Journal of Experimental Psychology:

Learning, Memory, and Cognition Copyright 1993 by the American Psychological Association, lnc
1993, Vol. 19. No. 1. 148-158 O278-7393/93/J3.0O

The Fate of Repetition Effects When Recognition Approaches Chance

Morris Moscovitch and Shlomo Bentin

We examined whether repetition priming effects remain above baseline when explicit recognition
is reduced to chance or near chance levels by forgetting. Subjects studied a set of words, and
memory was tested explicitly by yes/no (Experiments 1 & 3) or forced-choice recognition
(Experiment 3) after a 20-min delay filled with an interfering task. Memory was then tested
implicitly by perceptual identification (Experiment 3) or lexical decision (Experiments 1 & 2) for
words seen only at study, at recognition, or both. In all experiments, recognition d' was about 0.75,
and repetition effects remained above baseline and constant across conditions. At delays of 24 hr
(Experiment 4) yes/no recognition fell to near chance (d' < 0.6) levels in a third of the subjects.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Repetition effects as measured by lexical decision were not significantly above baseline for words
This document is copyrighted by the American Psychological Association or one of its allied publishers.

seen only once in those subjects and 4 other subjects from Experiment 1 who had comparable
recognition rates. Reducing recognition by forgetting to near chance levels in normal people is
very difficult but when it does occur, repetition effects in normal people, in contrast with those in
amnesics, are also significantly reduced or eliminated.

Explicit tests of memory, such as recognition and recall, explicit tests is steep in comparison with the decline on
require conscious recollection of the past, whereas implicit some implicit tests such as repetition effects in lexical de-
or indirect tests of memory do not (Graf & Schacter, 1985; cision tasks (Scarborough, Cortese, & Scarborough, 1977),
Moscovitch, 1984). Instead, on implicit tests, memory for a perceptual identification (Jacoby & Dallas, 1981), speeded
particular item or event is inferred from changes in perfor- reading of geometrically transformed script (Cohen &
mance on tests involving that item, with the subject often Squire, 1980; Kolers, 1976; Moscovitch, Winocur, &
being unaware that memory is being probed. On some im- McLachlan, 1986), word-fragment completion (Sloman,
plicit tests, memory is reflected in the increased speed or Hayman, Ohta, Law, & Tulving, 1988), word-stem comple-
accuracy with which a repeated item is processed compared tion with unique solutions (Graf, Squire, & Mandler, 1984),
with when it was first presented or compared with a newly and picture naming (Mitchell & Brown, 1988).
presented item. We label this effect stimulus-specific repe- The similarity in the pattern of performance on explicit
tition priming effect, or simply repetition effect. and implicit tests between normal people tested at long
Evidence that performance on explicit and implicit tests delays and amnesic patients has led to the tacit assumption
of memory can be dissociated one from the other has accu- that evidence from both can be used interchangeably to
mulated during the last decade. The most dramatic and, support or refute theoretical positions regarding the organi-
perhaps, the most compelling evidence comes from amnesic zation of memory. It is possible, however, that organic am-
patients who often score at chance or baseline levels on nesia and extensive, but normal, forgetting have different
explicit tests but perform normally on implicit tests of mem- effects on performance. For example, although amnesic pa-
ory (for reviews, see Moscovitch, 1982, 1984; Schacter, tients with virtually no explicit memory can still perform
1987; Shimamura, 1986). A comparable, but less dramatic, relatively well on some implicit tests, it is not well estab-
pattern has been observed in healthy subjects who are tested lished whether healthy subjects matched with severe amne-
at long delays from study. The drop in their performance on sics for performance on explicit tests of memory, such as
recognition, will perform up to the level of amnesics on
implicit tests.
By extrapolating from reports of a more rapid decline in
Morris Moscovitch, Department of Psychology, Erindale Col-
lege, University of Toronto, Mississauga, Ontario, Canada;
performance on explicit than on implicit tests of memory, it
Shlomo Bentin, Department of Psychology and School of Educa- is possible to conclude that normal people, like amnesics,
tion, Hebrew University, Jerusalem, Israel. will continue to perform well on implicit tests even if their
The contributions of all authors to this study were equal. The performance on explicit tests has dropped to chance or
order of authors was determined randomly. This study was sup- baseline levels. This conclusion, however, is not warranted
ported by a grant from the Israeli Academy of Science, the Fund for a number of reasons. First, with the exception of studies
for Basic Research to Shlomo Bentin, and by a Natural Science examining memory for unattended or poorly perceived ma-
and Engineering Research Council of Canada grant to Morris terial, which we discuss shortly, there are, to our knowledge,
Moscovitch.
We thank Maureen Patchett for help in preparing the manuscript.
no studies comparing performance on explicit and implicit
Correspondence concerning this article should be addressed to tests in normal people when performance on explicit tests
Morris Moscovitch, Department of Psychology, Erindale College, has been reduced to chance by forgetting. It is conceivable,
University of Toronto, Mississauga Road, Mississauga, Ontario, despite the rapid initial decay, that performance on recog-
Canada L5L 1C6. Electronic mail may be sent to momos@utorvm. nition remains above chance even after extremely long de-
bitnet. lays. Second, if performance on explicit tests can be reduced
148
 REPETITION EFFECTS WITH LOW STIMULUS RECOGNITION 149

to chance, then it is still possible that no savings would be show robust savings on implicit tests of memory. If they
apparent on implicit tests. would, it would strengthen the claim that memory as in-
An example from a widely cited study by Kolers (1976) ferred from performance on some implicit tests is more
on the longevity of repetition effects speaks to the first resilient to forgetting than memory when it is tested explic-
point. In that study, Kolers found that subjects reread pages itly. On the other hand, this claim would be placed in doubt
of geometrically transformed script faster the second time by a failure to find preserved memory on implicit tests in
than the first, even though 13 to 15 months elapsed between normal people when recognition induced by forgetting is
presentations. To quote Kolers (1976), "Although small near chance. Such an outcome would favor the view that
[about 5%], the [repetition] effect is reliable and its signif- performance on both implicit and explicit tests depends, in
icance lies with its occurrence, not its size" (p. 559). What part, on gaining access to a common trace or engram. To
has been often overlooked is that performance on an explicit anticipate the results, we note at the outset that we had great
recognition test for the same material also remained signif- difficulty in reducing recognition to near chance by forget-
icantly above chance: 65% of passages were classified cor- ting. This itself is an interesting finding when compared
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

rectly as old or new. Kolers's statement, therefore, can be with the ease with which encoding manipulations can
This document is copyrighted by the American Psychological Association or one of its allied publishers.

applied with equal force to performance on the explicit achieve these results. In those few subjects in whom we
recognition tests in normal people. In marked contrast, the succeeded in reducing recognition substantially, perfor-
performance of amnesic patients on a comparable explicit mance on implicit tests also dropped significantly. The im-
test drops to chance by 2 hr, whereas that on the comparable plications of these results for theories of normal and patho-
implicit test is normal (Moscovitch et al., 1986). logical memory and forgetting are discussed when we
Two other findings are worth noting in Kolers's study. present our findings.
When compared with the scores achieved on immediate The design and procedure was similar for all of the ex-
testing (Kolers, 1973), performance on both implicit and periments we conducted. Each experiment consisted of four
explicit tests declined noticeably, although that on the ex- successive phases: a study phase, an interpolated interfering
plicit test was steeper. Second, overall performances on the task, an explicit recognition test, and an implicit memory
two tests correlated poorly with each other. It is important to test in which a repetition effect was sought.
emphasize, however, that not all repetition effects are as
In the study phase, subjects made lexical decisions to 240
long-lasting as those reported by Kolers. Some, such as stem
strings of letters that were evenly divided into words and
completion with nonunique solutions, disappear within min-
nonwords. Depending on the experiment, the recognition
utes or hours (Diamond & Rozin, 1984; Squire, Shimamura,
& Graf, 1987), which is much quicker than performance on procedure was either yes/no recognition or modified forced-
explicit tests of memory. However, if one considers only choice. Memory was next tested implicitly in a lexical de-
long-lasting repetition effects, the data reported by Kolers cision or perceptual identification task. The words in the
are similar to those reported by others in the literature (eg. implicit test were encountered in the study phase, only in the
Komatusu & Ohta, 1984; Mitchell & Brown, 1988; Sloman recognition phase, in both the study and recognition phase,
et al., 1988). In all instances, repetition effects decline or were new words that had never been presented in the
slowly in comparison with recognition, but performance on experiment up to that point. Repetition effects, a measure of
tests of recognition remain substantially above chance when savings on the implicit test, were calculated by computing
tested at the same intervals. the difference in latency or identification accuracy between
old, repeated items and new items encountered for the first
In contrast, recognition can be reduced to chance by per- time.
ceptually degrading the stimulus at study (Bonnano & Still-
ing, 1986; Kunst-Wilson & Zajonc, 1980; Mandler, Naka- Although items were not counterbalanced within experi-
mura, & Van Zandt, 1987; Seaman, Brody, & Kauff, 1983; ments, different words from the same pool were chosen at
Seaman, Marsh, & Brody, 1984) or by diverting attention random across Experiments 1,3, and 4 to form the different
from it (Bentin, Kutus, & Hillyard, 1990; Eich, 1984; conditions. For example, the set comprising the repeated
Merikle & Reingold, 1991). Despite chance recognition, words in one experiment was different from the set in the
performance on implicit tests remained significantly above other experiments. The same is true of the other conditions.
chance and, in many cases (Merikle & Reingold, 1991; The only exception is that, in Experiments 1 and 2, the same
Seaman et al., 1983), remained stable in the face of wide items were used in corresponding conditions in both exper-
variations in recognition. Comparison of the effects of en- iments. The words in the different conditions were equated
coding manipulations and forgetting on implicit and explicit for frequency, imageability, and length.
tests of memory only serves to bring our question into relief: The first three experiments were devoted to assessing the
Would performance on implicit tests in normal people re- relative sensitivity of two different explicit tests of memory,
main stable and above baseline as recognition is reduced to yes/no recognition and modified forced choice, and two
near chance levels by forgetting? If not, what accounts for different implicit tests, improvement in lexical decisions
the differences between the two types of studies? and in identification accuracy with repetition. On the basis
The purpose of the experiments reported in this article, of this information, a fourth experiment was conducted in an
therefore, was to try to reduce, by forgetting, performance attempt to lower performance on the explicit test even fur-
on explicit tests of memory to chance or near chance levels ther and determine what effect that manipulation had on
in normal people to determine whether they would still performance on the implicit test.
 150 MORRIS MOSCOVITCH AND SHLOMO BENTIN

Experiment 1 for 1000 ms each and separated by 2,500 ms interstimulus inter-

vals. Responses were made throughout the study by pressing a
In previous studies, we had established that repetition button in a response box. The right hand was used for "word" re-
effects in a lexical decision task decline sharply, from a lag sponses and the left for "nonword" responses.
of 0 to a lag of 4 between repeated words, and then are An interference task was interpolated between study and rec-
virtually unchanged to a lag of 15 (Bentin & Moscovitch, ognition. The task lasted about 20 min, during which subjects
1988) and even longer (Bentin & Moscovitch, 1990). Rec- distinguished between words and numbers. The words in the in-
terference task were 80 regular nouns sampled from the same
ognition, on the other hand, decayed steadily over the same
original pool as the experimental words.
intervals but still remained well above chance, yielding a d'
In the explicit memory test that followed, 40 of the words en-
of about 1.8. Similar results were reported by Ratcliff, countered in the study phase were presented along with 80 lures.
Hockley, and McKoon (1985; but see Dannenbring & Bri- Each item appeared alone on the CRT screen and the subject had
and, 1982). Interpolating an interfering, mathematical task to judge whether it was an old item (from the study phase) or a
between the presentation of each word and extending the lag new one. The right hand was used to respond "old" and the left
to 29 items had no noticeable influence on the repetition hand, "new."
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

effect (Moscovitch, 1985). Recognition, on the other hand,

This document is copyrighted by the American Psychological Association or one of its allied publishers.

The implicit test was administered immediately afterwards.

did deteriorate but remained well above chance at about Subjects viewed 240 letter strings evenly divided between words
80%. Many amnesic patients, however, were scoring at and nonwords. Forty of the words were previously encountered
chance on recognition, even without the interpolated task, only in the study phase, 24 words were encountered only in the
but showed repetition effects that were no different from explicit phase (i.e., they were randomly sampled from the new
words in the recognition test), 24 were presented both in Phase 1
normal (Moscovitch, 1985). The purpose of Experiment 1,
and in Phase 2 (i.e., they were old items in the recognition task),
therefore, was to determine whether repetition effects in and 32 were new words. Subjects decided as quickly and as ac-
normal people could survive even further reductions in rec- curately as possible whether each item was a word or a nonword.
ognition. In addition, we wished to compare the repetition Subjects were told that some words might appear that they had
effect for words that were explicitly recognized with those seen before but that this fact was not relevant to the present task.
that were not, something the design of the previous exper- They were to treat those words as they would any others.
iment did not allow us to do.
Although stochastic independence between performance
on implicit and explicit tests of memory had been reported Results and Discussion
in previous studies (Jacoby & Witherspoon, 1982; Tulving,
Schacter, & Stark, 1982; for review, see Richardson- Explicit test (yes/no recognition). As Table 1 shows,
Klavehn & Bjork, 1988; Schacter, 1987, and Tulving, 1985), the proportion of words correctly judged as old or new was
we were interested in the effect that low overall recognition .60 and .71, respectively (d'= 0.86). Hits minus false
scores had on the phenomenon. alarms (FAs), another measure of recognition accuracy
(corrected for guessing), was 0.31. Reaction times (RTs)
were averaged separately for each of the four possible out-
Method comes: hit, correct rejection (CR), FA, and miss (see Table
1). A two-factor analysis of variance (ANOVA) showed
Subjects. Twenty-four undergraduate students, all native that the correct responses (hit and CR) were faster than in-
speakers of Hebrew, participated in Experiment 1 either for correct responses (miss and FA), F{\, 23) = 14.30, MSe =
course credit or payment. All had normal or corrected-to-normal 7,085, p < .001, and that the responses to new items (CR
vision. and FA) were as fast as to old items (hit and miss; F < 1).
Stimuli. Stimuli consisted of 232 Hebrew words (all nouns) The interaction between these two effects was not reliable,
and 160 pronounceable nonwords. In the absence of a reliable F(l, 23) = 2.63, MSe = 8,150, p < .12.
word-frequency count in Hebrew, word frequency was deter-
mined on the basis of subjects' ratings. In several previous stud- Implicit test (lexical decision). The results are pre-
ies, subjects rated the frequency of words on a 5-point scale (1 = sented in Table 2. At study, RTs to words (592 ms) were
low frequency, 5 = high frequency). The words included in Exper- longer than to nonwords (730 ms). RTs in the implicit test
iment 1 were sampled from that pool. Their average rated fre- were broken down according to whether the words hadap-
quency ranged between 2 and 3 points (M = 2.78). Both words peared only at study, only at recognition, on both occa-
and nonwords were between 3 and 5 letters long, which, because
the regular unvoweled print was used, were equivalent to 4- to
7-letter words in English. The words used in each phase of the Table 1
experiment were randomly selected, and there were no frequency Mean Reaction Times (RTs) and Percentages
or word-length differences among these groups. All stimuli were
computer-generated on a cathode-ray tube (CRT), and a standard
in the Explicit Recognition Test (Experiment 1)
set of Hebrew characters without the vowel dots was used. They Unrepeated words Repeated words
were presented in double-size and double-width format between
two lines of asterisks centrally located and subtending a visual Measure CR FA Hit Miss
angle of 7° X 4°. RT 832 867 796 890
Procedure. In the study phase, subjects made lexical deci- % 71.0 29.0 59.9 41.1
sions to 240 strings of letters evenly divided between words and Note. Mean RTs are expressed in milliseconds; d' = 0.86. CR :
nonwords and presented in a random order. Stimuli were exposed correct rejection; FA = false alarm.
 REPETITION EFFECTS WITH LOW STIMULUS RECOGNITION 151

Table 2 Experiment 2
Mean Reaction Times (RTs) in the Implicit Test
(Experiment 1) In Experiment 1, memory was tested explicitly by a
yes/no recognition procedure in which the ratio of new to
Seen once old items was 2:1. The purpose of Experiment 2 was to
before
determine whether a modified forced-choice recognition
In Seen twice procedure would prove to be a more sensitive, explicit test
recognition before of memory, perhaps a more worthy rival to the implicit test.
Study Unrepeated In
Measure list words study CR FA Hit Miss The modification consisted of presenting the subjects with a
printed list of 120 items of which they had to choose 40 that
RT 592 615 568 577 540 573 556 they believed they had seen before.
SEM 24 29 22 27 23 33 31
Note. Mean RTs are expressed in milliseconds. CR = correct
rejection; FA = false alarm. Method
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Subjects. Twenty-four undergraduates, all native speakers of

sions, or only at test. The RTs to words that had appeared Hebrew, participated in the experiment either for course credit or
at recognition were broken down further into hit, CR, and for payment. All had normal or corrected-to-normal vision. None
misses. of the subjects had taken part in Experiment 1.
A one-way ANOVA with repeated measures showed a Procedure. Except for the explicit recognition test, the stim-
significant main effect of condition", F(5, 115) = 6.96, MSe uli and procedure were identical to those in Experiment 1. As in
= 2,203, p < .0001. Post hoc Tukey-A comparison showed Experiment 1, assignment of words to each condition was ran-
that RTs to new words were slower than to words in any of dom. Modified forced-choice recognition was tested by request-
ing each subject to circle 40 words seen in the initial lexical de-
the other categories, except for words that were correctly
cision task from a printed list that included an additional 80
rejected at recognition. To control for the possibility that words as lures.
(despite the random selection of stimuli and equal word-
frequency in each stimulus category) the repetition effect
was caused by stimulus-specific factors, we have com- Results and Discussion
pared the RTs to repeated stimuli that were previously pre-
sented only in the study list with the RTs to words during Performance on the modified forced-choice recognition
study, and with RTs to new words in the implicit test. An tests yielded a a" of 0.84, a similar value to that of 0.86 in
ANOVA showed that the repetition effect was reliable, Experiment 1. By using the proportion of hits minus FA as
F(2, 46) = 7.39, MSe = 1,800, p < .002. Post hoc Tukey-A a corrected measure of recognition accuracy, we obtained a
comparisons (Winer, 1962, p. 198) revealed that the differ- value of .29, which was also similar to the value of .31
ence between the RTs to new words during study and to found in Experiment 1.
new words in the implicit test was not reliable. The RTs to As Table 3 shows, performance on the implicit lexical
repeated nonwords in the implicit test (670 ms) were sim- decision task was also similar to that of Experiment 1. A
ilar to new nonwords (667 ms). The number of errors in one-way ANOVA with repeated measures showed a signif-
the implicit test was too small to allow for any reliable sta- icant effect of stimulus category, F(5, 115) = 6.73; MSe =
tistical analysis. 1,160, p < .0001. Post hoc Tukey comparison showed that
The manipulations had the desired effect of lowering new words yielded longer RTs than did any of the categories
of repeated (old) words. RTs to old words did not differ
recognition. The mean a" dropped to 0.86 from a value of
significantly among themselves. A planned t test revealed
1.76 in a previous study when using the same procedure
that the RTs to new words in the implicit test were not
and materials (Bentin & Moscovitch, 1990). Although low,
reliably different from the RTs to words in the study list.
recognition was still substantially above chance and com-
parable to that reported in other studies in the literature
that used forgetting as the method for lowering recogni- Table 3
tion. Mean Reaction Times (RTs) in the Implicit Test
Performance on the implicit test was also consistent (Experiment 2)
with that of other studies. A significant repetition effect
that averaged 50 ms was found for words that had been Seen once
before
seen previously. No significant differences were found be-
tween any of the different conditions, indicating that the In Seen twice
before
repetition effect was not noticeably influenced by whether Study Unrepeated In
the word was recognized correctly or not (hit or miss, FA Measure list words study CR FA Hit Miss
or CR), whether it had been viewed once or twice, or
RT 591 611 569 579 577 565 562
whether it had been viewed recently or after a delay. These SEM 15 20 15 20 20 17 19
results, too, are in line with those reported by other inves-
Note. Mean RTs are expressed in milliseconds. CR = correct
tigators. rejection; FA = false alarm.
 152 MORRIS MOSCOVITCH AND SHLOMO BENTIN

The results of Experiment 2 indicate that, insofar as the Procedure. The stimuli were the same as in Experiments 1
materials and procedures that we used are concerned, mod- and 2, as was their random assignment to each condition. As be-
ified forced choice is no more sensitive than yes/no recog- fore, Experiment 3 consisted of a study phase, an interpolated
nition as an explicit test of memory. Because we did not use task, an explicit recognition test, and an implicit test. The interpo-
forced choice in an easier recognition condition, as we did lated task consisted of trials during which subjects tried to iden-
for old/new recognition (Bentin & Moscovitch, 1990), we tify perceptually degraded words to establish an exposure dura-
tion threshold of 50% correct identification. In each trial, a word
cannot tell whether performance on the two types of recog-
was exposed, immediately followed by a mask consisting of
nition tests decayed at the same rate. We can conclude, @@@@@@@ that lasted for 500 ms. Following each trial, the
however, that under the comparable conditions of Experi- subject identified the word by naming. To establish the threshold,
ments 1 and 2, recognition as measured by the two types of lists of 20 trials each were prepared. The exposure time of the
tests is roughly equivalent. words in the first list was 45 ms. In the second list, the exposure
It is also worth noting that a similar repetition effect of time was 15 ms. On the basis of the subject's performance on
40-50 ms was found in both Experiments 1 and 2 and was these two lists, the exposure time for words in the next four lists
was increased and decreased until identification was about 50%
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the same whether the words were seen only at study, only at
correct. The last list was used to test the reliability of the thresh-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

recognition, or at both instances.

old exposure time. This threshold exposure duration would then
be used in the implicit perceptual identification test. None of the
words used in this interpolated task appeared in any other part of
Experiment 3 the experiment.
Because we found no difference between old/new free recog-
Just as it is possible for one recognition test to be more nition and forced-choice recognition in Experiments 1 and 2, the
sensitive than another, so it is possible for one type of type of explicit recognition test administered in Experiment 3
implicit test to be more sensitive than another. In Experi- was distributed evenly among the subjects. In the implicit mem-
ments 1 and 2 we found that repetition effects in lexical ory test, 120 words were presented and followed by a mask, as in
decision, an implicit measure of memory, did not decay the interpolated task. On each trial, the subject attempted to iden-
noticeably over the intervals and interpolated task that we tify the presented word by naming it. The exposure duration for
used. Would the same be true of repetition effects in per- the words was determined for each subject individually and set at
ceptual identification, which is another widely used measure the measured threshold value (that which yielded 50% correct
of performance on implicit memory tests? Although there is identification in the interpolated task). The range of exposure du-
rations used was between 17 and 26 ms.
already substantial evidence in the literature that repetition
As in the previous experiments, the stimulus list comprised 40
effects in perceptual identification are long-lasting (Jacoby, words initially presented during study, 24 words initially pre-
1983; Jacoby & Dallas, 1981), there is no study, to our sented in the recognition test, 24 words initially presented at
knowledge, that compared decay of repetition effects in one study and repeated at recognition, and 32 new words.
type of test with decay of repetition effects in another. In this
case, the comparison cannot be direct because latency mea-
sures repetition effects in lexical decision, whereas it is Results and Discussion
accuracy that measures perceptual identification. Even if the Perhaps because the intervening task was a little more
measures were ostensibly the same in the two cases, it is demanding, the overall a" in Experiment 3 was 0.75 and the
unlikely that a direct comparison would be legitimate be- proportion of hits minus FAs was .26, both slightly lower
cause both the baselines and the metric would have to than in Experiments 1 and 2. The values of d' and propor-
equated. In addition, the underlying process that was being tion of Hits minus FA for the yes/no recognition condition
reflected in the two measures would have to be equivalent. were 0.86 and .28, respectively; for the forced-choice rec-
To avoid these problems, the comparison we propose is one ognition, they were 0.63 and .23, respectively.
between patterns of performance on the two types of im-
As Table 4 shows, the pattern of repetition effects ob-
plicit memory tests. As we noted earlier, the repetition effect
tained in perceptual identification resembled those in lexical
was equivalent for all repeated items regardless of whether
decision. The repetition effect was as large for items seen
they had been seen only once (either at the start of the
only at study as for items seen at study and at recognition.
experiment or at recognition) or twice (at both study and at
An ANOVA on the different categories of words showed an
the recognition test immediately preceding the implicit test).
overall main effect, F(3, 69) = 8.51, MSe = 109, p < .0001.
In short, there was no indication of a significant decay with
Post hoc Tukey-A tests showed that new items were iden-
time and interpolated activity over the intervals we sampled.
Would the same be true of repetition effects in the percep-
tual identification test? Table 4
Mean Percentages of Correct Responses in the
Perceptual Identification Test (Experiment 3)
Method Seen once
before
Unrepeated Seen twice
Subjects. The participants in the study were 24 undergradu- Measure words In study In recognition before
ates who met the same inclusion criteria as the subjects in Exper-
iments 1 and 2. None of them had participated in the previous ex- % 50.0 63.5 60.0 62.6
periments. SEM 2.3 2.6 2.9 3.1
 REPETITION EFFECTS WITH LOW STIMULUS RECOGNITION 153

tified more poorly than any of the repeated old items (p < mal and severely impaired memory can serve both as a
.01), with no significant differences among the old items. yardstick against which the performance of our subjects can
The results of Experiment 3 show that the pattern of be compared and as a guide for interpreting the results of
repetition effects on perceptual identification resembles that this experiment.
on lexical decision, despite the fact that the tests themselves
are not similar and that performance is measured by accu- Method
racy in one test and by latency in the other. In both tests, the
repetition effect for old words was not affected by recency Subjects. Participants in Experiment 4 were 24 undergradu-
or by the number of repetitions. The resemblance between ates at the Hebrew University who met the same inclusion crite-
the two tests was remarkable in that both showed the small- ria as subjects in Experiments 1, 2, and 3. None of these subjects
est repetition effect for words that were correctly rejected at participated in the previous experiments.
recognition. Although subsequent studies might find differ- Stimuli and procedure. The stimuli were identical to those
used in Experiment 1, and assignment to each condition was ran-
ences in sensitivity between the two tests, at the moment dom. The procedure was the same as that of Experiment 1, ex-
there is no reason to regard one as being a more sensitive
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

cept that the delay between the interpolated task and the recogni-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

test than the other. tion test was 24 hr.

The similarity in the pattern of repetition effects should
not be interpreted as evidence that performances on the two Results
tests are correlated with one another or that performance on
one test is statistically dependent on the other. Should a Explicit test (yes/no recognition). Despite the 24-hr
direct test of this prediction be conducted, it may well prove study-test interval the explicit recognition performance of
that stochastic independence holds for these two tests as it most subjects was as good as in Experiment 3. The aver-
does for other implicit memory tests (Witherspoon & Mos- age value of d' was 0.77, which was not significantly dif-
covitch, 1989). Instead, the results suggest that a common ferent, f(44) = 0.88, from the d' in Experiment 1 (which
component or components mediate performance on the two differed from the Experiment 4 only in the time elapsed
tests. At the moment, one can only speculate as to the from study to the recognition test). The proportion of hit
identity of those components. minus the proportion of FA was 0.32 (see Table 5). An
As noted in the introduction, not all implicit tests of ANOVA revealed that, as in Experiment 1, correct re-
memory produce comparable decay rates. Some decay sponses (hit and CR) were faster than incorrect responses
within minutes, others take hours, still others take days, (miss and FA), F(l, 23) =18.51, M5e = 2,283, p < .0003,
whereas some show only a slow decline whose limits have and responses to new words (hit and miss) were as rapid as
yet to be reached after more than a year. Experiments 1, 2, to old words (F < 1). Accuracy significantly interacted
and 3, therefore, have been successful in demonstrating (a) with stimulus type, F(l, 23) = 10.51, MSe = 5,634, p <
comparable levels of performance on two types of recogni- .004. Post hoc Tukey-A comparisons showed that hit re-
tion tests and (b) comparable patterns of performance on sponses were significantly faster than miss responses,
two types of implicit memory tests. In all cases, perfor- whereas CR and FA responses were equivalent.
mance was well above chance or baseline levels. The ques- Implicit test (lexical decision). The pattern of repeti-
tion still remains of whether performance on implicit tests tion effects observed in the implicit memory test was sim-
would remain above baseline if performance on either of the ilar to that observed in Experiments 1, 2, and 3 (see Table
recognition tests were reduced to chance or near chance 6).
levels. Experiment 4 brings us closer to an answer.
An ANOVA showed a significant main effect of repeti-
tion, F(3, 69) = 14.52, M5e = 620, p < .0001. Post hoc
Experiment 4 Tukey-A comparisons showed that RTs to new words were
slower than RTs to words in any of the other categories
To reduce recognition even more, the delay between study and that the number of repetitions or the recency of the ini-
and test in Experiment 4 was increased by 24 hr. Because tial presentation relative to test made no difference. The
there was no noticeable difference between the two types of RTs to repeated nonwords in the implicit test (574 ms)
recognition tests or between the two types of implicit tests were similar to RTs to new nonwords (577 ms).
that were used in Experiments 1, 2, and 3, we chose the
yes/no procedure for recognition and the lexical decision Table 5
task for the implicit test. We chose the lexical decision task Mean Reaction Times (RTs) and Percentages in the
for a number of reasons, not the least of which was that we Explicit Recognition Test (Experiment 4)
have had extensive experience in using the task to study
memory, reading, and other forms of priming. In addition, it Unrepeated words Repeated words
has been shown that even after 48 hr, repetition effects in Measure CR FA Hit Miss
lexical decision decline only moderately (Scarborough et
RT 810 803 756 848
al., 1977) and recognition remains above chance. Amnesic SEM 27 25 19 32
patients, on the other hand, demonstrate normal repetition % 63.5 36.5 64.8 35.2
effects on lexical decision even when their recognition is at
Note. Mean RTs are expressed in milliseconds; d' = 0.77. CR
chance (Moscovitch, 1985). These data on people with nor- correct rejection; FA = false alarm.
 154 MORRIS MOSCOVITCH AND SHLOMO BENTIN

Table 6 RT data from the two groups yielded a significant main ef-
Mean Reaction Times (RTs) in the Implicit Test fect of repetition, F(3, 66) = 12.34, MSe = 845, p < .0001.
(Experiment 4) The average repetition effect in the high and low d groups
Repeated words was 56 ms and 26 ms, respectively. A t test for indepen-
dent groups revealed that this difference was statistically
Measure Unrepeated words In study ]In recognition Both
reliable, f(22) = 2.46, p < .025. Subsequent comparisons
RT 556 527 518 512 between the two groups, in each of the conditions sepa-
SEM 13 12 13 10 rately, yielded significant group differences for items that
Note. Mean RTs are expressed in milliseconds. were previously presented only at study, f(22) = 4.08, p <
.05 and only at recognition, f(22) = 7.03, p < .01, but not
for the doubly repeated items (t < 1).
At first glance, it seems that the 24-hr lag between study Ideally, we would have preferred that recognition in the
and test did not significantly lower recognition and had lit- entire group be at least as low as in the sample of 12 sub-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tle influence on the pattern of repetition effects on the im- jects. However, lowering recognition to chance is very dif-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

plicit test. Closer inspection of the data, however, revealed ficult. Before interpreting these results, it is important,
that about one third of the subjects had a a" below 0.60 therefore, to establish that the performance of the 12 sub-
(hits = 0.56, FA = 0.44). It takes over 100 trials, or obser- jects with poor recognition was not artifactual. First, these
vations, to distinguish a a" of 0 from one of 0.05. Because subjects were selected at random from a pool of under-
our value of a" was based on only 40 trials, we did not graduate volunteers. Their performance was in other ways
know for certain whether these 8 subjects were performing comparable with that of the other subjects. Their mean RT
above chance. It is a reasonable assumption, however, that for lexical decisions at study (551 ms) was not reliably dif-
their performance was perilously close to it. Repetition ef- ferent from that of the other group (609 ms; p > .1). There
fects for these 8 subjects are presented in Table 7. is no reason to believe that something other than poor
An ANOVA showed a significant main effect of repeti- memory accounts for their performance, although the
tion, F(3, 21) = 3.27, MSe = 628, p < .05. Post hoc shorter RTs among the poor recognizers may suggest a re-
Tukey-A tests revealed, however, that only doubly re- lationship between rapid encoding and forgetting. Second,
peated items yielded significant repetition effects. RTs to it is noteworthy that the largest proportion of subjects with
words presented only at study or only at recognition were poor memory, at least twice that in any other condition,
not significantly shorter than RTs to new items (minimum was found in Experiment 4, in which delay and interfer-
significant RT difference = 41.46 ms). ence were highest. This suggests that factors affecting for-
To increase the power of the statistical tests, four addi- getting rather than only encoding are responsible for sub-
tional subjects with a" less than 0.61 were added from Ex- jects' poor memory.
periment 1, thereby forming a group of 12 subjects with Moreover, there is already ample evidence in the litera-
mean a" of 0.37 (SD = .15, range = -.03 to .61; see low d ture that poor encoding, to the point that the subjects may
group, Figure 1). The hit and FA rates were .53 and .40, re- not be aware that a stimulus was presented, results in nor-
spectively. The RTs in this group were analyzed by mal, often long-lasting, repetition effects on implicit tests
ANOVA, which confirmed the reliability of the trend ob- of memory. Our stimuli, on the other hand, were clearly
served in the smaller group: a main effect of repetition, visible and certainly of sufficient duration for the subjects
F(3, 33) = 3.35, p < .01, with post hoc Tukey-A tests to respond with nearly perfect accuracy in lexical decision
(honestly significant difference = 41.5 ms) indicating that on the initial presentation. Poor encoding is not likely to
only doubly repeated words yielded shorter RTs than did
be the cause of the poor repetition effects seen in our sub-
new words. Repetition effects for words seen only once
jects. The most parsimonious explanation for their poor
were not reliable (p > .05).
performance is that they were more prone than others to
As a final test, the group of low d subjects was com- forget the target items.
pared with a group of 12 subjects (6 from Experiment 1
The results of Experiment 4 suggest, therefore, that
and 6 from Experiment 4) whose d was above 1 (mean
when recognition approaches chance levels, long-lasting
d= 1.26, SD = .17, range = 1.1 to 1.6). Hit and FA rates
repetition effects are substantially reduced or even elimi-
were .70 and .24, respectively. An ANOVA performed on
nated. For the group of subjects with a low recognition
performance (d < 0.61), repetition effects were absent for
Table 7 words seen only once, either at study or at recognition. In
Mean Reaction Times (RTs) in the Implicit Test for contrast, a single previous presentation was sufficient to
Subjects With Low d (Experiment 4) produce repetition effects that were not significantly re-
Seen once before duced over a period of 24 hr in subjects with a higher rec-
Unrepeated Seen twice ognition level (d > 1). In a similar manner, presenting
Measure words In study In recognition before
items twice, at study and at recognition, led to substantial
RT 540 521 515 497 repetition priming effects, even in subjects with poor mem-
SEM 22 24 30 16 ory, presumably because those items were remembered
Note. N = 8. Mean RTs are expressed in milliseconds. better.
 REPETITION EFFECTS WITH LOW STIMULUS RECOGNITION 155

60-
Seen once in study
CO
50- Seen once in recognition
Seen twice

o
o
40- I
vt

20
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

£| 10-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

a
0
Low d High d
Figure 1. Mean differences in reaction time on the implicit test between new items and old items
seen at study, at recognition, or on both occasions for subjects with high d' and low d! scores on
explicit recognition tests in Experiments 1 and 4.

On the basis of evidence from our own study as well as The reduction in performance on implicit tests with poor
other studies, we can chart the time course of repetition ef- explicit recognition due to forgetting also contrasted with
fects for words relative to that for recognition. First, there findings from studies that manipulated encoding to lower
is a steep initial decline in repetition effects that occurs performance on explicit tests. As we noted, when perfor-
within seconds to minutes after stimulus presentation mance on explicit tests was reduced to chance by degrading
(Bentin & Moscovitch, 1988; Ratcliff et al., 1985; Sloman the stimulus at study or by diverting attention from it, rep-
et al., 1988; but see Dannenbring & Briand, 1982); this is etition effects were not significantly reduced but, as with
followed by a relatively long-lasting, stable period for rep- amnesics, remained reliably above baseline (Bentin in press;
etition effects in comparison with explicit recognition per- Bonnano & Stilling, 1986; Eich, 1984; Mandleret al., 1987;
formance, which gets steadily lower; finally, when recog- Merikle & Reingold, 1991; Seaman et al., 1983; Seaman et
nition approaches chance levels, repetition effects are also al., 1984).
noticeably reduced or fall to zero. Together, studies of normal and amnesic subjects suggest
that the following conditions must be met for explicit re-
membering: (a) The stimulus event must first be consciously
General Discussion and fully apprehended. The cited encoding studies suggest
that without full conscious apprehension of the stimulus at
The primary question addressed by our study was whether
encoding, explicit remembering is very ineffective. Con-
normal people would perform above baseline levels on im-
scious apprehension is not, however, a sufficient condition
plicit tests of memory when explicit recognition was re-
for explicit remembering, as the evidence from amnesic
duced to chance or near chance levels by forgetting. The
first point to note is that lowering recognition to that level in subjects indicates. Amnesic subjects can attend, perceive,
normal people proved to be difficult. This fact is not appre- and apprehend the stimuli, yet their conscious recollection is
ciated very much, as investigators are more impressed by severely impaired, (b) Clearly, an additional process is nec-
the initial rapid forgetting than by the longevity of memories essary to make this information available for long-term
as revealed by explicit recognition tests. This is a worthy explicit remembering. This process, often termed consoli-
topic for future research but not one that we can pursue in dation, is identified with the operation of the hippocampus
this article. and related limbic structures (Scoville & Milner, 1957;
Once recognition approaches chance levels, however, our Squire, 1987). In amnesic subjects, the deficit in conscious
evidence suggests that there are differences in implicit test recollection arises either because the apprehended informa-
performance between amnesic patients and normal people tion is not delivered to the hippocampal system or because
whose recognition was reduced by forgetting. Unlike am- that system is absent or malfunctioning. The information,
nesics who often have normal repetition effects in the ab- however, is not stored in the hippocampus, (c) Retrieval
sence of recognition (see, e.g., Moscovitch, 1985), normal cues are necessary to recover the consolidated information
subjects, who score at or near chance on recognition, have and deliver it to consciousness. Forgetting occurs either
repetition effects that are substantially reduced or even elim- because the consolidated information has decayed or be-
inated. cause retrieval cues are ineffective or both. These points and
 156 MORRIS MOSCOVITCH AND SHLOMO BENTIN

the ones made in the discussion that follows are elaborated By this account, with increased forgetting, performance
in Moscovitch (1989, 1992) and Moscovitch and Umilta on explicit and implicit tests becomes driven primarily by
(1991) where a neuropsychological model of memory is the same effective cue. When recognition approaches
presented. chance or near chance levels, it is either because the stim-
None of these processes is necessary for storage and ulus item alone is no longer an effective cue for evoking the
retention of information on implicit tests. For normal per- memory trace or because the trace itself may have dissipated
formance on implicit tests, it is only necessary that the or been altered with time. Our finding of a drop in repetition
stimulus be picked up by neural structures necessary for effects as recognition approaches chance is more consistent
decoding it. The mere pickup of information by these struc- with the second alternative. Because there is no reason to
tures, even without full conscious apprehension (as the en- believe that the effective stimulus for reactivating the record
coding studies show), is sufficient to form long-term records (i.e., driving performance on implicit tests) is altered, the
or engrams in memory. These records are not consciously drop in repetition priming effects as recognition approaches
accessible.' To become accessible, the input must be both chance probably occurs because of a change in the engrams
consciously apprehended and delivered to the hippocampal
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

that form the trace. This situation is different from that in

This document is copyrighted by the American Psychological Association or one of its allied publishers.

system. which chance performance on explicit tests results from

Consolidation, we suggest, involves the binding of en- poor encoding. In that case, repetition priming effects are
grams into memory traces by the hippocampus. The mem- obtained because testing occurs shortly after the study phase
ory trace consists of a collection of bound perceptual (and when the engram is still "fresh" and can support perfor-
possibly conceptual) engrams related directly to the stimu- mance on implicit tests. According to our proposal, substan-
lus, along with other engrams that encode semantic and tially increasing the study-test interval should reduce or
episodic information about the event in which the stimulus eliminate the repetition priming effect, even when encoding
was embedded. Performance on implicit tests requires only is poor.
that the repeated stimulus reactivate the long-term records The correlations between performance on some implicit
by engaging directly the same processes involved in their and explicit tests of memory, but not on others, has been
formation. In explicit tests it is necessary that the bound reported by Perruchet and Baveux (1989). Their point, how-
engrams that formed the memory trace be reactivated indi- ever, was to show that the pattern of dependence results
rectly through the hippocampal system. from the extent to which component processes are common
According to our framework, the records or engrams that to different tests. Our proposal goes further and suggests
mediate performance on implicit tests are a subset of the that even on tests that seem independent at high to moderate
engrams that form the memory trace that mediates perfor- levels of performance, at near chance levels performance
mance on explicit tests. Access to the record is achieved drops in both types of tests.
directly by the stimulus or indirectly, by the hippocampus, Our account clearly borrows some of Tulving's (1983,
through reactivation of the entire memory trace. That is, the 1985) ideas about ecphoric processes in recognition. There
effective cue for performance on implicit tests is the stim- are also similarities with Mandler's (1980) proposal that two
ulus itself or something closely related to it, whereas the component processes contribute to recognition: interitem
effective cue to the memory trace is richer, containing in- elaboration, which is context dependent, and intraitem in-
formation that was encoded along with the stimulus as well tegration, which leads to a sense of familiarity devoid of
as any other relevant information that can be brought to bear context. Mandler and others (Gardiner, 1988; Gardiner &
at the time of retrieval. Forgetting occurs either because the Java, 1990) have associated the latter process with perfor-
cue is no longer effective or because engrams or traces have mance on implicit tests of memory. According to this view,
decayed. Because the cues are different, memory for explic- long-lasting recognition is mediated by processes associated
itly remembered items is independent of memory for im-
plicitly elicited information. Item amnesia can occur with-
1
out concomitant effects on implicit tests for the same target. We use the term record (Kirsner & Dunn, 1985) to refer to a
The time course of repetition effects may also differ from modification of the neural circuitry in response to stimulation. The
that of recognition, not because the memory trace and en- modification contains information about the stimulating event and
has processing consequence so that identical, and perhaps related,
gram are different but because they are accessed differently. stimuli can be processed more efficiently and leads to above base-
The effective stimulus for reactivating the perceptual pro- line performance on implicit tests. We reserve the term engram to
cess presumably does not vary, thereby leading to repetition refer to the informational content of the record that can be accessed
effects that remain stable over prolonged periods.2 Whether or reactivated to contribute to recollection. Whether the engrams
semantic records are equally long-lasting is not known. The are purely perceptual or whether they also have semantic content
effective cue on explicit tests of memory, such as recogni- and show some contextual dependencies is currently disputed
tion, however, is not simply the stimulus itself but the com- (Masson & Freedman, 1990; Roediger, 1990; Tulving & Schacter,
plex event associated with that stimulus. With time, the 1990).
2
context becomes less readily available to act as a cue for the The rapid initial decline on implicit tests may have more to do
memory trace. Thus, performance on explicit tests comes to with recruitment of response factors and episodic memory factors
than with reactivation of perceptual records. That is, for a short
rely more and more on an impoverished cue that consists while after stimulus presentation, the subject may remember the
primarily of the stimulus item itself and the sense of famil- response associated with a particular stimulus, which would lead to
iarity to which it gives rise. faster or more accurate responses on repetition.
 REPETITION EFFECTS WITH LOW STIMULUS RECOGNITION 157

with implicit memory. What distinguishes our account from References

Mandler's is that we believe that both types of recognition
Bentin, S. (in press). Semantic processing of ignored stimuli: The
are mediated by the hippocampus and its related structures.
role of attention in memory. In C. Umilta & M. Moscovitch
Put in purely psychological terms, we believe (contrary to (Eds.), Attention and performance XV: Conscious and uncon-
Mandler) that recognition, even when it is based only on a scious processes in cognition. Cambridge, MA: MIT/Bradford
sense of familiarity, is an explicit test of memory. The Press.
process that gives rise to a sense of familiarity (insofar as Bentin, S., Kutus, M., & Hillyard, S. (1990). ERP evidence for
that sense is inextricably a product of the process rather than selective attention and task effects on semantic priming in au-
an inferred or attributed byproduct of it; Jacoby, 1983; ditory word processing. Bulletin of the Psychonomic Society, 29,
Johnston, Dark, & Jacoby, 1985) is different from the pro- 483 (abstract).
cess that is involved in producing repetition effects. Bentin, S., & Moscovitch, M. (1988). The time course of repetition
The distinction we have drawn between our proposal and effects for words and unfamiliar faces. Journal of Experimental
Psychology: General, 117, 148-160.
those of Mandler, Gardiner, and Jacoby is not trivial. It
Bentin, S., & Moscovitch, M. (1990). Psychophysiological indices
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

implies that recognition, even when it is based on familiar- of implicit memory performance. Bulletin of the Psychonomic
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ity, is mediated by processes that are distinct from those Society, 28, 346-352.
mediating performance on implicit tests of memory. At the Bentin, S., Moscovitch, M., & Heth, I. (1992). Memory with and
moment, we must admit that there is little evidence in the without awareness: Performance and electrophysiological evi-
literature to distinguish among the alternative proposals. dence of savings. Journal of Experimental Psychology: Learn-
Some findings, however, are more consistent with our pro- ing, Memory, and Cognition, 18, 1270-1283.
posal and we review these briefly. Bonnano, G. A., & Stilling, N. A. (1986). Preference familiarity
It is significant to note that performance based only on and recognition after repeated brief exposures to random geo-
familiarity was well above chance (Gardiner, 1988; Gar- metric shapes. American Journal of Psychology, 99, 403^15.
diner & Java, 1990). This was true also, and especially, Cohen, J., & Squire, L. R. (1980). Preserved learning and retention
of pattern analyzing skill in amnesia: Dissociation of knowing
when nonwords were used. By themselves, these results are how and knowing that. Science, 210, 207-210.
consistent with all of the proposals. When evidence from Dannenbring, G. L., & Briand, K. (1982). Semantic priming and
amnesic subjects is taken into account, however, it becomes the word repetition effect in a lexical decision task. Canadian
possible to distinguish between the predictions made by the Journal of Psychology, 36, 435-444.
various proposals. Performance on recognition tests can be Diamond, R., & Rozin, P. (1984). Activation of exiting memories
at chance in amnesics even when repetition effects are nor- in anterograde amnesia. Journal of Abnormal Psychology, 93,
mal for the same material. In addition, repetition effects for 98-105.
nonwords are very difficult to obtain in amnesic subjects. If Eich, E. (1984). Memory for unattended events: Remembering
processes leading to recognition judgments based on famil- with and without awareness. Memory & Cognition, 12, 105-111.
iarity were the same as those mediating repetition effects on Gardiner, J. M. (1988). Functional aspects of recollective experi-
implicit tests of memory, we would expect the performance ence. Memory & Cognition, 16, 309-313.
of amnesics to resemble that of normal people more than it Gardiner, J. M., & Java, R. I. (1990). Recollective experience in
does. Such similarity as does exist between familiarity ef- word and nonword recognition. Memory & Cognition, 18, 2 3 -
30.
fects in recognition and repetition effects on implicit tests of
Graf, P., & Schacter, D. L. (1985). Implicit and explicit memory
memory probably occurs because both are based on reacti- for new associations in normal and amnesic subjects. Journal of
vation of the same engram or record. Experimental Psychology: Learning, Memory, and Cognition,
Bentin and Moscovitch (1990; Bentin, Moscovitch, and 11, 501-518.
Heth, 1992) found electrophysiological evidence of repeti- Graf, P., Squire, R. L., & Mandler, G. (1984). The information that
tion effects in both implicit and explicit tests of memory. amnesic patients do not forget. Journal of Experimental Psy-
This result is consistent with the idea that a common record chology: Learning, Memory, and Cognition, 11, 501-518.
or engram is reactivated in both cases. The additional ob- Jacoby, L. L. (1983). Remembering the data: Analyzing interactive
processes in reading. Journal of Verbal Learning and Verbal
servation that repetition effects on explicit and implicit tests Behavior, 22, 485-508.
are associated with different components of the evoked re- Jacoby, L. L., & Dallas, M. (1981). On the relationship between
sponse suggests that different neural processes mediate their autobiographical memory and perceptual learning. Journal of
occurrence on the two types of tests. It would be important Experimental Psychology: General, 110, 306-340.
to determine whether one is associated with recognition due Jacoby, L. L., & Witherspoon, D. (1982). Remembering without
to familiarity and the other with repetition effects on im- awareness. Canadian Journal of Psychology, 36, 300-324.
plicit tests. Johnston, W. A., Dark, V. J., & Jacoby, L. L. (1985). Perceptual
To summarize, our finding, that performances on implicit fluency and recognition judgments. Journal of Experimental
and explicit tests are interdependent when explicit recogni- Psychology: Learning, Memory, and Cognition, 11, 3-11.
Kirsner, K., & Dunn, D. (1985). The perceptual record: A common
tion is very low, suggests that they are both mediated by
factor in repetition priming and attribute retention. In M. I.
common engrams. In implicit tests, only the record is ac- Posner & D. S. M. Marin (Eds.), Attention and performance XI
cessed by using the repeated stimulus as a cue. In the (pp. 447-566). Hillsdale, NJ: Erlbaum.
explicit test, reactivation of the entire memory trace is nec- Kolers, P. A. (1973). Remembering operations. Memory & Cog-
essary, a process which is performed by the hippocampal nition, 1, 347-355.
system and which requires conscious processing. Kolers, P. A. (1976). Reading a year later. Journal of Experimental
 158 MORRIS MOSCOVITCH AND SHLOMO BENTIN

Psychology: Human Learning and Memory, 2, 554-565. recognition. Journal of Experimental Psychology: General, 114,
Komatusu, S.-L, & Ohta, N. (1984). Priming effects in word frag- 435-150.
ment completion for short and long retention intervals. Japanese Richardson-Klavehn, A., & Bjork, R. A. (1988). Measures of
Psychological Research, 26, 194-200. memory. Annual Review of Psychology, 39, 475-543.
Kunst-Wilson, W. R., & Zajonc, R. B. (1980). Affective discrim- Roediger, H. L. (1990). Implicit memory: Retention without re-
ination of stimuli that cannot be recognized. Science, 207, 557— membering. American Psychologist, 45, 1043-1056.
558. Scarborough, D. L., Cortese, C , & Scarborough, H. S. (1977).
Mandler, G. (1980). Recognizing: The judgement of previous oc- Frequency and repetition effects in lexical memory. Journal of
currence. Psychological Review, 87, 252-271. Experimental Psychology: Human Perception and Performance,
Mandler, G., Nakamura, Y., & Van Zandt, B. J. S. (1987). Non- 3, 1-17.
specific effects of exposure on stimuli that cannot be recognized. Schacter, D. L. (1987). Implicit memory: History and current sta-
Journal of Experimental Psychology: Learning, Memory, and tus. Journal of Experimental Psychology: Learning, Memory,
Cognition, 13, 646-648. and Cognition, 13, 501-518.
Masson, M. E. S., & Freedman, L. (1990). Fluent identification of Scoville, W. B., & Milner, B. (1957). Loss of recent memory after
repeated words. Journal of Experimental Psychology: Learning, bilateral hippocampal lesions. Journal of Neurology, Neurosur-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Memory, and Cognition, 16, 355-373.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

gery, and Psychiatry, 20, 11-21.

Merikle, P. M., & Reingold, E. M. (1991). Comparing direct (ex- Seaman, J. G., Brody, N., & Kauff, D. M. (1983). Affective dis-
plicit) and indirect (implicit) measures to study unconscious crimination of stimuli that are not recognized: II. Effect of delay
memory. Journal of Experimental Psychology: Learning, Mem- between study and test. Bulletin of the Psychonomic Society, 21,
ory, and Cognition, 17, 224—232. 187-189.
Mitchell, D. B., & Brown, A. S. (1988). Persistent repetition prim- Seaman, J. G., Marsh, R. L., & Brody, N. (1984). Critical impor-
ing in picture naming and its dissociation from recognition mem- tance of exposure duration for affective discrimination of stimuli
ory. Journal of Experimental Psychology: Learning, Memory, that are not recognized. Journal of Experimental Psychology:
and Cognition, 14, 213-222. Learning, Memory, and Cognition, 10, 465-469.
Moscovitch, M. (1982). Multiple dissociations of function in am- Shimamura, A. D. (1986). Priming effects in amnesia: Evidence
nesia. In L. S. Cermak (Ed.), Human memory and amnesia (pp. for a dissociable memory function. Quarterly Journal of Exper-
337-370). Hillsdale, NJ: Erlbaum. imental Psychology, 38A, 619-644.
Moscovitch, M. (1984). The sufficient conditions for demonstrat- Sloman, S. A., Hayman, G. A., Ohta, N., Law, J., & Tulving, E.
ing preserved memory in amnesia: A task analysis. In N. Butters (1988). Forgetting in primed fragment completion. Journal of
& L. R. Squire (Eds.), The neuropsychology of memory (pp. Experimental Psychology: Learning, Memory, and Cognition,
104-114). New York: Guilford Press. 14, 223-239.
Moscovitch, M. (1985). Memory from infancy to old age: Impli- Squire, L. R. (1987). Memory and brain. New York: Oxford Uni-
cations for theories of normal and pathological memory. Annals versity Press.
of the New York Academy of Sciences, 444, 78-96. Squire, L. R., Shimamura, A. P., & Graf, P. (1987). Strength and
Moscovitch, M. (1989). Confabulation and the frontal system: duration of priming effects in normal subjects and amnesic pa-
Strategic vs associative retrieval in neuropsychological theories tients. Neuropsychologia, 25, 195-210.
of memory. In H. L. Roediger III & E. I. M. Craik (Eds.), Tulving, E. (1983). Elements of episodic memory. New York: Ox-
Varieties of memory and consciousness: Essays in honor of ford University Press.
Endel Tulving (pp. 133-160). Hillsdale, NJ: Erlbaum. Tulving, E. (1985). How many memory systems are there? Amer-
Moscovitch, M. (1992). Memory and working-with-memory: A ican Psychologist, 40, 385-398.
neuropsychological model of memory based on modules and Tulving, E., & Schacter, D. L. (1990). Priming and human memory
centrol systems. Journal of Cognitive Neuroscience, 4, 251-267. systems. Science, 247, 301-306.
Moscovitch, M., & Umilta, C. (1991). Conscious and noncon- Tulving, E., Schacter, D. L., & Stark, H. A. (1982). Priming effects
scious aspects of memory: A neuropsychological framework of in word-fragment completion are independent of recognition
modules and centrol systems. In R. G. Lister & H. J. Weingartner memory. Journal of Experimental Psychology: Learning, Mem-
(Eds.), Perspectives on cognitive neuroscience (pp. 229-266). ory, and Cognition, 8, 336-342.
New York: Oxford University Press. Winer, B. (1962). Statistical principles in experimental design.
Moscovitch, M., Winocur, G., & McLachlan, D. (1986). Memory New York: McGraw-Hill.
as assessed by recognition and reading time in normal and mem- Witherspoon, D., & Moscovitch, M. (1989). Independence of the
ory impaired people with Alzheimer's disease and other neuro- repetition effects between word fragment completion and per-
logical disorders. Journal of Experimental Psychology: General, ceptual identification. Journal of Experimental Psychology:
115, 331-347. Learning, Memory, and Cognition, 15, 22-30.
Perruchet, P., & Baveux, P. (1989). Correlational analysis of ex-
plicit and implicit memory performance. Memory & Cognition,
17, 77-86. Received April 22, 1991
Ratcliff, R., Hockley, W., & McKoon, G. (1985). Components of Revision received April 6, 1992
activation: Repetition and priming effects in lexical decision and Accepted April 8, 1992 •