:
, .,f(7-,..,
~!chologic.1
1993 Vol
VJP{lf
Bulletin
114. No.1.
('opyrightl993
by the
American
Psychologica.
3-28
AssociatioR.
0033-2909/93/$300
Inc.
SourceMonitoring
Marcia K. Johnson, Shahin Hashtroudi, and D. StephenLindsay
A frame~rk for understanding source monitoring and relevant empirical evidence is described,
and severalrelated phenomenaare discussed: old-new recognition, indirect tests,eyewitnesstestimony,misattributed familiarity, cryptomnesia, and incorporation of fiction into fact. Disruptions
in source monitoring (e.g.,from confabulation, amnesia,and aging)and the brain regions that are
involved are also considered, and source monitoring within a general memory architecture is
discussed. It is argued that source monitoring is based on qualities of experience resulting from
combinations of perceptual and reflective processes,usually requires relatively differentiated phenomenal experience,and involvesattributions varying in deliberateness.Thesejudgments evaluate
information according to flexible criteria and are subject to error and disruption. Furthermore,
diencephalic and temporal regions may play different roles in source monitoring than do frontal
regions of the brain.
Past experience affects us in many ways. It influences the
ease of identifying stimuli under degraded conditions (e.g.,Jacoby & Dallas, 1981), changes the probabilitytpat we will think
certain thoughts (e.g., Dominowski & Ekstrand, 1967; Kihlstrom, 1980), affects emotional responses such as preferences
(e.g., Johnson, Kim, & Risse, 1985; Zajonc, 1980), manifests
itself as expert or semantic knowledge (e.g., Chase & Simon,
1973; Tulving, 1983), and creates the potential for what we take
to be memories of autobiographical events from our personal
ul~~ ~ources through decision processes performed during re~~!!!berinl!.The ability to identify the source of remembered information
is critical for many cognitive tasks. In laboratory studies of
memor}\ for example, it helps subjects differentiate between
test items they recognize or recall from a study list and test
items that seem familiar or come to mind from other sources.
In everyday life, memory for source contributes to our ability to
exert control over qpr own opinions and beliefs; if you re-
past (e.g., Rubin, 1986).
This review focuses on expressions of memory that involve
judgments about the origin, or source, of information. The
term source refers to a variety of characteristics that, collectively, specify the conditions under which a memory is acquired
(e.g., the spatial, temporal, and social context of the event; the
media and modalities through which it was perceived). This
concept is closely related to, but somewhat more inclusive tha~,
that of memory for context. Ace~!ral claim of the source-!!!orntoring approach is that people do not typically directly retrieve
an abstract~
1abelftha~peCliieS"a-memoiry:s source, rath~r,
ac-fivated memory records are evaluated and attributed topartic-
member that the source of a "fact~ was a supermarket tabloid
such as the Nationa/Enquirer and not Consumer Reports, you
have information that is important for evaluating the veridicality of the purported fact. Perhaps most important, the subjective experience of autobiographical recollection-the
feeling of
remembering a specific experience in one's own life-depends
on source attributions made on the basis of certain phenomenal qualities ofre.membered experience. (When memory information enters consciousness without these qualitative characteristics, it is experienced as knowledge or belief.)
Inability to specify source information can be mildly d~sconcerting, as in not being able to remember whether the person to
--whom
you are about to tell a joke is the one who told you the
joke in the first place. Failures to remember source can also be
profoundly disruptive, as in delusions (e.g.,Oltmanns & Maher,
1988) and confabulation (e.g., Stuss, Alexander, Lieberman: &
Levine 1978). In fact, a severe disruption in remembering
source'is a salient feature of some; and perhaps all, forms of
.
Marcia K. Johnson, Department of Psychology,Princeton Uni~ersity; Shahin Hashtroudi, De.partmentof Psychology,Geor~e Was~lngton University; D. StephenLindsay,Depar~mentofPsychology,Umversity of,:,ictoria, Vict~ri~, British Columbia, Canada.
Shahin Hashtroudl dIed on February 24,1992...
Preparation
of
this
article
was
supported
by
Natlonallnstrtute
on
amnesia
Aging Grant I-ROI-AG09253. We would like to thank Carol Raye,
Sam Glucksberg, Colleen Kelley, Bill Hirst, the 1990-91 crew of the
Princeton memory lab (Chad Dodson, Allison Hermann, Tina Loose,
Kristi Multhaup, and Carolyn Weisz) and several anonymous reviewers for helpful comments on drafts of this article.
Correspondence concerning this article shoul~ be addre~sed.to
Marcia K. Johnson, Department of Psychology,Pnnc~ton Umversrty~
Princeton, New Jersey08544-1010 or to D. Stephen Lindsay, I:>epa~t
ment of Psycho.logy,University of Victoria, p.o. Box 3050, VIctOrIa,
British ColumbIa, Canada V8W 3P5.
(
H.
trs,
t
1982;
,
Mayes
,
Meudell
,
&
Pickering,
..
1985).
~o~~t~ri~~fers
to th.e .set or proces~es !nvol~ed~
making attributions about the orIginS of me~orles, knowled e,
an e Ie s as rou I, 0 nson, & hrosn~ak, 19 ; Johnson,
1988a, 1988b; Lindsay & Johnson, 1987; LIndsay, J~hnson,.&
Kwon, 1991). There has been a recent upsurge of interest m
such questions (e.g., R. E. Anderson, 1984; Eich & Metcalfe,
1989; Foley, Durso, Wilder, & Freidman, 1991; Hanley & ColI.
1989.lntraub & Hoffman 1992; Jacoby, Kelley, & Dywan,
1~:9 Joh~son & Raye, 1981; Kahan & Johnson, 1992; Masson,
3
M. JOHNSON, S. HASHTROUDI,
1989; Mcintyre & Craik, 1987; Rabinowitz, 1989; Schacter,
Harbluk, & Mclachlan, 1984; Schooler & Engstler-Schooler,
1990; Schooler, Gerhard, & Loftus, 1986; Slusher & Anderson,
1987; Voss, Vesonder, Post, & Ney, 1987; Zaragoza & Koshmider, 1989), including applications of ideas about source monitoring to a variety of phenomena (e.g., eyewitness memory,
persuasion, amnesia, and aging) and the development of special mathematical techniques for analyzing source-monitoring
data (Batchelder & Riefer, 1990). Such activity reflects a growing appreciation of the central role that source monitoring plays
in cognition. Source monitoring and failures in source monitoring constantly color one's memory for events and influence the
development and expression of knowledge and beliefs. In the
present article, we describe an integrative theoretical framework for exploring the critical cognitive function of source
monitoring and discuss recent empirical work in relation to
this framework.
We begin the first section by presenting the basic framework
and then reviewing some available evidence for it. In the second
section, we examine the relation between source monitoring
and a variety of phenomena-such
as recognition memory,
eyewitness testimony, and the incorporation of fiction into factual knowledge-and
illustrate how the source-monitoring
framework points to relations among such diverse phenomena.
In the third section, we examine the consequences of disruption in source-monitoring processes and discuss the brain regions that are implicated. In the fourth section, we briefly discuss how issues of source monitoring might fit within a more
general processing architecture of cognition and memor~
A Source-Monitoring
Framework
We build on previously presented ideas about source monitoring (Hashtroudi et al., 1989; Johnson, 1988a, 1991a, 1991b;
Johnson & Foley, 1984; Lindsay & Johnson, 1987; Lindsay et al.,
1991). A gen~ral o~erview is follo~ed by a discussion of evidence for major pOInts represented m the framework.
Overview
The source-monitoring framework is an extension of the reality-monitoring
framework proposed by Johnson and Raye
(1981). ~~~itoring
refers to discriminatinl! memories Qf
internally
generated
information
from
llVIIl memories
IIlCmUl:leS of
01 externall
ex~ernallY
d"'L~'. """.1
d . 6"1I",a."ut IlllVllllaLIVl1
h
d . t..
h.."
derived informatIon.
such as rlistinguishing
memories for
.1:
L.thougnts
--~
ana.
--;...monitoring)
~magmatlons.
from
memones.
for.
~erc~lved
tonn~
of
~Il ty~es
combination
~s based
on characteristics
of
memorle~
and temporal), semantic d~il, ~ffective information (e.g.,emotional reactions), and cognitive operations (e.g., records of organizing, elaborating, retrieving, ;nd identifying) that were established when the memory was formed!
~u~ce-monitoring
de~isions capitalize on average differences in characteristics of memories from various sources. For
example, compared with memories for imagined events, memories for perceived events tend to include more perceptual, spatial and temporal, semantic, and affective information and less
information about cognitive operations. Consequently, a memory with, say,a great deal of visual and spatial detail and very
little cognitive operations would be judged to have been externally derived. Decisions may also be made on the basis of a
~hbetween
the Qualities of memories and activated sche.}!!as that recresent carticular sources. For example, if the auditory quality in a memory of a statement matches your idea of
(or schema for) Sam's voice, you attribute the statement to Sam.
~
s~~rce-mo!!itoring decisions are made racidly and relatiyely nondeliberativel
on the basis of ualitative character istiCS 0 activated memories (e.g., amount or type of perceptual
detail). That is, often we Identify the sources of our memories
in the course of remembering them, without any awareness of
decision-making processes. Sometimes, however, source monit~g
involves more strategic crocesses. Such decisions tend to
be~ower a~d mo~ deliberate and involve retrieval of supPQrti~e.mor!es,
~?Jing o~ ~iscovering relations, and initiation of
~~e.g.,
"Does this seem plausible, given other things
that I know?"). Thus, you might correctly attribute a memory of
a conversation to imagination on the basis of the knowledge
that you are not acquainted with that person. Or you might
attribute a statement to a particular speaker on the basis of
general knowledge about them (e.g., "Sam was the only person
there who would have said this sort of thing, so he must have
said it").
m
,
.
.
.
f
.
generatedasopposed to externally derived origin ofsome information.
According to this definition, discriminating what one said aloud from
what one heard would be classified as reality monitoring, as would
discriminating what one only thought from what one heard. The second wayemphasizes the covert or nonpublic quality of mental events
such as imagination ~r inferen.ceas o~~sed ~o t.he.pu~lic Quality of
.other.events. Accor~mg to this definltl~n, dlscrlmma~mg what one
Imagined oneself doIng from whatone did (here called Internal source
...'.'.
would be an Instance
of reality monItorIng.
definitions
of reality monitoring
is more obviously
the other. The issue is whether in any particular investigation or theoretical analysis,one is emphasizing the self versus external source as
origin of information\or the actual (public) versus imaginal (private)
status of the information. Our more genel:al point is that there are
various types of source discrimination problems and that investigating
a varie.ty of them ca~ en:ich the unde~tan~ing of mechanis~s t~at
unde~llesource mo~ltorm~. The classification of source:momtormg
sltuatlons.use.dhere IS?ot Intended to be final or exhau~tlve.
Affective Information was added to the model later m responseto
I
..
Has ht TOUd.I, J0 hnson, & Chrosma. k ,
.I.
severa emplrlca I fi n dmgs
,e.g.,
1990; Suengas
& Johnson,
1988).
For a description
\
Neither
of
correct than
of this
I
memory
with judgment processes. ~mong the most !m-
characteristic, see Hashtroudi et aI.,1990.Note that thesecategoriesof
~t~oryc~ara~ter!st~csare
records.ofperce?tual inf~!.:
matlon (e.g., sound and color), contextual information (spatial
memory characteristics are analytically useful for what are probably
"fuzzy set" relations.
-
,
I Reality monitoring can be defined in two somewhatdifferent ways
(and we have used both). One way emphasizes the self, or internally
these
events. In addition to these Internal-external discrImInatIons,
two other source-monitoring situations are of particular inter.est: (a) discriminating
between externally derived sources (external source monitoring), for example, d-iscriminatlngmemories of statements that were made by person A from those made
by person B, and (b) discriminating between internally generated sources (internal source monitoring), for ex~ple, dls~riffil-nating memories of what one thought from memories of what
.I Accordmg
..2 to the present framework, source mollione said.
AND S. LINDSAY
!
"--
..-'-~
SOURCEMONITORING
t
;
t
Different conceptscould be invoked to characterize the nature of these two types of judgment processes.The first has
been called relatively "automatic,~ and the second has been
calle~ more "controlled~ (Hasher & Zacks, 1979; Norman &
Shalllce, 1986; Posner & Snyder,1975; Shiffrin & Schneider,
1977)or "analytic" (Jacoby& Dallas, 1981). The first also has
been called more "heuristic," and the second has been called
more "sy~te~atic" (e.g.,.Chaiken,Lieberm~n, & Eagly,1989).
The heurIstic-systematic contrast as descrIbed by Chaiken et
al. comes closestto ~hecontrast that Johnsonand Raye (1981)
proposed between Judgments that are based on qualitative
char~cteristic~of activated information (e.g.,amount of perceptual Information or match to a schema or template)and judgments that are basedon more extended reasoning(e.g.,retrieving additional information or discovering inconsistencies between what is remembered and what is otherwise known).
Chaiken et al. (1989) provided a multifaceted conceptualization of the cognition involved in evaluating information in a
5
mensions that might be used in any given decision(e.g.,weighting perceptual information as more important than affective
information), assigning confidence to different levels of this
weighted information, and assigning particular overt responses
(e.g.,yes in an old-new recognition task) to specific levels of
confidence. For example, one might require a substantial
amount of perceptual detail beforeaccepting an experienceasa
memory of an actual recenteventand require lessto accept it as
a memory of an actual long-agoevent. Becausesourcemonitoring depends on ongoing goals or agendas,it should be affected
by the sorts of motivational and social factors that influence
any goal-directed activit~ In particular, source monitoring will
be done more carefull~ with more stringent criteria, under
somecircumstances than under others. For example,we expect
people to be more careful about the origin of information when
they are testifying in court than when they are recounting
eventson a social occasion. What "more careful" means is using both heuristic and s stematic -=udmen roc sse- -
persuasioncontext:
t an one type alone and ti htenin he criteria used.
ource is not an either-or concept. Rather, source can be
[Systematic
processingis an] analyticorientationin which perspecified to differing degrees.For example, you may rememberceiversacc:ssand scrutinize.all.informationalinput.for its relethat Mary told you a fact and-whenand whereshe told you. Or
vanceand Importanceto their judgmenttask,and Integrateall
..
usefulinformationin forming their judgments.. ..When proyou may only remember that Mary c?nveyedthe Information,
cessingheuristically,peoplefocusonthatsubsetofavailable
inforbut not where,or when, or how (e.g.,m person, by letter, or by
mation that enablesthem to use simpleinferentialrules,sche- phone).Or you may remember that somebodytold you the fact
mat.a:or cognitiveheuristicsto form'.1late
their.jud.gmentsand
sometime recently but haveno idea who.Or you may remember
decIsions....~lth°u.gh. ..systematic proce~sl~g[ls)
gen~ral~y virtually no information about source as when you recognize
controlledand Intentional,the statusof heuristic processIngIS
..'.
.
lessclear. ..perceiverssometimes
useheuristicsinahighlydelibsomeone only as familiar but h~ve no Idea when or where yo~
erate,self-conscious
fashion,butatothertimestheymayuseheurmet that personbefore. Accordmg to the current source-mornisticsmorespontaneously,
with relativelylittle awareness
of havtoring framework source attn utlons ar
a e to differing
ingdoneso.(pp. 212-213)
egreesof specificity, with differing degreesof confidence,depending on so
theforth.
information available: criteria used, task- deIn the s~urce-monitoring fra~ework, both heuristic and sy~- -§as,-and
~matic Erocessesrequire setting crite~ia for ~a!inga ju~gment and proceduresfor comparin activated information with
~~
or example,heuristic judgments involve criteria such
as "if the familiarity level is above X, the event probably happened," or "if the amount of perceptual detail exceedsX, the
eventwasprobablyperceived.~Criteriaforsystematicprocesses
might include, for example, limits on the degree of inconsistency betweenthe known and rememberedthat will be acceptable.
~ce
judr.ments are ~y.pical~ made.!!eu~i~ti~all~_;
sy~!~m-=~roc~sses
areeng~ed lessQftenand te!!Q.to ~~slowe! an.9
~~e subjectto disruption. Both can provide important checks
Becausesourcemonitoring dependson the information available from activated memor records it r I' fundam
C
t e quality of the information recorded about events initiall .c"l
' .
~sememory records are the pr~uct of the specific per~
""'i'..iJ~.,:
tual and reflective processesengageddurin the initial ex eri- br'~ ~ \~.,~
ence 0 nson,
; 0 nson & Hirst, 1991; Johnson & Mul- J ",'" ~
thaup, 1992). ~~hing that preventsa pers~~ from fully con- Je.&-t".4;~"
t~tualizing infor~ationat acqu~sj!iQ!L(i.e.,
creatinga~ "event~) .~ ./,~
will reduce encodIng of potenttally relevant source mforma- re c"
~.
~. ror example,stressor. divided attention ~ay ~isruPt ?or- {~c.t.,<. ..{~"
mal perceptual and reflective processes,resultIng m relatively I (e. cO(,.4«impoverished encoded information from which source could 6\..;(.1.:~~
on each other. For example, systematic processescan, on the
basis of implausibility, provide a challenge to memories that,
say,passeda heuristic check for reality monitoring becauseof
high vividness. Conversely,heuristic processescan challenge
recolle~tions(onb~sessuchaslackofsensorYdetail)th~twoul?
otherwise be readily accepted because they fitted with ones
generalknowledgeand beliefs (Johnson,1988a,199la).
~!~ ~euris!ic ~nd systematic process~sinvolve a range of
types of criteria and can be influenced b biases,metamemor
~mDtions, and current r.oals~nd agendas.For example,.you
might identify a vague recollection as a memory for a prevIous
imagination ifyo~ believed that you would reme.mber~he~v~nt
more clearly if it had actually happened. SettIng crIterIa mcludes a number of important aspects:assigningweights to di-
be later derived. Similarly, any factors that reduce the likeli- ~I.Jo-SIOe.'C~
hood that an event will become embedded in other events ,<. .fe~{r
should reducethe amount of potentially usefulsourceinforma- .../
tion. For example,imagine that at a cocktail party you overhear ~ ~~'
a remark, but the c~nvers~tion~n which you are engaged pre- ~\<;~\~~~
vents you from turnIng to Identtfy the speakerand remember- c.(!..~\\.., ~..{t,1
ing a related statement made by them earlier, and fro~ co~- rS?-c.~ ~
sidering the implications of this particular person makIng this \ ~~~ ~
remark at a cocktail party. The statementhas not beencontex~.J;.
tualized. ~nthe extreme ca:e, the ,?ccasionof,overhearing ~his &~<.v
~~,
remark will not become an event, and you will later havelittle ~\~
~
source-spec~fyinginformation!or the i~e~expressed:Nonethe- ..'~ j~ ~
less, you might later find the Idea familiar or have It come to e-s\ o'¥-=
mind relatively easily as a consequenceof this experience. (.v""-.;\';:V
~"v
()
uo(Il'-
~~
~«'3-<':' .
:"),'
l~~~~A+t
I" 1G~"- ~\J<""?e,~~.~
~~
~ \;;,.<" ~..~..;},&-
~ ~0 "" '"' ~,\..-...<:.L,,-'
6
M, JOHNSON, S. HASHTROUDI, AND S. LINDSAY
Becausesource monitoring depends not only on the quality
of the information as encoded but also on the quality of the
~clslon processes.~he~ source~~onitoringjudgmen~ are
made,anything limiting these decision processesat test should
~so disruPt source monitoring. For example, time pressure,
severestress, distraction, or alcohol should decreasepeople's
b o l O,
.
fl .,
I I
I ' I
a I Itles to engageIn re ectlve processes,partlcu ar y re atlve y
deliberate processes(e.g.,retrieving related information). Evid~ce for the dependenceof source monitoring on the quality
of the information available and the quality of the decision
processesapplied is discussed in the following sections.
The presentframework is in the spirit ofTulving's (1983)call
for work on recollective experience: "In theories of episodic
memory, recollective experience should be the ultimate object
of interest, the central aspectof remembering that is to be explained and understood" (p. 184). He lamented the fact that
researchershave"evadedproblemsentailed in recollectiveexperience" (p. 185). Tulving suggestedthat recollective experience
may range from clearand preciseto vagueand fuzzy. He speculatedthat"thesubjectivefeelingsofpastnessandveridicalityof
memory must be provided by intrinsic properties of ecphoric
information" (p. 187). (Ecphory is a hypothetical process by
which cues combine with trace information to yield a combined product called ecphoric irlformation.)Tulving also noted
that considering recollective experience raises the issueof the
distinction between memory and decision processes(which in
his General Abstract ProcessingSystem[GAPS] framework are
part of. th,emore g.eneralconversionprocessesthat operate on
ecphorlc information).
Thus, Tulving has emphasized the i.<:!~~
that information is
~t simpl~ acc~sed,. bu~ co.mbinesw~th~ue informa~ion ~ a~~!nt product that is the input to further processinR.The
source-monitoring framework takes this notion as axiomatic
and further attempts to characterize the qualitative nature of
this ecphoric product and the factors that enter into its conversion into a feeling of pastness,veridicality, and belief in source.
As the following review illustrates, there is lesscauseto lament
than Tulving once had; there is a growing body of researchon
which to draw in understanding recollection.
In summary,according to the source-monitoring framework,
there are at least three important types of source monitoring:
external source monitoring, internal source monitoring, and
internal-external reality monitoring. In all three situations
there are multiple cues to source. We categorized thesecuesas
-sensory/perceptual
information, contextual (spatial and tem~
p~ral) information, semantic de~ail,a~ect, and cognitive operatlons. The easeand accuracywith which the sourceof a memory is identified is determined by severalfactors: (a)the type
.and the amount of these memory characteristics included in
activated memory records (or in the ecphoric information), (b)
howunique thesecharacteristicsare forgiven sources(the more
similar the memory characteristics from two or more sources,
the more difficult it will be to specify the source correctly), and
(c)the efficacyof the judgment processesby which sourcedecisions are made and nature of the criteria used. These attribution processesvary in the extent to which they might be characterized aslessdeliberative (heuristic) or more deliberative (systematic). In general,source-monitoring attributions should be
relativelyeasyand accurate whenthe event memory in question
\:..
is richly detailed, its attributes are uniquely characteristic of its
source,and appropriate decisionprocessesand criteria are used
during remembering. In the next section, we examine evidence
for the approach to source monitoring that we haveoutlined.
E .d
VI ence fior
the Baslc
.1:1
rramewo~
k
Role of memory characteristics in source monitoring.
Memories for perceived events include more perceptual and
.§te~~al
info!mation than m_e!!!oriesfor imagineg e'len~s,
and differences in the amount of these memory characteristics
ma be use
is for realit -monitor in decisions Ha~
troudi, Johnson,& Chrosniak, 1990; Johnson, Foley,Suengas,
& Raye,1988; Johnson, Raye, Foley,& Kim, 1982; Schooler et
aI" 1986; Suengas& Johnson, 1988). For example, Johnson,
Foley,Suengas,and Raye (1988)asked subjects to remember
past actual eventsand pastdreams or fantasiesand to rate them
on a number of memory characteristics. In a separate study,
different subjectsindicated how they knew that a remembered
autobiographical event was real or imagined. Subjects in the
first study rated real eventsas having clearer temporal and spatial information and greater perceptual detail. Subjects in the
second study offered such information asevidence that particular remembered events were real (e.g.,"[I know it really happened because] I can remember what the dentist's office
looked like.").
In addition, confusion between memories of perceived and
~e~
inform~!ion Increaseswith decre~se~in the InIo[ma:tlon about co nitive 0 erations characteristicall associated
with imagination (.ourso & Johnson, 1980; Finke, Johnson,&
ShYi,1988; Foley et al., 1991; Intraub & Hoffman, 1992; Johnson, Finke, Danzer,& Shyi, cited in Johnson, 1991b; Johnson,
Raye, Foley,& Foley,1981; Rabinowitz, 1989). In a study by
Finke et al. (1988),for example,subjects'ratings indicated that
it was easierto imagine half of a form as complete when the
form was symmetrical about the vertical axis than when it was
symmetrical about the horizontal axis, suggesting that completing forms about the vertical axis requires fewer cognitive
operations than completing forms about the horizontal axis, In
a second reality-monitoring study, subjects at test indicated
which ofa set of whole forms had earlier been shown in complete or incomplete versions. In relation to controls, subjects
who had imagined vertical forms as complete had more difficulty in reality monitoring than did subjectswho had imagined
horizontal forms as complete. This outcome is consistent with
~he i~ea that ~~corQsofc~gnit~e operat!Qns can be used to
~ntifv oneselfas the oril!in of!! memor~
There is also evidence that £-onfusionis increased by percep~~l similaritv hetween memorie~ from external ann intprnal
w~(Johnson,
Foley;& Leach, 1988; Johnson, Raye, Wang,
& Taylor, 1979) or between two external sources (Ferguson,
Hashtroudi, & Johnson,1992; Lindsay et al., 1991). In an experiment by Johnson, Foley,and Leach (1988), subjects who had
imagined themselvessaying some words and had heard a confederatesayother words werequite good at later discriminating
the words that they had imagined saying from the words that
the confederate had actually said. If, however, subjects were
askedto imagine hearing the words in the confederate'svoice,
they had much more difficulty discriminating what they had
~
J
SOURCE
heard
from
what
semantic
they
had
similarity
(Johnson
et al., 1981)
1990;
Lindsay
found
that
made
by one
1991)
were
speaker
the same
Likewise,
external
or between
et al.,
subjects
described
imagined.
between
two
more
events
than
when
the
they
ent eve~ts.
~he
of
tonng
IS also
errors.
For example,
that
they
had
about
that
they
seen
in the
were
much
event
read
about
an event
speakers
differ-
that
teristi~s
&
often
gives
source
of.me~or~
sItuation
solidating
and
qualitative
event
text
Belli,
with
rise
Role
to err
by saying
they
than
that
(1992)
had
to err
they
actually
McCarthy,
related
to give
reading
to the
rise
&
monitoring
is based
records,.it
is especi~lly
integrative
this,
event,
to imagined
to be
brain
various
of decision
processe.s.
below
aging,
im~rta~ce
the
various
here.
that
divided
characteristics
attenof events
monitoring.
that source
or attributio-n-processes
monitoring
1978;
& Johnson,
example,
the
1989;
more
times
comes
Raye,
Johnson,
subjects
had
from
generated
~
~~
" ""
during
a study
phase,
the more
times
they later
theY had seen it (Raye et al., 1980).
However,
when
cr"
between
a limit
to their
0 and
generated
memories
are
made
the
idea
-vf3'i
--,~\
.~t
when
tend
action
effect"
items
with
are
in the
falsely
performed
during
falsely
than
the
(Jacoby;
full
&
time
to their
Raye,
monitoring
of
recognized
also
to sources
through
graphical
and
support
both
more
from
memories
the "it-had-to-be-
action,
Source
a probe
from
sIgnal
attention
The
at the time
1989;
retrieval
Foley,
Kelley
hypothesis
evaluation
of source
to res.pond
of test
& Lind-
that
source
of characteristics
and
reasoning
tions
with
various
Eviden~e
fron:z
situations
but
monitoring
Suengas,
of autobio&
Raye,
1988),
in
that
internal
under
in
of group
Reardon,
&
J
Jolly,
demonstrate
a variety
deficit
should
the
of
source-
not be taken
tasks
of
more
sources
pattern
,I
than
of
memories
important
across
in others
imagine
yourself
6-year-olds
doing
and
monitoring),
of
pe~s?n
they
mlslde~tlfy
your
had
whIch
of
of
wIth
Foley
words
memories
et al. ~1983;
that
children
such
see also
had
as was
or actions
from
pressed
(referred
to
suggest
that
finding
this
et al.,
F~ley,
spe~lal
those
as realization
may
for
that
what
they
anmore
~rformed
monItorIng
of spo.
~
S?pasakls,
1n.In~ernal
dIff~rent~atlng
a more
1989)
source
expressed
Imagined
judgments).
reflect
of
I ?8.3).
were
of
(internal
ha~
dlffic~lty
requIred
wIth
to confuse
nor did
~antlm,
!
y?u
memories
.doing
but
Recent
general
';
i
memoor did
lI~ely
conf~der~tes
(Foley
Foley
Lindsayet
Compared
actually
particularactlons(externals.ourcemomtonng).~sl~llardevelopmental
pattern
was obtained
for source
imagined
j
.yo~ng
1987;
nose,
to confuse
monitoring),
two
f1act,
investigated
your
n.ot more
done
-In
19~3;
nose?").
likely
were
(reality
,
monitoring.
& Gutch,
touch
memories
don~
~
source-monItorIng
Johnson
really
they
!
trace but
rat
processes,
source-
Foleyet~l.,
and
far more
yut
what
h~d
Aman,
1985;
touching
were
;;
"
of source
In some
(Foley,
you
exter-
of memories
are not
properly
by popula-
studies
Foley
(e.g., "Did
in-
two
source-monitoring
by a memor
attribution
as adults
& Johnson,
cognitive
deficits.
rovlded
complex
For example,
just
ken and
(e.g.,
in distinguishing
which
aspects
or weighted
developm.ental
as well
How-
differentially
for distinguishing
deficits
~
entirely
processes.
tasks
source-monitoring
not
'1
1
as evi-
involve
or judgment
source-monitoring
us identify
reactivated,
Foley
for actions
thoughts
(Foley
was disrupted
IIhiliti~~ ~hrullrl
ti~v~lnn
w;th ~e.
liP
~--~
--:'
~-
ries
noted
but
internal
instances
1991)
.
Y
other
found
difficulty
source-monitoring
perform
ofte~
in
1
~{If"'"
"]
If source
IS not slm I
is a result of relatively
other
subjects
in
In
(1985)
other
Delys,
characteristics
can help
encoded,
source
&
{...Ii
in iln-
monitoring
(Durso,
characteristics
the
monitoring
re-
Shore,
of selective
tasks
being
adults,
and
~t~
impaired
difficulty
performance
,~
~ne.\
monitoring
have
monitoring
external
sources),
memories
also
their
These
are ordinarily
imagining
to an-
~onitoring
and
& Kelle~
1991).
items
with the cues available
Kahan,
& Raye, 1984),
heuristic
(Johnson,
example,
own
systematic
analyses
deci-
new
to divided
Lane,
For
support
source
different
et al., 1991;
that
recognized
1981).
Woloshyn,
&
ihvolves
memories
ceives
Jacoby,
Zaragoza
on
chIldren
sug-
attributions
remembering.
between
as opposed
1991;
say, 1993;
reduced,
inappropriate
that
as different
~~.n.ltonng~tI~s
internally
but
investiga-
1989).
Harvey
\
\"
et al.,
et al., 1983).
not
were
1
clas~Lof
source
source
monitoring.
memory
mnnitnrinp
number'
out
source
(Johnson,
Kounios,
& Reeder,
1992),
wh ile making
test judgments
(Johnson,
with
that
wer ~' 7
was
gate
are attributed
to attribute
rather
item
adults
particular
did
idea
performance
in external
manics
different
al., 1991).
biases
(Johnson
' improves
and
~
new
an
between
events
selectively
1980).
thought
subjects
to use
nal
criterion.
memories
processes
you
a stricter
asked
of confusion
presented
could
of systematic
that
subjects
other's
externally
with
by being
amount
subjects
Findings
~6.'
10, the
and
that
judgment
st~es
& Taylor,
Foley
patterns
insofar
draw
not
~f co.mparing
different
:1
inter-
sItuations.
that
ever,
poor
external
had
but
The
performance
(Hashtroudi
Reardon,
monItorIng
ternal
~\O(t~
Lindsay
Durso,
dence
stud-
and
systematic
poor
older
and
children
1985;
operations
Evidence
monitoring
in
These
In keep-
indicates
and
for example,
monitoring
source
that
of con-
the
mean
monitoring
1985;
and
of these
monitorin
deserve
necessarIly
study,
Johnson,
source
some
distin2uishahle
that
not
young
external
that source
monitoring
depends
on the criteria
that
adopt
(Dodson
& Johnson,
in press; Hasher
& Griffin,
sion
&
Foley,
on amnesia
monitoring
source
by evide?ce
monitoring
charac-
kind
produces
presented
qualitative
in source
In the
discussed
damage,
in deficits
~mporta.nt
that
of memories
from
on qualitative
to engaging
processing
evidence
(e.g.,
gesting
constitut~
reality
studies,
1985;
data
of source
.mo~itoring
thought-disordered
snowing
su ~ects
e;\
types
external
source
in
source
.C,~
(e.
(-. given
among
does
"ii1ieriial
Johnson,
We discuss
monitorin
In one
not
by saying
had
Presumably,
imagery
that
actually
decision
'
l
1991a).
&
clinical
,
I.S s~~ported
other.
McCarthy
involves
J~
from
that
t~on
,
a.;~For
(Foley
and
belo\J.:
realit
and
1980).
to
IS conducIve
in encoding
results
("
that
~ltuatlon
Windschitl,
is less likely
characteristics
disruption
~
ies in detail
In source-monItorIng
that
the
Johnson,
an event
encoding
tion)
1983),
(Johnson,
differences
Given
studies
Raye,
~urce
reading.
ing
&
Distinguishing
had
for source.mo.m-
likely
in the
(see also
Durso
viewing
more
describing
event
Gales,
something
something
visual
1992;
whereas
LIndsay;
in a visual
in a text
had
charac~erl~tlcs
~y ~symmetnes
BellI,
subjects
read
memory
revealed
seen
Winfrey,
et al.
had described
developmental
...:ourc.e-momtonngproblem~
~mportance
noted
from
Johnson,
statements
two
7
confabulation
Lindsay
to misattribute
when
the
sources
sources(Lindsa~
confusion.
likely
another
internal
external
increases
to
increasing
and
MONITOi~iNG
not ~xstudies
problem
.~
']
'1
1
8
M. JOHNSON,
that
young
sources.
children
Lindsay
have
in coping
et al. (1991) found,
with
S. HASHTROUDI,
similarity
across a variety
among
AND
Hirst,
of source-
S. LINDSAY
Phelps,
Shimamura
Multhaup,
& Squire,
& Volpe, 1992; Schacter
1991).
Recognition
et al., 1984;
scores can be near
monitoring
tasks, that a~ differences
in source monjtori~g
tended to be greater when (be to::be-discnmlnated
sources
perfect while source-monitoring
performance
is at chance (Kahan & Johnson, 1990). Manipulations
that affect source-moni-
were highly
toring
similar.
For example,
8-year-old
children
also had
performance
may have no effect on old-new
recognition
difficulty
discriminating
actions that they imagined
another
person doing from actions that they saw that person do (see also
(e.g., Johnson & Raye, 1981; Lindsay et al., 1991), and manipulations that improve ~ecognition
may impair source monitoring
Markham,
(Lindsay
become
ful
1991). One idea to be explored
more similar
on one dimension,
source-rel.evant
Children
information
may have difficulty
is that as sources
other potentially
becomes
managing
more
multiple
use-
important.
two
cues to source
& Johnson,
Our claim
fundamentally
different
t~~~§~lli~r
sources, recognition
more
msource
likely
to require
reasoning,
more deliberative
the types of processes
veil, 1985).
Earlier
sZudies
of source
focus on relatively
recent
have been investigated
temporal
monitoring.
work,
aspects of source
& Malmi,
retrieval
later (Fla-
Although
for many years in a variety
(Underwood
1971) discrimination,
or extended
that develop
here
monitoring
S"fiIdied items
istic and systematic
Block,
list;
memories
cf. J. R. Anderson
&
of
Bower,
as drawing
toring
on the same family
ofheur-
are not different
in recognition
at a general
and source moni-
level. But across situations
Summers, 1973; Winograd,
1968), intrusions
in recall or false
recognition
of associates of studied words (e.g., Cramer, 1965;
both vary in the range and type of information
used and in the
specific decision processes applied; in a particular
experiment,
Deese, 1959; Underwood,
1965) or of tacit implications
of sentences (e.g., Bransford & Franks, 1971; Corbett & Dosher, 1978;
the kinds of memory information
and decision processes
to discriminate
between old and new items (recognition)
Johnson, Bransford, & Solomon, 1973), intrusions
of schemabased knowledge (see Alba & Hasher, 1983, for a review and
be the same as or different
from those used to discriminate
among old items from various sources (source monitoring).
In
Hirt,
1990, and M. Ross, 1989, for recent treatments),
for surface details
(e.g., Craik & Kirsner,
1983; Geiselman
& Belleza,
memory
the following
1974; Fisher & Cuervo,
1976, 1977; Hintzman,
Block,
&
paragraphs,
ences in heuristic
sions, then discuss
Bjork,
nition
attempting
to understand
interference
was of interest
effects (e.g., Abra,
in
1972;
should
situation,
the similarities
for recognition
the roles of systematic
two types of decisions.
The observed relation
list differentiation
we explore
processes
Inskeep, 1972; Kolers, 1976), and context reinstatement
effects
(e.g., Eich, 1980; Godden & Baddeley, 1975; Smith, Glenberg, &
1978). For example,
depend
to which,
they use the same or different
poral
could
panel of Figure lA indicates
that a complex
creates a complex pattern of initial activation,
fit easily within
the source-monitoring
framework.
tion attributable
to memory
give rise to different
Relation
to Other
Memo
In this section, we examine
itoring
the relation
and several different
how these diverse
monitoring
ry
Phenomena
memory
phenomena
fit within
and recog-
in a particular
as input.
differences among tasks in
I. The shading in the first
records.
stimulus
including
usually
activa-
stimuli
would
Different
levels and kinds
of initial
activation.
The
second an~ t~ird panels in Figure lA.sh~w that over the course
of a few milliseconds
or seconds, activation
may become more
between source mon-
phenomena
potential
in Figure
deci-
in these
information
One way of characterizing
the input used is shown
findings
and differ-
between source monitoring
on the extent
and describe
the general
source-
framework.
differentiated,
that
is, may yield
memory.
In connectionist
mutually
support
each other's
settle into groupings
more
terms,
certain
activation
(indicated
specific
attributes
activated
units
of
might
and, hence, cohere
by squares,
circles,
or
and trian-
From the beginning
of our work on source monitoring,
we
have emphasized
that source identification
and old-new recog-
gles in the second and third panels) that correspond
to memory
-characteristics
such as perceptual,contextual,andsemanticde
.1 I
d ' ffi
..
d
tal.
ncreaslng
I erentlatlon
oes not re.er to an Increase 0 f a
single type of information
(e.g., strength or activation level). A
manipulation
might, for example, increase undifferentiated
ac-
nition
tivation
S
II
.to '
l".lOm
nng an
ource
-AT
may draw on different
different
may be equated
ing different
age groups
Lindsay
Valdiserri,
W
Reco
,a nz
0
'
on the specifics
& Raye, 1981; Raye, 1976).
on old-new
ences in source-monito~ing
1985;
e
t on
'
.
z
aspects of memories
processes, depending
(e.g., Johnson
groups
dO [lV4
1d
and involve
of the two tasks
For example,
re.cognition
subject
yet show differ-
accuracy, as in the case of compar(Ferguson
1991) or amnesics
and
Kaszniak,
normal
Kihlstrom,
controls
&
(Johnson,
"
without
increasing
characteristics
and activation
low levels
might
et al., 1992; Foley & Johnson,
et al., 1991; Schacter,
I
used
may
and source
processes
Winograd,
1968). Studies of list differentiation
demonstrated
the importance
of factors such as semantic similarity
and temseparation of two lists in determining
whether subjects
correctly identify the list from which a word came. These
I
processes.
Thus, the processes involved
&
~
1974). Both source monitoring
and old-new recognition
can be
described within the framework
we use for source monitoring,
and both can be viewed
(Rothkopf,
are
and old items on recognition
(i.e., subjectS must attribute
to the-study
monitoring
(Hintzman,
and recognition
On the contrary,
~o the subi~
from ex!raexDeri;n~~ta!
judgm~nts
typically
involve some de~e
of contexts:
1978) and spatial
list differentiation
we
monitoring
processes.
cause both the new (nonstudied)
simultaneously
(see also Ferguson et al., 1992). Another possibility is that more difficult
source-monitoring
problems
may be
and
1991).
is not that source
differentiation.
may revive
of differentiation,
whereas
Figure
IB represents
activation
tion of degraded
stimuli
might
rates,
at relatively
from
others
very rapidly:
the idea that
average differentiation
memory
at different
might asymptote
reach high levels of differentiation
have different
Different
and differentiate
from some stimuli
.
different
memory
requirements.
benefit
from
tasks
Identifica-
activation
of rec-
7
\
G
i
i
SOURCE MONITORING
I
I
I
9
i
the source of perceived items. Assuming that memories for
.',
" .:. ,.:::~';'~:::~
I
'
,
: ".
:~~f:~:f'~"~::'\~':'
:.~~~;'~:~,.~j:::~~;~~::~:':',
'.' ."'}-"" .."-'~,,-,, , ..mation
..'.
.!'~,~~~~~;t'-*:~~';:
,,' "':,:;:~:;;":-,Y'"
1A.
!
,
I
'
c
.g
.~
~
~
E
0
::,.:
'.
imagined items include more information about cognitive
--.operations
and memories for perceiveditems include more perceptual detail, the pattern suggeststhat in this situation, inforabout cognitive operations is more salient or revives
more quickly than ~oes In!ormati~n about perceptual detail,
We are currently using thIs techmque to explore the rate at
which different memory attributes become differentiated.
It is impOrtant not to lose sight of the fact that both source
'.
and old-new decision processescan be flexibly directed to
various points in the revival functions relating various memory
characteristics to time. As illustrated by the overlap in the circles in Figure IB, in any particular situation, the information
used for.a recognition responsecould be quite different from
(e.g.,noting perceptual fluency)or quite similar to (e.g.,noting
~erce~tual detail) that us~d in source monitoring for the same
situation. In fact, corr~lations betweensource monitoring and
~co~ni!ion ~ary considerably from situation to situation (e.g.,
see Craik, 1989),~ wo~ld be exDectedif the information pach
Source
Monitoring
ad/New
Recognition
Priming
~~
on w~s not fixed but was somewhatflexible and overlaD-
ping.
Time
18.
--m-summary, heuristic source-monitoring processessetdecision criteria to evaluatethe amount and nature of various types
, memory
J
(A) Actlvato bec
..mcreasmgy
dI ffieren t.latoo
"'
. Id
ing
r, gu
,~.
characteristics
I n (e.g.,
omes
perceptual
or Icontextual);
(B) .Yle
Differ-
of information in activated memory records (e.g.,criteria to
enttaskstypicallyrequiredifferentdegreesof differentiation.
lar sources).Heunstlc old;.;.newrecognition judgments can use
similar criteria and can use much less differentiated input to
evaluatefami liarity derived from such information asdegreeof
activation of individual representations (Mandler, 1991), fluency (Jacoby& Dallas, 1981), amount of associativespread of
activation (Gillund & Shiffrin, 1984), or "echo" intensity
(Hintzman, 1988),4Speed-accuracytrade-off is one promising
technique for mapping these attributional processesin more
detail.
The class of heuristic processesused in memory decisions
can, of course, serve functions other than recognition and
source-monitoring judgments; they assesspreferences,persuasiveness,similarity, and so forth. Clearly, heuristic decision
making is quite flexible in the type of information that is taken
as input and in the type of evaluations,attributions, and so on
'
rn
'
'
'
ords of prior exposure (priming) whether or not the activation
achieves higher levelsof differentiation} Old-new recognition
~~
r~uires more differenti"tinn th"n rlnpcpriming.a~<!
~ou!ce monitoring requires evenmore so.This is not to saythat
old-new decisions do not benefit from more differentiated information (as indicated by the overlap in circles designating
recognition and source-monitoring tasks in Figure IB) but simply that they can be made at lower differentiation levels than
1 can sourcedecisions. In general,t~ phenomenologicalexpe~
'~ ~ i .~c~ ~f con~ci2usrememberin2 is associatedwith hi2her leve_ls
£W\~.
."
of differentiation, which ive memories an eventlike qualit~
ecent work by Johnson, Kounios, and Reeder (1992)illustrates the point that source monitoring typically depends on
."
hIgher
...e
levels of d~fferentlatlon
than
does recognItIon.
~hey
used a response-signal,spe~d-accuracy trade-.off te.chmque
(Dosher,1984; Reed,1973; Wlckelgren, 1977)to investigatethe
time course of source monitoring and old-new recognition for
previously perceivedor imagined pictures, Acrossone seriesof
experiments,the signal to respond varied from 300 to 1,500ms
after the presentationof a test word. For each test word, subjects were required to indicate whether it corresponded to a
previously perceived picture, a previously imagined picture, or
a new.lt~m. Above ~hancesou~cemonlt?nng~ In th~scasedlffer~ntlatlng m~monesof p~rcelved from Imagl.n~dplctur~s, requlred more time than dId old-new recognItIon, consistent
with the idea that source monitoring typically requires more
complete or differentiated information than does recognition.
Perhaps even more interesting was that subjects were able to
identify the sourceof imagined items before they could identify
evaluate
degreer~cord
of match
the pr~pertles
qualitative of
characterIStiCSof athe
memor~
andbetween
the kno",:~
partlcu-
3Th
' .
concept 0fd 'I ffierentlatlon
specificity. Specificity is illustrated
"
. he d firom t ha t 0 f
can be d Istmguls
by the fact that priming is usually
better whenmodalityof input staysthesamethan whenit changes
(Kirsner& Smith,1974).Differentiationisa psychological
conceptto
expressvariationsin phenomenal
experience.
4Dual.processmodelsof old-new recognition(Atkinson& Juola,
1973;Jacoby& Dallas,1981;Mandler,1980,1991;seealso Gardiner,
1989)in~ludeboth a heuris~icand.syste~aticco~po~ent. H~v:e~er,
theytYPIcallytreatonly undIfferentIated
InformatIon(I.e.,famlilanty)
as a potentia! input to heuris~icrecognitionjudgments.The second
process,
retneval
or recollectIon,
tends
to be
asand
a search
or
retrievalwith
anall-or.none
outcome.
Both
thetreated
heuristic
systematiccomponentsof dual-process
theorieswouldneedto beelaborated
alongthe linesIhat wedescribeto accountfor sourcemonitoring,as
wellasto betterdescribeold-newrecognitiondecisions.The present
discussionsuggestswaysin whichthe notionsof familiarity and retrievalcanbeincorporatedinto amoreembellished
andgeneralframework for memoryattributions.
2:
10
M. JOHNSON,S. HASHTROUDI, AND S. LINDSAY
that is produced as output. In addition, the same information
may sometimesbe used for different purposes. For example,as
Jacoby has also argued (e.g.,Jacoby & Dallas, 1981),subjects
might use perceptual fluency not only as an index of familiarity, but. as an index for such diverse judgments as preference
(Kunst-Wilson & Zajonc, 1980; Seamon,.Brody,& Kauff, 1983),
darkness or lightness of stimuli (Mandler, Nakamura, & Van
Zandt, 1987), or easeof problem solution (Jacoby& Kelley,
1990; Kelley & Jacoby,1993).
Similarly, both recognition and source monitoring can draw
on more systematic processes,suchas retrieving additional information, discovering relatiops, initiating new searchstrategies, and so forth. Important questions include whether the
probability of using particular systematicprocesses(e.g.,plausibility checks)differs for old-new recognition and sourcemonitoring and how the criteria applied might differ. Like heuristic
processes,systematicprocessesare recruited for a wide rangeof
activities in a I Ion 0 rec gm Ion an source monItorIng,
such as memorizing text ntons- eterson
myt, 1 7)and
~~aluating persuasiveness~fcommunications (Chaiken et al.,
1989). Like heuristic cognition, systematic cognition is quite
flexibl~'ln the mtormatl~n usedasmputand the typesofevalu~
attr1l:ji:itio.ris,
.and so forth-~~t aregroducedasoUtPut.
Tne source-monitoring framework focuseson how these heuristic and systematiccognitive processesoperate in sourcemonitoring. (For further discussion about how a flexible cognitive
systemwith a limited setof components might be configured to
perform a wide rangeof tasks,seethe last section,SourceMonitoring Within a General Memory Architecture, and Johnson,
1991b; Johnson& Hirst, in press;Johnson& Multhaup, 1992.)
Source Monitoring and Indirect Tests of Memory
Dissociationsbetweendirect andindirect tests. Recently,considerable researcheffort has focused on dissociations between
direct (recall and recognition) and indirect (especiallypriming)
testsof memory (e.g.,Richardson-Klavehn & Bjork, 1988;Roediger, Weldon, ~ Challis, 1989; Sc~acter,I~~7a). Priming occurs when a prIor exposure to an Item facilItates subsequent
performance on that item in tasks that are not presented to
subjectsas memory tests-for example, when prior exposure
increasesthe probability that a word will be identified under
degradedstimulus conditions; the speedwith which a word can
be respondedto in a lexical decisiontask; or the likelihood that
it will be given as a response in free association,word stem
completion, or homophone spelling. Amnesic patients who are
impaired in recall and recognition show intact priming in a
variety of tasks (Schacter,1987a;Shimamura, 1986,1989;Tulving & Schacter,1990; Warrington & Weiskrantz, 1968,1970).
In normal subjects,priming may occur even when recognition
is at chancelevels(Eich, 1984).Certain experimental manipulations that improve recall and recognition (e.g.,elaborativeprocessing)are less likely to affect priming (Graf & Mandler, 1984;
Hashtroudi, Ferguson,Rappold, & Chrosniak, 1988; Jacoby&
Dallas, 1981; but see Rappold & Hashtroudi, 1991; Toth &
Hunt, 1990).Changesin modality of presentationdecreaseperformance on priming tests but have less effect on recall and
recognition (Graf, Shimamura, & Squire, 1985; Jacoby,1983;
Roediger& Blaxton, 1987a,1987b).
According to the source-monitoring framework, one reason
~diSSOciations betweendirect and~indirect tests arise is ~~useorimingtvDicallvreflectsfacilitatednrocessing(e.g.,facilitated perceptual processing)?Ian undifferentiated sort (see
Figure IB) a~~~need not en~agethe source-monitoring 12r9~essesthat we have described, whereas direct memory tests
~ays involve processesthat result in some soecification fif
source.Amnesic subjectsare not impaired in priming because
the~ary
memory deficit in these subjects is in reflective
processesthat are not required in priming but that establish
differentiated information critical for source monitoring (Johnson, 1990; Johnson& Hirst, 1991; see also the Disruption of
Source Monitoring section). Elaborative manipulations do not
'
a~t priming becauseelaborative_Erocessingcontextualiz~s coif c.J..r
information and becausedifferentiated contextual infOrmation
is not necessarym most priming tas s. hanges in modality
reduce priming becausethey ~duce-fhe chances that test
probes will activate previous records (Bransford, Franks,
Morris, & Stein, 1979; Kolers & Roeqiger,1984).
Indirect tests ofsourceinformation. Elsewherein this review,
we are primarily concerned with memory for source as a phenomenal experience (i.e.,the subjective experience of rememi
~eri~g source). However,evidence o.fsour~e-specific.informat
lIon m memory records can be obtained WIthoutsubjectsphe:
nomenally experiencing the source. Greater priming or
recognition(Craik& Kirsner, 1974; Kirsner, 1974)whenpresen- c{.
tation modality is constant betweenstudy and test provide indi- vq.r\ ~~
rect indexesof source information (seealso Kelley et al., 1989). fil'\t oJih~
The advantageof source constancy may occur even when subjects have difficulty in directly specifying the source(cf. Jacoby
& Brooks, 1984).
This d!ss~ciatio!!between direct and indirect testsof source
information can be described in terms of the framework used
here f~derstanding
source monitoring. Enhanced recogni~n !!!emory when modality is constant b~tweenstudv~§!
lCor:
can be mediated
relativel un ifferentiated inti rmation CI. <:w
e.g.,fluency in perceptual processing)that contributes input to
i
~ he~istic-recognition judgment .process.Di!~c~.d~~isio~
tl::
gardmg source, however, usuall mv lve an attrIbutIon that.s
I'
~~
on more differentiated information as inout to heuristic
;
~d, per~aps,more systematic orocesses.
!
Direct and indirect testsof memory will not alwaysbe disso!
ciated as both types of tests may be basedon the sameinformai
tion (seeFigure IB). For example, Kelley, Jacoby,and HollingsI
head's (1989)subjectsread some words and heard others. In a
i
subsequentcombined visual-perceptual identification and modality-judgment task, compared with items not perceptually
identified, identified items were more likely to be called ~read"
than ~heard" regardlessof whether the item had actually been
read, heard, or was new: Kelley et al. suggested that fluent
perceptual pro<;:essing
is experienced by subjectsas familiarity
and that more familiar items will seemto have beenpresented
in the same modality at testas at stud)! In a secondexperiment,
during the study phase,some subjects thought about positive
aspectsof read items and negative aspectsof heard items (or
vice versa).In the subsequentperceptual identification and modality-judgment tasks, the bias to respond ~read»to perceptually identified words was markedly reduced. Evidently,subjects in the elaboration group relied less on perceptual fluency
SOURCEMONITORING
f?r ma~ing ~odality j~dgme~ts becauseth~y ~ad an a~ternatlve.basIs f?r judgment. the d~fferentsema~tlc mformatlo~ as~oclated with r~~d an~ heardIt~ms: Thus, like source monltorI~g an? recognition, ~Irect and mdlrec~ testsma~ or may?ot be
?Issoclated,
dependIng on the type of InformatIon that ISused
In the two tests.
something that was presentedduring the experimental session
and, becausethey are using a familiarity criterion, indicate that
they remember seeing that item in the original event. On a
forced-choice test, the suggesteditem in a test pair may seem
more familiar than .rthe ori ginal item for a numbe 0 f reasons
(e.~., recency
I
11
or salience).
Thus,
s~nd~rd
testing
procedures
..wlt~
EyewItness TestImony
.memor
both yes-n~ and forced-choice recognition tests mav lead
s~cts t~ ~aseju~gments on familiari~~nd. hence:toignQre
information about the sour
.,..
.
.The regardIng
fact t~at subjects
susceptible
to misleading
suggestlons
recentlyare.
witnessed
events
is of considerable
practical and theoretical interest (e.g.,Loftus, 1979a;D. F: Ross
Read,~ ~~glia, in press~Wells & Loftus, 1984).The standard
(cf. Raye
th m~k e fiewererrors I'f~ey
.' 1976). SubjectsshouId .u.s
are orIented toward source-mon.rt.on.n.ud ments e.g.,Raye&
:~7~:~n, 1980) rather than familIarIty judgments (e.g..Loftus,
suggestibility procedure Involvespresenting visual information
~ollowed by verbal misinformation and then later testing subjects' memory of the visual information. Subjects often claim
to have seenthings that were presentedonly in the verbal suggestions. Recentdebatehasfocusedon whethermisleadingsuggestions impair subjects' memorial recordsof the visual information (e.g" Belli, Windschitl, McCarthy, & Winfrey, 1992;
Ceci, Ross,& Toglia, 1987; Chandler, 1991; Lindsa)\1990;Loftus, Donders, Hoffman, & Schooler,1989; McCloskey & Zaragoza, 1985; Tversky & Tuchin, 1989). Although the memory
impairment question is important, the question of whether
misled subjects rememberseeingsuggesteddetails in the event
is equally important (Lindsa)\ in press; Lindsay & Johnson,
1989; Zaragoza& Moore, 1990).The source-monitoring framework suggestsguidelines for predicting the likelihood of such
source confusions.
According to the source-monitorin h
sis me ry
errors may occur at test whenmisled subjectserroneouslyidentlfy memorIes enved from the misleading information as
~n;~;i~~-d~~iv~d fr~mthe witnessed event itself. Subject~
sOUrcejudgments should be affected by the decision-making
processesand criteria they adopt on the test (Hasher& Griffin,
1978; Johnson,1988a;Raye et al.. 1980).The criteria that subjects use to attribute a memory to a particular source will vary
with factors suchasthe purpose of the remembering,the biases
that are active, and the weight given to plausibility (Johnson,
1988a). Under some conditions, it might be sufficient that an
item is vaguelyfamiliar and fits with other details derived from
that source; under other circumstances, other information,
suchas a perceptuallydetailed recollection, might be required,
The recognition tests that are t icall use'
eyewitness su gestibility may tacitl induce sub' cts to make
~urce-monitoring errors on th~ crit~cal ite~s:In the standard
procedure,subjectsreceivea seriesof recognition trials consisting of items from the original eventand newdistractor items (in
the form of either forced-choice pairs or individual yes-no
items). The critical test items, in which the suggestedobjects
appear,are embedded among these filler items. Becausemost
~~~~
re.auir~~!!?i~~ts:~-~i~~r~~~~.~~~,
~:~~~~~
jects presented in the visual eventand completely new distract~~ ;~bj~~t; ~a ado t an undifferentiat~d faiiiiliarit., criienon Atkinson & Juola,1973; Jacoby& Dallas, 1981; Mandler.
1980) early on in the test and sto attendin to tentiall useful information about the sourcesof their memori s (e.g..the
amount and nature of perceptual detail). When a suggested
item is encountered on the test, subjectsma recognize it as
.
Lmdsay and Johnson ~1989)and Zara~ozaan~ Kos~mid~r
(198~)~ecentlyfoun? evIdence that was m keepIng wIth this
p~edlctlo~. In. the Lmd~ayand J?hnson studies, subjects first
viewed PI~tonal ta.rget..~forma!lon and. then .receivedverbal
posteventmformatl~n wIth o~ without mIsleadIngsuggestions.
Late~,half of the su?je~tsrecelv~da.yes-no recognition test that
r~qulred them t.o IndIcate whIch Items they had seen in the
picture and. w~lch they had not.. The other half received a
source-m~nltonng test that .required them to indicate the
sourc~ (pIcture, verbal narratIve, or both) of their memory of
each Item that .they recognized as old. As is typically found,
~hentes~~d with the recognition procedure, misled sub~s
o!!:~n clai~ed that thev had seen thing~ th:lt h:lti nnl¥ ~"~n
suggestedm the_text. !n contrast. there was no sueeestibility
e~0~2
subjects tested with th~ ~rulrrp_mnnitoring t"st:
These.subjects correctly attributed suggested details to the
narratIve.Zaragozaand Koshmider(1989) found similar results
with a slightly different source-monitoring test. These results
are consistent with the idea that source attributions are the
r~ult- ~f de~isioii processesthat are sub'ect to crit ri effi
Subsequentexperiments have demonstrated that even subjects given a source-monitoring test sometimes claim to have
seensuggesteddetails (e.g..Lindsay,1990; Zaragoza& Muench,
1989). In the Lindsay (1990)study,subjects in the easysourcemonitoring condition received misleading suggestions 2 days
after viewing the event, minutes before taking the test, and
under conditions that differed from those in which they viewed
the event.Subjectsin the difficult condition, on the other hand,
receivedthe misleading suggestionsminutes after viewing the
event, under very similar conditions, 2 days before taking the
test. Before taking a recall test, subjects were correctly informed that anything mentioned in the posteventnarrative that
was relevant to any test question was wrong. That is, subjects
were explicitly told not to report anything that they remembered from the posteventnarrative (an adaptation of Jacoby's
"logic of opposition," e.g.,Jacoby,Woloshyn,& Kelley, 1989).
Subjects in the easy condition showed no tendency to report
suggesteddetails, indicating that subjects understood and attempted to follow the injunction againstreporting information
from the posteventnarrative. Nonetheless,subjects in the difficult condition quite often reported seeingthe suggesteddetails.
Taken together. two points are clear from thesestudies: (a)~
~.Iihood- o~rce
~.isattribut!ons 5!~~~~_d:
g~...~~e_~~!t~
subjects adopt in making source-mo'
.'ud
ments see
a so 0 son 0 nson, m press;Hasher& Griffin, 1978; Raye
I' .
12
c.'.~~
rQ
I
cJ";, ~
M. JOHNSON,S. HASHTROUDI, AND S. LINDSAY
et
al., quite
1980),
and (b)criteria.
source misattributions may
under
stringent
-~occur even
1989;
Raye
al., 1980;
also
see Hasher
1978).
In
keeping
withetthis,
Jacoby,
Kelley;
Brown, &
andGriffin,
Jasechko
(1989)
'7
In addition to helping to characterizecriteria effects in eyewitness testimony, the source-monitoring framework suggests
other factorsthat will influence such memory errors. ForexampIe,sourcemisattributions should be more likely whensubjects
incidentally visualize misleading information (e.g.,Durso &
Johnson, 1980; Finke et al., 1988; Intraub & Hoffman, 1992;
Zaragoza, 1991)or imagine suggestedutterances in one of the
actor's voices (e.g., Johnson, Foley,& Leach, 1988). Subjects
should also be more likely to make source confusions when
they consider misleading information tangential and do not
contextualize it as part of an event distinct from the original
event.This may be one reasonthat it is easierto induce subjects
to make errors on peripheral than on central details (Loftus,
1979b; Tousignant, Hall, & Loftus, 1986). Source misattribut.i~nssh~l!ld ~Iso~~
~ely. ,:"hensub~e~~~~r
~IS~ra.~t;dat the tlm; tha.tthe misInformation ISIntroduced (or
tested) bec.ause,a~in, they would be less likely to en~a~e in
found that discrimination between famous and nonfamous
names was better when the test list included names that had
been presented earlier than when it did not. Thus, when subjects realized that they might be confused about the sourceof a
familiarity response,they would tighten their criteria for calling a name famous. For example, they might not saysomeone
was famous unless they could remember what he or she did.
Limitin reflective activity at either encodin or source jud ments will also produce impoverished memories or judgment
~~~s~es t!!at may lead to e~rorsin sourcemonit~~i;~ (J;~~G~1991).Jacoby,Woloshyn,and Kelley (1989)varied whether subjects simply read names in the first phase(full-attention condition) or at the sametime listened for runs of three odd numbers
in a tape-recorded series of random digits (divided-attention
conditi~n). Divided attenti~n reduc.edold-new recognition but
not the Incrementof error m fame judgments caused by recent
exposure.This wastrue even when subjectswere informed that
'
processesthatwouldproducepotentialsourc~cues.fort~emis,1~ (h'.,1\...
~.
~~(Za~ago.za
& Lan~,1991).in keeping ~iifi -e~~d~~~e
r ~~
that source-monItorIngprocessesdevelopand change with age
(seesections in this article on development,p. 7, and aging,p.
16),there are age-relatedchanges in children's (Ceci & Bruck,
1993)and elderly adults' (Cohen& Faulkner, 1989)susceptibility to misleading suggestions.
.tion.
(,(s~
Qmt."
Misattributing Familiarity
'.
""
.In
Failures m source monItorIng maygive riseto the feelIng that
~.!!~
a faCtrfiat IstamII;ar t;o~ ~ ;~~~~t-~~~~;~';;;;~
~n~n P~~lously(e.~., Begg, Robertson, Gruppuso, Ana~
Needham,1992; Jacoby,Kelley, Brown, & Jasechko,1989; Kelley & Lindsay,1993; Neely& Payne,1983;Schacteretal., 1984).
For example,Jacobyand colleaguesreported a numberof interesting experiments all involving two phases.In PhaseI, subjects
read a list of nonfamous names(e.g.,SebastianWeisdorf), and
in PhaseII, subjectsmade fame judgments for a list containing
both Phase I nonfamous names, new nonfamous names,and
.famous names. Before they made their fame judgments, subjects were told that all Phase I names were nonfamous. On an
immediate test, subjectswere less likely to call Phase I nonfamous names famous than to call new nonfamous names famous. When Phase II was delayed, however, subjects were
more likely to call Phase I nonfamous names famous than to
call new nonfamous names famous, suggesting that subjects
based fame judgments on overall familiarity of the name and
had forgotten, or were not accessingor using,source informaThus, they misjudged the basi~for the fa!!!iliarity of~hase
I items as preexperimental. These findings illustrate that peopIe sometimes recogn~formation
but not its source (e.g.,
Johnson& Foley,1984; Johnson& Raye,1981; Lindsay& Johnson, 1987)and that recognitionand identification of origin may
be basedon different aspectsof memory or different relative
contributions of heuristic and systematic processes(e.g.,Foley
et al., 1983; Johnson,1985; Johnson& Raye,1981; Lindsay &
Johnson,1991; ~aye & Johnson,1980).
The c~iteria su~jectsuse will affect the ex~entto which they
confuse Information from two sources(e.g.,Lmdsay& Johnson,
all names onathe
initial
and that
they
remembered
name
fromlist
thew.er.e.no~famous
ImtiallIst, they should
sayifthat
it
was nonfamous. Thus, when study was under conditions of
divided attention, subjectswere unable to "correct" the senseof
familiarity (that might be misattributed to fame) with further
information, presumablybecauseunder divided-attention conditions, they werelesslikely to engagein the additional processing necessaryto establish information that could later be used
to id~ntify th~ sourceof the familiari~y response.
mterpr.etmg.the~eresults,J!~~b~and colleaguessu~~ested
~at such~Isa_ttnbutlonsof familIarity to oast knowledl!e were
an example of a more generalclass of misattribution~ which
~s ba~~n "flue~
of orocessing (Jacob:,;K~lley, &Oywan, 19!59;Jacoby,Woloshyn,& Kelley,1989; Kelley & Jacoby,
1990). For example,briefly flashing a word immediately before
its presentationas a recognition probe increasesthe likelihood
that subjects will identify that word as a word from the study
list; presumablythe flashed previewof the word facilitates processingof the recognition probe, and that fluent processing is
takenasevidenceof having previouslyencounteredthe word on
the list (Jacoby& Whitehouse,1989). Likewise, prior exposure
I
to solution words facilitates solving anagrams, which in turn
I
leads subjects to underestimate the difficulty of those anagrams for others (Kelley & Jacoby,1993).
Jacoby'sideasand thosepresentedhereare not contradictory,
but they do differ substantially in emphasis.Like the view expounded here, Jacoby and colleagues(e.g., Jacoby & Dallas,
1981; Jacoby;Kelley & Dywan, 1989)hold that the subjective
experience of rememberin .result
of some i";trinsic
qua Ity of "memory traces" but rath
ectsthe 0 ration of
a.de~isionproc~ssthat ~ttributes current mental eventsto oarti~ular sources(e.g.,memory). Jacoby's work has focused primarily on attributions that are basedon fluenc}( The feeling of :f3 cOhi
!~~~!i~r~~ is said to arise from an undifferentiated-1seediscus- (fl."~"1-:o
slon of Figure I) .?!..glo.bal~ssessm~ntof processing ease2r "A"A.':.t'~'"
fluency; specific qualitative characteristics of memories do not 5\4\...tS.
~~itluencY
iudl!ments.Our work on sourcemo~it~;i~;:
~.. \ ~.f'
on t~eotherhand, hasfocusedon attrihution~ th:\t :\rpha~~on
,,~
ations, contextual mform:\tion anti "ffpl't
\i...r""~i
I
tc
..}
t
13I
SOURCE MONITORING
'
,
Jacob~contrasts j~dg~ents that are based on familiarity; an
automat~cprocess,,;It.hJudgmentsthat ~re basedon controlled
recollection. !acob~s Id~~ th~!re~oll~c!lon ca~ c~u~teractthe
results of a fluency-base?familiar.ity judgment is similar to our
IdeathatmoresystematIcprocessln cancounteractJu gments
that might oth~rwlse be based on heuristic roc~ssin e.g;,
ohnson, 199Ia.,Johnso~ & R~ye,1981). However, In Jacoby--s
.w~~~coll;ctIo~
.of prIor eplso~eshas often be~n treat~d as
something that either occurs or does not occur. Researchon
'h J
s~
monitonn~, In contr~t, has emphaSizeathe idea th~t
() 1Jo'1
~lectionoccu~.
in d~greesa~d p~~uces yariations in _~h~
i
phenomenal qualities of memOrIes.Finally, the term recollecij
~n
~a~~bY's.wo~k,u~ually refers.to _directretrieva.lo~~~;
cords of ast episodes. Source monItorIng processesinclude
t
not only retrieval but ~Is~ interpretation 01 wnat ~sretnev~.
WO"rIcon source monrtoring addressesthe question of what
gives rise to the feeling that a specific past episode has been
retrieved. T~e s_ource-moni!oringframeworkattempts to ch~!:acterizethe various.di~ensions of .henome~al ex erience th
are releva~~for at~lbutl~~ ~emones t~ partIcular so!!rces,the
way conditions affect which as ects 0
r will be criterIa , and the way such information interacts with decision~i!juhon) processes.
Cr
~
~
.
Cryptomnesia)
Cryptomnesia is inadvertent la iarism that occurs when a
II person pr uces somethin and believes tha i i
ri inal
~- (;0 -se -generatedproduct of the moment when in fact it was er,
celve or enerate earli r. A. S. Brown and Murphy (1989)
.obtained
cryptomnesia experimentally; Small groups of subjects took turns generatingexemplarsof categories.Later, sub!
jects tried to recall the items they had themselvesgeneratedand
to produce new items not given before by anyone.Cryptomnesia was obtained from most subjects in that during the initial
generation phase they gave items given earlier by others and
also in that they later claimed to havegeneratedexemplarsthat
'\
\'
had been produced by others in the group. Furthermore, sub-
""
.111
:..' ~.!
.,
"
.(,
J ,~."\'('0,1""
.\
.~:"',
:1'
"'," !
can become the bases for particular decisions, depending on
task demands (seealso Hasher, Go,ldst~in, & Toppino, 1977;
Mandler et al., 1987), especially when other kinds of source,
information are not readily available. If the task is to recognize
famous names,the familiarity that comes from recent activatio~ can be inter~reted heuristica~IYas familiarity from preexpenmental experIence.If the task ISto generate new items (A. S.
Brown & Murphy, 1989; Kubovy, 1977),to solveproblems (KeIley & Jacoby,1993), or to answer questions (Begg & Armour,
1991; Kelley & Lindsay,1993),the availability that comes from
recent activation may be interpreted as (unaided) generation of
the moment. Subjects are "fooled" in both casesbecausethe
discrimination is difficult; the familiarity from recent activation of a newname maybe indistinguishable from the familiarity from pre~~perimentalexperience of the sort that createsa
vaguely familiar name. The false-fame effect should thus be
much reduced if the familiarity of the famous names on the test
liSt is increased,becausethis should increasethe criterion sub~_use for ,judging a nam~ famo~s from prio!: ~xperienc~.
Similarly; it is hard to distinguish the cognitive operations
that go into generating from those that go into inadvertent plagiarizing in the category-exemplar-generationtask. In both
cases,subjects presumably retrieve exemplarsactivated by the
category name that have little other operations information to
distinguish
them.
1989) found
that compared
et al. (1981; see also
with high-frequency
Rabinowitz,
instances,
sub-
jects were better able to identify whether they had generated or
the experimenterhad presentedlow-frequencyinstancesof categories, presumably because lower frequency instances took
more cognitive operations to generate, which could then be
usedascuesto the origin of the information. From the Johnson
et al. (1981) finding, we would expect that subjects in the A. S.
Brown and Murphy (1989)procedure would make fewer cryptomnesia errors if instructed to generate low-frequency category instances. Although A. S. Brown and Murphy did not
specifically manipulate instance frequency in their experiment, in keeping with what we would expect, they reported
that errors were more likely on the higher frequency exemplars
jects were m?re likely to p~agiarize from subjects ,,:,ho came
before them In the generation order than from subjects who
given by subjects in their experiment.
came after them. A. S. Brown and Murphy suggestedthat subjects were especiallylikely to be thinking of their next response
just before it was time for them to give it and probably processedthe eventsjust preceding their turn lesscompletely than
the eventsafter their turn (Brenner,1973).Thus, subjectsmight
havefailed to establishmemories with the kind of perceptual,
contextual, and reflectively generated detail on which correct
sourceattributions depend. The source-monitoring framework
also suggeststhat subjects may have in fact covertly generated
an item that was subsequentlygiven by another subject before
they had a chanceto report it. The discrimination (thought and
heard vs. thought and said)would be even harder in suchcases.
Considered together with the Jacoby,Kelley, Brown, and Jasechko (1989)and Jacoby,Woloshyn,and Kelley (1989)"falsefame" studies,the resultsreported by A. S. Brown and Murphy
(1989)suggestthat phenomenal differences between material
Incorporating Fiction as Fact
Another clearly important source-monit ri
r I
cerns the extentto which the fictional accounts that we read or
~i;~i~~~~~i~d,-il~;;-g w~r;-e
acco!,lJ:!~
~~r
own Irec experIences,into our eneralknowled e and beliefs.
MOVies, televIsIon, books, magazines, newspapers-all are
sourcesof fictional information that may, under some circumstances,be treated asreliable information. A particularly striking example of this phenomenon was illustrated in a "60 Minutes" program broadcast on CBS television that showed thenpresidentRonald Reaganrecounting a story to Navy personnel
about an actof heroism that he attributed to a real U.S. pilot but
that bore an uncanny resemblance to a scene from a Dana
Andrews movie releasedin the I 940s. According to t.he report
of this incident, no record ofa similar, real act of herOIsmcould
that has been recently experienced and material that has not
been recently experienced (e.g.,differences in perceptual fluency,differences in familiarity, and differences in availability)
be found.
As with other source-monitoring situations, we would expect
~~n
_etweenact a~- ctlon t~ e ~elated~actors~
i
I
~-:';C"."..""";
Johnson
,-,,,.:.~~--
~
~-
-
SAY
affect the quality
of source-specifying
information
encoded
cesses
acquIsItIon
ater. an
Some
to evidence
actors t at
in asupport
ect source-mon
of this general
comes
from-reports
fictional
of attempts
accounts
into
Ross, 1980; Gerrig
at
idea
to study the incorporation
general
knowledge
& Prentice,
1989). For example,
?ut ~ource-morntonng
&
InevItable.
subjects
real or that it was fictional.
"false"
to statements
such as "The
pukeko
has feathers."
cerned
story
who were told
questions
I(most
concepts,
to be studied
Later, subjects responded
is a bird"
of the verification
response
times
that the material
did not concern
pukeko
or
or "The
questions
was artificial.
con-
way that it would,
like semantic
on any particular
Potts et al. (1989
knowledge
context
further
su
cognItIon,
so errors
,\re
so.clal, and em?tlonal
In courtroo~
hallucInations,
testImony
and faIlure
delu~lons,
confabulation,
and am-
nesl~ and as a cons.equ.ence of agIng. The b~e.akdowns.that
~orntonng.
provide
under
po~ent~ally
mg processes
standIng
such c?ndltlon~
valuable
of sour.ce-morntonng
(and,
about. the
Ind~ed,
as well ~s a~ opportunIty
of the so~r.ce-morntonng
disrupted
oc-
of disrupted
Infor~atlon
processes
m general),
to
c~rcumstances,
~eo~le sufIn source mo.rntonng-as
morntort~ evaluate
framework
m under-
cognItIon.
was real
had fast response times in the nons tory context because they
had integrated the material into general knowledge
in such a
dependent
s Ideas). Un~er so~e
prof,!un~
dlsruptlo.ns
the u~eful~ess
was real. Potts et al.
practIcal,
credIt another
~er even ~ore
cognItion
the story, response times were faster
are Imperfect,
~e.~.,errors
In ~ypnosls,
p~rt of normal
many of these e!rors ar~ of no great c~nse-
cons~quences
~ature
If most of the
that subjects who were told that the material
processes
Althoug~
cur I~ .source
were faster for subjects
for subjects who were told that the material
suggested
was
"true"
Monitoring
IS an Integral
quence, some have Impo~tant
in
the Potts et al. experiment
read passages including
information
about a bird called a "takehe" and its ancestor the "pukeko:'
Subjects were told either that the information
of Source
mornt?nn.g
of
(J. R. Anderson
1990, 1991; Potts & Peterson,
1985; Potts, St. John, & Kirson,
Disruption
Source
in general,
be less
for its retrieval.
Delusions
Hallucinations
..
and
As we argue here and elsewhere
son & Raye, 1981), reality
ested that subjects
had some
Confabulations
memory
(e.g., Johnson,
is not given directly
representations
1988a; John-
in perceptual
but is a product
and
of judgment
pro-
degree of strategic control over whether information
would be
cesses. The characteristics
of mental experience
that provide it
~rporated
into their know~se
or comE~rtmentalized.
.with
the quality of reality are similar
for perceptions,
event
We interpret
compartmentalized
ing information
is available.
al. had in mind
that establish
presumably
complex
to mean that source-specify-
The strategic
if~ubjects
trieve additional
between fictional
tional
t~
study
of
and Prentice
how interesting
records
confusion
might
more
likely
effort
ing framework,
formation
~~
rom the stor
rellective
required
Subje~s
w~are
in related
might
also be expected
interested
allow
b
them
in
~~d,
information
and therefore
(Pratkanis,
1988), long retention
to increase
in-
not interested
be more influenced
& Baumgardner,
and
special
its source when subjects
As in the "sleeper effect"
Leippe,
were
be the sorts of reflec-
ratings and that would
than are more
wald,
information.
was the
les~ ~i~~IY to en~a~e in s~c.h ~ct~ity
later might
subjects.
were
In the source-monitor-
effort would
later help identify
the information.
~ho.~ld.~e
some
by it. Gerrig
embed the information
their opinion
paradoxically,
indi-
opinions
interesting
affected
with interest.
the special
that could
to discount
task, subjects
of topics, including
Particularly
Reality
toring
of memories
the influence
testing
operations
of ongoing
and
Green-
of fictional
memories,
absence of con-
producing
perception
beliefs
in spatial
in supporting
affect; and the relative
the event or
and reality
are complex
teristics
of memories
clear
and of intact
in considering
in Oltmanns
1977; Horowitz,
1988), hallucinations
held
false
as delusions
of evidence
(Oltmanns
under ordinary
velopment
& Maher,
of delusions
voice) that even nondelusional
find
control
difficu.lt
to distinguish
from
Leach,
Delusions
over thoughts
automatically,
and typically
or unbidden.
tive, cognitive-operations
easy to confuse
1980; Finke
with
Because
taken as evidence
gas, &
involve
information
external
this
(e.g.,
someevents
a loss of
thoughts
that come
produce
little
(e.g., Durso
connections
may lead
events as perceived.
involve
reflec-
& Johnson,
1991 b). Furthermore,
among
of supporting
that an event occurred
Raye, 1988),
people
and so are particularly
stimuli
the retrieval
moni-
1988a). De-
information
real perceptual
Such thoughts
et al., 1988; Johnson,
arate events.
imagined
sensory
Foley, &
1988).
are
in the de-
(Johnson,
another
(Johnson,
in reality
are implicated
imagined
beliefs
1988). Many of the
to be important
circumstances
and maintenance
personal
to involve
times
& Wil-
to the contrary
lusions are likely
person's
(see
(Hilgard,
(Baddeley
or implausible
in spite
factors that we have postulated
toring
is espe-
as delusions
1991a; Moscovitch,1989;Stussetal.,1978).
significant
that are firmly
classified
& Maher,
pro-
in some
charac-
processes
such phenomena
1978), and confabulations
son, 1986; Johnson,
Clinically
judgment
moni-
judgment
cesses that are subject to error and are more difficult
situations than in others. The importance
of qualitative
sional person may find elaborate
intervals
embeddedness
embeddedness
of the cognitive
belief.
chapters
read a story and rate
filler
that compartmentalization
that would
were making
fiction,
who rated a story as less interesting
that was correlated
tive activities
and
in the stories (e.g., that choco-
to have their opinions
suggested
fact
in the story. Subjects'
be carcinogenic).
may be avail-
sensory detail;
context;
and beliefs;
sciousness
cially
between
about a number
that subjects
Prentice
(e.g., re-
later.-
a brief
by false facts introduced
finding
reflectivelr
(1990) had subjects
that had been discussed
late
1983, 1992). For
of this reflection
decisions
it was. After
cated their opinions
affected
information
and beliefs:
and temporal
knowledge,
information
and note potential
differences
birds and real birds when presented with fic-
information),
Gerrig
activities
of the sort that form
(e.g., Johnson,
process
able for source-monitoring
In another
records
memories,
that Potts et
were the sorts of reflective
memory
the basis for source monitoring
example,
activities
(Johnson,
to incorrectly
Finally;
cognitive
a delumany sep-
memories
is
Foley, Suenidentifying
habits
or the
,
--
SOURCE MONITORING
Amnesia
I
There is considerable evidence that failure to remember
source may bean important component of amnesia(J. Brown &
Brown, 1990; Graf, Squire,& Mandler, 1984; Hirst, 1982; Hirst
& Volpe, 1982,1984; Huppert & Piercy;1976; Johnson& Hirst,
in press; Mayes et al., 1985; Meudell, Mayes, Ostergaard, &
Pickering, 1985; Pickering, Mayes,& Fairbairn, 1989; Schacter
et al., 1984; Schacter& Tulving, 1982; Shimamura & Squire,
1987,Smith & Milner, 1989; Weinstein, 1987,cited in Johnson,
1990). Various studies have shown disruptions in amnesics for
memory for temporal (Squire, 1982)and spatial (Mayes,Meudell, & MacDonald, 1991; Shoqeirat & Mayes, 1991)information and modality of presentation (Pickering et al., 1989).
S.chacteret al. (1984)investigated source monitoring in amneSICS
by meansof a memory-for-trivia-facts paradigm. They pre~enteda~nesic patients with statements that included fictional
informatIon about well.,known people (e.g.,"Bob Hope's father
was a .fireman"). After brief ~nte~vals,amnesic patients could
some~lmesproduce the studIed Information when cued with
questIons but often could not remember whether they learned
those facts in t~e. experiment or from an extraexperimental
source (e.g.,televIsIon or newspapers).Schacteret al. (1984)refer~,edt? this finding ~ "source amnesia," ~hich th~y d~fined
as retrIeval of expen~entally pre~ented informatIon In the
~s~;~~. of a corresponding recollectIon of how it wasacquired"
!
i
i
I
I
~-
t
fi
ea.
t I (1984)d...
tt .
h
IstInguls
S
e
.
d
b
etween
source
.
amnesia
h
er
0
d
~ects
su
en
w
t rememberw h eth eran .Item was Iearned .In th eexpenment
.
or outside
the experiment. Source forgetting occurs when subjects can remember that an item was learned in the experiment
but do not remember the specific source (e.g.,which of two
speakerssaid it). From the present perspective, source is a
shorthand way of referring to a variety of types of information
(e.g.,spatial, temporal, and voice)that may be accessibleto different degreesor with varying probability and thus may yield
source identifications of varying degreesof specificit)CThe issues here are analogous to those concerning the relation between old-new recognition and source monitoring. From the
source-monitoring point of vie\.\;source forgetting and source
amnesia are both examplesof source-monitoring failures that
simply reflect differences in either the amount or the specificity
of information available about various external sources. In any
case,the fact that some amnesic subjects have so little source
information available after so short a retention interval is a
striking phenomenon.
These findings suggest that those aspectsof memory that
would ordinarily support source monitoring are profoundly
impaired in amnesia. Thus, severelyamnesic subjectsmay have
accessto relativelylittle differentiated information about qualitative characteristicsof memories suchas perceptualdetail (e.g.,
the voice in which something was said). They also may have
little accessto recordsof prior reflective activity ~uchas noting
relations among events. Without such source-relevantinformation, they may base their source-monitoring judgments in the
contextoftriviaquestionsoncombinationsofcuesandcriteria
that lead to many errors (e.g.,"Why would 'fireman' come to
occurs
the Idea that realIty monitoring Involves at least two kinds of
..'
..no
Jud~ment
processes.on~ t~at ISbas~don a ~ondellb.eratIve
evaluatlon of the characteristIcsof actIvate? informatIon, such ~s
the type and amoun~of perceptual~etall, and the o.therthat IS
based on ~ mor~ dellber~te e.valuatl?nof the meamngf~1content of actIvatedinformatIon In the lIght of other memonesand
knowledge (Johnson,199Ia).
Patients who have had a cingulectomy (to treat obsessivecompulsive behavior)sometimesexperiencereality-monitoring
failures that appear to be the consequenceof unusually vivid
mental experiences. They will report something as real, but
they may catch themselves(e.g.,in answer to a question, one
patient ~aid "I have been having tea with my wife" but then
said, "Oh, I haven't reall)CShe'snot been here toda)C. ..The
sceneoccurs so vividly; I can seethe cups and saucersand hear
her pouring out" (Whitty & Lewin, 1957,p. 73). It is as if vivid
imaginations initially passa heuristic reality-monitoring check
on the basisof qualitative characteristics and are accepted as
real until caught by more systematic processe~.In contrast,
patients with frontal damagewho confabulate may experience
reality-monitoring failures that appearto be the consequenceof
errors resulting from disruption in more extended reasoning
processes.Thesepatientssometimesmake bizarre or implausible statements.For example,one frontal patient claimed that he
was shotand killed by a young woman during World War II but
that the surgeonsbrought him back to life (Stuss et al., 1978).
More systematicprocessesare responsiblefor catching implausible ideas that may passa more heuristic reality-monitoring
check.
ac
d
with
amnesia
consIstent
ource
generally
mg.
are
source
they
an
because
.
c
interestIng
b
S h
...'.
tIcularly
orge
intermixture of cognitive and motivational effects(e.g.,Johnson
& Sherman,1990)mayinduce delusionalpeopleto useinappropri~tely lax .criteria for evaluating mental experiences.As desC~lb~dearlier, e~en ~or~al people, when induced to use lax
crIterIa, make mlsattrlbutlons about the sourceof information
th~t t.heywou.ld not make if they were using more stringent
CrIteria (e.g.,Lindsay & Johnson,1989; Raye et al., 1980).
Similar points can be made about the reality-testing errors
that produce hallucinations (Perky;1910; Segal,1970). That is,
hallucinations can be analyzedaccording to many of the same
factors that have beenshown to influence reality monitoring in
normal individuals (e.g.,Bentall, 1990; Horowitz, 1978; Johnson, I 988a). Kunzendorf (1985-1986) specifically suggested
that certain hallucination phenomena occurring in hypnosis
result from the suspensionof a reality "monitor" (p. 258). An
important difference betweenKunzendorfs notion of a reality
monitor and our notion of reality-monitoring processesis that
Kunzendorfs monitor is a relativelysimple device for reading
an unambiguous "tag" (p. 257)that specifiesthat a givensensation is an image or a percept. In contrast, we assume that
various 'qualitative characteristics of activated information are
evaluated and that the decision processesare more complex
than simply "reading out" a single index of source information
Patientswho havesufferedorganic brain damagesometime~
make false statements,apparently without intending to deceive
which arec~assifiedas ~on!abulations(Talla?d, 196.1;Whitlock:
1981).PublisheddescrIptIonsof confabulatIng patIentsare par-
15
--~
16
M. JOHNSON,S. HASHTROUDI, AND S. LINDSAY
mind when I think about Bob Hope'sfather unless I had read or
heard about it someplacebefore; it must be true but I don't
rememberthe specificsof learning it, and I'm not very good at
learning new things,so I probably learned it a long time ago";cf.
Begg & Armour, 1991; Kelley & Lindsay,1993).
Some further clues about the nature of amnesicsource monitoring might be obtained by comparing amnesic patients' performance in various source-monitoring situations. That is, am-
mation. Burke and Light suggested that the inability to use
contextual information may be a key mechanism involved in
agedeficits in memor)( This idea is consistent with the finding
that agedifferencestend to be greatestwhen testing procedures
provide the least amount of contextual support for retrieval, as
in free recall, in which subjects must reconstruct the original
events (Craik, 1984; Hultsch & Dixon, 1990; Light & Singh,
1981).
nesics may not be disrupted equally in all source-monitoring
situations. To investigate this possibility, Johnson, Hirst,
Another line of evidence comes from more recent studies in
the source-monitoring tradition. McIntyre and Craik (1987),
Phelps, Multhaup, and Volpe (1992) tested 5 mixed-etiology;
nonalcoholic amnesics(and 5 control subjects)in two tasksthat
required subjects to identify who said wha~. In one, subjects
watcheda video in which an experimenteraskedquestions (e.g.,
"Name a type of coin that usedto be larger than a quarter") and
2 people alternated giving the answers(e.g.,a penny). Subjects
later were read the words given as answersand had to indicate
who said each item (external source monitoring). In the other
task, I person gavesome of the answers(again on video), and
the subjectwas required to give the others(cued by first letters).
Here, too, subjectswereaskedwho said eachword (aninternalexternal discrimination, or reality monitoring). Amnesics
showed source-monitoring deficits for both external source
monitoring and reality monitoring, but the deficit in external
source monitoring appeared to be greater than the deficit in
reality monitoring. Evenso, in the reality-monitoring task, amnesicswere particularly poor at identifying the sourceof items
they had generated(also the case in a follow-up experiment).
These results,although preliminary; suggestthat amnesicsmay
be disproportionately disadvantaged in remembering what
they have generatedcompared with what they have perceived.
Assuming that source identification of self-generated items
drawson recordsof cognitive operations more than doesidentification of perceived items, these findings are consistent with
the suggestionthat amnesicsare more disrupted in self-generated, reflective memory than in perceptual memory processes
(Hirst et al., 1986; Hirst, Johnson, Phelps& Volpe,1988;Johnson, 1983,1990; Johnson& Hirst, 1991).
More generally,systematic comparisons of which aspectsof
source (e.g.,temporal, spatial, modality of presentation, and
person's voice)are particularly difficult for various memorydisordered subjectswould.be especially ~seful in clarifyi~g the
role of source memory In memory dIsorders (cf. Smith &
Milner, 1984).As is discussed in the Brain RegionsImplicated
in SourceMonitoring section, various aspectsof source may be
differentially disrupted by damage to various regions of the
brain.
using a procedure similar to Schacteret al.'s(1984),found that
after a week, older adults remembered trivia facts as well as
young adults but had difficulty determining whether a particular fact was learned in the experiment or outside the experiment. Older adults also had trouble remembering presentation
modality (experimenter vs. an overhead projector). Similarly~
Dywan and Jacoby (1990), using the fame judgment task described earlier(Jacoby;Kelley, Brown, & Jasechko,1989),found
that older subjectswere more likely than young subjectsto mistakenly call the nonfamous names from an earlier presentation
phasefamous.
Rabinowitz (1989)reported that older subjectshad more difficulty discriminating between previously read and previously
generated(e.g.,A_CO_OL) words than did young subjects. In a
conceptually similar study (subjectsread some entire sentences
and filled in the last word for others), Mitchell, Hunt, and
Schmitt (1986)did not find agedifferences in remembering the
source of information but noted that a ceiling effect may have
maskeda source-monitoring deficit among the older subjects.
In a study by Cohenand Faulkner (1989),older and younger
subjectswatched, performed, or imagined themselvesperforming a series of simple actions (e.g.,"Put the spoon next to the
toothbrush"). At test, older subjects were more likely than
young subjectsto say that imagined actions had been watched
and that watched actions had been performed. Cohen and
Faulkner also showed that older subjects were more often
misled by false information than were younger subjects in an
eyewitness-testimonyparad igm (Loftus, 1919a).In a study by
Kausler, Lichty, and Freund (1985), however, older adults did
not havedifficulty discriminating betweenmemories for activities (e.g.,card sorting) that were planned and those that had
beenperforme~. Si~ilarly, in anot~er study (~uttentag ~ H~n~,
1988),only a mInorIty of older subjectshad dIfficulty dlscnmlnating betweenperformed and imagined actions.
Thesestudies suggestthat there ~re agedifferences in source
monitoring, but taken together, they also suggest that these
differences depend on the type of source-monitoring task a
A gln g
person confronts. Direct evidence for this point was reported
by
Hashtro~di
al.of(1989).
Older
and
~<?un~eradults
:-vere
presented
with aet.
lIst
common
words
originating
from dlffer-
One line of evidence for age deficits in remembering source
comes from studies demonstrating that older adults have difficulty in remembering contextual information (Burke & Light,
1981). Older adults have trouble in remembering spatiallocation (e.g.,Light & Zelinski, 1983; Park, Puglisi, & Lutz, 1982;
Perlmutter,.Metzger, Nezworski, & Miller, 1981), sexof voice
(Kausler & Puckett, 1981), upper versus lower case format'
(Kausler & Puckett, 1980), and list membership (Zelinski &
Light, 1919[cited in Burke & Light, 1981])of presentedinfor-
ent sources.Compared with young adults, older adults had difficulty discriminating memories of words they had said from
memories of words they had imagined saying (internal source
monitoring) and difficulty discriminating which of two other
people had said particular words (external source monitoring;
see also Schacter et al., 1991). In contrast, they did not have
more difficulty in discriminating words they had said from
words someone else had said or words they imagined saying
from words someoneelse had said (reality monitoring). (Real-
"'"
.-
i
I
,
J
t'
t
I""""""
'i~"!'
rr';
SOURCEMONITOR[NG
17
'-monitQring'difficulties have been demonstrated in older
bjects iti other situations, e.g.,Hashtroudi, Johnson,& Chrosak, 1990; Rabinowitz, 1989.)
Hashtroudi et al. (1989)suggestedthat the particular pattern
source..monitoring deficits obtained provided a clue about
1ich memory characteristics contributed to the age deficit in
eir experiment. On the basisof previous evidence (Johnsonet
, 1981; Rabinowitz, 1989), it was assumed that with verbal
aterial, the amount and kind of information about cognitive
>erationswere especially salient cues for source monitoring.
the Hashtroudi et al. study, older adults did not have diffiIlty discriminating either what they said aloud or what they
ought from what they heard another personsay.In thesecon.
h
f
...
bl
Id
tlons, t e amount 0 cogmtrve operations presuma y cou
: used as a cue that an item was generated rather than per:ived.This suggeststhat older adults usedthesecuesas well as
)ung subjects.In contrast, the cognitive operations involved in
.ying and thinking are likely to be similar, thereby reducing
Ie effectivenessof cognitive operations as a cue to source in
iscriminating what one said from what one only imagined
lying. Presumably,subjectsthen haveto rely either on memory
the hypothesis that older adults have difficulty encoding or
using multiple cuesto source.
Thesestudies demonstrate how predictions generatedby the
source-monitoring framework can be used to explicate the
mechanisms of source-monitoring deficits. Given that source
monitoring is basedon memory characteristics,one cansystematically examine the contribution of various types of memory
characteristics to source-monitoring deficits. This method of
examining cue effectivenessof various memory characteristics
provides a powerful tool for understanding source-monitoring
processes.
Frontal areas. A number of lines of evidence suggest that
frontal-lobe dysfunction may produce deficits in source monitoring. Patients with frontal damage may recall or recognize
information at normal levels yet show disrupted memory for
temporal order (Milner, 1971; Milner, Petrides,& Smith, 1985;
Shimamura, Janowsk)\& Squire, 1990). Disruptions in temporal-order judgments that are disproportionate to disruptions
.1j
;j
f
1
c~
'.}
:1
"
Ir motoric and kinesthetic information or on memory for spefic perceptual information about voice qualit)( In this situaon, older adults seemed to have difficulty in discriminating
)urce. Similarly,discriminating which of two other peoplesaid
>methingshould also depend on evaluating perceptual infor-
in item recognition or cued recall have also been shown in
Korsakoff patients (Huppert & Piercy, 1976; Meudell et al.,
1985; Shimamura, Janowsky,& Squire, 1991; Squire, 1982;
Squire, Nadel, & Slater, 1981). Patients with Korsakoff syndrome are often impaired in tests such as the Wisconsin Card
1
:~
J
1
"
lation such as specific voice quality, becausethe amount or
ind of cognitive operations would not provide a reliable cue to
)urce. Again, older adults had difficulty in this condition.
'hus the results of this experiment provided a hint that older
dults havethe most trouble in conditions in which perceptual
1formation is particularly important.
In a second study, Hashtroudi et al. (1990)directly examined
ge-relateddifferences in various memory characteristics. The
uestion addressedwas whether some memory characteristics
:.g., perceptual and contextual) were affected more by aging
hanothers. In this study,all subjectsparticipated in someeverylay situations (e.g.,packing a picnic basket) and imagined
hemselvesparticipating in other situations. On a second day,
ubjectsrecalled all they could remember about eachsituation.
lecollections were evaluated for mention of information such
IScolors, objects, spatial references,and thoughts and feelings.
)Ider adults had particular difficulty in remembering percepual and contextual (spatial)information. In addition, in a realty-monitoring test given 3 weeks later, older adults had lower
Sorting Test (Heaton, 1981) and the Benton Verbal Fluency
Test (Benton, 1968), which are sensitiveto frontal-lobe pathology (e.g.,Squire, 1982)and sometimesshowcomputed tomography scans indicating frontal-lobe damage (Kopelman, 1989;
Shimamura, Jernigan,& Squire,1988).Similarly, other populations in which there may be frontal damage, namely Parkinson'sdiseaseand multiple sclerosis patients, also show deficits
in temporal judgments (Beatty & Monson, 1991; Sagar,Cohen,
Sullivan, Corkin, & Growdon, 1988; Sagar,Sullivan, Gabrieli,
Cork in, & Growdon, 1988). Within subjects, impaired temporal-order scoresmay be correlated with scoreson neuropsychological tests sensitiveto frontal damage (Squire, 1982), but
note that such a correlation is not always found (Kopelman,
1989;Shimamura et al., 1990).
Disrupted memory for the sourceof trivia facts also appears
to be related to measuresof frontal-lobe dysfunctions in some
studies of amnesics (Schacteret al., 1984)and is also found in
frontal-lobe patients who do not show disrupted memory for
the facts themselves (Janowsky,Shimamura, & Squire, 1989).
\
I
"
.,
'
B"" zn Re'o zo
'"
'
0
'
T
/. t d . So ce
ns 1mp zca e m
U1'
II
.t
.
lY1om Dring
)erformance than young adults. Becauseperceptual informaion is likely to be a salient cue for reality monitoring, Hash:roudi et al. suggested that reality-monitoring difficulties
might be related to older adults' difficulty remembering seniOry-perceptual information.
Further evidencethat older subjectshave difficulty encoding
perceptualcuesis that an age deficit in identifying which of two
femalespeakershad said particular words waseliminated when
one speakerwasa female and one a male (Fergusonet al., 1992;
see also Lindsay et al., 1991). Ferguson et al. also found that
There is also evidence from both physiological and behavioral
studies that the frontal cortex is particularly sensitive to the
effects of aging (Albert & Kaplan, 1980; Woodruff, 1982). On;
the basis of these findings, McIntyre and Craik (1987) suggested that memory deficits in remembering source in o~der
normal subjects may be linked to frontal-lobe dysfunction.
Some evidence for this suggestion was provided in a recent
study by Craik, Morris, Morris, and Loewen (1990), wh~ re~rted a correl~t~onin older subjects betweenso~rce-monrto~mg scores(decldmg whether a fact was learned m the expe~l-
increasing the salience of the spatial cues in the male-female
condition by having the two speakerssit in distinctive spatial
locations improved young adults' source-monitoring pe~ormance but not that of older subjects.Weare currently exploring
ment or outside the experiment)and performance on the W1Sconsin C~r.dSort.ing T~sta~d the Yerbal F!ue~cy Test. ...
In addition, dlsrUPtion.sm. reall~y momtor~ng described I.n
casereports of confabulation m patients suffering from organIc
i
,;
;
;
,:1
;!
:
"
I'-,
;
»
~
ti
I
..
,'I
-.-,
18
I
M. JOHNSON,S. HASHTROUDI, AND S. LINDSAY
..brain disease form a systematic and suggestivepicture (Johnson, 1991a). The relatively controlled lesions made in anterior
cingulectomy patients produce dramatic but temporary confabulation, lasting severaldays (Whitty & Lewin, 1957,1960).
Damage to the basal forebrain region produces confabulation
that may last weeks to months (Damasio, Graff-Radford, Eslinger, Damasio,& Kassell, 1985),and damageto various areas
in the frontal lobes can produce confabulation that may last
months to years(Stusset al., 1978).Stuss et al. made a persuasive argument for the role offrontal-lobe damage in confabulation. They reported five patients with demonstrable frontallobe lesionswho all showed Sl::>ntaneous,persisting confabulation (seealso Baddeley & Wilson, 1986; Kapur & Coughlan,
1980; Moscovitch, 1989). At least temporary, and sometimes
long-lasting, confabulation is found in Korsakoff patients,who
often show prefrontal symptomatology as well as damage to
thalamic nuclei that project to prefrontal and anterior cingulate
cortex. The fact that confabulators often show recognized
symptoms of frontal damage suchas perseverativetendencies,
difficulty in shifting response sets,and lack of concern about
inappropriate behavior also suggests frontal involvement. In
summary, disruption of areas immediately adjacentto the frontal lobes producesmarked but transienteffects.Structural damage to the frontal lobes themselvesproduces more permanent
effects. Thus, overall, there is a fairly consistent pattern of evidence pointing to confabulation as a potential consequenceof
disruption of frontal-lobe functioning.
Taken together, these observations of relations between
disruption in frontal-lobe functions and disruptions in source
monitoring in amnesic patients and older people and, particularl)\ confabulation in organic-brain-diseasepatients strongly
implicate the frontal lobes in normal source-monitoring processes.As the next section argues,however,this does not mean
that the frontal lobes are uniquely responsiblefor source memor~
Temporal and diencephalic areas. Some investigatorshave
considered the possibility that the type of memory disruption
that is associatedwith damage to the medial temporal lobes or
diencephalon is primarily a problem of disrupted source
(usually called context) memory (J. Brown & Brown, 1990;
Hirst, 1982,1989; Huppert & Piercy,1976; 1978; Mayes, 1988,
1992; Mayes et al., 1985; Schacter,1987b;Stern, 1981).According to this view, the disrupted memory for content shown in
anterograde amnesia is secondary to a more fundamental
disruption in source memory (e.g.,context no longer provides
an effective recall cue).One conceptual problem with this idea
is that the cognitive processes(e.g.,reactivating information and
elaborating relations among items) that contribute to source or
contextual information contribute to factual or content memory as well (cf. Johnson,1992; Mayes et al., 1985). For example,
improving organization among items on a particular list should
both boost item recall and improve subjects' ability to correctly
attribute items to that list (assuming the same organizational
categorieswere not used for other lists). Thus, from a processing point of view,it is hard to differentiate thosecognitive activities that produce good memory for content from those that
produce good memory for context (Johnson,1992).5
Others have argued that the memory disruption associated
with temporal/diencephalic damage does not involve disrup-
tion of source memory but rather that disruptions in source
memory occur asa consequenceof additional frontal-lobe damagefound in some temporal/diencephalic patients (e.g.,Shimamura et al., 1990; Shimamura & Squire, 1987; Squire, 1982).
According to this view, memory for content is a function of
temporal/diencephalic regions, and memory for source is a
function of frontal regions. Again, this is based on too simple a
notion of the processesthat are involved in both memory for
content and memory for source.
We think it more likely that although the frontal lobes contribute to memory for source, they contribute to memory for
contentaswell. Similarly, although the temP9rallobes and diencephalonare critical for memory for content, they contribute to
memory for sourceaswell. Given the likely overlap in processes
that promote memory for source and memory for content (see
also Mayeset al., 1985),patients who havedisruptions of source
memory should show disrupted memory for content under
some circumstances and vice versa. In keeping with this, frontal patients show disruption in content recall if they are tested
in multitrial free-recall learning-a type of memory test more
dependent on reflective organizational processes than the
more commonly used, single-trial free-recall, cued-recall, or
recognition tests,on which they show"intact" content memory
(Janowsky et al. [cited in Shimamura et al., 1991]; also see
Moscovitch, 1989;and Smith & Milner, 1984). Similarly, amnesics, even those without obvious frontal damage,would have
poor source memory if they were directly compared with controis under the same acquisition and test conditions.
Usually,amnesicsand controls are not compared under the
same acquisition and testconditions. A common methodological practice is to equate patientsand controls on recognition or
cued recall and then compare source-monitoring scores (e.g.,
Cave& Squire,1991; Meudell et al., 1985;Pickering et al., 1989;
Schacteret al., 1984;cf. Hirst et al., 1986; Hirst et al., 1988). For
example,amnesics might be given longer to study the items, or
controls might be given a longer retention interval. Using this
procedure, it is tempting to interpret a deficit in performance
on a source task that is disproportionate to a deficit on the
content task as demonstrating that source memory is
disrupted. However, it does not follow that either content or
source memory has been uniformly disrupted or that they have
beenlocalized to particular brain areas.Sucha result only implies that some processor type of information is involved to a
greater extent in one task than in the other. For example, predictions for a situation in which, say, recognition is equated
between groups and in which source monitoring is tested
would depend on the type of information used by subjects in
evaluating content and source(seeSourceMonitoring and OldNew Recognitionsection). Suppose,in this particular test, both
amnesics and controls used only perceptual information for
making both recognition and source-monitoring judgments. If
~~er
reasonto nottakeasfundamentalthe contentversuscontext distinction but ratherto treat themas heuristiccategoriesis that
they maybe impossibleto separateon a priori grounds.Whatis core
contentand whatis incidentalcontextmay vary with the situation.
especiallywith a person'sactivatedgoals(e.g.,Allport, 1955;Johnson.
1992;Menzel& Johnson,1976).
\
i
!
!
i
.i
I
,
...
\
"',"i,
CC'
.";\
j!"
,
SOURCEMONITORING
the groups were equate~ o~ recognition, they would also be
equated on source monItorIng. Nevertheless,this would not
me~n that the groups would remain equ~ted in a source-moniton~g task that demanded more reflectl~ely~Iaborated informatlon for accurateperformance (e.g.,WeInsteIn,1987,cited in
J?hnson, 1990). In short, whether amnesics show disproportlonate deficits in source monitoring in relation to controls depends on what types of information the two groups use in the
content and source tests. Similarly, the correlation within a
~roup betweenrecognition (orcued recall)andsource-monitormg
scoresshould.I depend on the extent to which subjects use
..."
sImIlar
mlormatlon
decisions, or do not use heuristic and systematic processesto
check eachother. Suchdisruptions
can come about, for examh
...
d "
when
..
subjects
about exactly which of these factors might operate under
various conditions and involve various brain areas. New developments in neuroimaging techniques provide increasinglyprecise and detailed anatomical information about cognitively impaired patients (e.g~Squire, Amaral, & Press,1990).To maximally exploit theseexciting technical developments we need to
use a correspondingly sophisticated characteriza~ionof the
mental processesinvolved in the tasks we stud~
Source Monitoring
.In a Genera
W .th
1M
emory
A h.
rc
Itecture
for the two types of tests. Because criteria
are flexible and individually determined and the number of
subjects per group is typically small, the size of these correlations may not be consistentacrosslabs or even within labs.6
Linking aspects of source memory to different brain areas.
Although bothtemporal/diencephalic and frontal areascontribute to memory for source, they presumably do so in different
ways. One possibility is that temporal/diencephalic areas are
involved in consolidating (e.g~Milner, 1970;Squire,1987;ZolaMorgan & Squire, 1990)memory characteristicssuchas perceptual and spatial detail through a processof reactivating memory records (Johnson,1992; Johnson& Hirst, 1991;cf. Spear&
Mueller, 1984)and that frontal areasare involved indiscoverin g
relations among events,strategic retrieval, and setting criteria
in memory tasks (e.g.,Johnson,1990; Moscovitch, 1989).As we
!lave argued, all of these processesare important for both content and source memory. The fact that they may be important
to different degrees in different situations produces the variable relation betweencontent and sourceand betweenpatterns
shown by memory-disordered patients with and without frontal damage. A more general point is that given the complex
processesinvolved in establishing particular memory characteristics relevant to source (e.g.,perceptual, temporal, spatial,
and affective;seethe next section,SourceMonitoring Within a
General Memory Architecture) and the complex processesinvolved in source monitoring, it is unlikely that any single brain
area is alone responsiblefor memory for source (see,also Mayes
et al., 1985; Mishkin, 1982).
In summary,the source-monitoring framework suggeststhat
disruptions in source monitoring can occur for a number of
reasons,which fall into twObroad categories: (a)disruption in
the encoding of events initially-for example, disruption in
processesthat might limit encoding or consolidation of some
types of perceptual, contextual, affective,semantic,and cognitive operations information-or disruption in retrieving and
noting relations between prior and ongoing events that would
severelylimit relational information and (b) disruption in heuristic or systematicsource-monitoring judgment processes-for
example, when subjectsdo not accessavailable relevant information, use lax or inappropriate criteria for making source
~Ie,
19
at elt
er .acqulsltlon
or test
are .presse
lor
time, stressed,depressed.,distracted, or under the Influence ?f
alcohol or other drugs or lfthey havesuffereddamagetocertam
areasof the brain. Given these more specific hypothesesabout
the cognitive mechanisms underlying source monitoring, it
should be possible to design studies that are more analytic
Our primary goal in this article is to illustrate that a wide
variety of phenomena can be usefully discussed in a common
setof terms, which suggestadditional directions for empirical
research.For the future, greater theoretical specificity in characterizing the memory records and processes involved in
source monitoring will require embedding discussions of
source monitoring within the context of a specific processing
model of memory. We briefly consider what the outlines of such
an approach might look like, using the Multiple-Entry, Modular
(MEM) memory framework (Johnson,1983,1991a;Johnson&
Hirst, 1991, in press; Johnson& Multhaup, 1992).
The aim of the MEM framework is to describe a relatively
small setof cognitive processes(the component processes)that
are jointly sufficient to account for changes in memory and
performance from cognitive activit~ That is, any task involves
some combination of these component processes,and ~
is viewed as a record of the operation of the component processes.At present, MEM consistS-of16 component processes,
T:ar-geiY
derived from prior findings or concepts in the cognitive
literature. These component processesare grouped into four
proposed functional subsystems,two (P-I and P-2)that arecomposedof perceptualprocessescontributing to memory records
(e.g.,those involved in seeing)and two (R-I and R-2) that are
composed of reflective processes,such as the self-generated
activities involved in organizing and elaborating. The component processesof each subsystemare P-I (locating, resolving,
extracting,and tracking), P-2 (placing, identifying, examining,
and structuring), R-I (reactivating,refreshing,shifting, and noting), and R-2 (retrieving,rehearsi(lg,initiating, and discovering).
Although specific processesfrom two or more subsystems
may be analogous,in relation to P-I and R-I, the processesof
P-2and R-2 operate on more complex data structures (Johnson
& Hirst, in press).So,for example,locatingcan be applied to an
undifferentiated external stimulus (Weiskrantz, 1986)whereas
placing can compute relative positions of two differentiated and
usually identified objects. Similarly, noting can compute overlapping relations from associationsactivated by two items (e.g.,
dog and cat both activate animal), whereas discoveringfinds
relations that are lessdirect, for example,relations that depend
..
6SIml 1ar I y, Sagar an d co II eagues
Sagar, Gabrieli,
Sullivan,
& Corkin,
(S
D
& Mayes, 1991.,
agar,
ownes,
1990) have discussed
the possibil-
ity thatsourcejudgmentsmightbe madein differentways-depending on suchfactorsas the severityof a patient'srecognitiondeficit,
etiologyof theamnesicdeficit,andthenatureof the taskandmatching
procedureused (cf. Schacter,Kaszniak, I<;ihlstrom,& Valdiserri,
1991).
-
M. JOHNSON,S. HASHTROUDI, AND S. LINDSAY
on other relations as in computing analogies (e.g., Gentner,
1988).The two reflectivesubsystemsalso include the capability
of "executive" controJ and monitoring functions (e.g.,Miller,
Galanter, & Pribram, 1960; Norman & Shallice, 1986;Stuss&
Benson, 1986). Control and monitoring functions are drawn'
from setsof potential agendas.An agendacomprisesa recipe or
program for cognitive action consisting of component subprocesses.Agendasare often learned ana are activated by ongoing
stimulus conditions, including ongoing cognitive activities.
These component processesand their proposedstructure serve
as a basis for modeling a variety of memory phenomena(JohnI son, 1983, 1990,I991a, 1991b,1992; Johnson& Hirst, 1991,in
relations among them) combine to determine what characteristics are encoded in memory. For example,affective information
mayarise from perceptualprocesses(e.g.,feararising from suddenly locating a stimulus in your peripheral vision) or from
more reflective processes(e.g.,fear arising from discoveringan
inconsistency in your argument just before it is your turn to
speak; Johnson& Multhaup, 1992).Similarly, temporal aspects
of memory might arise from perceptual records of identified
objects(e.g.,a moon in the memory implies that it was night) or
from reflective records of noted relations (e.g.,"I remember
thinking that I also sawhim as I left the seminar last week,and
the seminar meetson Wednesdays").With respectthen to tem-
press; Johnson& Multhaup, 1992).
Further researchmay necessitateadding additional component processes,and much more work will be neededto clearly
define the nature of particular componentsand to unambiguouslydifferentiate among them (e.g.,to know whethera particular performance involves locating or placing). Despite its provisional nature,a component-processesapproachto memory has
many potential benefits, especially compared with simple dichotomies such as conceptualizing processesas automatic o~,
controlled, or data driven or conceptually driven or conceptualizing the content of memories as procedural or declarative,or
i semantic or episodic (Johnson & Hirst, in press). A relatively
small increasein theoretical complexity in moving from anyof
poral information, records of perceptual activity may be sufficient to indicate that an event took place at night, but recordsof
reflective activity may be necessaryfor specifying that an event
took place on Wednesdayor after another particular event (cf.
Tzeng,Lee, & WetzeJ,1979).
In short, there is not a single mental computation, identifiable with a single brain region, that is responsible for affect or
for temporal information (cf. alton, 1989),spatial information
(cf. Smith & Milner, 1984)and so forth. Furthermore, the MEM
framework can be usedto help characterizethe likely complexity of the mental computations involved in encoding, reactivating, and using various memory characteristics for various purposes.An example is our earlier suggestionthat frontal damage
these dichotomies to a component-processesframework, such
as MEM, provides substantial increasesin both specificity and
I generality.
I
MEM provides a mid-level vocabulary for integrating work
from more specific theoretical and empirical efforts. For exampIe, locating is explicated by researchon visual-spatial attention (e.g.,Yantis & Johnson, 1990), identifying by researchon
object
(e.g.,effects
Biederman,
1987),react.ivating
by researchidenti~cation
on reInstatement
(e.g~Rovee-Colller
& Hayne,
disrupts processessuch as retrievingand discoveringrelations
that would help specify that an eventoccurred on Wednesdayor
after another related event, whereas temporal/diencephalic
da!Dagedisruptsreactivationprocessescriticalfor the consolidation evenof within-event characteristics suchas the perceptual
information indicating that an event occurred at night. This
would account for the profound absenceof source-specifying
information
in patients
age
and frontal
damage:with combined medial/temporal dam-
",,!
,~c"'1
'": !
I
'
c
I' i".:
1987),and rehearsingby researchon working memory (Baddeley,1986). At the same time, MEM componentsthemselvescan
be used to decompose other concepts. For example, organization and elaboration (e.g.,Tulving, 1962) can be described in
Another aspectof MEM that is relevant to conceptualizing
source attribution, and decision processes in general, is that
MEM divides reflection into two subsystems,R-I and R-2, each
capable of executivecontrol and eachable to recruit processes
1
terms of specific component processesofMEM suchas refreshing, shifting, no~ing,and reactivating ~Johns.on,
.1990, 1992).
MEM also provides a vocabulary for dIscussingIssuessuch as
I the cognitive activities required for the binding and consolida-
for prolonging activation (rt:freshingand rehearsing),for introducin~ change in an .activ~tion pattern (shifting and initiating),
for going backto earlier objectsof perceptionand thought (reactivating and retrieving),and for creating relations among events
f';:
i
tion of central and contextual information (Johnson, 1992),
processesfundamental to determining cue effectiveness.MEM
I not only offersparsimonious language for describing a rangeof
phenomena,it also provides rich enough detail to make predictions about the breakdown of functioning from brain damage,
aging, and psychological distress (Johnson,1990; Johnson &
Hirst, 1991,in press).
With respectto source monitoring, MEM can be useful in
conceptualizing both the memory characteristicsand the deciprocesses
discussedaffective,
here.Thesemantic,
memoryand
characteristics
(i.e.,
I sion
perceptual,
contextual,
cognitive oper-
(noting and discovering).Compared with R-I processes, R-2
processesare more controlled or deliberate (systematic).The
more a task draws exclusivelyfrom the set of R-I components
(and the fewer the components engaged),the lessdeliberate it
will seem; the more it draws from R-2 components (and the
more components it engages),the more deliberate it will seem.
However, R-l and R-2 do not map simply onto two discrete
categories,nondeliberative and deliberative (or automatic and
controlled), or to any simple dimension. As noted by Chaiken
et
al. (1989),
a "strict many
automatic-controlled
dichotomy
too
restrictive
to capture
phenomena of interest"
(p. is213).
ations)postulated by the source-monitoring frameworkaresummary labels for the outcomes of combinations of underlying
cognitive activiti~s. Within M.EM: vari.ou.scomponent.sub~rocessesof perceptIon(e.g.,locating,Identifymg,and trackmgstlm.uli and placing objects in spatial relations) and of reflection
.(e.g;, rehearsing,reactivating,and retrieving events and noting
7Of course,medial/temporal
and frontalareasoflhebrainarecomplexstructures;differentcomponentprocesses
of MEM verylikelyare
differentiallysupportedby ordistributedacrossvarioussubstr
uctures
of thesemoregeneralregions.
I
1
'
i
i
!'
~
"
l
I
i
i
J
.~~
: -~c,.~'~:~;;;:
~_c
'",,"""
'.~
..
SOURCE MONITORING
Thus, rather than using terms that imply a unidimensional analysisof cognition, R-I and R-2 can be usedasshorthand terms to
refer, respectively,to prototypical types of heuristic (relatively
automatic) and systematic (relatively controlled) processing
drawn from a multiply determined family of processingpossibilities (e.g.,Johnson,1991a;Johnson& Hirst, in press; Kahan
& Johnson,1992).An advantageof the terms R-I andR-2 is that
they are not proposed as primitives and can remind us that the
task ahead is to further decomposeconcepts suchas "control"
into component processessuchas those specified in MEM.
Dividing reflection into two cohesivesubsystems(R-I and
R-2) has a number of advantagessimilar to advantagesgained
from computer software designed to execute as cooperating
processes:It easily allows two agendasto be engagedat once.
For example,source monitoring can simultaneouslyinvolve assessing qualitative characteristics of activated information
through R-I processesand retrieving additional information to
check plausibility through R-2 processes.Dividing reflection
into two subsystems provides a way of understanding how
some aspects of reflection might be disrupted without fully
disorganizing thought (e.g.,Johnson, 1991a; Johnson& Hirst,
in press). Hence, the disruption of R-2 processesevident in
divided attention tasks or in some frontal patients may disrupt
some aspectsof source monitoring without disrupting other
aspectsof memor)( An equally important idea represented in
MEM is that R-I and R-2 also normally interact, that is, they
all on each other's agendasand evaluatethe outcomesof processes.Thus the outcomesofheu~i~tic and system~ticprocesses
can correct eachother,as when VIVidmental experiencesdo not
passa plausibility check in source monitoring. More generally,
working together,.R-l and R-2 yield the ex~eri~nceofthinki.ng
about thinking. Such a conceptual organIZatIon of reflectIve
processesprQvidesa way of discussing how the phenomenal
experience of control, intention., and agency might arise and
suggests mechanisms underlYIng self-awareness (Johnson,
1991b; Johnson& Hirst, in press).
We believe that this component-processapp~oac~hasadv~ntages over both structural models and processIngdichotomies.
The brief overviewofMEM offered here is not intended to fully
explicate the model but rather to communicate the generalflavor of the component-processapproach. Interested readersare
referred to Johnson(1990,1991a,1991b, 1992)and Johnsonand
Hirst (in press)for more detailed presentations.
21
work for understanding the processesinvolved in source monitoring, A consideration of recent findings and theorizing in a
number of domains suggeststhat this framework can point to
connections among diverse phenomena. Ultimately, however,
the valueof sucha framework lies in the extent to which it helps
explain otherwise confusing results and guides future research
in a meaningful direction.
We believe that the framework has been helpful in accounting for disparate results. For example, the framework can explain why under somecircumstances performance on recognition memory and source tests are dissociated, whereas under
other conditions they are correlated. In addition, it predicts the
dissociations as well as parallel effects found in studying the
relation betweendirect and indirect testsof memor)( Similarly,
it can account for why in some situations memory-disordered
patients have difficulty in remembering both source and content, whereasin other situations the source deficit occurs in the
absenceofa deficit in remembering content. The framework
also suggestsfactors that determine whether misleading information will affect memory for recently experiencedeventsand
whetherpeople will misattribute familiarity to prior knowledge
(e.g.,the false-fame effect), will show cryptomnesia, or will be
influenced by fictional information.
As a general researchstrategy,the framework suggestsconsidering both memory characteristics and decision processesin
source monitoring and encourages investigating different
source-monitoring situations under comparable conditions or
in the same su~jectpopulations. It also suggestsseveralsp.e~ific
researchquestIons. For example, can the boundary condItIons
for appropriating fictional information into one's knowledge
?aseor u~ing it late~be.related t? specific component processes
Involved m processIngInformation? To what extentdo developmental changes in source monitoring reflect growth of more
compl~~ heuristic a~d s.ystemati~reflective processing,s':lchas
the abIlity to combmemformatlon from two or more dlmensions or the ability to engagein strategic retrieval? Cana sourcemonito.ring analy~i~~ontribut~ to th.eun~erstand.ingof differences m susceptIbIl!ty to misleadIng Information between
young adults and chIldren or between young adults and older;
adults?To what extent do deficits in source monitoring experienced by older adults arise from difficulties in encoding or
reviving particular aspects of memories or from the criteria
they adopt in source-monitoring decisions?Can.differences in
source-monitoring problems among various braIn-damaged or
populations be identified and classified?
And can differences betweenpatients in source amnesia,types
.memory-disordered
Conclusions
Source monitoring is a critical everyday memory function.
The feeling that a memory hasa specific sourceis a compelling
reason for experiencing it as belonging to our personal past.
Furthermore, source-monitoring processesdifferentiate fact
from fantasy in remembering (e.g.,whether your neighbors
yelled at you or you only imagined they might), reliable from
unreliable sourcesof information (e.g.,your doctor vs. the Nationa/ Enquirer for nutrition tips), and actions from intentions
of delusions, confabulation, source misattribut!ons, and oth~r
source-monitoring phenomena be understood m terms o~d!fferent combinations of disruption of memory charac~nstlcs
and decision processes?Finall~,can pr~gres~be ma~e m relating factors that.co~nitive studle~ have ~dentIfiedas Important
f~r source mo.nltonng to underlYIng.braIn ~t~ucturesor mech~msms?Questions suchas thes.e.provlde~x~ltmgne~ opportumties for understanding the cntIcal cognitive functIon of source
(e.g., taking pil.ls vs: only think.ing ~bou~ it).. ~isruPti<:>n many
or all of these situations has senous ImplicatIons both m potential impact on event memory and on develQpment and use of
monItorIng.
knowledge and beliefs.
In this article, we outlined an integrative theoretical frame-
Abra,J.C. (1972).List differentiationandforgetting.In~. P.Dun~a~,
L. Sechrest,& A. W Melton (Eds.),Humanmemory.FestschriftIn
...\
1
i
:i
;
;
i,
1
~
1
'1
~
i
I
'1
~
i
!;
oj
:!
'
j
1
.
~
t
~
I
t
l: '
I
.j
"'::
'
j
i
'
:
,'
;
,
§
~
"
~:
~;
References
.~
~
~
I.~,
I.