When False Recognition Meets Metacognition: The

Journal of Memory and Language 46,782–803 (2002)doi:10.1006/jmla.2001.2822, available online at http://www.academicpress.com on

0749© 200All rig

When False Recognition Meets Metacognition: The Distinctiveness Heuristic

Chad S. Dodson and Daniel L. Schacter

Harvard University

We investigated the contribution of a distinctiveness heuristic to rejecting false memories. Individuals studiedwords, pictures, or both types of items and then completed a recognition test on which the studied items appeared

lsely rec-ures pro-absence the in-uristic.udiedected re-pectively.’s old-cognitiver reduc-

A arsh,

Granenceon th

AdDodverscdod

once, whereas the new words appeared twice. Participants who had studied pictures were less likely to faognize repeated new words than were participants who had studied words. We argue that studying pictvides a basis for using a distinctiveness heuristic during the recognition test; participants infer from the of memory for expected picture information that a test item is “new.” These experiments also investigatedfluence of two variables—diagnosticity and metacognitive control—on the use of the distinctiveness heWe examined the role of diagnostic information in eliciting the heuristic by varying the proportion of stitems that appeared as pictures. Compared to a word encoding condition, participants successfully rejpeated new words after studying 50, 25, and 33% of the items as pictures in Experiments 1, 2, and 3, resThus, the distinctive information need not be completely diagnostic (i.e., perfectly predictive of an itemness) for participants to use the heuristic. We also show that the distinctiveness heuristic is under metacontrol such that it can be turned on or off depending on participants’ expectations about its usefulness foing memory errors. © 2002 Elsevier Science (USA)

Key Words:false recognition, metacognition, distinctiveness heuristic, and false memory.

lthough memory is often durable and accu-Roberts, & Seamon, 1997; Hicks & M
eu
oeda

a

l

al,r,

6;r-

re-

,inodem-ke

gni-emo-

e to

theeri-n

rate, it is also subject to various types of forgting and distortion (Schacter, 1999, 2001). Ding the past several years, increasing experimtal and theoretical attention has focused on mattribution errors that occur when some formmemory is present but is attributed to an increct time, place, or source (for reviews, sJohnson, Hashtroudi, & Lindsay, 1993; Roeger, 1996; Schacter, Norman, & Koutsta1998). Demonstrations of errors and distortioin remembering raise a question with importatheoretical and practical implications: How cmemory misattributions be reduced or avoideSeveral studies have shown that a number ofcoding and retrieval manipulations can produreliable reductions in memory errors such false recall and false recognition (e.g., Gal

This work was supported by National Institute on Agi

-596X/02 $35.002 Elsevier Science (USA)hts reserved.

78

byd bye

m,

on-

t AG08441 and a grant from the Human Frontiers S Program. We are grateful to Carrie Racine for her his project.dress correspondence and reprint requests to C

son, Department of Psychology, P.O. Box 400400, Unity of Virginia, Charlottesville, VA 22904-4400. E-mail:[email protected].

2

t-r-en-is-ofr-ei-l,nsntnd?en-ceaso,

1999; Kensinger & Schacter, 1999; KoutstaSchacter, Galluccio, & Stofer, 1999; MatheHenkel, & Johnson, 1997; McDermott, 199McDermott & Roediger, 1998; Schacter, Vefaellie, Anes, & Racine, 1998).

We recently suggested one mechanism forducing misattribution errors that we call the dis-tinctiveness heuristic(Dodson & Schacter2001; Schacter, Cendan, Dodson, & Clifford,press; Schacter, Israel, & Racine, 1999), a mof responding in which people expect to remeber vivid details of an experience and marecognition decisions based on this metacotive expectation. When a novel event or itlacks the expected distinctive information, peple can use this absence of critical evidencreject the item.

We provided evidence for the operation of distinctiveness heuristic in three sets of expments using a procedure originally developedDeese (1959) and later refined and extendeRoediger and McDermott (1995). In thDeese/Roediger-McDermott (DRM) paradigparticipants hear lists of words (e.g.,candy, sour,sugar) that all are semantic associates of a n

gci-elp

hadi-

E

o

sfireio alie

hrr

olscc

ats o-th&uoae

imreo

ti

cthh

y8ythede-

senc-oid

i-nc-en-onssre-ngub-vi-rets

diedrds.nti-aidub-n.

son de--

-id atemab-ro-notbyi-

ntsordsistsnot

est is

mcon-en-sup-is-r-onot

However, there is a confounding feature of

THE DISTINCTIV

presented theme or lure word (e.g.,sweet). Whenlater given an old–new recognition test that ctains studied words (e.g.,sour), new unrelatedwords (e.g.,point), and new related lure word(e.g., sweet), participants frequently and condently claim that they previously studied the lated lures. This robust false recognition effhas been documented and explored in varlaboratories (e.g., Gallo et al., 1997; Mather et1997; Norman & Schacter, 1997; Payne, EBlackwell, & Neuschatz, 1996; Schacter, Verfalie, & Pradere, 1996; Smith & Hunt, 1998).

Schacter et al. (1999; see also Israel & Scter, 1997) modified the DRM procedure by psenting each word in an associated list auditoalong with a picture of the item. Compared tcondition in which participants studied onwords (in both visual and auditory modalitiefalse recognition of related lures was redudramatically following pictorial encoding. Schater et al. (1999) argued that the reduction in frecognition was attributable to participanmetacognitive expectation that they shouldable to remember the distinctive pictorial infmation. Thus, the absenceof memory for this distinctive information provides evidence that test item is new (cf. Rotello, 1999; Strack Bless, 1994). By contrast, participants who stied words would not expect detailed recollectiof studied items and, hence, would not brecognition decisions on the presence or absof memory for such distinctive information.

Dodson and Schacter (2001) reported a slar reduction in false recognition of related luafter participants said aloud target words study lists compared to when they heard theget items (participants also saw the studwords in both conditions). Dodson and Schanoted that earlier studies provide evidence people expect to remember information tthey have generated themselves (ConwayGathercole, 1987; Foley, Johnson, & Ra1983; Hashtroudi, Johnson, & Chrosniak, 19Johnson, Raye, Foley, & Foley, 1981; KelleJacoby, & Hollingshead, 1989). They suggesthat in the DRM procedure, participants wsaid words at study employed a distinctivenheuristic during the recognition test; they
manded access to the distinctive “say” inform
NESS HEURISTIC 783

n-

-e-ctusl.,e,l-

ac-e-ily ay),ed-lse’

ber-

e

d-nssence

i-sn

ar-edterat

at&

e,9;,

edoss

tion in order to judge an item as “old.” Becaurelated lure words were never said, the distitiveness heuristic helped participants to avfalsely recognizing them.

Although our previous studies provide evdence consistent with the operation of a distitiveness heuristic, they leave open a fundamtal question: What are the necessary conditifor eliciting or “turning on” the distinctivenesheuristic? Schacter et al. (1999) reported duced false recognition after pictorial encodicompared to word encoding in a between-sjects design. However, they observed no edence of reduced false recognition for pictulists compared to word lists in a within-subjecdesign where some associate lists were stuas pictures and others were studied as woDodson and Schacter (2001) reported an idecal pattern—reduced false recognition for slists compared to heard lists in a between-sjects design but not in a within-subjects desig

As noted by Schacter et al. (1999) and Dodand Schacter (2001), in a between-subjectssign, distinctive information is perfectly predictive or diagnosticof prior study. If participants remember having seen a picture or having saword aloud, then they can be certain that the iappeared on the study list. Conversely, the sence of the expected distinctive information pvides diagnostic evidence that the item did appear in the list. In a within-subjects design,contrast, distinctive information is no longer dagnostic of prior study. Because participastudied some lists as pictures and others as w(Schacter et al., 1999) or said aloud some land heard others (Dodson & Schacter, 2001),remembering distinctive information about a titem does not necessarily mean that the itemnovel; it might mean only that the item was froone of the lists presented as words. Thus, the trasting patterns of false recognition in betweand within-subjects designs can be taken as port for the idea that participants rely on the dtinctiveness heuristic only when distinctive infomation is diagnostic of prior study and abandthe heuristic when distinctive information is ndiagnostic of prior study.

a-the DRM procedure that creates two different

D

gioinpingaeo rr

lid-a

thpennh

ssro

ec rathneinnenrgehinyrit

lythe).en

ro-edal.hendtsryct

g-Mm-ss

ine

ad-on-ndcicvalem-

is--

alat-ent

n-ryalcecy.elsp-yls,i-ch

o,&of-

784 DODSON AN

interpretations of the aforementioned findinof no difference between the false recognitrates for the picture and word lists in the withsubjects design. As we just noted, one intertation is that participants abandon the disttiveness heuristic because it is no londiagnostic. Alternatively, participants may be tempting, but failing, to apply the distinctivness heuristic selectively to the test items frthe lists presented as pictures. Consider, forample, a participant who studies pictures cosponding to the heard words butter, dough, milk,and so forth and then encounters the relatedword bread on the recognition test. To avomaking a false alarm to breadbecause of the absence of expected pictorial information, the pticipant would have to be able to remember butter, dough, milk, and other associates apeared with pictures in the study list. Howevif the participant cannot remember the presetion mode of the study list—and, thus, canremember which of the lists for which theuristic would be useful—then he or she mapply the heuristic globally to all of the teitems. This global use of the distinctiveneheuristic would produce the finding of no diffeence in false recognition rates to lures assated to the picture and word lists.

This latter point is important theoretically bcause it implies that failure to observe redufalse recognition for pictures or said wordsthe within-subjects version of the DRM padigm might not provide evidence supporting importance of diagnostic information in turnion the distinctiveness heuristic. Instead, thdata may speak to the difficulty of remembermode of list presentation when making recogtion judgments about related lure words. IndeDodson and Schacter (2001) noted evidefrom their experiments and those of Schacteal. (1999) suggesting that participants mihave been trying to use the distinctivenheuristic in the within-subjects version of tDRM paradigm, even though the distinctive formation was not diagnostic of prior studThey observed that in the within-subjects padigm, levels of false recognition in all condtions were suppressed compared to the “s
dard” or nondistinctive condition in the
SCHACTER

sn-

re-c-ert--m

ex-e-

ure

r-at-r,ta-oteayts-ci-

-edin-e

gsegi-d,ce ethtsse-.a--an-

between-subjects design (i.e., the word-onpresentation in Schacter et al., 1999, and heard condition in Dodson & Schacter, 2001This cross-experiment observation has beconfirmed by Schacter et al. (in press), who pduced in a single experiment the trends obtainin separate experiments by Schacter et (1999) and Dodson and Schacter (2001). Tfact that no additional suppression was foufor pictures or said words in the within-subjeccondition may be attributable to the memolimitations noted above rather than to the fathat distinctive information was no longer dianostic of prior study. Thus, data from the DRparadigm cannot speak conclusively to the iportance of diagnosticity for the distinctiveneheuristic.

In the experiments reported here, we examthe role of diagnostic information in elicitingthe heuristic. The current experiments also dress the more general question of the relatiship between the distinctiveness heuristic afamiliarity and recollection of source or specifiitem information. The distinctiveness heuristis one instance of a general class of retriestrategies that participants use to assess rembered information. (we further consider the dtinctiveness heuristic in relation to similar retrieval strategies later in the GenerDiscussion). As a retrieval strategy for evaluing remembered information, our view of thdistinctiveness heuristic is very much consistewith the source-monitoring framework of Johson et al. (1993) and the constructive memoframework of Schacter, Norman, and Koutsta(1998), both of which emphasize the importanof evaluative mechanisms for memory accuraHowever, in the context of dual process modof memory, the distinctiveness heuristic concetually differs from the processes of familiaritand recollection. Specifically, in most modefamiliarity is based on a unidimensional varable that is produced by a variety of factors suas the overall similarity of the familiar item tother items in memory (e.g., Gillund & Shiffrin1984; Hintzman, 1988; Humphreys, Bain,Pike, 1989; Murdock, 1982), the frequency prior exposure of the familiar item (e.g., Atkin
son & Juola, 1974; Underwood, 1972), and the

EN

.ggho

ntr

;9osmy

rre Inccessp

m

,tras&mm

iepentii

mwerhrda

reob

ngrtldenl(

n-ndnitt at

atstheer-as testrdsls,

ci-ord

il-iarity—derived from previous exposure on the

etld

ssoidhe

ar-r-re-edstly. Ifro-c-ionusngssu-a-ir

od-heu-

n-si-

ic-

THE DISTINCTIV

fluency of processing the recognized item (eJacoby & Dallas, 1981). Recollection is thouto involve memory for more specific item source information, such as multiple featuresan event, or a remembered event and the coin which it occurred (e.g., Anderson & Bowe1972, 1974; Gillund & Shiffrin, 1984Humphreys et al., 1989; Johnson et al., 19Mandler, 1980). There is sometimes an opption relationship between familiarity and meory for more specific item information (Jacob1991) such as when familiar items can bejected when more specific information is membered (cf. Dodson & Johnson, 1996).this article, we provide evidence that the distitiveness heuristic constitutes a separate prothat contributes to performance when peoconfront a familiar test item but fail to remeber expected source information about it.

To address the aforementioned questionsuse a repetition lag paradigm that was induced by Underwood and Freund (1970) modified by Jennings and Jacoby (1997; also Fischler & Juola, 1971; Koriat, Ben-Zur,Sheffer, 1988). In the repetition lag paradigparticipants study a list of words and then coplete a recognition test in which the studwords appear once but the new words aptwo or more different times. Underwood aFreund (1970) used a forced choice recognitest in which each studied word was paired wa new word that either occurred for the first tior was repeated. Recognition performance substantially lower when studied words wpaired with repeated new words than when twere paired with once-presented new woJennings and Jacoby (1997) changed the pdigm in two ways. First, all test words were psented individually, and a judgment of “old”“new” was made for each. Second, the lag tween the first occurrence of a new word andrepetition was systematically varied. Jenniand Jacoby observed that even though papants were specifically instructed to say “oonly to words from the study list and not to nwords that are repeated, after sufficiently lolags, some participants (especially older adumade false alarms to repeated new words
similar results with the same paradigm, se
ESS HEURISTIC 785

.,t

rofext,

3;i-

-,e--n-

Dywan, Segalowitz, & Webster, 1998). As Jenings and Jacoby (1997) argued, familiarity arecollection of more specific item informatioexhibit an opposition relationship when comes to responding to new words that repeadifferent lag intervals. When a new word repeafter just a few test items (i.e., a short lag),repeated new word’s familiarity can be countacted by recollecting that the new word wseen earlier on the test—a clear sign that theitem must be new because only the new worepeat. By contrast, at the longer lag intervathere is an increasing likelihood that partipants fail to recollect that the repeated new wwas encountered before and mistake its fam

le-

weo-ndee

,-

dar

dontheaseeys.ra--

re-itss

ici-”

wg

ts)for

test—for prior presentation in the studyphase.

EXPERIMENT 1

We began by asking whether studying targitems as pictures, rather than words, wouallow participants to invoke a distinctiveneheuristic on the recognition test and, thus, avmaking false alarms to repeated new words. Tgeneral idea is that after studying pictures, pticipants would expect to remember this infomation during the recognition test. Because peated new words lack this expectinformation, by relying on the distinctivenesheuristic, participants should be able to greareduce false alarms to repeated new wordssuch an outcome were observed, it would pvide evidence for the generality of the distintiveness heuristic by demonstrating its operatbeyond the DRM paradigm used in previostudies. We replicated the procedure of Jenniand Jacoby (1997) by presenting the words vially only, whereas the pictures were accompnied by the auditory presentations of thenames. Because any effects of distinctive encing could be a result of either the picture or tauditory information, we refer to the picture/aditory encoding condition as the distinctive en-codingcondition.

We also examined the role of diagnostic iformation in turning on the distinctivenesheuristic. Consider a situation where particpants study a list in which half the items are p
etures and half are words (i.e., 50% distinctive

D

teaig d tmteicteoio,em

a icpbnrt

ss s

s nniv

dp.dm

iteLhcu

yrere- in-r-ossand then-

nc-e-

eesinasionsech?” re- byextly

of of

ei-vi-hep-asesechn-ldwmsrd0s

nedeo-Un-ts

786 DODSON AN

condition) and are then given a recognition comprised of all words, with new words repeing at varying lags. Because the test item mhave appeared as a word at study, failing tomember distinctive information is no longer agnostic of prior study, as was the case inwithin-subjects version of the DRM paradigused by Schacter et al. (1999). However,repetition lag paradigm provides a more dirtest of the role of diagnostic information in eliting the distinctiveness heuristic than does DRM paradigm because the former procedurnot confounded by the “memory for mode list” presentation issue that blurs interpretatof results from the latter procedure. That isthe repetition lag procedure, repeated nwords are not associated with any specific itefrom the study list in the same sense that rellure words are linked to specific study liststhe DRM procedure. After studying both ptures and words, it seems reasonable to exthat participants will not heavily weight the asence of memory for distinctive informatiowhen making recognition judgments. Therefothey should fail to reduce false recognition raof repeated new words relative to participantthe 100% distinctive encoding condition. Theresults would clearly support the importancediagnosticity in turning on the distinctiveneheuristic.

Method

Participants. A total of 48 paid volunteerwere recruited from the student populationHarvard University. There were 16 participain each of the three conditions: word encodi100% distinctive encoding, and 50% distinctencoding.

Design and materials. The stimuli consisteof 120 Snodgrass and Vanderwart (1980) tures and their corresponding verbal labelstotal of 60 items were studied and also servethe old items on the test. The remaining 60 itewere the new items on the test. Each new repeated at either Lag 4, Lag 12, Lag 24, or 48. The 120 stimuli were divided into eiggroups of 15 items. The groups were balanso that they had similar mean ratings for pict
familiarity (range 5 3.5–3.6), picture complex-
SCHACTER

stt-htre-i-he

hect-he isfninws

tedin-ect-

e,es ine

ofs

attsg,e

ic- A as

sm

agtedre

ity (range 5 2.6–2.7), and word frequenc(range 5 34.2–34.9). Four groups of items wepresented at study, and four groups were psented as new items on the test at the four lagtervals. Eight different counterbalancing fomats rotated the groups of items so that acrparticipants, each group appeared at study also was presented as a new word in each oflag conditions at test. This rotation also guarateed that across participants in the 50% distitive condition, each group of items was prsented as a word and picture at study.

An Apple G3 computer presented all of thstimuli in the center of the screen. The picturwere approximately the same size and fit witha 6- 3 6-in. area of the screen. Each picture walso accompanied by the auditory presentatof its name. The words appeared in lowercaletters in size 48-point Geneva font. For eastudy item, the phrase “How many syllablesappeared at the bottom of the screen. After asponse, the screen cleared and was followeda 1-s delay before the presentation of the nstudy item. The 60 study items were randomintermixed with the restriction that each third the study list contained an equivalent numberitems from each group.

The test items were visual words only that ther corresponded to the names of the preously studied pictures or exactly matched tpreviously studied words. Each test item apeared in the center of the screen in lowercletters in size 48-point Geneva font. The phra“Old or new?” appeared 1.5 cm beneath eatest item. The order of the test items was radom, with the restriction that no more than 3 oor new items could occur consecutively. All newords repeated after either 4, 12, 24, or 48 iteseparated the initial occurrence of the new wofrom its repetition. In total, the test was 18items long: 60 study items and 60 new itemthat repeated at one of the lag intervals.

Procedure. Participants were assigned to oof three different study conditions: the worcondition, the 100% distinctive condition, or th50% distinctive condition. In contrast to the prcedure of Jennings and Jacoby (1997) and derwood and Freund (1970), our participan
were given incidental instructions and were told

EN

om

esont

ndnth

onor uu ahaeatenntic ret s

ndi-t

ah

s

edeirlinedy-fouratest,n-re-

es toressese on

the

fbe-as--thei-fortestre-

gni-

THE DISTINCTIV

that they would have to enter as quickly as psible the number of syllables in the study iteNo mention was made of a later memory tThe people in the 50% distinctive conditiwere told that they would see items preseneither as pictures or as words. Both these ividuals and the participants in the 100% distitive condition were told that they would hear name of each picture.

Immediately after the study phase, everyreceived the test instructions. They were tthat the test was based on their memory foof the studied items. The participants who stied pictures were informed that the test wocontain names of the picture study items. Indition, all of the participants were instructed tthe test would contain new words and that thnew words would repeat so that they would pear two different times. They were instructo respond “old” to the studied items only a“new” to the new items by pressing the “a” a“;” keys, respectively. We emphasized to paripants that they should respond “new” to thepeated new words. They were warned nomistake repeated new words for studied item

Results and Discussion

Table 1 displays the probabilities of respoing “old” in the three different encoding condtions to studied items, new words, and repeanew words at the four lag intervals. There two notable patterns in these results. First,rates to studied items and false alarm rate
once-presented new words were comparab ent
“Distinctive” refers to conditions in which participants havtions of pictured studied items.

ESS HEURISTIC 787

s-.t.

edi-

c-e

eldalld-ldd-tsep-ddd--

to.

-

edreittole

lar overall item recognition. Second, comparto the false alarm rates to new words on thfirst occurrence (hereafter referred to as basefalse alarm rates), participants in the word stucondition greatly increased their “old” responses to repeated new words across the lag intervals (i.e., 16% baseline false alarm rvs 32% false alarm rate at Lag 48). By contraparticipants in the 100% and 50% distinctive ecoding conditions were more successful at jecting the repeated new words.

For item recognition, we examined hit ratto studied items, baseline false alarm ratesnew words, and corrected recognition sco(i.e., hits minus baseline false alarms). Analyof variance (ANOVAs) confirmed that therwere no differences between the conditionsany of these measures, all Fs (2, 45) , 1.09. Wealso conducted signal detection analyses ofhit rates and baseline false alarm rates with d8 asa measure of sensitivity and C as a measure obias. There were no significant differences tween the conditions on either of these meures,Fs , 1.00. Thus, overall old–new discrimination and bias were comparable across conditions. Within the 50% distinctive condtion, however, recognition rates were higher test items earlier studied as pictures than for items earlier studied as words (78% vs 61%,spectively),F(1, 15) 5 27.59,MSe5 .008,p ,.0001.

Figure 1 presents the corrected false recotion rates of the repeatednew words in the threedifferent conditions. These scores repres

rds

rvals in

tectionitems.

across the different conditions, indicating simi-false recognition rates to repeated new wo

TABLE 1

Probabilities of Responding “Old” to Studied Items, New Words, and Repeated New Words at Each of the Lag InteExperiment 1

Repeated new words

Study condition Studied New d8 C Lag 4 Lag 12 Lag 24 Lag 48

Word .72 .16 1.77 .27 .18 .22 .26 .32100% distinctive .71 .11 1.88 .37 .08 .12 .11 .1650% distinctive .70 .15 1.70 .32 .14 .17 .17 .22

Note. The “new” column contains false alarm rates to the first presentations of the new words. The signal demeasures,d8 (discrimination) andC (bias), are derived from the “old” responses to studied and once-presented new

e studied pictures, with the percentages referring to the propor-

7

fla

thtbnn

,

nlsov

%

e-

Asthea-

ore-here-d

ar-e-

est

x-degofcter

ach ofthe four lag intervals in Experiment 1. Lag refers to the number of test items separating the initial occurrence ofthe new word from its repetition. Vertical lines depict standard errors of the means.

after subtracting out the false recognition ratenew words on their first occurrence. The linezero, for example, is the point at which thereno difference between the false alarm rate repeated new words and the baseline false arate. As expected, participants in the word ecoding condition were likely to falsely recognize repeated new words, especially those repeated at the longer lags. Participants in 100% and 50% distinctive encoding groups,contrast, show considerably lower false recogtion rates to repeated new words. We examithe corrected false recognition rates with a(Condition) 3 4 (Lag) ANOVA that yielded asignificant effect of condition,F(2, 45) 5 3.55,MSe5 .031,p , .05; a significant effect of lagF(3, 135) 5 10.40,MSe5 .005,p , .001; andno significant interaction. Planned comparisoconfirmed the pattern in Fig. 1; corrected farecognition rates were higher in the word encing condition than in either the 100% distinctiencoding condition,F(1, 45) 5 6.23,p , .05, orthe 50% distinctive encoding condition,F(1, 45)
5 4.23,p , .05. There were no differences be
toatisorrmn--at

heyi-ed3

sed-e

tween the 100% distinctive encoding and 50distinctive encoding conditions,F(1, 45) ,1.00. The significant effect of lag reflects thgreater probability of falsely recognizing the repeated new words at the longer lag intervals. Jennings and Jacoby (1997) discussed, at longer lag intervals, there is a heightened probbility that participants will fail to recollect thatthe repeated new word was encountered befand will mistake its familiarity for prior presentation in the study phase. By contrast, at tshort lags, such as Lag 4 and Lag 12, the peated new word’s familiarity can be countereby recollecting that the new word was seen elier in the test. Because only the new words rpeat, such recollection is a marker that the titem must be new.

There are two important results from this eperiment. First, our finding of lower correctefalse recognition rates in the 100% distinctivencoding condition than in the word encodincondition conceptually replicates the results Schacter et al. (1999) and Dodson and Scha

88 DODSON AND SCHACTER

FIG. 1. Corrected false recognition rates in the three study conditions to the repeated new words at e

-(2001) in the DRM paradigm. As we argued in

E

ao

etiioootc idst torp

ci

h0

u

seio

a

u

n

ic,nndattes%

y-e-notnsethe

scores did not differ among the three condi-

usecedew

iss;

thelly inatn re-owtive

Weent

iederex-d-eets

ds.mss ase-ls

tic, re-tes

THE DISTINCTIV

the DRM paradigm, studying all of the itemspictures reduces the tendency to falsely recnize repeated new words because it providbasis for invoking the distinctiveness heurisParticipants in the distinctive encoding conditexpect to remember distinctive informatiabout studied items. When these individuals cfront a test item that is familiar (e.g., a repeanew word) but that fails to elicit memory for pitorial information, they infer that the test itemnew. However, as we will discuss in greater tail later, the distinctiveness heuristic appearbe used primarily for new words that repeathe longer lags, when participants are likelyfail to recollect encountering the words earliethe test. The second important result is that ticipants still use the distinctiveness heuristicsuccessfully identify repeated new words “new” in the 50% distinctive encoding conditioeven though the heuristic is no longer perfediagnostic of the items’ oldness. That is, partpants persist in using the heuristic even wheis not a reliable indicator that an item is new.

It might appear surprising, in light of thwell-known picture superiority effect, that wfound no difference in item recognition (i.e.,rates and false alarm rates) between the 1distinctive and word encoding groups. Howevto our knowledge, there is only one other stthat has examined recognition memory for ptures and words using a between-groups deand a verbal test. Jenkins, Neale and D(1967) observed no difference in recognitmemory between participants who had studpictures and participants who had studied wo(i.e., the names of the pictures). By contrmany studies have found more accurate recotion memory for pictures than for words onverbal test when the pictures and words wstudied together in a single list such as our 5distinctive condition (e.g., Dewhurst & Conwa1994; Madigan, 1983; Rajaram, 1993; Scarbough, Gerard, & Cortese, 1979). Similar finings of comparable memory for two classesmaterials when the stimuli are presented in atween-subjects manner, but not in a within-sjects manner, have been found by many searchers examining free recall performa
(e.g., McDaniel, DeLosh, & Merritt, 2000).
NESS HEURISTIC 789

sg-

s ac.nnn-

ed-se- toat

With respect to the distinctiveness heuristhowever, the lack of overall difference iold–new discrimination between the word adistinctive encoding conditions indicates ththe reduced corrected false recognition rafor repeated new words in the 100% and 50distinctive encoding conditions is not a bproduct of generally increased memory. Morover, the lower corrected false recognitiorates in the two distinctive conditions cannbe a result of a more conservative respobias in these conditions, as compared to word encoding condition, because the bias (C)

inar-

toasntlyci-n it

eeit0%

er,dyic-ignnon

iedrdsst,gni- aere0%y,or-d- ofbe-b-re-ce

tions.

EXPERIMENT 2

Experiment 1 has demonstrated that the of a distinctiveness heuristic produces redufalse recognition responses to repeated nwords, even when the distinctive informationnot completely diagnostic of the items’ oldnesparticipants seem to ignore the base rate ofdistinctive information. These results naturalead us to question the role of diagnosticityactivating the distinctiveness heuristic. At whlevel of unreliability will participants abandothe distinctiveness heuristic as a strategy forsponding to test items? More concretely, hmany of the studied items need to be distincbefore the heuristic is invoked?

We address this question in Experiment 2. used the same basic procedure as in Experim1 except that we varied the number of studitems that appeared as pictures. There wthree different groups of participants in this eperiment. As in Experiment 1, one group stuied only words—an exact replication of thword encoding condition in Experiment 1. Walso included two other groups of participanthat studied a mixture of pictures and worSome participants saw 25% of the studied iteas pictures, and others saw 10% of the itempictures. These latter two conditions were dsigned to reveal the point at which individuawould abandon the distinctiveness heuristhus producing false recognition rates to thepeated new words that are similar to those ra
in the word encoding condition.

D

sio 1

-2

15re

na t 1uore.,

tivompn ie

teowreveec

tehdic

dy

ano-edi-endiedtedni-

er-wem

od-

nd-d,on-lseewte,n-te the

ned

tly

790 DODSON AN

Method

Participants. A total of 48 paid volunteerfrom the Harvard University student populatparticipated in this experiment. There wereparticipants in each of the three conditions.

Design and materials. The stimuli were identical to those used in Experiment 1. The 1stimuli were divided into eight groups of items. Four of the groups of items were psented at study, and four groups served asitems that repeated at the different lag intervThe word encoding condition was identicalthe corresponding condition in Experimentparticipants were presented with 60 words ding the study phase. In the 25% distinctive cdition, one of the four study groups was psented as pictures and the rest as words (i.epictures and 45 words). In the 10% distinccondition, 6 items were randomly selected frthe four study groups of items to appear as tures with the constraint that no more thaitems could come from the same group. AsExperiment 1, each picture was accompanwith the auditory presentation of its name.

The study lists and test lists were construcin the same manner as in Experiment 1. Hever, in the 10% distinctive condition, thewere no pictures within the first or last fistudy positions because primacy/recency effmight increase the salience of the item.

Procedure. We used the same study and instructions that we used in Experiment 1. Tpeople in the 25% and 10% distinctive contions were told that they would see either p

ticipants have studied pictures, with the percentages refe

SCHACTER

n6

0

-ewls.o;r-n--15

e

ic-2ind

d-

ts

stei--

relative proportion of picture and word stuitems.


Because the word encoding condition is exact replication (i.e., identical stimuli and prcedure) of this condition in Experiment 1, wfirst compared performance in these two contions with the goal of combining them. Thword encoding conditions in Experiments 1 a2 are characterized by similar hit rates to studitems, false alarm rates to new words, correcrecognition rates, and corrected false recogtion rates to the repeated new words, all Fs ,1.40. To increase our power of detecting diffences between conditions in Experiment 2,merged the two word encoding conditions froExperiments 1 and 2 into a single word encing condition.

Table 2 presents the probabilities of respoing “old” to the different test items (i.e., studienew, or repeated new) in the three different cditions. We observed a large increase in farecognition responses to the repeated nwords, relative to the baseline false alarm rain both the word and 10% distinctive study coditions. Interestingly, the false recognition rato repeated new words was suppressed in25% distinctive condition.

In terms of overall item recognition, aANOVA performed on the hit rates to the studiitems yielded a significant effect,F(2, 61) 54.16,MSe5 .025,p , .05. A Newman–Keulstest determined that hit rates were significan

n

rvals in

false

ich par-

tures or words. No mention was made about thehigher in the word encoding condition than i

TABLE 2


Repeated new words

Study condition Studied New d8 C Lag 4 Lag 12 Lag 24 Lag 48

Word .72 .19 1.64 .20 .21 .23 .27 .3225% distinctive .58 .11 1.61 .58 .13 .13 .10 .1810% distinctive .63 .17 1.42 .35 .18 .23 .20 .30

Note. The “word” condition includes data from the identical condition in Experiment 1. The “new” column containsalarm rates to the first presentations of the new words. The signal detection measures,d8 (discrimination) and C (bias), arederived from the “old” responses to studied and once-presented new items. “Distinctive” refers to conditions in wh

rring to the proportions of pictured studied items.

N

eehs

nicye

isr

o

rei

edceofnit

is-y

ni- of

s

-er-ewohe

THE DISTINCTIVE

the 25% distinctive encoding condition. No othcomparisons attained significance. There wno differences between the conditions in eitthe baseline false alarm rates to the new wordthe corrected recognition rates,Fs(2, 61) , 1.62.In addition, there were no differences in thed8(discrimination) scores,F(2, 61) 5 1.02. How-ever, an analysis of the bias (C) scores yielded asignificant effect,F(2, 61) 5 3.48,MSe5 .223,p , .05. Participants in the 25% distinctive codition were more conservative than were partpants in the word condition, as determined bNewman–Keuls test. No other comparisons wsignificantly different.

For item recognition in the 25% and 10% dtinctive conditions, participants recognized motest items studied previously as pictures thanwords. This pattern was confirmed by a 2 (Cdition: 25% vs 10%) 3 2 (Item: picture vs word)ANOVA of the recognition rates of the pictuand word study items in the 25% and 10% p

the four lag intervals in Experiment 2. Lag refers to tthe new word from its repetition. Vertical lines depict

ESS HEURISTIC 791

rreer or

-i- are

-easn-

c-

of condition; a main effect of item,F(1, 30) 533.22,MSe5 .013,p , .0001; and a marginallysignificant interaction,F(1, 30) 5 3.96,p , .06.Although picture study items were recognizmore often than word study items, the differenin the recognition rates of these two types items was larger in the 25% picture conditio(75% hit rate for picture study items vs 53% hrate for word study items) than in the 10% dtinctive condition (73% hit rate for picture studitems vs 62% hit rate for word study items).

Figure 2 presents the corrected false recogtion rates of the repeated new words at eachthe lag intervals. A 3 (Condition) 3 4 (Lag)ANOVA of the corrected false recognition rateyielded a significant effect of lag,F(3, 183) 510.77,MSe5 .009,p , .01, and no other significant effects. Planned comparisons detmined that false recognition of repeated nwords did not differ between participants whhad studied words only and participants in t

h of

ture conditions. There was no significant effect10% distinctive encoding group,F(1, 61) ,

FIG. 2. Corrected false recognition rates in the three study conditions to the repeated new words at eac
he number of test items separating the initial occurrence ofstandard errors of the means.

D

olsth

n tlardeth

ld

ex-ri-uli

og-we asn-ted

heheords

792 DODSON AN

1.00. However, participants in the word encing condition exhibited higher corrected farecognition rates than did participants in 25% distinctive encoding condition,F(1, 61) 54.28,MSe5 .024,p , .05.

These results indicate that participants dorequire a high degree of predictability to usedistinctiveness heuristic. That is, only a retively small amount of distinctive study infomation (25% pictures) seems to be require“turn on” this heuristic. After encoding somdistinctive information, participants appear weight heavily its presence or absence w

t gh ubaa

ntiti

reimls tD g

thth.gs

invvhivso

Weus

atedten onse a

-ed

seig-ionedon

edrdsher-

rolfor%

thenoPar-

old

he

see

the

making subsequent old–new judgments.

EXPERIMENT 3

The previous experiments have documenthat studying 25% or more of target itemspictures produces a reduction in false recotion responses to repeated new words. We interpreted these results as reflecting the usemetacognitive heuristic. Specifically, we argthat people infer from the absence of rememing distinctive information about a test item ththe item is novel. However, it is possible ththis reduction in false recognition rates doesreflect the use of a strategic metacogniheuristic. Instead, the reduction might reflecrelatively automatic consequence of encodpictorial information; perhaps encoding pictusimply renders old and new items more discrinable, thereby producing a reduction in faalarms to repeated new items (we considerissue at greater length later in the General cussion). If this were so, then there would beneed to postulate the involvement of metacotive or heuristic processes during retrieval.

A fundamental question, then, is whetherdistinctiveness heuristic is characterized byproperties of a metacognitive mechanism (eBarnes, Nelson, Dunlosky, Mazzoni, & Naren1999; Nelson & Narens, 1990, 1994). Forstance, one essential attribute of a retriestrategy is that it is susceptible to cogniticontrol (e.g., Dodson & Johnson, 1996). If tdistinctiveness heuristic is a metacognitmechanism, then there should be situationwhich people can turn this strategy on or
based solely on their expectations regarding t
SCHACTER

d-ee

othe-

- to

oen

edasni-aveof aeer-ttot

ve angs-

ehisis-noni-

ee.,,-aleeein

ff

kinds of information they believe they shouremember.

To assess this possibility, we developed anperimental design in which the critical compason holds the nature of the study and test stimconstant and varies only participants’ metacnitive expectations. During the study phase,used the identical instructions and procedurein the previous experiments. However, in cotrast to the preceding experiments, we tesparticipants’ memory for only a subset of tstudy items. As with the prior experiments, tstudied items appeared once and the new wappeared twice at either Lag 24 or Lag 48. did not use the two shorter lags from previoexperiments because the earlier results indicthat at the relatively short lags participants ofrecollected having seen the new word earlierthe test and, thus, would have no reason to udistinctiveness heuristic.

There were four different conditions in Experiment 3. One group of participants studionly words. Results from thisword encodinggroup should replicate our prior findings; falrecognition of repeated new words should snificantly exceed the baseline false recognitrate. A second group of participants studi33% of the items as pictures and was testedboth picture and word study items. Thisstan-dard distinctive group should also replicatprior findings and exhibit a lower correctefalse recognition rate to the repeated new woas compared to the word encoding group. Tremaining two groups were critical for answeing our question about metacognitive contand were identical in every respect exceptthe test instructions. Both groups studied 33of the items as pictures but were tested onword study items only; the test containeditems that had been seen earlier as pictures.ticipants in theuninformed distinctivegroup re-ceived standard test instructions; they were tthat their memories would be tested forall ofthe study items. By contrast, participants in tinformed distinctivegroup were told that thetest included only the word study items; theindividuals were told that they would not basked about the picture study items on

herecognition test.

nen

n

nicr3

or

do

hpfh

urtt

tu

stu nes

a dpf

oais L

easingeethes

thatture at

tedrdsas

asrddoherar-se-thecal

s ex-g,veed

-re forot

ad-ld

uld theyri-is-ey

e-r-a-noine

THE DISTINCTIVE

The primary comparison in this experimethen, is between the two conditions—informdistinctive and uninformed distinctive—iwhich participants studied 33% of the itemspictures and then either are or are not informthat the test includes the word study items oIf the suppression effect is entirely attributabto studying a sufficient number of pictures adoes not reflect a strategic process, then partpants in both of these conditions should perfosimilarly to each other and to the standard 3distinctive study condition; test instructionshould have no effect on performance. All these distinctive study groups should exhibit duced false recognition rates to the repeanew words relative to the word encoding contion. On the other hand, if the test instructimanipulation abolishes false recognition supression in the informed distinctive group—tgroup that does not expect to be tested on ture study items—then this would be powerevidence that participants are using a mecnism that is under metacognitive control. Thwe predicted that false recognition rates to peated new words would be higher when paripants do not expect to be tested on items appeared as pictures during the study phaseinformed distinctive group) than when particpants do expect to be tested on the picture sitems (uninformed distinctive group).

Method

Participants. A total of 75 paid volunteerwere recruited from the Harvard University sdent population. There were 15 participantsthe word encoding condition and 20 participain each of the standard distinctive, uninformdistinctive, and informed distinctive condition

Design and materials. The stimuli were 125Snodgrass and Vanderwart (1980) pictures their corresponding names. The stimuli werevided into five groups of 25 items. The grouof items had similar mean ratings for picture miliarity (range 5 3.5–3.7), picture complexity(range 5 2.6–2.7), and word frequency (32 feach group). Three groups of items were rdomly intermixed and constituted the study lThe remaining two groups of items servednew items on the test that repeated at either
24 or Lag 48. In the standard distinctive, unin
NESS HEURISTIC 793

t,d

asedly.ledi-m%sf

e-tedi-np-eic-

ula-

s,e-ic-hat(thei-dy

-intsd.

ndi-s

a-

rn-t.asag

formed distinctive, and informed distinctivgroups, one of the three groups of items wpresented at study as pictures and the remaintwo groups were presented as words. All thrgroups of items were presented as words in word encoding condition. The groups of itemwere counterbalanced across participants so each group was presented at study as a picand word and as a new word that repeatedeach of the two lag intervals during the test.

Participants studied 75 items but were teson only 50 of them. Specifically, in the standadistinctive condition, the test contained 25 itemstudied as pictures and 25 items studied words. The test contained 50 items studiedwords in the remaining three conditions (woencoding, uninformed picture, and informepicture). In all of the conditions, the test alscontained 50 new words that repeated at eitLag 24 or Lag 48. Across participants, the pticular subset of studied items that were subquently tested was counterbalanced. Overall,lengths of the study and test lists were identiacross all conditions.

Procedure. The study and test instructionwere the same as those used in the previousperiments. Individuals in the word encodinstandard distinctive, and uninformed distincticonditions were told that the test would be bason memory for all of the studied items. Participants in the informed distinctive condition weinstructed that their memories would be testedonly the items seen as words; they would nhave to remember the picture study items. In dition, everyone was informed that the test woucontain new words and that the new words worepeat. We emphasized that they should press“old” key for the studied items and the “new” kefor the new words. As in the previous expements, the participants were warned not to mtake repeated new words for old words; thshould call the repeated new words “new.”


Table 3 presents the proportions of “old” rsponses to the different items in the four diffeent conditions. Item recognition was comparble across the conditions; there were differences in hit rates to studied items, basel
-false alarm rates to new words, or d8 and C

794 DODSON AND SCHACTER

ticl

dd

ticnrth

e lit

o,

tiathronc

ve

rbn

o

og-ngf

erinerr.i-

ewiny,edey

s iny

-edntsdctwt isi-p-r-

tegicoff

rvals in

n meas-istinc-In the un-ord study, whereas

scores. As in the previous experiments, parpants in the word encoding condition falserecognized more repeated new words thanparticipants in the standard distinctive contion. There was, however, a dramatic effecttest instructions on the false recognition ratethe repeated new words. As predicted, parpants in the informed distinctive conditioshowed higher false recognition rates to peated new words than did participants in uninformed distinctive condition.

Overall item recognition was similar in thvarious conditions; there were no differenceseither the hit rate to studied items, the basefalse alarm rate to new words, or the correcrecognition rates, all Fs(3, 71) , 1.52. In addi-tion, there were no differences between the cditions in d8 (discrimination) or C (bias) scoresFs(3, 71) , 1.25.

Figure 3 presents corrected false recognirates to the repeated new words. Corrected frecognition rates were considerably lower in standard distinctive condition than in the woencoding condition. Critically for the issue metacognitive control, corrected false recogtion rates were higher in the informed distintive condition than in the uninformed distincticondition. A 4 (Condition) 3 2 (Lag) ANOVAproduced a significant effect of condition,F(3,71) 5 6.17,MSe5 .021,p , .001, and no othesignificant effects. Because of the large numof pairwise comparisons, we used NewmaKeuls tests to analyze the effect of conditi

i-yidi-oftoi-

e-e

inneed

n-

onlseedfi--

er–

n.

These tests indicated that corrected false recnition rates in the informed distinctive conditiodid not differ from those in the word encodincondition. However, participants in both othese conditions exhibited significantly highfalse recognition rates than did participants either the standard distinctive condition or thuninformed distinctive condition; these lattetwo conditions did not differ from each otheThus, we replicated previous results in that indviduals falsely recognized more repeated nwords in the word encoding condition than the standard distinctive condition. Importantlparticipants falsely recognized more repeatnew words when they were informed that thwould not be tested on the picture study itemcompared to the otherwise identical conditionwhich participants were informed that thewould be tested on all of the studied items.

Although participants in the informed distinctive condition falsely recognized the repeatnew words at comparable levels as participain the word encoding condition, the informedistinctive participants did not exhibit an effeof lag (i.e., higher false recognition rates to newords repeating at Lag 48 than at Lag 24) thatypically observed in the word encoding condtion. We have no ready explanation for this aparent anomaly and will explore the issue futher in future experiments.

Overall, the data from Experiment 3 indicathat the distinctiveness heuristic is a stratemechanism that participants can turn on or

TABLE 3


Repeated new words

Condition Studied New d8 C Lag 24 Lag 48

Word encoding .63 .28 1.01 .14 .35 .44Standard distinctive .63 .23 1.15 .21 .26 .23Uninformed distinctive .55 .21 1.00 .36 .20 .22Informed distinctive .57 .21 1.07 .35 .31 .31

Note. The “new” column contains false alarm rates to the first presentations of the new words. The signal detectioures,d8 (discrimination) and C (bias), are derived from the “old” responses to studied and new items. In the standard dtive condition, participants study 33% of the items as pictures and are tested on both picture and word study items. informed and informed distinctive conditions, participants study 33% of the items as pictures but are tested on the witems only. The uninformed group was incorrectly told that the test was based on both picture and word study itemsthe informed group was told the true nature of the test.

THE DISTINCTIVENESS HEURISTIC 795

h

p thdh

ristit

lsrees,achardns

n-ns

ent 3. Verti-

depending on their expectations of whether twill or will not be tested on the picture studitems. Because all features of the two key eximental conditions were held constant exceptthe test instructions, we can be confident the distinctiveness heuristic operates unmetacognitive control and, consequently, tfalse recognition suppression is not an aumatic consequence of encoding pictures duthe study phase. Studying pictures createbasis for invoking the distinctiveness heurisbut only when test instructions indicate that i
a
FIG. 3. Corrected false recognition rates to the repeated new words at the two lag intervals in ExperimLag refers to the number of test items separating the initial occurrence of the new word from its repetition.cal lines depict standard errors of the means.

ppropriate to activate the heuristic.

r

Aion o

ni-

ic-

EXPERIMENT 4

We have suggested that memory for pictoinformation drives the use of the distinctiveneheuristic under appropriate test conditions. we mentioned earlier, however, this conclusis not entirely clear-cut because participawho studied pictures also heard the nameeach picture. Thus, it is possible that mem

eyyer-forat eratto-ng ac,is

ialsssntsofry

for auditory information, rather than pictoriainformation, underlies the distinctivenesheuristic. To address this issue, we tested thdifferent groups. One group studied only wordone studied pictures and heard the name of epicture, and one studied words and also hethe name of each word. If auditory informatiois critical for activating the distinctivenesheuristic, then this latter word 1 sound groupshould perform similarly to the basic picture ecoding group. However, if auditory informatiois not critical for activating the distinctivenesheuristic, then the word 1 sound group shouldbehave like the word encoding condition.

Method

Participants. A total of 48 paid volunteersfrom the Harvard University student populatioparticipated in this experiment, with 16 particpants in each of the three groups.

Design and materials. The stimuli consistedof 120 Snodgrass and Vanderwart (1980) p

D

n r

Lon

-e

nmdnt t

sithe i

nemni

the

ionand

et of

er re-d athat inhe

he

ssonc isiedteMrds, istic

796 DODSON AN

tures and their verbal labels. As in Experime1 and 2, participants studied 60 items. Themaining 60 items were presented as new iteon the test that repeated at either Lag 24 or 48. The stimuli were divided into four groups 30 items. The groups had similar mean ratifor picture familiarity (range 5 3.5–3.6), picturecomplexity (range 5 2.6–2.8), and word frequency (33 for each list). Four different countbalancing formats rotated the groups of itemsthat across participants, each list was preseat study and also was presented as a new iteeach of the lag intervals. There were three ferent study conditions. The word encoding apicture encoding conditions were identical those conditions in Experiment 1. In the word1sound encoding condition, people heard name of each word and saw the visual formthe word on the screen.

Procedure. The study instructions were thsame as those used in the prior experimentthe word 1 sound encoding condition, particpants were informed that they would hear name of each word that appeared on the scrThe test instructions were identical to thosethe earlier experiments. Participants were formed that the test would contain old items anew items that would repeat.


Table 4 presents the probabilities of respoing “old” to the studied, new, and repeated nwords in the three different conditions. Iterecognition rates were similar across conditioThere were no differences in hit rates to stud

Note. The “new” column contains false alarm rates to thures,d8 (discrimination) and C (bias), are derived from the “

SCHACTER

tse-msagfgs

r- soted at

if-d

o

heof

e. In-een.inn-nd

d-w

s.ed

rates to new words,F(2, 45) 5 1.82; or cor-rected recognition rates,F(2, 45) , 1.00. In ad-dition, there were no differences between conditions in either d8 scores,F(2, 45) 5 1.62,or C scores,F(2, 45) 5 1.05.

Table 4 also shows that the false recognitrates to the repeated new words at Lag 24 Lag 48 are similar in the word and word 1sound encoding conditions. A 2 (Lag) 3 3(Condition) ANOVA of the corrected falsrecognition rates produced a significant effeclag,F(1, 45) 5 13.42,MSe5 .006,p , .001; amarginally significant effect of condition,F(2,45) 5 2.58,MSe5 .026,p , .09; and no signif-icant interaction. Participants exhibited highfalse recognition rates to the new words thatpeated at Lag 48 than to those that repeateLag 24. Planned comparisons confirmed tcorrected false recognition rates were lowerthe picture encoding condition than in either tword encoding condition,F(1, 45) 5 3.97,MSe5 .026,p 5 .05, or the word 1 sound encodingcondition,F(1, 45) 5 3.76, MSe5 .026, p ,.06. False recognition rates did not differ in tlatter two conditions,F(1, 45) , 1.00. The highfalse recognition rates in the word 1 sound con-dition demonstrate that the distinctiveneheuristic is not based on auditory informatiabout the studied items. Instead, the heuristiactivated only after participants have studpictorial information. However, we must noone qualification to this conclusion. In the DRparadigm, we have shown that saying woaloud at study, rather than hearing the wordssufficient to activate the distinctiveness heuris

e to

rvals in

items, F(2, 45) , 1.00; baseline false alarmbecause there is a lower false recognition rat

TABLE 4


Repeated new words

Study condition Studied New d8 C Lag 24 Lag 48

Word .62 .18 1.30 .34 .27 .30Word 1 sound .61 .17 1.31 .34 .24 .31Picture .62 .12 1.64 .49 .11 .18

e first presentations of the new words. The signal detection meas-old” responses to studied and once-presented new items.

o

0ha

ci-este

reu-

THE DISTINCTIVE

the critical lures in the say encoding conditithan in the hear encoding condition (for resuand discussion, see Dodson & Schacter, 20Thus, auditory information does not drive tdistinctiveness heuristic when it has been p
sively heard, rather than self-generated, duri
nrrn

nette

t4utteaWouao ni

aortiomeo

re ta

a

iv

i-dsasatdyss

d ast,etic-re-hatni-

offut

ltsrtis-s).d-e-rdi-teserg.ren-eatsrtoiveedes-n

in-llyitss

the study task.

GENERAL DISCUSSION

In four experiments, we examined the distitiveness heuristic with a repetition lag padigm. Participants studied words, pictures, omixture of both and then completed a recogtion test in which studied items appeared oand new items appeared twice. All of the expiments demonstrated that participants in word encoding condition were vulnerable falsely recognizing the repeated new words,pecially when the words repeated at one oflonger lag intervals such as Lag 24 or Lag Studying pictures, however, produced a redtion in false recognition rates to the repeanew words. This finding conceptually replicathe results of Dodson and Schacter (2001) Schacter et al. (1999) in the DRM paradigm. interpret this suppression effect as the outcof a metacognitive heuristic in which individals infer that a test item is novel from the sence of memory for expected distinctive infmation. These experiments also investigatedinfluence of two variables—diagnosticity ametacognitive control—on the use of the dtinctiveness heuristic.

Consider first the role of diagnostic informtion, which we examined by varying the proption of studied items that appeared as pictuCompared to a word encoding condition, paripants successfully rejected repeated new wafter studying 50%, 25%, and 33% of the iteas pictures in Experiments 1, 2, and 3, resptively. Thus, the distinctive information need nbe completely diagnostic—(i.e., perfectly pdictive of an item’s oldness) for participantsuse the heuristic. Instead, a relatively smamount of distinctive study information appeto be sufficient to “turn on” this heuristic.

The second critical issue we examinedwhether or not individuals exert metacognit
control over the distinctiveness heuristic. We i
NESS HEURISTIC 797

nlts1).es-

ng

c-a- ai-

cer-

heos-

he8.c-

edsnde

me-b-r-theds-

-r-es.c-rdssc-t-oll

rs

ise

vestigated this issue by manipulating partipants’ expectations about whether or not the twould contain items studied as pictures. Wcompared two different conditions that weidentical in every respect except for the particlar test instructions; participants in both condtions studied a mixture of pictures and worbut were tested only on the items studied words. When participants were informed ththey would not be tested on the picture stuitems, they did not engage the distinctiveneheuristic and, consequently, falsely recognizenumber of the repeated new words. By contrawhen they were (incorrectly) informed that thtest was based on all of the studied items, paripants used the heuristic and rejected the peated new words. These results indicate tthe distinctiveness heuristic is under metacogtive control such that it can be turned on or depending on participants’ expectations aboits usefulness for reducing memory errors.

These results are consistent with other resufrom our lab showing that participants exesome degree of control over the use of the dtinctiveness heuristic (Schacter et al., in presIn that study, we used the DRM paradigm anfound that participants could turn the distinctiveness heuristic on or off depending on the dmands of the retrieval test. With a standaold–new recognition test, we replicated prevous results and found that false recognition rato semantically related lure words were lowafter picture encoding than after word encodinHowever, this suppression effect after pictuencoding was nearly eliminated with a “meaing” recognition test on which participants werinstructed to respond “old” to any test items thmatched the theme or gist of studied item(Brainerd & Reyna, 1998). Under this lattecondition, there is no reason for participantsconsider whether test items possess distinctproperties that are characteristic of studiitems. Thus, these results indicate that supprsion of “old” responses to related lures is not aautomatic consequence of encoding pictorialformation. Instead, the effect depends criticaon retrieval conditions that do or do not elicthe strategy that we call the distinctivene

n-heuristic.

D

ltse

n

s b

enh

nt in

ited

n

l r

nro

5pttn

hd

e

ive-teder-

inc-n

bu-y-is-ctsotedct

ng in-inges It inewnewre-nd,l-ctewuch

ectallewig-et atzedtheusehenottici-fa-heat to

n atsenny

798 DODSON AN

One further intriguing feature of our resuconcerns whether or not there are costs to uthe distinctiveness heuristic. If individuals rspond on the basis of the absence of memorypictorial information about a test item, then omight expect that using such a criterion woulower recognition rates of test items that westudied as words because memories for the witems would not contain the critical pictorial information (at least not to the same extentwould memories for the picture study itemSpecifically, one reasonable prediction wouldthat participants who used the distinctiveneheuristic would exhibit lower hit rates to thitems studied as words than would participawho did not use the heuristic. To examine tissue, we focused on the two conditions froExperiment 3—the informed distinctive and uinformed distinctive groups—that were idencal in every respect except that participantsone condition were informed that the test cludes only word study items and participantsthe other condition were not so informed. Athough participants in the informed conditiodid not use the heuristic and showed signcantly higher corrected false recognition rathan did participants in the uninformed contion, the two conditions yielded highly similaresponses to the word study items. As seeTable 3, hit rates (.57 vs .55) and baseline faalarm rates (.21 vs .21) were nearly identicaboth conditions. These data suggest that theno cost to using the distinctiveness heuristic.

The apparent lack of a cost in item recogtion from using the distinctiveness heuristic pvides one explanation of why people persistapplying the heuristic even when it is not highdiagnostic of an item’s oldness (e.g., when 2or 33% of the studied items appeared as tures). On the one hand, it might seem counproductive to use the distinctiveness heurisand thus heavily weigh memory for pictorial iformation, when only 25% of the studied itemwere seen as pictures. On the other hand, if this no cost to invoking the heuristic then it migindeed be reasonable to apply it even when tinctive information is not highly diagnostic because false alarms to repeated new words ar
duced with no cost to hit rates. These resu
SCHACTER

sing- foreldreord-as).e

ss

tsism-

i-in-

inl-nfi-s

i-r inlseine is

i--

inly%ic-er-ic,-seretis-- re-

suggest that participants can use the distinctness heuristic in a relatively specific or targemanner that does not merely result in a genally more conservative response bias.

Consider next the distinctiveness heuristicrelation to issues concerning familiarity and reollection of source information. The repetitiolag paradigm illustrates the separate contritions of familiarity, recollecting source-specifing information, and the distinctiveness heurtic. Familiarity contributes to the correrecognition of earlier studied items. But it alleads to the incorrect acceptance of repeanew words when participants fail to recollespecific item information about encounterithe new words earlier in the test and do notvoke the distinctiveness heuristic. Recollectitem-specific information, of course, contributto the correct recognition of studied items.also can underlie a “recall-to-reject” processwhich participants correctly reject repeated nwords because they recollect seeing the word earlier on the test list (for discussion of call-to-reject processes, see Clark & Gronlu1996; Rotello & Heit, 1999; Rotello, Macmilan, & Van Tassel, 2000). This recall-to-rejeprocess probably occurs frequently for nwords that repeat at one of the shorter lags sas Lag 4 or Lag 12. Use of the recall-to-rejprocess can account for the finding that in conditions, the false recognition rate for the nwords that repeat at Lag 4 or Lag 12 is not snificantly different from the baseline falsrecognition rate. The new words that repeathe longer lags, however, are falsely recognisignificantly above the baseline rate (when distinctiveness heuristic is not invoked) becaparticipants fail to recollect encountering tnew words earlier on the test list (i.e., are able to use a recall-to-reject process). Parpants, therefore, misattribute the new words’miliarity to their having been studied. Thus, tlag effect in the word encoding conditions—this, greater corrected false recognition ratesnew words repeating at long lags rather thashort lags—is primarily attributable to failurein recollecting that the new words were seearlier in the test. According to this account, a
ltsinconsistencies in the magnitude of this lag ef-

d

elitu iae

v

nt

eai.fthmy

ioEneg

lioen

ano athiean

m

s

eveni-ing in-m)ms er-c-elity

ent

eri-inedna-ob-oryiedountonen-oreenni-atewing).al-ntedereer inrall

nc-d onog-h

ressnewi-fterndfor

THE DISTINCTIVE

fect across experiments are attributable to ferences in the use of this recall-to-rejeprocess.

Finally, participants activate the distinctivness heuristic when they encounter a famitest word (e.g., a repeated new word), have sied distinctive information, and believe that ituseful to invoke the heuristic. It is in this sitution that participants infer that the failure to rcall expected distinctive information signifiethat the test item is new. Importantly, the actition of the distinctiveness heuristic is contingeon failing to recollect item-specific informatioabout seeing the new word earlier in the tesduring the study phase.

From the above description of the distinctivness heuristic, our account predicts that the falarm rates to once-presented new items (the baseline false alarm rates) will be lower apicture encoding than after word encoding. Tis, because new test items will not evoke meory for pictorial information, even though thewill occasionally pass a familiarity criterionthis absence of expected distinctive informatshould be used to reject the new test items. dence for this effect, however, is mixed. In coditions in which participants studied all pictur(i.e., the 100% distinctive [picture] encodinconditions in Experiments 1 and 4), the basefalse alarm rate is lower than in the correspding word encoding conditions. However, whparticipants studied a mixture of pictures awords, the results are inconsistent; sometimthere is a lower baseline false recognition rafter picture encoding than after word encodisuch as performance in the 25% distinctive cdition in Experiment 2, and sometimes thereno difference in the baseline false alarm rate function of the encoding condition, such as 50% distinctive condition in Experiment 1. Thinconsistency might be attributable to a floor fect in the baseline false alarm rates that mdifferences between encoding conditioNonetheless, further research is needed toamine the issue.

One potential problem with using the terdistinctiveness heuristicis that it is difficult topredict a priori what participants will view a
distinctive or subjectively memorable. Alterna
NESS HEURISTIC 799

if-ct

-ard-

s--

sa-nt

or

-lsee.,erat

-

,nvi--s

nen-ndesteg,n-iss ae

sf-sks.ex-

tively, it might be preferable to refer to a pictureheuristicand avoid the potential pitfalls with thconcept of “distinctiveness.” However, we hapreviously found that patterns of false recogtion reduction observed after pictorial encodare also seen with another type of distinctiveformation (saying words aloud vs hearing the(Dodson & Schacter, 2001). Therefore, it seeneedlessly narrow to use a label such as thepic-ture heuristicfor a strategy that appears to opate the same way with different kinds of distintive information. Thus, we believe that thdistinctiveness heuristic captures the generaof processing that occurs after studying differkinds of distinctive material.

We believe that the data from these expments and others from our lab are best explaby a distinctiveness heuristic. Are there altertive accounts that could also explain the served findings? For instance, can “memstrength” differences between the items studas pictures and those studied as words accfor our finding lower corrected false recognitirates after picture encoding than after word coding? If the items studied as pictures are mfamiliar than the items studied as words, ththere should be a similar effect on false recogtion rates to new items—a mirror effect—so ththere are lower false recognition rates to nitems after studying pictures than after studywords (e.g., Glanzer, Kim, & Adams, 1998The main problem with this account is that though false alarm rates to the once-presenew words (baseline false alarm rates) wlower after studying all pictures than aftstudying all words, there were no differenceshit rates, nor were there differences in ovediscriminability, as measured by d8 or correctedrecognition scores. Hence, use of the distitiveness heuristic does not appear to depenstrength differences because overall item recnition did not vary across conditions in whicpeople either suppressed or did not suppfalse recognition responses to the repeated words. An alternative possibility is that particpants are generally more conservative astudying pictures, or a mixture of pictures awords, than after studying all words. Except
-the 25% condition in Experiment 2, false recog-

D

sea

tslr

cigt deo8e aenynreoova

t aarir anaac

tcr)i

inedd

rshnt

nts’er- is,ries as anothf aionarthe

tothedan-lo,rwerardber ofch a thein-y of ton,f.w oraresslo-

re”lsoltsis-

nceory

ithte-

dis-tor

to

800 DODSON AN

nition suppression did not correspond with uof a more conservative criterion. In short, nther strength nor criterion shift accounts creadily explain our results.

Our view of the distinctiveness heuristic fiwell with a basic assumption of Johnson et a(1993) source monitoring framework that paticipants can recruit a variety of different desion strategies when making memory judments. A number of studies have documenthat people use a strategy, comparable to thetinctiveness heuristic, when they attribute titems to a particular source (e.g., Anders1984; Foley et al., 1983; Hashtroudi et al., 19Hicks & Marsh, 1999; Johnson et al., 1981; Kley et al., 1989). For instance, Johnson et(1981) observed that individuals who had genated some words and heard an experimepresent others during the study phase displaa marked bias on a later test when they idefied the origin of various test items (i.e., wethey new words or were they previously genated or heard?). Participants who falsely recnized new words were nearly three times mlikely to respond that the items had been preously heard spoken by the experimenter ththat they had been previously generated byparticipant—what Johnson et al. referred tothe “It had to be you” effect. This decision biappears to reflect participants’ metamemobeliefs that self-generated information is momemorable than heard information (JohnsonRaye, 1981). Therefore, familiar test items thfail to evoke any source-identifying informatioare more likely to be judged as heard thjudged as generated given the expectation ththe items had been generated, they would tainly be remembered as such.

Hicks and Marsh (1999) demonstrated thasimilar metacognitive strategy reduces the reof false memories in the DRM false memoparadigm (e.g., Roediger & McDermott, 1995They observed that participants who had studlists of related words (e.g.,tired, bed, dream)from two different sources, such as generatwords from anagrams and hearing words, wless likely to falsely recall a critical lure wor(e.g., sleep) than were participants who ha
studied the lists of words from the same sour
SCHACTER

ei-n

.’s---

edis-

stn,9;l-l.r-teredti-er-g-rei-n

hes

sale&t

nt ifer-

aally.ed

gre

(e.g., hearing all of the words). Hicks and Ma(1999) argued that studying items from differesources reduces false recall when participarecall judgments are contingent on remembing source information about the items. Thatparticipants are able to reject false memothat lack expected source information suchmemory for having generated the items fromanagram during the study phase. Thus, bHicks and Marsh’s proposal and our notion odistinctiveness heuristic depend on decisrules in which the failure to retrieve particulinformation about an event is diagnostic of event’s nonoccurrence.

The distinctiveness heuristic is also similara metacognitive heuristic, referred to as “lack of knowledge” inference by Collins ancolleagues, that contributes to how people swer questions (e.g., Collins, Warnock, Aiel& Miller, 1975; Gentner & Collins, 1981). Foexample, most people can confidently ansthat they have never shaken hands with RichNixon because they would expect to rememmeeting someone so well known. The lackknowledge about what is assumed to be sumemorable event is used as evidence thatevent did not occur. A similar metamemorial ference, based on the perceived memorabilittest items, affects participants’ tendenciesfalsely recognize new words (e.g., BrowLewis, & Monk, 1977; Strack & Bless, 1994; cRotello, 1999; Wixted, 1992). That is, newords that are judged as particularly salientmemorable, such as a participant’s name,more often correctly rejected than are lesalient or memorable new words (for an anagous “I would have seen it if it had been thestrategy with memory for visual scenes, see aBrewer & Treyens, 1981). In sum, the resufrom these different paradigms point to the extence of a fundamental metacognitive infereprocess that is based on the absence of memfor expected distinctive information.

The present experiments, in combination wother results from our lab, indicate that a stragic retrieval process that we refer to as the tinctiveness heuristic represents a third facaffecting recognition performance in addition
cefamiliarity and recollection of item-specific in-

E

esnhcatrbd l

wi

e

D.-

d

icu-

s,

msry.

t Ies

rce

THE DISTINCTIV

formation. The central findings of the presexperiments—that the distinctiveness heuriis used even when distinctive information is perfectly diagnostic of an item’s oldness, tthe heuristic reflects the operation of expetions that are under metacognitive control,that using the heuristic appears to entail licost—raise questions concerning the propeof the “it had to be you” heuristic described Johnson et al., (1981) and the lack of knowleinference described by Collins (e.g., Collinsal., 1975) and others. Do these heuristics,the distinctiveness heuristic, operate under cditions of imperfect diagnosticity? Are they, tosusceptible to metacognitive control? And these heuristics have few costs associated them, in line with our findings regarding the dtinctiveness heuristic? We think that these promising questions that merit investigationsfuture studies that examine the relation betw

n

nn

s,

,e I

hs

a

y

ina

n-

on.

-for

e-oid

m

is-

d

).

he

y,

og-

nt

metacognition and false recognition.

REFERENCES

Anderson, J. R., & Bower, G. H. (1972). Recognition andtrieval processes in free recall. Psychological Review,79,97–123.

Anderson, J. R., & Bower, G. H. (1974). A propositiotheory of recognition memory. Memory & Cognition,2, 406–412.

Anderson, R. E. (1984). Did I do it or did I only imagidoing it? Journal of Experimental Psychology: Geeral, 113,594–613.

Atkinson, R. C., & Juola, J. F. (1974). Search and deciprocesses in recognition memory. In D. H. KrantzC. Atkinson, R. D. Luce, & P. Suppes (Eds.),Contem-porary developments in mathematical psychology(Vol.1). San Francisco: Freeman.

Barnes, A. E., Nelson, T. O., Dunlosky, J., Mazzoni, G.Narens, L. (1999). An integrative system of metamory components involved in retrieval. In D. GopherA. Koriat (Eds.), Attention and performance XVI.Cambridge, MA: MIT Press.

Brainerd, C. J., & Reyna, V. F. (1998). When things tnever happened are easier to remember than thingdid. Psychological Science, 9, 484–489.

Brewer, W. F., & Treyens, J. C. (1981). Role of schematmemory for places. Cognitive Psychology, 13,207–230.

Brown, J., Lewis, V. J., & Monk, A. F. (1977). Memorabilitword frequency, and negative recognition. QuarterlyJournal of Experimental Psychology, 29,461–473.

Clark, S. E., & Gronlund, S. D. (1996). Global matchmodels of memory: How the models match the d
Psychonomic Bulletin & Review, 3, 37–60.
NESS HEURISTIC 801

ntticotatta-ndtletiesygeetikeon-o,do

iths-are inen

re-

al

e-

ionR.

&m-&

at that

in

,

gta.

Collins, A., Warnock, E. H., Aiello, N., & Miller, M. L.(1975). Reasoning from incomplete knowledge. In Bobrow & A. Collins (Eds.),Representation and understanding: Studies in cognitive science. New York:Academic Press.

Conway, M. A., & Gathercole, S. E. (1987). Modality anlong-term memory. Journal of Memory and Language,26,341–361.

Deese, J. (1959). On the prediction of occurrence of partlar verbal intrusions in immediate recall. Journal of Ex-perimental Psychology, 58,17–22.

Dewhurst, S. A., & Conway, M. A. (1994). Pictures, imageand recollective experience. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 20,1088–1098.

Dodson, C. S., & Johnson, M. K. (1996). Some problewith the process dissociation approach to memoJournal of Experimental Psychology: General, 125,181–194.

Dodson, C. S., & Schacter, D. L. (2001). “If I had said iwould’ve remembered it”: Reducing false memoriwith a distinctiveness heuristic. Psychonomic Bulletin& Review, 8, 155–166.

Dywan, J., Segalowitz, S. J., & Webster, L. (1998). Soumonitoring: ERP evidence for greater reactivity to notarget information in older adults. Brain and Cognition,36,390–430.

Fischler, I., & Juola, J. F. (1971). Effects of repeated testsrecognition time for information in long-term memoryJournal of Experimental Psychology, 91,54–58.

Foley, M. A., Johnson, M. K., & Raye, C. L. (1983). Age-related changes in confusion between memories thoughts and memories for speech. Child Development,54,51–60.

Gallo, D. A., Roberts, M. J., & Seamon, J. G. (1997). Rmembering words not presented in lists: Can we avcreating false memories? Psychonomic Bulletin & Re-view, 4, 271–276.

Gentner, D., & Collins, A. (1981). Studies of inference frolack of knowledge. Memory & Cognition, 9, 434–443.

Gillund, G., & Shiffrin, R. M. (1984). A retrieval model forboth recognition and recall. Psychological Review, 91,1–67.

Glanzer, M., Kim, K, & Adams, J. K. (1998). Response dtribution as an explanation of the mirror effect. Journalof Experimental Psychology: Learning, Memory, anCognition, 24,633–644.

Hashtroudi, S., Johnson, M. K., & Chrosniak, L. D. (1989Aging and source monitoring. Psychology and Aging,4, 106–112.

Hicks, J. L., & Marsh, R. L. (1999). Attempts to reduce tincidence of false recall with source monitoring. Jour-nal of Experimental Psychology: Learning, Memorand Cognition, 25,1195–1209.

Hintzman, D. L. (1988). Judgments of frequency and recnition memory in a multiple-trace memory model. Psy-chological Review, 95,528–551.

Humphreys, M. S., Bain, J. D., & Pike, R. (1989). Differe
ways to cue a coherent memory system: A theory for

D

re

oro

bea

ti

roio

3

g

Ae

)c-

ru

e

Ald

d.r

ou

lueue

).ca

on

me

oofrn-

ria:

re-

o-.),

tera

ontics

J.g,

ans

gd ing,

or

oft?

).C

c-ion

hts

he

E.on:

Schacter, D. L., Israel, L., & Racine, C. A. (1999). Suppress-

802 DODSON AN

episodic, semantic, and procedural tasks. Psychologi-cal Review, 96,208–233.

Israel, L., & Schacter, D. L. (1997). Pictorial encoding duces false recognition of semantic associates. Psycho-logical Bulletin & Review, 4, 577–581.

Jacoby, L. L. (1991). A process dissociation framewSeparating automatic from intentional uses of memJournal of Memory and Language, 30,513–541.

Jacoby, L. L., & Dallas, M. (1981). On the relationship tween autobiographical memory and perceptual leing. Journal of Experimental Psychology: General, 3,306–340.

Jenkins, J. R., Neale, D. C., & Deno, S. (1967). Differenmemory for picture and word stimuli. Journal of Edu-cational Psychology, 58,303–307.

Jennings, J. M., & Jacoby, L. L. (1997). An opposition pcedure for detecting age-related deficits in recollectTelling effects of repetition. Psychology and Aging, 12,352–361.

Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (199Source monitoring. Psychological Bulletin, 114,3–28.

Johnson, M. K., & Raye, C. L. (1981). Reality monitorinPsychological Review, 88,67–85.

Johnson, M. K., Raye, C. L., Foley, H. J., & Foley, M. (1981). Cognitive operations and decision bias in rity monitoring. American Journal of Psychology, 94,37–64.

Kelley, C. M., Jacoby, L. L., & Hollingshead, A. (1989Direct versus indirect tests of memory for sourJudgments of modality. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15,1101–1108.

Kensinger, E. A., & Schacter, D. L. (1999). When trecognition suppresses false recognition: Effectsaging. Cognitive Neuropsychology, 16,399–415.

Koriat, A., Ben-Zur, H., & Sheffer, D. (1988). Telling thsame story twice: Output monitoring and age. Journalof Memory and Language, 27,23–39.

Koutstaal, W., Schacter, D. L., Galluccio, L., & Stofer, K. (1999). Reducing gist-based false recognition in oadults: Encoding and retrieval manipulations. Psychol-ogy and Aging, 14,220–237.

Madigan, S. (1983). Picture memory. In J. C. Yuille (EImagery, memory, and cognition: Essays in honouAllan Paivio(pp. 65–89). Hillsdale, NJ: Erlbaum.

Mandler, G. (1980). Recognizing: The judgment of previoccurrence. Psychological Review, 87,252–271.

Mather, M., Henkel, L. A., & Johnson, M. K. (1997). Evaating the characteristics of false memories: Rember/Know judgments and memory characteristics qtionnaire compared. Memory & Cognition, 25,826–837.

McDaniel, M. A., DeLosh, E. L., & Merritt, P. S. (2000Order information and retrieval distinctiveness: Reof common versus bizarre material. Journal of Experi-mental Psychology: Learning, Memory, and Cogniti,26,1045–1056.

McDermott, K. B. (1996). The persistence of false meries in list recall. Journal of Memory and Languag,
35, 212–230.
SCHACTER

-

k:ry.

-rn-

al

-n:

).

.

.al-

. e:

eof

.er

),of

s

-m-s-

ll

o-

McDermott, K. B., & Roediger, H. L. (1998). Attempting tavoid illusory memories: Robust false recognition associates persists under conditions of explicit waings and immediate testing. Journal of Memory andLanguage, 39,508–520.

Multhaup, K. S. (1995). Aging, source, and decision criteWhen false fame errors do and do not occur. Psychol-ogy and Aging, 10,492–497.

Murdock, B. B., Jr. (1982). A theory for the storage and trieval of item and associative information. Psychologi-cal Review, 89,609–626.

Nelson, T. O., & Narens, L. (1990). Metamemory: A theretical framework and new findings. In G. Bower (EdThe psychology of learning and motivation(Vol. 26).New York: Academic Press.

Nelson, T. O., & Narens, L. (1994). Why investigametacognition? In J. Metcalfe & A. P. Shimamu(Eds.),Metacognition: Knowing about knowing. Cam-bridge, MA: MIT Press.

Norman, K. A., & Schacter, D. L. (1997). False recognitiin young and older adults: Exploring the characterisof illusory memories. Memory & Cognition, 25,838–848.

Payne, D. G., Elie, C. J., Blackwell, J. M., & Neuschatz,S. (1996). Memory illusions: Recalling, recognizinand recollecting events that never occurred. Journal ofMemory and Language, 35,261–285.

Rajaram, S. (1993). Remembering and knowing: Two meof access to the personal past. Memory & Cognition,21,89–102.

Roediger, H. L., III. (1996). Memory illusions. Journal ofMemory and Language, 35,76–100.

Roediger, H. L., III, & McDermott, K. B. (1995). Creatinfalse memories: Remembering words not presentelists. Journal of Experimental Psychology: LearninMemory, and Cognition, 21,803–814.

Rotello, C. M. (1999). Metacognition and memory fnonoccurrence. Memory, 7, 43–63.

Rotello, C. M., & Heit, E. (1999). Two-process models recognition memory: Evidence for recall-to-rejecJournal of Memory and Language, 40,432–453.

Rotello, C. M., Macmillan, N. A., & Van Tassel, G. (2000Recall-to-reject in recognition: Evidence from ROcurves. Journal of Memory and Language, 43,67–88.

Scarborough, D. L., Gerard, L., & Cortese, C. (1979). Acessing lexical memory: The transfer of word repetiteffects across task and modality. Memory & Cognition,7, 3–12.

Schacter, D. L. (1999). The seven sins of memory: Insigfrom psychology and cognitive neuroscience. Ameri-can Psychologist, 54,182–203.

Schacter, D. L., (2001). The seven sins of memory: How tmind forgets and remembers. Boston: Houghton Mifflin.

Schacter, D. L., Cendan, D. L., Dodson, C. S., & Clifford,(in press). Retrieval conditions and false recognitiTesting the distinctiveness heuristic. Psychonomic Bul-letin & Review.

ing false recognition in younger and older adults: The

THE DISTINCTIVENESS HEURISTIC 803

distinctiveness heuristic. Journal of Memory and Lan-guage, 40,1–24.

he

Cc

han

in o

li

Snodgrass, J. G., & Vanderwart, M. (1980). A standardizedset of 260 pictures: Norms for name agreement, imageagreement, familiarity, and visual complexity. Journal

d

s:

nd

in

or,

Schacter, D. L., Norman, K. A., & Koutstaal, W. (1998). Tcognitive neurosciences of constructive memory. An-nual Review of Psychology, 49, 289–318.

Schacter, D. L., Verfaellie, M., Anes, M. D., & Racine,(1998). When true recognition suppresses false renition: Evidence from amnesic patients. Journal ofCognitive Neuroscience, 10, 668–679.

Schacter, D. L., Verfaellie, M., & Pradere, D. (1996). Tneuropsychology of memory illusions: False recall recognition in amnesic patients. Journal of Memoryand Language, 35, 319–334.

Seamon, J. G., Luo, C. R., & Gallo, D. A. (1998). Creatfalse memories of words with or without recognitionlist items: Evidence for nonconscious processes. Psy-chological Science, 9, 20–26.

Smith, R. E., & Hunt, R. R. (1998). Presentation modaaffects false memory. Psychonomic Bulletin & Review,5, 710–715.

.og-

ed

gf

ty

of Experimental Psychology: Human Learning anMemory, 6, 174–215.

Strack, F., & Bless, H. (1994). Memory for nonoccurrenceMetacognitive and presuppositional strategies. Journalof Memory and Language, 33, 203–217.

Underwood, B. J. (1972). Word recognition memory afrequency information. Journal of Experimental Psy-chology, 94, 276–283.

Underwood, B. J., & Freund, J. S. (1970). Testing effectsthe recognition of words. Journal of Verbal Learningand Verbal Behavior, 9, 117–125.

Wixted, J. T. (1992). Subjective memorability and the mirreffect. Journal of Experimental Psychology: LearningMemory, and Cognition, 18, 681–690.

(Received December 18, 2000)(Revision received April 27, 2001)(Published online January 10, 2002)

When False Recognition Meets Metacognition: The

Documents

Transcript of When False Recognition Meets Metacognition: The