as afunction of dyad composition or error type.Recall-Own. Again, two types of error can be committed on this task.

When instructed to recollect their own responses, participants can either

CONTEXTS OF CRYPTOMNESIA 283

produce items previously provided by their partner (i.e., part

ner-plagiarism), or else they can produce entirely new items that they er

roneously believe to be old (i.e., new-errors). A 2 (group: same-sex or

mixed-sex) x 2 (error type: partner-plagiarism or new-error)

mixed-model ANOVA revealed a main effect of error type on participants' task performance, F(l, 26) = 49.84, p < .0001. As expected, how

ever, this effect was qualified by a significant group x error typeinteraction, F(l, 26) = 11.24, p < .003. An analysis of simple effects confirmed the predicted pattern of results. Whereas new-errors (i.e., intru

sions) were not affected by the composition of the dyad [F(l, 44) < 1, ns],

dyad composition did affect participants' tendency to take credit for

their partner's responses, with higher rates of plagiarism emerging in

the same-sex than the mixed-sex groups, F(l, 44)= 9.77, p < .003. In addi

tion, partner plagiarisms were more abundant than intrusion errors in

both groups [same-sex, F(l, 26) = 53.95, p < .0001; mixed-sex, F(l, 26) =

6.77, p < .02]. Closer inspection of the reproductive errors in the

mixed-sex dyads revealed no difference in men and women's tendencyto steal items from a partner of the opposite sex (respective Ms: male =

15%, female = 14%). As expected, however, women in the same-sex

dyads (M = 24.4%) were more likely than women in themixed-dyads (M= 14.8%) to steal an item from their partner, F(l,19)

= 6.18, p < .03.

Generate-New. In this task, participants could err either by plagiarizingtheir partner's efforts (i.e., partner-plagiarism) or by duplicating their

own previous responses (i.e., self-plagiarism). A 2 (group: same-sex or

mixed-sex) x 2 (error type: self-plagiarism or partner-plagiarism)mixed-model ANOVA revealed only a main effect of error type on performance on this task [F(l, 26) = 5.49, p < .03], with participants more

likely to plagiarize their partner's responses than to indulge in

self-plagiarism (respective Ms: 5.4% vs. 2.6%). Sex composition of the

dyad did not exert a reliable effect on performance of this task (p > .55).

These results, then, support the prediction that source confusion in

creases the incidence of cryptomnesia, but only with respect to certain

forms of inadvertent plagiarism (Landau & Marsh, 1997). On the two

generative tasks (i.e., generate items; generate new), the composition of

the groups had little impact on the incidence of cryptomnesia. As ex

pected, however, quite different effects emerged on the reproductivetask (i.e., recall-own) where participants were required to produce their

previous responses. As predicted by Johnson et al.'s (1993)

source-monitoring framework, levels of cryptomnesia were elevated

when participants performed the initial generation task in same-sex

rather than mixed-sex groups (Lindsay et al., 1991). The results con

firmed that female participantsweremore likely to plagiarize a responseif it was previously provided by a woman than by a man, thereby con-

284 MACRAE, BODENHAUSEN, AND CALVINI

firming that source confusion exacerbates perceivers' tendency to take

credit for other people's efforts. But just how robust is this finding?Would other manipulations of source confusion prompt comparable

cryptomnesic effects? In our second experiment, we addressed this

question.

EXPERIMENT 2

DISTRACTION AND CRYPTOMNESIA

In Experiment 1, source confusion was elicited by manipulating the gender composition of the groups that performed the category-exemplar

generation task (Lindsay et al., 1991). As an expansive literature testifies,

however, target confusion could also have been induced by means of a

variety of other manipulations (Johnson et al., 1993; Johnson & Raye,1981). Because source monitoring depends upon the quality of stored

memorial representations, anything that prevents perceivers from bind

ing memory details to one another, and to the memory trace itself,

should promote source confusion (Chalfonte & Johnson, 1996), hence er

rant source attributions. Thus, factors such as stress, distraction, and al

coholic inebriation, to name but a few, should all attenuate perceivers'

source-monitoring performance. As Johnson et al. report, "anything that

prevents a person from fully contextualizing information at acquisition(i.e., creating an event) will reduce encoding of potentially relevant

source-information" (1993, p. 5). By implication, of course, these factors

should also have a significant impact on perceivers' propensity to plagiarize others. Thus, if distraction prompts source confusion, then it

should also increase the likelihood that perceivers will erroneously believe that other people's effortswere in fact self-generated. In our second

experiment, we investigated this prediction.

METHOD

Participants and Design. Twenty-eight undergraduates (14 women and14 men) were paid 2 ($3.20) for their participation in the experiment.The experiment had a single factor (distraction: present or absent) be-

tween-participants design.StimulusMaterial and Procedure. Participants arrived at the laboratory in

mixed-sex pairs, were greeted by a female experimenter, and randomlyassigned to one of the treatment conditions. The experimentwas identicalto Experiment 1, except for two procedural modifications: First, all participants performed the category-exemplar generation task in mixed-sex

pairs. Second, the manipulation of source confusion was introduced

CONTEXTS OF CRYPTOMNESIA 285

TABLE 2. Percentage of Incorrect Responses as a Function of Task, Error Type, and

Distraction (Experiment 2)

Task

Genera te-Items Recall-Ovvn Genera te-New

Self-

Plagiarism(%)

Partner-

Plagiarism(%)

Partner-

Plagiarism(%)

New-

Errors

(%)

Self-

Plagiarism(%)

Partner-

Plagiarism(%)

Distraction

Present

Absent

0.6

0.8

0.8

0.8

23.5

12.7

9.1

9.7

2.3

2.0

5.2

3.5

while participants performed the item-generation task. Specifically, for

half of the participants (i.e., distraction condition) a radio was ostensibly

playing in the laboratory during the task. In reality, however, it was a

tape-recorded excerpt from a radio show featuring a series of topicalnews items. This excerptwas selected becausewe believed itwould be in

teresting (hence distracting) to undergraduates. The radio was already

playing when participants arrived at the laboratory and it was switched

off by the experimenter when the generation task was completed. In all

other respects, this experiment was identical to Experiment 1.

RESULTS AND DISCUSSION

As in Experiment 1,the proportion of errors produced on each taskwere

calculated and analyzed separately in a series of 2 (distraction) x 2 (error

type) mixed-model ANOVAs (see Table 2 for treatment means).

Generate-items. As in the previous experiment, errors on this taskwere

extremely uncommon and the error rate was not affected by the experimental manipulations.Recall-Own. There were reliable main effects of both factors on partici

pants' recall-own task performance [distraction, F(l, 26)= 5.47, p < .03;

error type, F(l , 26) = 15.42, p < .0006]. As expected, however, these effects

were qualified by a significant distraction x error type interaction, F(l,

26) = 5.27, p < .03. Simple effects analyses confirmed our experimental

predictions. Whereas new errors were not affected by cognitive distrac

tion [F(l, 52) < 1, ns], this manipulation did affect participants' tendencyto take credit for their partner's responses, with higher rates of plagiarism emerging when participants were distracted during the generation

task, F(l, 52) = 10.73, p < .002.! In addition, under conditions of cognitive

1 None of the plagiarized items in the distraction-present conditionwerewords that ap

peared in the radio segment.

286 MACRAE, BODENHAUSEN, AND CALVINI

distraction partner-plagiarisms were more abundant than new-errors,

F(l, 26) = 19.35, p<.0001.

Generate-New. The ANOVA revealed a significant effect of error type on

task performance [F(l, 26)= 4.43, p < .05], with participants more likely to

plagiarize their partner's responses than to engage in self-plagiarism (re

spective Ms: 4.4% vs. 2.2%). However, distraction did not produce a reli

able effect on error rates in this generate-new task (p > .29).

Replicating the results of Experiment 1, these findings provide further

evidence that source confusion exerts quite specific effects on the inci

dence of cryptomnesia (Landau & Marsh, 1997). Under conditions

where source monitoring was compromised through cognitive distrac

tion, although an increase in cryptomnesia was observed, this was only

apparent for reproduction errors (i.e., recall-own task). That is, whilst

distraction increased the likelihood that participants would take credit

for their partner's previous efforts, it did not affect any of the generativeerrors also symptomatic of inadvertent plagiarism.

EXPERIMENT 3

RETRIEVAL CONTEXT AND CRYPTOMNESIA

Corroborating Landau and Marsh's (1997) findings, the results of Experiments 1 and 2 confirmed that source confusion can increase the inci

dence of cryptomnesia, at least with respect to perceivers' tendency to

believe that other people's efforts were in fact self-generated. Of generaltheoretical interest, in each of these experiments, source confusion was

induced by a manipulation that compromised the encoding of

source-relevant material during the item-generation task. Thus, when

the protagonists were both women (i.e., Experiment 1 ) or the generationtask was performed under distracting conditions (i.e., Experiment 2),source confusion exacerbated participants' propensity to take credit for

their partner's efforts. As it happens, however, source monitoring can

also be impaired by manipulations that obstruct a variety of

post-encoding operations (Johnson et al., 1993; Johnson & Raye, 1981).As Johnson et al. note, "because source monitoring depends not only on

the quality of the information as encoded but also on the quality of the

decision processes when source-monitoring judgments are made, any

thing limiting these decision processes at test should also disrupt source

monitoring" (1993, p. 6). Thus, stress, distraction, and attentional depletion are all test factors that reliably impair source-monitoring performance (Johnson et al., 1993). To this list, one can also append the

characteristics of the retrieval context, as this is also known to play a

prominent role in source monitoring (Jacoby & Kelley, 1987; Johnson et

CONTEXTS OF CRYPTOMNESIA 287

al., 1993). But what impact might this factor have on perceivers' propensity to plagiarize others?

One intriguing possibility is that levels of cryptomnesia may be ele

vated when the context in which information was generated is rein

stated at a later date. As a graphic illustration of this possibility, consider

Jacoby, Kelley, Brown, and Jasechko's (1989) cautionary observation:

Many of us have had the experience of presenting an idea to a colleague onlyto have the colleague thoroughly reject the idea. In a later conversation,

however, the colleague reintroduces the earlier rejected idea as an insightthat he or she just had. ...Nature may be so perverse as to make it likely that

we will present a stolen idea as being our own to the very person from

whom we stole it. The cues offered by his or her appearance and the content

of a current conversation may be similar to those present during the priorconversation and serve as excellent cues for the unconscious retrieval of the

idea. (p. 41)

If operating, such a process would have a considerable impact on the

incidence of cryptomnesia, as participants typically generate and retrieve

items in the company of their partner.When present, one's partnerwould

obviously cue many responses that he or she produced during the initial

generation task (Jacoby et al, 1989). If, of course, these items are accompanied by source-specifying information, then there is little cause for con

cern, as the items will be construed as recollections and attributed to the

appropriate source (i.e., Self or partner). When this information is less ac

cessible, however, problems may begin to emerge. Specifically, when

faced with the puzzle of establishing the status and origin of an item in

consciousness, one may be duped into believing that either it is the original product of one's current cognitive processing (i.e., is a novel item), or

alternatively that it was formerly self-generated. Either way, an

attributional error would occur. When, however, one's partner is absent

during the retrieval process, source confusion should be less pro

nounced, as fewer partner-cued items would be expected to pop into con

sciousness. Somewhat ironically, therefore, the presence of one's partner

during the recall-own taskmay actually increase one's propensity to pla

giarize his or her previous responses (Jacoby et al., 1989).

Given the structure of the retrieval context in experimental investigations of cryptomnesia, however, it is also possible that the presence of

one's partner could have precisely the opposite effect on task performance that is, it could reduce levels of inadvertent plagiarism. If an

item is produced in the initial generation task (by either Self or other),

then through its enhanced accessibility in memory (Marsh & Landau,

1995) this will increase the likelihood that the item will appear in con-

288 MACRAE, BODENHAUSEN, AND CALVINI

TABLE 3. Percentage of Incorrect Responses as a Function of Task, Error Type, and

Retrieval Context (Experiment 3)

Task

Generate-items Recall-C>wn Generate-New

Self- Partner-

Plagiarism Plagiarism(%) (%)

Partner-

Plagiarism(%)

New-

Errors

(%)

Self-

Plagiarism(%)

Partner-

Plagiarism

(%)

Retrieval Context

Partner Present 0.8 0.3

Partner Absent 0.6 0.3

9.7

21.4

8.1

7.9

1.1

2.3

3.8

3.1

sciousness during the recall-own task. If so, then again one would be

confronted with a tricky attributional dilemma; specifically, from where

did the item originate? If one's partner is present, however, it is conceiv

able that he or she may function as something of an aide-de-memoire, pro

viding valuable source-related cues to the origin of the item (Johnson et

al., 1993). Thus, on generating a particular exemplar, the presence of

one's partner may remind one that she previously produced the item.

When one's partner is absent, however, these source-specifying cues

would be unavailable, hence one might expect levels of cryptomnesia to

be elevated under these conditions. The question of interest, then, is as

follows: does the presence of one's partner during the recall-own task in

crease or decrease the incidence of cryptomnesia (Jacoby et al., 1989)? In

our third experiment, we investigated this issue.

METHOD

Participants and Design. Twenty-eight undergraduates (14 women and14 men) were paid 2 ($3.20) for their participation in the experiment.The experiment had a single factor (retrieval context: partner-present or

partner-absent) between-participants design.StimulusMaterials and Procedure. Participants arrived at the laboratory

in pairs (female-male), were greeted by two female experimenters, and

randomly assigned to one of the treatment conditions. The experimentwas identical to Experiment 2, apart from the manipulation of source

confusion. On this occasion, rather than varying the difficulty of the en

coding process, we manipulated the retrieval conditions under which

participants performed the final two cryptomnesic tasks (i.e., re

call-own, generate-new). Specifically, whereas half of the participantsperformed the recall-own and generate-new tasks in the company of

their partner, the others performed these tasks while their partner was

absent. Following the item-generation task, to minimize the impact of

CONTEXTS OF CRYPTOMNESIA 289

contextual cues on task performance, all participants were taken to a different room (or rooms) to complete the experiment. In each condition

(i.e., partner-present vs. partner-absent), two people remained in the

room during the recall-own and generate-new tasks (i.e., participant +

partner or participant + new experimenter). In all other respects, this ex

periment was identical to Experiment 2.

RESULTS AND DISCUSSION

As in Experiments 1 and 2, the proportion of errors produced on each

task were calculated and analyzed separately in a series of 2 (retrieval

condition) x 2 (error type) mixed-model ANOVAs (see Table 3 for treat

ment means).

Generate-items. As in each of the previous experiments, error rates on

this task were uniformly low and were not influenced by experimentalcondition.

Recall-Own. Therewere reliablemain effects of both experimental fac

tors on participants' recall performance [retrieval condition, F(l,26) =

13.16, p < .002; error type, F(l,26)= 15.43, p < .0006]. These effects were

qualified, however, by a significant retrieval condition x error type in

teraction, F(l,26) = 9.62, p < .005. Simple effects analyses revealed the

following results. Whereas new-errors were not affected by retrieval

condition [F(l,50) < 1, ns], partner-plagiarism was influenced by this

manipulation [F(l,50) = 22.05, p < .002], with higher rates of plagiarism

emerging in the partner-absent than the partner-present condition. In

addition, partner plagiarisms were more abundant than intrusion er

rors in the partner-absent condition, F(l,26) = 24.71, p < .0001.

Generate-New. There was a marginal effect of error type on task performance [F(l,26) = 3.57, p < .07], with participants more likely to plagiarizetheir partner's responses than to indulge in self plagiarism (respectiveMs: 3.4% vs. 1.7%). Retrieval context did not have an effect on error rates

in this task (p > .75).

Although employing a different manipulation of source confusion,

these results directly replicate those observed in Experiments 1 and 2.

Whereas source confusion had no effect whatsoever on the incidence of

cryptomnesia in the two generative tasks (i.e., generate items; gener

ate-new), quite different effects emerged on the reproductive task (i.e.,

recall-own) where participants were required to produce previous re

sponses. Specifically, levels of cryptomnesia were elevated when the retrieval context did not contain any source-related cues (i.e.,

partner-absent condition). That is, participants were more likely to steal

290 MACRAE, BODENHAUSEN, AND CALVINI

their partner's responses when he or shewas absent, rather than present,

during the recall-own task.- Of course, it remains possible that levels of

cryptomnesia could be elevated when one's partner is present duringthe retrieval task (Jacoby et al., 1989). The deciding factor, we suspect,

may be how many items one's partner has previously generated. In the

present paradigm, the production ofmultiple items may have rendered

one's partner a valuable source-specifying cue. That is, through the re

peated production of items, strong associative links may have been

forged in memory between one's partner and his or her responses (Johnson et al., 1993). As such, when present, one's partner may have func

tioned as a useful source-specifying retrieval cue. If, however, only a

single item had been produced in the generation task, it is unlikely that

such associative links would have been formed, hence under these con

ditions the presence of one's partner may actually elicit bouts of

cryptomnesia (see Jacoby et al., 1989). One useful task for future research

will be to investigate this possibility.

GENERAL DISCUSSION

Taken together, the results of the present studies support the notion that

impairments in source monitoring can increase the incidence of at least

certain forms of inadvertent plagiarism (Johnson et al., 1993; Landau &

Marsh, 1997; Marsh et al., 1997). In the context of a category-exemplar

generation task (Brown & Murphy, 1989), when: (1) the exemplars were

generated by similar targets (i.e., Experiment 1); (2) the generation task

was executed under distracting conditions (i.e., Experiment 2); or (3)source-related cues were not present at the time of retrieval (i.e., Experiment 3), source confusion exacerbated participants' tendency to believe

that other people's efforts were in fact their own. In the discussion that

follows, we consider the theoretical and practical implications of these

findings.

2. Another possible explanation can be offered for the low levels of plagiarism that were

observed in the partner-present condition. When participants in the partner-present con

dition were completing the recall tasks, it is possible that they may have believed that their

responses would later be shared with the other person, thereby prompting them to use

more stringent decision criteria than their colleagues in the partner-absent condition

(Johnson et al., 1993). If operating, such an effect would reduce levels of cryptomnesia in

the partner-present condition. Whilst such an explanation is possible, it is worth notingthat during the debriefing session no participant expressed the belief that his or her re

sponses would be seen by the other person. Thus, it is unlikely that the present effects origi-nated via this route.

CONTEXTS OF CRYPTOMNESIA 291

ON DUPLICATION: YOUR IDEA OR MINE?

Given the documented existence of unintentional plagiarism, it is obvi

ously important to delineate the task conditions under which this worri

some phenomenon is likely to occur. For example, if one has the

potential to steal a colleague's latest research idea, it is useful to identifyfactors that may precipitate this heinous action. Following others, we

speculated that cryptomnesia may reflect a basic malfunction of source

memory specifically, perceivers' capacity to reconnect a recollection

with its rightful owner (Johnson et al., 1993; Landau & Marsh, 1997).

Thus, when unable to discern the identity of an item in consciousness,

perceivers may be inclined to believe that the item was self-generatedwhen, in fact, it was produced by someone else. Our findings confirmed

this prediction.Moreover, whether source confusionwas elicited byma

nipulations that impaired the encoding (i.e. Experiments 1 & 2) or re

trieval (i.e., Experiment 3) of source-related information, a reliable effect

emerged source confusion exacerbated perceivers' tendency to take

credit for other people's efforts (Johnson et al., 1993; Landau & Marsh,

1997;Marsh et al., 1997). As expected, however, source confusion did not

increase the incidence of other types of cryptomnesia. Specifically, generative errors were not affected by manipulations of source confusion.

This finding confirms Landau and Marsh's (1997) observation that the

recall-own task differs from the two generative activities (i.e., gener

ate-items, generate-new) in terms of the decision processes it requires.Whereas the recall-own task demands extended source-monitoring (i.e.,

was it your response or mine?), the two generative activities simply re

quire that participants establish the novelty of each exemplar generated

(i.e., is the item old or new?). Given these processing differences, onlythe recall-own task is affected by standard manipulations that increase

the difficulty of source monitoring (Johnson et al., 1993). It remains possible, however, that in other experimental paradigms even the two generative activities may be disrupted by manipulations of source

confusion. One task for future research on this topic will be to consider

this possibility.

Despite using the same generative task (i.e., category-exemplar gener

ation), the present findings diverged from those observed by Brown and

Murphy (1989) in a couple of interesting ways. First,whereas Brown and

Murphy observed significant rates of plagiarism in the generation task,

such errors were almost entirely absent in the present research. Second,

in the experiments reported herein, plagiarism rates in the recall-own

task were apparently much higher than those reported by Brown and

Murphy (1989). Our intuition is that these differences may be due to the

292 MACRAE, BODENHAUSEN, AND CALVINI

procedural alterations we made to the original paradigm in the present

research. Specifically, whereas Brown and Murphy had participants

perform the item-generation task in groups of four, in the present work

participants completed this task in pairs. One consequence of this modi

fication was that fewer items were elicited during the generation task

(i.e., 48 vs. 64 items). This difference, we suspect, may account for the

elimination of cryptomnesia in the initial generation task. The reduction

in size of the experimental groups would also be expected to influence

task performance on the recall tasks, as the decision criteria that partici

pants recruitwould be different in groups of four than in dyads (Johnsonet al., 1993). As for the increased rate of inadvertent plagiarism observed

on the recall-own task, this we believe is a function of the manipulationsof source confusion employed. In the control conditions (i.e., no confu

sion), rates of cryptomnesia were comparable to those reported byBrown and Murphy (1989).

Although performance on the two generative activities was unaf

fected by manipulations of source confusion, these tasks nonetheless

help to elucidate another important aspect of cryptomnesia namely,what exactly is it that perceivers plagiarize? For example, do perceivers

engage in bouts of self-duplication, unwittingly repeating their own

previous responses (Skinner, 1983), or do they prefer instead to steal

other people's efforts (Jacoby & Kelley, 1987)? This turns out to be an im

portant question, as clearly these two varieties of cryptomnesia have

quite different implications for perceivers.Whereas self-plagiarism typ

ically elicits nothing more sinister than feelings of embarrassment or re

gret, copying other people's outputs (e.g., ideas, songs, poems) can have

decidedly disturbing consequences for perceivers, such as an appear

ance in court. So which form of inadvertent plagiarism ismost prevalent,

partner plagiarism or self-duplication? At least with respect to performance on the generate-new task, the present findings tell a pretty consistent tale. Across all three experiments, participants were more likely to

plagiarize their partner's responses than they were to engage in bouts of

self-duplication. So why might this be so? Why do perceivers prefer tosteal their partner's responses?To answer the previous question, one must consider the dynamics of

the initial item-generation task. In furnishing items on this task, not onlydo participants represent their articulated responses in memory, but so,

too, they store a record of the cognitive operations associated with the

generation of each item (Johnson et al., 1993; Johnson & Raye, 1981). For

example, if an item was particularly difficult to generate, was affectivelysalient, orwas personally relevant in someway or another (e.g., it cued a

plethora of autobiographical recollections), then this information is also

stored in memory together with a record of the item itself. The same can

not be said, of course, for partner-generated responses, as obviously one

CONTEXTS OF CRYPTOMNESIA 293

has no idea how these items were generated or what associations they

may have evoked (see Marsh & Hicks, 1998). Thus, because

self-generated items are highly differentiated in memory, one is less

likely to plagiarize them at a later date (Johnson et al., 1993). The representational similarity of partner-generated items, however, make them

likely candidates for appropriation. Somewhat disturbingly, therefore,in demonstrating a predilection for partner plagiarism rather than

self-duplication, perceivers display the variety of cryptomnesia that is

ultimately most damaging.But from the plethora of partner-generated items that reside in mem

ory, which ones get stolen? According to Marsh and Landau (1995), the

answer lies in the post-generational accessibility of items in memory.

Following their initial generation, items are believed to retain different

activation strengths, and it is these differences that are thought to deter

mine which responses are likely candidates for theft. To evaluate this

prediction,Marsh and Landau made a simple modification to the Boggle

procedure. Following the item-generation task, participants performeda lexical-decision task (LDT), where they were required to reportwhether a series of letter stringswere legal "English" words. During this

task, participants were presented, among other things, with the solu

tions that had previously been provided to the Boggle puzzle. On com

pletion of the LDT, participants then performed the conventional

recall-own and generate-new tasks. The question of theoretical interest

was as follows: Would the relative accessibility of items in the LDT de

termine the extent to which they were plagiarized by participants, with

highly accessible items comprising the most likely candidates for theft?

As it turned out, this is exactly what Marsh and Landau found, prompt

ing them to conclude that, "Items with greater strength are more likely to

be later plagiarized" (1995, p. 1575). Much like other memory illusions,

therefore, basic cognitive principles appear to determine both the condi

tions under which cryptomnesia is likely to occur and items that are

probable candidates for theft (Jacoby & Kelley, 1987; Johnson et al., 1993;Landau & Marsh, 1997; Marsh & Landau, 1995). In this respect, the

source-monitoring framework advanced by Johnson and her colleagues

(Johnson, 1988; Johnson et al., 1993; Johnson & Raye, 1981) provides a

useful theoretical framework for understanding cryptomnesia and its

associated effects in mental life.

CRYPTOMNESIA AND EVERYDAY LIFE

The existence of inadvertent plagiarism poses some potentially anxi

ety-provoking questions for social perceivers. When we bask in the pridethat derives from ourmost recent creative product be it a research idea, a

new recipe for cheesecake, or a romanticaria can we be sure that our feel-

294 MACRAE, BODENHAUSEN, AND CALVINI

ing is justified? After all, perhaps the glory belongs elsewhere, as we have

been the unwitting victim of cryptomnesia. This sudden realization that

one's cherished ideas are actually someone else's is an experience that has

troubled some of psychology's leading thinkers. In describing how he de

veloped his theory of original bisexuality, for example, Freud (1901/1960)noted that a colleague, Fliess, claimed to have furnished him with the idea

several years earlier. Although, Freud completely rejected this claim, he

subsequently remembered their original encounter and correctly traced

the origin of the idea to his colleague. But just how ubiquitous is this typeof source error? Dowe plagiarize others on a regrettably regular basis, or is

inadvertent duplication a relatively infrequent affair? Is cryptomnesia

something we should worry about?

Inspection of the modest literature on this topic confirms the potential

perils of cryptomnesic thinking (Betz, Skowronski, & Ostrom, 1996;

Bink, Marsh, Hicks, & Howard, in press; Wicklund, 1989; Wicklund,

Reuter, & Schiffmann, 1988). Betz et al. (1996), for example, provide a

provocative demonstration of what happens when a memory and its

source become dissociated (Johnson et al., 1993). Following the presentation of a story, participants were required to complete a multiple-choice

recognition test assessing their memory for the piece. After respondingto some of the questions, however, participants were shown other peo

ple's bogus responses to the same test items. Later, participants' recol

lections of the story were reassessed, this time with a cued-recall task.

The question of interest was as follows: To what extent would partici

pants' memories be shaped by other people's bogus recollections? The

results were intriguing. Participants frequently appropriated other people's recollections, particularly when these memories were seemingly

highly credible. Similar effects have also been reported in a recent studyby Bink et al. (in press). When asked to produce novel ways to reduce

traffic accidents, participants were more likely to plagiarize an idea if itwas previously produced by a high- than a low-credibility source (i.e.,traffic planner vs. college student). Thus, not only do perceivers have a

worrying propensity to purloin other people's behavioral products, butit would appear that they prefer to steal items from the most reliable,

trustworthy, and prestigious of sources.

Although source-monitoring errors are by no means an inevitable

consequence of memory function (Johnson et al., 1993), the characteris

tics of everyday interaction may ironically serve to promote unwantedbouts of cryptomnesia. Of relevance in this regard are the decision criteria and processing strategies that perceivers routinely employ duringsource monitoring (Chaiken et al., 1989; Johnson et al., 1993). In a rangeof social settings, it is probable that perceivers will employ stringent

source-monitoring criteria, thus preventing the occurrence of

cryptomnesia. Take, for example, task contexts where perceivers believe

CONTEXTS OF CRYPTOMNESIA 295

their outputs will be scrutinized or evaluated by others. Under condi

tions such as these, cryptomnesia is unlikely to emerge as perceivers will

be highly motivated to avoid source-monitoring errors (Johnson et al.,

1993; Johnson & Raye, 1981). Generally speaking, whenever perceiversare encouraged to pay close attention to their behavioral products, inad

vertent plagiarism is unlikely to occur (see Marsh et al., 1997). Unfortu

nately, these conditions may be the exception rather than the rule in

daily life. As Marsh et al. argue, "In the generation of ideas, and perhapsin many other real-world situations of interest...people's decision crite

ria may be quite different. Most likely, people fail to apply sufficiently

stringent criteria to avoid source-monitoring errors" (1997, p. 895). This

failure to scrutinize source-relevant informationmay be due to a varietyof factors, although most bouts of cryptomnesia can probably be traced

to the fact that in many task contexts perceiversmay simply lack the mo

tivation or cognitive capacity to employ the systematic processing strat

egies that would ensure source-monitoring success (Johnson et al.,

1993). In others words, the characteristics of everyday task environ

ments may serve to promote unwanted bouts of cryptomnesia.

Notwithstanding the previous discussion, unconscious plagiarism is

not necessarily an undesirable mental phenomenon. Rather, inadvertent

plagiarism simply reflects the intermittent failure of a source-monitor

ing system that typically serves us well (Johnson et al., 1993). The same

processes that enable us to keep track of the origin of our recollections

also periodically persuade us that an insight or thought is original,when

in fact itwas generated by someone else. Sometimes, of course, these fail

ures in source monitoring can have rather unpleasant consequences,such aswhen one steals amelody ormovie plot from its creator. On other

occasions, however, it is possible that perceivers may accrue tangiblebenefits through cryptomnesic thinking, such as creative insights, con

versational fluency, and facilitated learning.What is certain, however, is

that perceivers lack the self-monitoring skills needed to detect the some

times subtle influences of inadvertent plagiarism. As such, it is probablethat cryptomnesic effects shape a variety of our behavioral products, we

are simply unaware that these unconscious forces are operating. Indeed,

given people's propensity to plunder other people's efforts, perhaps

kleptomnesia3 is a more appropriate label for this particular memory

lapse.Of course, where some of our behavioral outputs are concerned, we

would be well advised to consider the possibility of "cryptomnesic con

tamination." Consider, for example, academic life, where the intellec-

3. Thanks to Dan Gilbert for coining this term and to DanWegner for suggesting that we

(intentionally) steal it.

296 MACRAE, BODENHAUSEN, AND CALVINI

tual ownership of ideas is a cornerstone of the scientific enterprise. Here,

too, implicit cryptomnesic forces may operate. As Marsh et al. speculate,"As we repeat the ideas of our colleagues each day in teaching our

courses and in our casual conversations often without properly credit

ing the source do these failures to engage source-monitoring processescombined with these repeated exposures lead to a greater level of inad

vertent appropriation than many of us would ever imagine (1997, p.

896)?" Our suspicion is that they probably do, and in the present re

search we have attempted to demonstrate how standard manipulationsof source confusion (Johnson et al., 1993) can exacerbate people's pro

pensity to plagiarize others. Given that these manipulations simulate

processing contexts routinely encountered in life outside the laboratory

(e.g., target similarity, cognitive distraction), one may reasonably sur

mise that cryptomnesia is a potent force in shaping our behavioral products. To avoid such an influence, of course, the recommended course of

action is straightforward,monitor the origin of one's recollections (Johnson etal., 1993; Johnson & Raye, 1981). Or as Obi-Wan Kenobi should per

haps have said "may the source be with you."

REFERENCES

Betz, A. L, Skowronski, J. J., & Ostrom, T. M. (1996). Shared realities: Social influence and

stimulus memory. Social Cognition, 14, 113-140.

Bink, M. L., Marsh, R. L, Hicks, J. L.,& Howard, J. D. (in press). The credibility of a source

influences the rate of unconscious plagiarism. Memory.

Broun, A. S., & Halliday, H. E. (1991). Cryptomnesia and source memory difficulties.

American journal of Psychology, 104, 475190.

Brown, A. S., & Murphy, D. R. (1989). Cryptomnesia: Delineating inadvertent plagiarism.Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 432-442.

Chaiken, S., Lieberman, A., & Eagly, A. H. (1989). Heuristic and systematic information

processing within and beyond the persuasion context. In J. S. Uleman & J. A. Bargh(Eds), Unintended thought (pp. 212-252). New York: Guilford.

Chalfonte, B. L., & Johnson, M. (1996). Feature memory and binding in young and older

adults. Memory and Cognition, 24, 403-416.

Ferguson, S., Hashtroudi, S., & Johnson, M. K. (1992). Age differences in usingsource-relevant cues. Psychology and Aging, 7, 443-452.

Freud, S. (1960). The psychopathology of everyday life. In J. Strachey (Ed. and Trans.), Thestandard edition of the complete works of Sigmund Freud (Vol. 6). London: HogarthPress. (Original work published in 1901).

Hovland, C, & Weiss, W. (1951 ). The influence of source credibility on communication ef

fectiveness. Public Opinion Quarterly, 15, 635-650.

Jacoby, L. L., & Kelley, C. M. (1987). Unconscious influences of memory for a prior event.

Personality and Social Psychology Bulletin, 13, 314-336.

Jacoby, L. L, Kelley, C. M Brown, J., & Jasechko, J. (1989). Becoming famous overnight:Limits on the ability to avoid unconscious influences of the past. Journal ofPersonal

ity and Social Psychology, 56, 326-338.

Johnson, M. K. (1988). Reality monitoring: An experiential phenomenological approach.Journal of Experimental Psychology: General, 117, 390-394.

CONTEXTS OF CRYPTOMNESIA 297

Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. PsychologicalBulletin, 114,3-28.

Johnson, M. K., Nolde, S. F., & De Leonardis, D. M. (1996). Emotional focus and source

monitoring. Journal ofMemory and Language, 35, 135-156.

Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. Psychological Review, 88, 67-85.

Landau, J. D.,& Marsh, R. L. (1997). Monitoring source in an unconscious plagiarism para

digm. Psychonomic Bulletin and Review, 4, 265-270.

Lindsay, D. S. (1990). Misleading suggestions can impair eyewitnesses' ability to remem

ber event details, journal ofExperimental Psychology: Learning,Memory, and Cognition,16, 1077-1083.

Lindsay, D. S., Johnson, M. K., & Kwon, P. (1991). Developmental changes in memory

source monitoring. Journal of Experimental Child Psychology, 52, 297-318.

Macrae, C. N., Shepherd, J. W., &Milne, A. B. (1992). The effects of source credibility on the

dilution of stereotype-based judgments. Personality and Social Psychology Bulletin,

IS, 765-775.

Marsh, R. L., & Bower, G. H. (1993). Eliciting cryptomnesia: Unconscious plagiarism in a

puzzle task, journal of Experimental Psychology: Learning, Memory, and Cognition, 19,

673-688.

Marsh, R. L., & Hicks, J. L. (1998). Test formats change source-monitoring decision pro

cesses, journal of Experimental Psychology: Learning, Memory, and Cognition, 24,

1137-1151.

Marsh, R. L., & Landau, J. D. (1995). Availability in cryptomnesia: Assessing its role in two

paradigms of unconscious plagiarism, journal of Experimental Psychology: Learning,

Memory, and Cognition, 21, 1568-1582.

Marsh, R. L., Landau, J. D., & Hicks, J. L. (1997). Contributions of inadequate source moni

toring to unconscious plagiarism during idea generation, journal of Experimental

Psychology: Learning, Memory, and Cognition, 23, 886-897.

Pratkanis, A., & Aronson, E. (1992). Age ofpropaganda: The everyday use and abuse ofpersuasion. New York: Freeman.

Pyszczynski, T., & Greenberg, J. (1987). Toward an integration of cognitive and motiva

tional perspectives on social inference: A biased hypothesis-testing model. In L.

Berkowitz (Ed.), Advances in experimental social psychology (Vol. 20, pp. 297-340). San

Diego, CA: Academic Press.

Raye,C. L., Johnson, M. K.,& Taylor, T. H.(1980). Is there something special aboutmemoryfor internally generated information? Memory & Cognition, S, 141-148.

Schacter, D. L., HarblukJ. L, & McLachlan, D. R. (1984). Retrieval without recollection: An

experimental analysis of source amnesia, journal ofVerbal Learning and Verbal Behav

ior, 23, 593-611.

Skinner, B. F. (1983). Intellectual self-management in old age. American Psychologist, 38,239-244.

Taylor, F. K. (1965). Cryptomnesia and plagiarism. British journal of Psychiatry, 111,

1111-1118.

Taylor, S. E., Fiske, S. T., Etcoff, N. L., & Ruderman, A. J. (1978). Categorical and contextual

bases of person memory and stereotyping. Journal of Personality and Social Psychol

ogy, 36, 778-793.

Wicklund, R. A. (1989). The appropriation of ideas. In P. Paulus (Ed.), Psychology ofgroupinfluence (2nd ed pp. 393-423). Hillsdale, NJ:

Erlbaum.

Wicklund, R. A., Reuter, T., & Schiffmann, R. (1988). Acting on ideas: Appropriation to

one's self. Basic and Applied Social Psychology, 9, 13-31.