ArticlePDF Available

Recollective performance advantages for implicit memory tasks

Taylor & Francis
Memory
Authors:

Abstract and Figures

A commonly held assumption is that processes underlying explicit and implicit memory are distinct. Recent evidence, however, suggests that they may interact more than previously believed. Using the remember-know procedure the current study examines the relation between recollection, a process thought to be exclusive to explicit memory, and performance on two implicit memory tasks, lexical decision and word stem completion. We found that, for both implicit tasks, words that were recollected were associated with greater priming effects than were words given a subsequent familiarity rating or words that had been studied but were not recognised (misses). Broadly, our results suggest that non-voluntary processes underlying explicit memory also benefit priming, a measure of implicit memory. More specifically, given that this benefit was due to a particular aspect of explicit memory (recollection), these results are consistent with some strength models of memory and with Moscovitch's (2008) proposal that recollection is a two-stage process, one rapid and unconscious and the other more effortful and conscious.
Content may be subject to copyright.
PLEASE SCROLL DOWN FOR ARTICLE
This article was downloaded by:
[Canadian Research Knowledge Network]
On:
25 April 2011
Access details:
Access Details: [subscription number 932223628]
Publisher
Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK
Memory
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713683358
Recollective performance advantages for implicit memory tasks
Signy A. M. Sheldona; Morris Moscovitchb
a University of Toronto, Canada b University of Toronto, and The Rotman Research Institute, Toronto,
Canada
First published on: 17 August 2010
To cite this Article Sheldon, Signy A. M. and Moscovitch, Morris(2010) 'Recollective performance advantages for implicit
memory tasks', Memory, 18: 7, 681 — 697, First published on: 17 August 2010 (iFirst)
To link to this Article: DOI: 10.1080/09658211.2010.499876
URL: http://dx.doi.org/10.1080/09658211.2010.499876
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Recollective performance advantages for implicit
memory tasks
Signy A. M. Sheldon
University of Toronto, Canada
Morris Moscovitch
University of Toronto, and The Rotman Research Institute, Toronto, Canada
A commonly held assumption is that processes underlying explicit and implicit memory are distinct.
Recent evidence, however, suggests that they may interact more than previously believed. Using the
rememberknow procedure the current study examines the relation between recollection, a process
thought to be exclusive to explicit memory, and performance on two implicit memory tasks, lexical
decision and word stem completion. We found that, for both implicit tasks, words that were recollected
were associated with greater priming effects than were words given a subsequent familiarity rating or
words that had been studied but were not recognised (misses). Broadly, our results suggest that non-
voluntary processes underlying explicit memory also benefit priming, a measure of implicit memory. More
specifically, given that this benefit was due to a particular aspect of explicit memory (recollection), these
results are consistent with some strength models of memory and with Moscovitch’s (2008) proposal that
recollection is a two-stage process, one rapid and unconscious and the other more effortful and conscious.
Keywords: Implicit memory; Explicit memory; Recollection; Familiarity; Priming.
Long-term memory is traditionally divided into
declarative, explicit memory and non-declarative,
implicit memory (Graf & Schacter, 1985; Schacter,
1987; Squire, 2004). Although past research
focused on their independence, recent studies
suggest that explicit and implicit memory may
interact more than was previously believed
(Kinoshita & Wayland, 1993; Schacter, Dobbins,
& Schnyer, 2004). The present article will examine
how aspects of explicit memory, specifically pro-
cesses associated with recollection, are related to
performance on implicit memory tasks.
Initial support for a distinction between implicit
and explicit memory came from several studies
that reported dissociations between the two types
of memory at a functional, behavioural level in
healthy participants (for review, see Roediger &
McDermott, 1993). At a neuropsychological level,
there was evidence of preserved performance on
implicit memory tasks by individuals with amne-
sia, despite their impaired performance on explicit
memory tests (for reviews, see Bowers & Schacter,
1993; Moscovitch, Vriezen & Goshen-Gottstein,
1993; Shimamura, Salmon, Squire, & Butters,
1987). These findings led to the suggestion that
the medial temporal lobes (MTL) and related
structures in the diencephalon that support ex-
plicit memory are not necessary for normal
performance on implicit memory tasks.
This dissociation was questioned by a number
of investigators (see review in Butler & Berry,
2001) with much of the early criticism being
#2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
Address correspondence to: Signy Sheldon, University of Toronto, Department of Psychology, Sidney Smith Hall, 4th Floor,
University of Toronto, 100 St. George Street, Toronto, Ontario, Canada M5S 3G3. E-mail: signy.sheldon@utoronto.ca
The authors would like to thank Keith Horton for graciously providing us with the stimuli used in Experiment 2 and Marilyne
Ziegler for her help with programming. The authors would also like to thank NSERC for funding.
MEMORY, 2010, 18 (7), 681697
http://www.psypress.com/memory DOI:10.1080/09658211.2010.499876
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
concerned with the methods used to assess
implicit and explicit memory (e.g., Hintzman &
Hartry, 1990; Ostergaard & Jernigan, 1993;
Shimamura, 1985). Some investigators noted
that priming was impaired in patients with
amnesia on a variety of measures (e.g., Schacter,
Church, & Bolton, 1995; Squire, Shimamura, &
Graf, 1987) such as perceptual-identification tasks
(Yang et al., 2003), stem-completion tasks (Graf
& Schacter, 1987), and stimulus-specific priming
(e.g., Kinoshita & Wayland, 1993; Schacter
et al., 1995) under some but not all conditions
(Cermak, 1993; Gabrieli et al., 1994: Goshen-
Gottstein, Moscovitch, & Melo, 2000; Musen &
Squire, 1992). To reconcile these discrepant
findings, Schacter, Wig, and Stevens (2007; see
also Shacter, et al., 2004) proposed two different
types of priming: one that relies on cortical-
perceptual representations, independent of the
MTL or explicit memory, and another that relies
on MTL-based operations which are involved in
establishing new associations.
Building on this proposal, Moscovitch (2008)
suggested that all forms of priming can also be
influenced by some explicit memory processes
mediated by the MTL at an unconscious, auto-
matic level. Supporting this more interactive view
are neuro-imaging studies that have shown that
similar brain regions, including the MTL and the
hippocampus in particular, are active during both
implicit and explicit memory tasks (e.g., Kirchoff,
Wagner, Maril, & Stern, 2000; Turk-Browne, Yi &
Chun, 2006). Consistent with this neuro-imaging
evidence, Westmacott and Moscovitch (2002,
2003) used behavioural procedures to show that
processes associated with recollection contribute
to tests that are seemingly semantic and lexical,
both of which figure prominently in studies of
implicit memory. They asked healthy participants
to perform tests of fame judgement and speeded
reading of names of famous people that were
associated pre-experimentally with either high or
low recollective experience (R). For example, the
name of Princess Diana was associated with a
memory of where the participants were and how
they felt when they heard of her fatal accident. By
contrast, George Bush Sr, though equally familiar,
evoked no recollection in most participants. West-
macott and Moscovitch found that not only was the
performance on tests of recall and recognition
(episodic based tests) better for high R names than
low R names, but performance was also better
(faster and more accurate) for high R names than
low R names on the tests of fame judgement and
speeded reading, which can be considered as
conceptual and perceptual implicit memory tasks,
respectively. In a later study, Westmacott, Black,
Freedman, and Moscovitch (2004) found that the
advantage for high R names was absent in patients
with focal MTL lesions or degeneration (Alzhei-
mer’s disease), whose recollection is severely
compromised, but not in patients with semantic
dementia whose MTLs, and recollections, are
relatively preserved. They concluded that the
contribution of recollection on these tasks is
dependent on the MTL (for reviews, see Aggleton
& Brown, 1999; Eichenbaum, Yonelinas, &
Ranganath, 2007).
Based on these findings, Tulving’s (1983)
description of ecphory and retrieval, and the
component process model (Moscovitch, 1992),
Moscovitch (2008) proposed that recollection is
a two-stage process: The first stage (ecphory) is
fast, relatively automatic, operates outside con-
scious awareness, and is hippocampus dependent;
the second stage is slower, requires conscious
awareness, and depends on the interaction of the
prefrontal and parietal cortices with the hippo-
campus (see reviews by Cabeza, Ciaramelli,
Olson, & Moscovitch, 2008; Ciaramelli, Grady,
& Moscovitch, 2008). According to this proposal,
a target associated with a prior event is broadcast
to all systems, including those that support
recollection. The interaction between the stored
information and its output occurs rapidly, auto-
matically and outside conscious awareness (ecph-
ory), and can influence other ongoing processes
related to the target. At a later stage, conscious
awareness of the output results in the experience
of recollection. Thus the initial, rapid process
reflects access to recollected information. Subse-
quent, slower processes make that content avail-
able to consciousness.
According to this two-stage model, the recollec-
tion advantage evident on tests of semantic mem-
ory and lexical processing should be observed not
only for memories acquired pre-experimentally as
in the Westmacott and Moscovitch studies (2002,
2003), but also for memories acquired in tradi-
tional laboratory settings. Thus we conjecture that
processes that support rapid recollection should
influence performance on a variety of implicit
memory tasks (Moscovitch, 1992; Moscovitch
et al., 1993; Schacter et al., 2004). An altern-
ative possibility is that the strength of the memory
trace influences performance on both explicit and
implicit tests of memory. Since recollection is
typically associated with greater memory strength
682 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
than is familiarity, the underlying trace may also
contribute to performance on implicit tests. For
ease of exposition, we focus on the two-stage
model and will return to the strength account in
the General Discussion.
Of note, the two-stage model is similar to the
concept of fluency reprocessing proposed by
Jacoby and Dallas (1981), which states that an
initial experience with a stimulus will result in
more efficient processing when the stimulus is
encountered later, both in terms of implicit and
explicit decisions. In later work, Jacoby (1991)
developed the process dissociation procedure
(PDP) that enabled researchers to separate the
later strategic, conscious recollection component
from the non-conscious aspects of memory. How-
ever, in doing so, the possibility was not consid-
ered that using the PDP may preclude the
discovery that early, non-strategic stages of re-
collection can influence implicit memory as
suggested by the two-stage model of recollection
(see General Discussion for an elaboration of this
point).
It is also important to note here that, according
to this proposal, implicit memory is not contami-
nated by explicit memory in the sense that volun-
tary, conscious retrieval of information influences
performance on implicit tests (see MacLeod, 2008;
see also Gardiner, Richardson-Klavehn, Ramponi,
& Brooks, 2001, for a treatment of this issue), but
rather that automatic processes associated with
recollection benefit performance on such tests.
Indeed, the influence of these rapid processes
should be evident on the very implicit tests that
have been shown to be immune to the deliberate,
contaminating effects of explicit memory.
To test our hypothesis that processes under-
lying recollection benefit performance on implicit
memory tasks, we performed two experiments
using the type of implicit memory tests that have
been shown to be relatively resistant to contam-
ination by voluntary retrieval of explicit memory
(Goshen-Gottstein, & Moscovitch, 1995: Horton,
Wilson, & Evans, 2001).
The procedure used in both experiments had
four phases: A study phase in which words were
presented, a phase in which a distractor task was
presented, an implicit memory test for those
words, and a subsequent recognition test phase
that used the rememberknow procedure (R/K;
Tulving, 1985). In the first experiment the implicit
test was lexical decision, whereas in the second
experiment it was speeded stem completion. In
both cases we investigated whether performance
on the implicit test varied according to whether
the words were subsequently judged to be recol-
lected, familiar, or new via the R/K procedure
(Tulving, 1985). We predicted that words that
were subsequently recollected would be asso-
ciated with greater priming effects (faster perfor-
mance to previously studied items) on the implicit
memory tasks than would words that were sub-
sequently given a familiarity rating or those that
had been studied but not recognised (misses). All
of the words in the above categories, of course,
would be expected to elicit better performance
than new words.
We are mindful that the rememberknow
procedure is controversial at both a methodolo-
gical and theoretical level (e.g., see Dunn, 2004).
As we noted earlier, dual- or single-strength
models have been proposed to account for the
data derived from studies using the remember
know procedure without appealing to threshold
dual process accounts of familiarity and recollec-
tion (e.g., Rotello & Macmillan, 2006; Wixted,
2007). Before we consider which of the various
theories best accounts for the data, it is important
to establish empirically that a relation exists
between performance on implicit memory tasks
and recollection as indexed by remember re-
sponses. For ease of exposition we adopted the
dual-stage framework to provide a theoretical
rationale for our studies, but have tried to be as
operational as possible in describing the proce-
dures and results. We defer a discussion of
alternative interpretations of our findings to the
end of the paper.
EXPERIMENT 1A
In this experiment we used the rememberknow
procedure (Gardiner, 1988; Tulving, 1985) and a
lexical decision task to examine the influence of
the processes underlying recollection and famil-
iarity on tests of implicit memory. We chose the
lexical decision task because it is likely that
the rapidity of making such decisions precludes
the possibility of contamination by voluntary,
explicit retrieval. Although the R/K procedure
does not map directly onto recollection and
familiarity, it was chosen because it provides a
tolerable first approximation of these processes
by capturing the subjective experiences of recog-
nition: participants classify old items as remember
(R) if they consciously recollect details or the
context of their occurrence during the study
RECOLLECTIVE PERFORMANCE ADVANTAGES 683
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
phase, or as know (K) if they cannot recollect
specific information associated with the occur-
rence of the item. According to the dual-process
model of recognition memory, R responses are
based on the process of recollection in episodic
memory, whereas K responses are based on the
process of familiarity (Tulving, 1985). According
to the two-stage model of recollection (Moscov-
itch, 2008), we hypothesised that words that are
later recollected (R responses) will be associated
with faster lexical decision response times
(greater priming) and, by inference, better im-
plicit memory than words that are later recog-
nised on the basis of familiarity (K responses).
An important note is that we refer to familiarity
(K responses) as explicit because, as measured by
the R/K procedure, participants are stating that
they explicitly recognise an item as old when they
give it a K response. Further, we consider the
possibility that processing fluency underlies this
decision in the discussion to Experiment 1a and
test it in Experiment 1b.
Method
Participants. A total of 18 people between the
ages of 19 and 28 years (M21.4) participated in
this study in exchange for course credit or an
honorarium of $10. Of the participants, 11 were
female. All were right-handed, had normal or
corrected to normal vision, and were free from
neurological or psychiatric illness, and English
was their primary language.
Stimuli. The stimulus set consisted of two lists
of 120 words. All of the words were monosyllabic
nouns, four to six letters long, low frequency
(between 2 and 10 occurrences per million words;
Kucera & Francis, 1967), and highly concrete
(400700; Coltheart, 1981).
Procedure. Each list was used in an experi-
mental cycle that consisted of four phases: study
phase, distractor phase, lexical decision phase,
and recognition test phase. All words were
presented in 25-point Courier font in the centre
of a computer screen, approximately 60 cm away
from the participant.
For the study phase participants were in-
structed to study 40 words that appeared one at
a time for 2000 ms. A 500-ms fixation cross
appeared during the inter-word interval. During
the distractor phase participants completed
3 minutes of a task in which they were to
decide which of two mathematical equations
would result in a larger value. Participants
completed as many sets of this problem as
possible within the time limit. Participants then
completed the lexical decision phase, in which
they were instructed to indicate by a key press
whether the stimulus that appeared in the
centre of the computer screen was a real
word. The words were presented one at a time
with an inter-word 250-ms fixation cross. The 40
words from the study phase, 40 new words, and
80 nonwords made up the stimulus set. The
nonwords were taken from the ARC nonword
database (Rastle, Harrington, & Coltheart,
2002). The participants were told to respond
by pressing the ‘‘1’’ button if what they saw was
a word and the ‘‘2’’ button if it was not. They
were told to use only their dominant hand and
to respond as quickly and as accurately as
possible. Finally, participants completed the
recognition test phase. Participants were told
that words would be presented one at a time in
the centre of the computer screen and they
were to decide whether the word was one that
they recollected from the study phase, was one
they knew from the study phase, or was one
that they believed was new. They responded by
pressing the ‘‘1’’ button if the word they saw
was one they recollected (R response), the ‘‘2’’
button if the word they saw was one they knew
(K response), and the ‘‘3’’ button if the word
they saw was one they believed was new (N
response). They were told to use only their
dominant hand and to respond as quickly and
accurately as possible. The participants were
given sufficient definitions and examples of
remember, know, and new responses and were
reminded several times that these judgements
were to be made on only the studied words. The
words presented in this phase consisted of the
40 studied words (phase 1), the 40 words that
appeared only in the lexical decision phase
(phase 3), and 40 new words.
The appearance of a word as a studied word,
lexical decision new word, or test phase new word
was randomised across participants. The presen-
tation of words within each phase was rando-
mised for each participant. Each participant
completed two cycles of the above experiment
with a simple problem-solving distractor phase in
between cycles to minimise carry-over from one
experimental cycle to the other. The order of lists
for the two experimental cycles was counter-
balanced across participants.
684 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
Results
Recognition accuracy. As Table 1a shows, the
proportion of responses that were correctly iden-
tified as either old (hits) or new (correct rejections)
via the R/K procedure was significantly different
from chance, t(17)17.97, pB.001. Further, the
difference between R responses to old items (hits)
and to new items (false alarms) was greater than
this difference for K response items. This impres-
sion was confirmed by a repeated-measures AN-
OVA that examined the proportions of responses
for old and new words in each of the test response
categories (R, K, N). There was no difference in
the proportion of responses between old and new
words, F(1,17)0.46, p.05, but there was a main
effect of response, F(2, 34)4.89, pB.05, and an
interaction between the word type and response
given, F(2,34)61.40, pB.001. Specifically, there
were more R responses given to old words than to
new words, t(17)8.20, pB.001 (Cohen’s d
6.34) and more N responses given to new words
compared to old words, t(17)9.43, pB.001
(Cohen’s d6.98), but there was no difference in
the proportion of K responses between old and
new words, t(17)1.37, p0.05 (Cohen’s d
0.75).
Because new words were of two types, those that
appeared in the lexical decision phase for the first
time and those that appeared only in the test phase,
we divided them into these two categories and
entered these values in a subsequent analysis (see
Table 1b). There was a significant difference across
the response types, F(2, 34)47.81, pB.001, and
a significant interaction between the type of new
word and the response given, F(2, 34)82.86,
pB.001. Simpler comparisons revealed that there
were more R responses (R false alarms) and K
responses (K false alarms) for the new words that
appeared in the lexical decision phase than the test
phase new words, t(17)5.67, pB.001, Cohen’s
d3.58; t(17)6.57, pB.001, Cohen’s d3.91,
respectively, but more N responses (correct rejec-
tions) for the test phase new words than the lexical
decision phase new words, t(17)14.20, pB.001
(Cohen’s d5.59).
Comparing the proportion of responses across
R, K, and N response types and the three word
types (old, lexical decision new, test new), we
found that there was a significant interaction,
F(4, 68)64.12, pB.001. Both the lexical
decision new words and test new words had
significantly fewer R responses compared to
old words; t(17)5.68, pB.001, Cohen’s d
4.01; t(17)9.59, pB.001, Cohen’s d8.01,
TABLE 1A
Means estimates of hits, misses, false alarms, correct rejections, and d prime for all experiments
Hits Misses False alarms Correct rejections d?
Experiment 1a
Overall 0.80 (0.02) 0.20 (0.02) 0.42 (0.04) 0.58 (0.04) 1.04
R responses 0.46 (0.04) 0.11 (0.02) 1.13
K responses 0.35 (0.03) 0.31 (0.02) 0.11
Experiment 1b
Overall 0.80 (0.02) 0.20 (0.03) 0.46 (0.04) 0.54(0.04) 1.02
R responses 0.51 (0.03) 0.17 (0.03) 0.98
K responses 0.29 (0.03) 0.29 (0.03) 0.0
Experiment 2
Overall 0.84 (0.02) 0.16 (0.02) 0.12 (0.03) 0.88 (0.03) 2.17
R responses 0.46 (0.03) 0.02 (0.01) 1.95
K responses 0.39 (0.03) 0.10 (0.01) 1.00
Means estimates of hits, misses, false alarms, correct rejections, and d prime for Experiment 1a, Experiment 1b,
and Experiment 2 (standard errors are shown in parentheses next to the corresponding means).
TABLE 1B
Proportion of R, K, and N responses for new words:
Experiments 1a and 1b
R response K response N response
Experiment 1a
Lexical Decision New 0.18 (0.03) 0.41(0.02) 0.40 (0.04)
Test New 0.04 (0.01) 0.21 (0.03) 0.76 (0.04)
Experiment 1b
Lexical Decision New 0.28 (0.03) 0.31 (0.03) 0.40 (0.03)
Test New 0.10 (0.02) 0.23 (0.04) 0.68 (0.05)
The proportion of R, K, and N responses for the two types
of new words (lexical decision new words and test new words)
for Experiment 1a and Experiment 1b (standard errors are
shown in parentheses next to the corresponding means).
RECOLLECTIVE PERFORMANCE ADVANTAGES 685
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
respectively. There were fewer K responses,
t(17)3.18, pB.005, Cohen’s d3.58, for test
phase new words than old words (but not for
the lexical decision new words). Also, old words
were given significantly fewer N responses (misses)
than both lexical decision new words and test
new words: t(17)4.71, pB.00, Cohen’s d
3.73; t(17)12.27, pB.001, Cohen’s d9.11,
respectively.
Reaction time. To examine the effect of word
type (old versus new) on response time (RT)
during the lexical decision phase, we calculated
each participant’s mean RT to make a lexical
decision. To ensure that lexical decision RT was
accurately reflected by the mean RT, we set a
response criterion of at least 10 correct responses
in each of the critical experimental categories. This
involved testing 26 participants to get 18 who met
this criterion (only the data for the 18 participants
who met the response criterion were included in
the above recognition accuracy analyses). Further-
more, RTs that were too fast (less than 350 ms) or
too slow (greater than 3500 ms) were eliminated,
as were RTs associated with incorrect lexical
decision responses. This eliminated less than 8%
of the total responses across all participants.
The mean RT to old words was 680 ms (SD
110 ms), which was faster than the mean RT to new
words (730 ms, SD119). A paired sample t-test
confirmed that this difference was significant,
t(17)4.31, pB.001 (Cohen’s d0.97) indicating
that there was a priming effect in this experiment.
Given that participants completed two study
test cycles, a repeated measures ANOVA was run
on the mean lexical decisions RTs for old and new
words with list order as the within-participants
factor. Not surprisingly, there was a main effect of
word type, F(1, 17)18.97, pB.001; however, the
effect of list order was not significant, F(1, 17)
1.49, p.05, nor was the interaction between
word type and list order, F(1, 17)0.09, p.05.
This confirms that list order did not affect the
pattern of results for this study.
Mean RTs to words presented in the lexical
decision phase were then categorised according to
their associated response in the test phase. There-
fore we had a total of six experimental categories:
old words given a remember (R), know (K), or
new (N) response, and new words given a remem-
ber (R), know (K), or new (N) response. The
corresponding means are reported in Table 2.
Since there were few participants who reached
the appropriate response criterion (10 responses)
for new words given R responses (recollection
false alarms) and old words given N responses
(misses), we first will focus on the three condi-
tions in which the associated test response was
correct (old words with a R response, old words
with a K response, new words with a N response).
From Table 2 it is evident that RTs to old
words with a R response and to old words with a
K response were significantly faster than to new
words with a N response. Furthermore, RTs to old
words with a R response were significantly faster
than to old words with a K response. A repeated-
measures ANOVA confirmed this pattern. RTs
were significantly different across the three types
of words, F(1.57, 26.83)11.69, pB.001; degrees
of freedom corrected using the Huynh-Feldt
estimates of sphericity. Simple planned compar-
isons revealed that RTs to old words with a R
response were faster than RTs both to old words
with a K response, t(17)3.41, pB.005 (Cohen’s
d0.74) and to new words with a N response,
t(17)4.68, pB.001 (Cohen’s d1.67). RTs to
old words with a K response were also faster
than to new words with a N response, t(17)2.98,
pB.01 (Cohen’s d0.95).
RTs to words with the appropriate R, K, and N
responses were also compared to the overall mean
RT for all old words and all new words. A repeated-
measures ANOVA revealed significant differences
between these responses, F(4, 68)11.21, pB.001.
RTs to new words with a N response were
not significantly different from RTs to all new
words, regardless of their subsequent memory
response, t(17)1.51, p.05 (Cohen’s d0.35).
Furthermore, RTs to old words with a R response
were significantly faster than the RTs to all new
words, t(17)5.91, pB.001 (Cohen’s d1.58), as
TABLE 2
Mean lexical decision response times: Experiments 1a
and 1b
Word type R response K response N response
Experiment 1a
Old 656 (21.8) 692 (26.9) 703 (32.9)
New 721 (36.2) 713 (25.3) 751 (35.4)
Experiment 1b
Old 696 (26.7) 681 (34.5) 657 (27.4)
New 670 (30.7) 699 (35.0) 697 (34.3)
Mean lexical decision response times (in ms) as a function
of study status and recognition response (standard errors are
shown in parentheses next to the corresponding means) in
Experiment 1a (experimental condition) and Experiment 1b
(control condition).
686 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
were RTs of old words with a K response, t(17)
3.26, pB.01 (Cohen’s d0.74). RTs to old words
with a R response were also significantly faster
than the mean RTs of all old words, t(17)2.73,
p.014 (Cohen’s d0.56), but that was not the
case for RTs to old words with a K response,
t(17)1.54, p.05 (Cohen’s d0.22).
One question that was important to address is
whether RTs to items that were studied but not
recognised (misses) were faster than the RTs to
items that were not studied, and whether misses
differed from old words given correct R and K
responses. To examine this, we performed a
repeated-measures ANOVA on RTs to all old
words categorised by their subsequent memory
response (R, K, N) and to new words correctly
recognised as N. A note of caution is that, in this
analysis, some participants did not contribute an
adequate number of responses (i.e., 10 or greater)
to the old words given an N response condition
(misses). Nevertheless, these analyses found a
significant difference across the four responses,
F(3, 51)9.49, pB.001. Simple planned compar-
isons revealed a significant difference between
RTs to old words given a R response and to those
given a N response, t(17)2.42, pB.05 (Cohen’s
d0.86), no significant difference between RTs to
old words given a K response and those given a N
response, t(17)0.62, p.05 (Cohen’s d0.18),
and a significant difference between RTs to old
words given a N response and new words given a N
response, t(17)2.62, pB.05 (Cohen’s d0.70).
Discussion
In Experiment 1a we found that studied words
later recognised via R responses were associated
with faster lexical decision RTs (more priming)
than both studied words later given a K response
and new words. Furthermore, studied words that
were given a R response were significantly faster
than those given a N response (misses). This was
not the case for old words given a K response as
compared to misses, although both had faster RTs
compared to new words correctly labelled as new
(N response). This suggests that processes under-
lying explicit memory per se do not confer a benefit
to priming because priming associated with K
responses (presumably based on familiarity) is no
greater than that for studied items that were
missed (but see below). It is only the processes
underlying R responses (presumably recollection)
that contribute to priming beyond what typically is
reported on implicit memory tasks.
These priming results need to be interpreted
in light of the accuracy performance on the
recognition test. Accuracy was greater for old
words given R responses than for K responses.
Hits given R responses were significantly greater
than R response false alarms (FA), but hits and FA
were equivalent for K responses. As well, the
proportion of FA was greater for new words that
appeared in the lexical decision phase as compared
to words that appeared for the first time at test.
Overall, these findings suggest that when assigning
a K response, participants could not distinguish old
from new items, although they could do so on the
basis of recollection. These results may reflect K
responses relying primarily on item memory.
Participants had seen the targets and many of the
lures during the priming phase, thus making item
memory a poor basis for distinguishing between
them.
In light of the poor accuracy performance of K
responses, the equivalent RTs to K responses to
words recognised (hits) and words not recognised
(misses), both of which were significantly faster
than to new words, suggest that priming for K
responses is independent of explicit memory, and
vice versa. Had fluency of processing influenced
recognition memory for K responses, then it
would have paralleled the priming results, and
better memory would have been observed for old
items than for new ones, which was not the case.
There remains the possibility that the particular
words presented in the experiment have attributes
that led to both greater priming and recollection.
That is, some underlying item characteristic may
have influenced both RTs during the lexical
decision phase and performance on the recogni-
tion test. Similarly, it is possible that studied words
with a fast RT induced a R response on the
subsequent test, or led to greater fluency of
processing on the subsequent test which influenced
both R and K responses, although as we noted the
likelihood of this affecting K is low. Thus, before
accepting the above conclusion that recollection
benefits priming disproportionately, it is necessary
to examine these alternative hypotheses. To do so,
we conducted Experiment 1b.
EXPERIMENT 1B
The purpose of Experiment 1b was to rule out the
possibility that the relation observed in Experiment
RECOLLECTIVE PERFORMANCE ADVANTAGES 687
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
1a was determined by item characteristics that
influence both lexical decision RTs and item
memorability (Balota, Cortese, Sergent-Marshall,
Spieler, & Yap, 2004) or the possibility that the
items associated with the fastest RTs on the
lexical decision task provoked R responses at
recognition. To test this we replicated Experiment
1a, but with the experimental phases presented in
a different order. Participants first made lexical
decisions, then studied a list of words that
included those presented in the lexical decision
phase, and then were tested for their recognition
of the studied words. If lexical decision speed or
other item characteristics are a factor in deter-
mining the subsequent memory of a later studied
word, then items with the fastest lexical decision
RTs should be the ones given R responses rather
than those given K responses at recognition. On
the other hand, if lexical decision speed or other
item characteristics are not a determining factor,
then there should be no relation between the
lexical decision RTs and studied words given a R
or a K response.
Method
Participants. A total of 18 people selected from
the same participant pool as in Experiment 1a
completed this experiment. This involved testing
25 participants to get 18 participants who appro-
priately met the criterion stated in Experiment
1a. The participants mean age was 21.0 years and
13 were female.
Stimuli and procedure. The stimulus set was the
same as that used in Experiment 1a. The experi-
mental task differed only in terms of order of
tasks (phases). In this experiment, participants
completed (1) the lexical decision phase, (2) the
study phase, (3) the distractor phase, and then (4)
the test phase.
Results and discussion
Recognition accuracy. Shown in Table 1a, the
proportion of responses that were correctly iden-
tified as either old (hits) or new (correct rejec-
tions; CR) was significantly different from chance,
t(17)19.60, pB.001. A repeated measures AN-
OVA examining the proportions of responses
given to old and new words in each of the response
categories (R, K, N) showed that there was no
difference in proportion between old and new
words, F(1,17)1.00, p.05, no difference in the
proportion of responses, F(2, 34)1.57, p.05,
but an interaction between word type and
response given, F(2, 34)66.45, pB.001. Simpler
comparisons showed that there were more R
responses given to old words than new words,
t(17)8.15, pB.001 (Cohen’s d5.24), more N
responses given to new words compared to old
words, t(17)9.63, pB.001 (Cohen’s d6.08),
but no difference in the proportion of K responses
between old and new words, t(17)0.12, p.05
(Cohen’s d0.05).
As in Experiment 1a, we investigated the
influence exerted by the two types of new words
on the proportion of responses (Table 1b). There
was a significant difference across the response
types, F(2, 34)48.17, pB.001, and a significant
interaction between the type of new word and the
response given, F(2, 34)8.26, p.001. There
were more R responses (FA) for the new words
that appeared in the lexical decision phase than
for the test phase new words, t(17)4.02, pB.001
(Cohen’s d3.54), a similar number of K re-
sponses across the type of new words, t(17)0.69,
p.05 (Cohen’s d0.66) and more N responses
(CR) for the test phase new words than the lexical
decision phase new words, t(17)3.20, p.005
(Cohen’s d2.77).
Importantly, comparing the proportion of re-
sponses across R, K and N response types and the
three word types (old, lexical decision new, test
new), we found a significant interaction, F(4, 68)
23.13, pB.001. Both the lexical decision new words
and test new words had significantly fewer R
responses than old words, t(17)3.41, pB.005
(Cohen’s d3.32; t(17)10.89, pB.001 (Cohen’s
d6.63), but there was no difference in the
proportion of K responses compared to old words,
t(17)0.17, p.05 (Cohen’s d0.05); t(17)
1.28, p.05 (Cohen’s d0.64). Also, old
words were given significantly fewer N responses
than both lexical decision new words and test
new words, t(17)4.68, pB.001, (Cohen’s d
4.20); t(17)10.05, pB.001 (Cohen’s d6.50),
respectively.
Reaction time. As in Experiment 1a, the mean
RTs to all old words and new words were first
compared. It is important to note that since the
lexical decision phase preceded the study phase,
old words were not actually ‘‘old’’ when the
participants made the lexical decision; however,
to make appropriate comparisons to Experiment
1a, we will use this terminology. The mean RT
to old words (682 ms, SD105 ms) was not
688 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
significantly different from the mean RT to new
words (mean686 ms, SD124); t(17)0.31,
p.05 (Cohen’s d0.08). This was expected
since participants were seeing both old and new
words for the first time in this experiment.
Following Experiment 1a, mean RTs to words
presented in the lexical decision phase were then
categorised according to their associated re-
sponses in the test phase (R, K, N). These mean
RTs are reported in Table 2. A repeated-measures
ANOVA indicated that RTs were not significantly
different across the three types of correct re-
sponses, F(2, 34)0.14, p.05. Again, none of
the simple comparisons that were significant in
Experiment 1a was significant in this experi-
ment*Old-R responses: Old-K responses,
t(17)0.48, p.05, Cohen’s d0.25; Old-R re-
sponses: New-N responses, t(17)0.09, p.05,
Cohen’s d0.05; Old-K responses: New-N re-
sponses, t(17)0.36, p0.05, Cohen’s d0.18.
RTs to old words with R and K responses and
to new words with N responses were also com-
pared to the overall mean RTs for all old words
and all new words. A repeated-measures AN-
OVA did not reveal significant differences be-
tween these responses, F(4, 68)0.17, p.05.
Again, none of the simple planned comparisons
that were significant in Experiment 1a was
significant in this experiment.
We also performed a repeated-measures AN-
OVA on the RTs to all old words categorised by
their subsequent memory response (R, K, N) and
to new words correctly recognised as N (see note
of caution from Experiment 1a), and did not
obtain a significant difference across the four
responses, F(3, 51)0.92, p.05. Not surpris-
ingly, there was no significant difference between
RTs to old words given a R response and to those
given a N response, t(17)1.9, p.05 (Cohen’s
d0.77), between RTs to old words given a K
response and to those given a N response, t(17)
0.85, p.05 (Cohen’s d0.43), nor between RTs
to old words given a N response and to new words
given a N response, t(17)2.08, p.05 (Cohen’s
d0.63).
To assess further the possibility that some
features of specific items led to fast lexical
decision RTs as well as an increased likelihood
that the item would be recollected, we used the
average RTs for each word in Experiment 1b as a
baseline measure to adjust the RTs from Experi-
ment 1a in an item-specific manner. That is, for
each participant in Experiment 1a, we subtracted
the baseline calculated from the RTs collected
Experiment 1b for each item. Then we did the
same analysis as stated previously wherein we
compared the average mean RTs to items given
R, K and N responses. If there were an item-
specific contribution to the recollection super-
iority effect, this procedure would nullify it.
Importantly, we found the same pattern as in
Experiment 1a. That is, there was a significant
difference between RTs to old words given a R
response, old words given a K response, missed
old words and new words given a N response,
F(3, 51)6.99, p.001. Simple comparisons
revealed that RTs to old items given a R
response, a K response, and a N response (missed
items) items were faster than RTs to correctly
labelled new items, t(17)4.65, pB.05 (Cohen’s
d1.03); t(17)2.46, pB.05 (Cohen’s d
1.04); t(17)2.32, pB.05 (Cohen’s d0.80),
respectively.
Importantly, RTs to old items given a R
response were faster than to old items given a K
response, t(17)2.15, pB.05 (Cohen’s d0.62)
and to missed old items t(17)2.21, pB.05
(Cohen’s d0.80), but RTs to old items given a
K response were not different from RTs to missed
old items, t(17)0.53, pB.05 (Cohen’s d0.19).
This demonstrates that there is not some item
characteristic that is leading to fast lexical decision
RT and better recognition memory.
EXPERIMENTS 1A AND 1B:
DISCUSSION
Experiment 1a suggests that processes associated
with R responses can benefit performance on
implicit memory tasks. Experiment 1b rules out
the possibility that lexical decision speed or other
item characteristics were the cause of the effect
found in Experiment 1a because there was no
relation between RT and item memorability when
the lexical decision RTs were measured before the
study phase.
Overall, the results of these experiments are
consistent with the hypothesis that recollective
processes underlying R responses that are pre-
sumed to be involved exclusively in explicit
memory also influence performance on tests of
implicit memory. We believe that this finding is
unlikely to be related to contamination of implicit
memory performance by voluntary retrieval of
information from episodic memory (see MacLeod,
2008). First, the R advantage was found on a test of
lexical decision, which has been shown to be
RECOLLECTIVE PERFORMANCE ADVANTAGES 689
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
insensitive to deliberate contamination (Goshen-
Gottstein, & Moscovitch, 1995). Second, the mean
lexical decision RT, which was on the order of
760 ms, was much faster than the mean RT to
make a recognition response, which was on the
order of 1200 ms, a decision that requires con-
scious retrieval. Third and more importantly, the
traditional priming advantage for unrecognised
studied items (misses) over unstudied items, which
constitutes the basis of uncontaminated priming,
was no different than the priming advantage for
familiar words (K responses). These familiar
words are also considered part of explicit memory
(because they are recognised on the memory test)
and are therefore as open to ‘‘contaminated’’
priming as recollected words. However, the prim-
ing augmentation pertained only to R responses.
It is important to note that although the
proportion of R response hits was greater than R
response FA, this was not the case for the
difference between K response hits and FA. This
indicates that participants could distinguish be-
tween targets and lures on the basis of recollection,
which capitalises on context or source information,
but not on the basis of familiarity, which depends
only on item information. Thus, whereas priming
seems to benefit from explicit memory for words
associated with R responses, equivalent priming
was obtained for K hits and misses. This suggests
that priming for the latter cases (K responses), but
not the former case, is similar to that obtained by
patients with amnesia whose explicit memory is
severely impaired. The significance of this finding
for strength theories of memory will be considered
in the General Discussion.
Because of the implications these findings have
for theories of explicit and implicit memory, it is
important to determine whether the R advantage
seen here generalises to other implicit memory
tasks. We conducted a second study using word
stem completion; one of the most commonly used
implicit tests of memory.
EXPERIMENT 2
To examine the effects of processes underlying
recollection and familiarity on word-stem comple-
tion we chose the Horton et al. (2001) procedure
for administering this task because it has been
shown to be minimally affected by explicit, con-
scious retrieval. Thus we can assess whether
automatic processes associated with R responses
affect performance on implicit tasks without a
voluntary, conscious retrieval component and
thereby establish the generality of the R advan-
tage on implicit tasks.
In the Horton et al. (2001) procedure partici-
pants study a list of words, and then receive a
practice stem completion test in which none of
the stems corresponds to the studied words, and
during which participants respond as quickly as
possible. The inclusion of this task minimises any
bias participants may have to adopt the strategy
of using explicit memory of studied words to
complete the task. After the practice test, parti-
cipants are given the stem completion phase in
which half of the stems can be completed with the
studied words. Again, they are encouraged to
complete the stems as quickly as possible. Horton
et al. (2001) found that participants under this
procedure completed critical stems more quickly
than participants who performed this task but
were asked to switch to a conscious retrieval
strategy. The faster RTs during the speeded
variation of the task compared to the conscious
retrieval strategy variation are thought to repre-
sent the exclusion of conscious, deliberate retrie-
val from the speeded version.
Following the results of Experiment 1a and 1b,
we hypothesised that even in the case where
explicit retrieval is not emphasised, automatic
processes associated with R responses will benefit
performance on the word stem completion task.
To test this we modified the above-described
speeded variation of the stem completion task
so that we could assess subsequent memory using
the R/K procedure, as described in Experiment
1a. We expected that stem completion RTs would
be faster for words associated with a R response
than for words associated with a K response.
Method
Participants. A total of 28 people selected from
the same participant pool as in Experiment 1a
completed this experiment. Of these, 16 met the
set response criterion as described in Experiment
1a, and were included in this experiment. The
mean age was 18.9 years and nine of the
participants were female.
Stimuli. The stimulus set used in this experi-
ment consisted of 180 words and their respective
stems chosen from a normative set of word stem
responses (Horton, 1989). The words were chosen
so that their stems had a 0.10 to 0.30 probability
of being completed with the critical word.
690 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
Furthermore, all stems were unique within the
stimulus set list.
Procedure. Participants completed five pha-
ses: study phase, practice stem completion phase,
distractor phase, critical stem completion phase,
and test phase. Words were randomly assigned to a
phase for every participant. For the study phase,
participants were asked to rate the pleasantness of
60 words on a scale of 1 to 5, indicating their
responses by key presses. The practice phase
consisted of 60 stems in which the participants
were instructed to complete the stem as quickly as
possible with the first word that came to mind.
None of these stems could be completed with the
studied words. Their response times were recorded
via a microphone. As well, the experimenter
recorded the responses. In the distractor phase,
participants completed a 3-minute distractor task
in which they had to decide which of two mathe-
matical equations would result in a larger value.
Participants completed as many of these problem
sets as possible within the time limit. The critical
implicit phase consisted of the same stem comple-
tion task as the practice phase; however, the
stimulus set for this task consisted of the 60 words
from the study phase (critical words) and 60 new
stems. The R/K recognition test phase was the
same as that described in Experiment 1a. The
words presented in this phase consisted of the 60
studied words (phase 1) and 60 words that could be
used to complete the new stems seen in the critical
stem completion phase.
Results
Recognition accuracy. As shown at the bottom of
Table 1a, the proportion of responses that were
correctly identified as either old or new was
significantly different from chance, t(15)34.68,
pB.001. A repeated-measures ANOVA that ex-
amined the proportions of responses given to old
and new words in each of the response categories
showed that there was no significance in propo-
rtions between old and new words, F(1,15)0.00,
p.05, but a significant difference in the propor-
tions across responses, F(2, 30)61.74, pB.001,
and a significant interaction between the word type
and response given, F(2, 30)266.31, pB.001.
There were more R responses given to old words
than given to new words, t(15)13.35, pB.001
(Cohen’s d13.48), more K responses given to old
compared to new words, t(15)8.96, pB.001
(Cohen’s d6.46), and more N responses given
to new words compared to old words, t(15)23.50,
pB.001 (Cohen’s d18.94). Unlike Experiment
1a and 1b, there was only one type of ‘‘new’’ word.
For old words, there was a significant difference
in the proportion of responses given in each of the
three response types, F(2, 30)24.38, pB.001.
There were significantly more R responses and K
responses than N responses, t(17)6.74, pB.001
(Cohen’s d6.88; t(17)6.97, pB.001 (Cohen’s
d5.69), respectively, but no difference between
the proportion of K responses and R responses,
t(17)1.41, p.05 (Cohen’s d1.92). For new
words, the proportion of responses in each re-
sponse type also showed a significant difference,
F(2, 30)358.53, pB.001. Significantly more N
responses were given than R or K responses,
t(15)28.56, pB.001 (Cohen’s d33.34), t(15)
15.83, pB.001 (Cohen’s d19.20), respectively,
and more K responses than R responses, t(15)
4.17, p.001 (Cohen’s d3.47).
Reaction time. To compare our results to those of
Horton and colleagues (2001), we first compared
the average mean response times (RT) to the stems
associated with studied words (regardless of
whether the studied word was used to complete
the stem) and the RTs to the stems associated with
the new words presented in the critical stem
completion phase. To be consistent with the
statistics used in Experiment 1a and 1b, the
average mean RT was used in this experiment
rather than the average median RT that was used
by Horton and colleagues, although the same
pattern of results is obtained if the median RTs
are used. Replicating Horton et al. (2001), studied
words had significantly faster stem completion
times (1053 ms, SD270 ms) compared to new
words (1128 ms, SD263 ms), t(15)3.37, pB.01
(Cohen’s d0.63) (removing stems that were
incorrectly completed or took over 3500 ms to
complete). Both had significantly faster stem
completion times compared to the practice stem
completion phase (1356 ms, SD407 ms); t(15)
3.53, pB.01 (Cohen’s d2.00); t(15)2.78, pB
.05 (Cohen’s d1.52); old and new words, respec-
tively. Therefore we are confident that our meth-
ods are similar to those of Horton and colleagues:
Participants were not using conscious explicit
retrieval to complete the stems in our study, just
as they were not in the Horton et al. (2001) study.
Similar to Experiment 1a and 1b, the stems were
then classified according to their recognition
response (R or K for old words). We could not
include the RTs for misses (old words given a N
RECOLLECTIVE PERFORMANCE ADVANTAGES 691
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
response) in this experiment because many parti-
cipants did not have any miss responses. A
repeated-measures ANOVA revealed a significant
difference between the stems completed with
studied words given a R (923.6 ms, SD160 ms)
or K response (973.9 ms, SD172 ms) and the
mean RT to new stems (1128 ms, SD263 ms),
F(2, 30)22.80, pB.001. Planned comparisons
revealed that RTs to stems completed with stu-
died words given a R response were significantly
faster than to those given a K response, t(15)
3.31, p.005 (Cohen’s d0.67), and to new words,
t(15)5.08, pB.001 (Cohen’s d2.93). Stems
completed with studied words given a K response
were also significantly faster than to new words,
t(15)4.50, pB.001 (Cohen’s d2.52).
Discussion
In Experiment 2 we found that studied words that
received a R response were associated with faster
word stem completion times than were words that
received a K response or new words. Also, studied
words given a K response were faster than new
words. Because there were such few misses, we
could not determine whether completion to
words that were missed differed significantly
from those that were recognised correctly. Also,
unlike Experiment 1a and 1b, the proportion of
both R and K hits significantly exceeded that of
FA of each type. Here, memory based on K
responses is well above chance, yet priming for R
responses, as measured by time to complete the
stems, still exceeds that for K responses. Overall,
these findings are consistent with those from
Experiment 1a and support the hypothesis that
some processes underlying R responses benefit
performance on implicit memory tasks.
We think that our results are not due to the
contamination of consciously retrieved memories.
As noted earlier, we chose Horton and collea-
gues’ (2001) speeded word stem completion task
because it has been shown to be free from
contamination by conscious retrieval. In fact,
previous studies suggest that when conscious
retrieval is used in a word stem completion task,
RT suffers (see Richardson-Klavehn & Gardiner,
1995, 1996, 1998; Toth, 1996). Our pattern of
results was opposite to this in that RTs were faster
to words given R responses than to words given K
responses, reinforcing our conjecture that partici-
pants were not using conscious retrieval of the
studied words to complete the stems. Moreover,
the R advantage was weakened when accuracy
rather than RT was used as a measure. In contrast
to speed, accuracy provides a measure of all
influences on performance, both conscious and
unconscious, without distinguishing between
them. Words receiving K responses were also
completed more accurately than misses, suggest-
ing that the contribution of consciously retrieved
information to accuracy is not insignificant.
Overall, our finding that words given a R
response were associated with faster stem com-
pletion performance supports the two-stage ac-
count of recollection. Information associated with
recollection can be retrieved from episodic mem-
ory during an initial phase rapidly, relatively
automatically, and without conscious awareness
(Moscovitch, 2008).
GENERAL DISCUSSION
The goal of the current study was to explore the
relation between processes underlying recollec-
tion and performance on two implicit memory
tasks: lexical decision and speeded word stem
completion.We hypothesised that processes asso-
ciated with recollection would contribute to
performance on these non-conscious, implicit
memory tasks. Our primary finding was in line
with this hypothesis. Studied words that were
later given a R response were associated with
greater priming than studied words that were
later given a K response. No difference was found
between recognised words given a K response
and missed words, although both showed priming
effects. Our results establish the basic phenom-
enon that some non-voluntary processes asso-
ciated with explicit memory can contribute to
performance on tests of implicit memory.
Alternative explanations of the R
advantage
Before describing possible interpretations of our
findings, we consider a number of alternative
explanations for our results in light of our own
data and previous findings.
Contamination of implicit memory tasks by
voluntary or involuntary retrieval of explicit
memory. As we noted earlier, we chose implicit
memory tests that are likely to be resistant to
contamination (Goshen-Gottstein & Moscovitch,
692 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
1995: Horton et al., 2001; see discussion in
MacLeod, 2008). For both tasks, latencies for
completing the implicit memory task (lexical
decision760 ms; word stem completion
1014 ms) were significantly faster than the recog-
nition judgement response times, a task that does
require the retrieval of specific conscious explicit
memories (1224 ms and 1575 ms for the lexical
decision task and word stem task, respectively). In
addition, numerous studies have found that when
conscious retrieval contaminates implicit memory
tasks, the process is slowed (Horton et al., 2001;
Toth, 1996), but here we found the opposite effect.
Items given R responses were associated with
more rapid priming than those that were pre-
sented but not recognised at all. Even the items
associated with K responses were not slowed with
respect to misses, although both were faster than
the RTs to new items.
It is possible that awareness of an item from
the study phase arose spontaneously during the
implicit memory task, accounting for enhanced
priming effects. The idea that involuntary con-
scious memory affects priming is a conte-
ntious issue with some researchers endorsing it
(e.g., Kinoshita, 2001) and others not (e.g.,
Richardson-Klavehn & Gardiner 1996; Roediger
& McDermott, 1993). Although we are not
opposed to this idea in principle, we think it
cannot account fully for our results, as it would
also predict a benefit for all memorable words of
which the participants were aware, including
those judged as familiar, in comparison to words
that were presented but not recognised (misses).
Contrary to this prediction, we found no differ-
ence in priming between words receiving a K
response and missed words in Experiment 1a.
This interpretation, however, is mitigated by the
fact that there was no significant difference
between the proportion of K hits and FA.
Using the PDP, Jacoby and colleagues (Jacoby,
1991; Toth, Reingold, & Jacoby, 1994) argued that
performance on indirect tests of memory may be
contaminated by explicit memory. They note that
applying the PDP can remove the source of
contamination, leaving a measure of implicit
memory that is not influenced by variables that
benefit explicit memory, such as depth of proces-
sing or generation. While this may be true,
applying such reasoning (or the PDP) to our
study effectively would remove most of the items
associated with recollection, given that recollec-
tion is most influenced by those variables. By
removing all the recollected items, the PDP does
not allow one to answer the question we have
posed*do processes associated with recollection
influence performance on implicit tests before or
without the participant’s awareness of the explicit
memory for the item? Our tests suggest that a
crucial part of the interaction between explicit
and implicit memory occurs before participants
become aware of their explicit memory.
Greater priming leads to greater recollection.As
a first step, Experiment 1b ruled out the possibi-
lity that it was lexical decision times (or item
characteristics associated with lexical decision
times) that influenced subsequent memory for
words. When lexical decision times were mea-
sured prior to learning, there was no relation
between RTs and subsequent memory.
One might still argue that implicit memory
itself may be causally contributing to recollection,
as it does on some tests in which explicit memory
for the studied item is weak and benefits, by
inference, from priming or processing fluency
(see Kinoshita, 2001; Masson & MacLeod,
1997). If true, this finding itself would be novel
and noteworthy, but we believe that this inter-
pretation is unlikely for the following reasons.
First, there is evidence from Wagner, Maril, and
Schacter (2000) to indicate that priming leads to
worse recognition memory for primed as com-
pared to unprimed items (see also Stark, Gordon,
& Stark, 2008). This pattern is the opposite of the
one that we obtained making the hypothesis that
priming contributes to recollection suspect.
Further to this point, in the present study the
amount of time spent processing highly primed
words (those with the fastest RTs) during the
implicit memory phase was significantly less than
the amount of time spent processing the words
that were not primed as much, or not primed at
all. Since explicit memory is known to vary with
study time for the same item (items typically have
better memory when they are studied longer),
one would expect that memory would be better
for an item when it is more poorly primed than
when that item is more highly primed because of
the additional processing the poorly primed item
would receive during the implicit memory task.
On an item-level basis, we did not see a relation
between amount of processing time and explicit
memory. That is, explicit memory was no better
for a particular item that was allotted more
processing time (i.e., more poorly primed) for a
particular participant than when it was a hig-
hly primed word for another participant.
RECOLLECTIVE PERFORMANCE ADVANTAGES 693
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
Last, in Experiment 1a we found significant
priming for old words regardless of whether they
were recognised correctly or missed. This is
particularly telling for proportion of K response
hits, which were equivalent to that of K response
FA. If priming had influenced recognition via
fluency, one would have expected that K response
hits would have exceeded K response FA since
the former were primed but the latter were not.
Implications and interpretations
As stated previously, our results broadly support
the notion that explicit and implicit memory share
underlying processes, consistent with findings from
other researchers (e.g., MacLeod & Masson, 2000;
Masson & MacLeod, 1997). We have theoretically
framed our study with a dual-process model of
recognition memory to investigate the specific
benefit of recollection to implicit memory. Early
dual-process views described familiarity as a con-
tinuous, strength variable and recollection as a
categorical, high-confidence (threshold) variable
(Yonelinas, 2002). An alternative interpretation,
however, is that R responses and K responses do
not reflect recollection and familiarity, but reflect
different degrees of memory strength. Items that
receive R responses have a stronger memory trace
than items that receive K responses (Dunn, 2008;
Wixted & Stretch, 2004). In fact, single-system,
signal detection strength models have been pro-
posed to account for the relation between explicit
and implicit memory in a number of tasks (see
Berry, Shanks, & Henson, 2008). The pattern we
obtained in Experiment 1a resembles the pattern
of RTs that Berry and colleagues observed in their
own study and in others that they reviewed for
priming of items associated with hits, misses, FA
and CRs.
However, behavioural (Diana, Reder, Arndt, &
Park, 2006; Yonelinas, 2002) and neuropsycholo-
gical evidence (Diana, Yonelinas, & Ranganath,
2007; Eichenbaum et al., 2007; Rugg et al., 1998;
Wagner, Gabrieli, & Verfaellie, 1997) regarding
explicit memory, have led many investigators to
favour a dual-process model of recognition. In
light of such evidence, some investigators, combin-
ing both viewpoints, have proposed dual-process,
aggregate strength models to account for recogni-
tion performance. In contrast to dual-process
threshold models, the aggregate strength models
posit that both recollection and familiarity vary
continuously in strength (Mickes, Wais, & Wixted,
2009; Rotello & Macmillan, 2006; Rotello, Mac-
millan & Reeder, 2004; Wixted, 2007). To our
knowledge, such dual-process, strength models
have not been applied to deal with the relation
between explicit recognition and priming,
although it is conceivable that they can do so.
With respect to our own findings, such models may
predict that priming, like recognition, would be
determined by the aggregate strength drawn from
recollection and familiarity. Because recollection
is typically associated with greater strength than
familiarity, one would expect that priming should
be better for items associated with R than with K
responses, as was the case.
However, the equivalent RTs for K response
hits and misses in Experiment 1a, and the greater
RTs for misses compared to CR, are problematic
for strength models. This problem is easily resolved
if we assume that strength for K response hits is
very weak, which it seems is the case in our study
as they are indistinguishable from K response FAs,
and that misses are based on stronger traces than
are CR. To test these theories against each other
and against our own model (see below), future
investigations should collect data on confidence
levels associated with each of these responses.
The two-stage model of recollection
Related to the hypothesis we set out to test, and
taking into account the other alternatives we have
considered, we believe our findings are consistent
with the two-stage model of recollection (Mos-
covitch, 2008). The first stage is automatic,
dependent on the MTL, and can be implemented
without conscious awareness. The second stage is
slower, dependent on the interaction of prefron-
tal and parietal cortex with the MTL and, as a
result, is associated with conscious awareness.
Evidence for the first stage of recollection
comes from the present study in which there was
a priming advantage for items that were also
recollected as well as from other studies that have
found a recollective advantage on tasks thought not
to rely on explicit/episodic memory (Westmacott &
Moscovitch, 2002, 2003; Westmacott et al., 2004).
In these cases the retrieval required is not specifi-
cally linked to recovering details from the past at
the moment of retrieval. That is, the demands at
retrieval are fairly minimal, such as those during
lexical decision or word stem completion. In other
situations when task demands are minimal, such as
in particular old/new recognition tasks, RTs are
694 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
faster for those items that are recollected than for
those that are recognised on the basis of familiarity
(Dewhurst & Conway, 1994; Dewhurst, Holmes,
Brandt, and Dean, 2006; Diana et al., 2007;
Gardiner, Gregg, & Kariyianni, 2006), paralleling
the priming results that we found.
The second stage of recollection is involved
when retrieval of specific details from past
episodes is accompanied by conscious awareness
of the episode and decisions must be made on
this basis. That is, if the task demands are great,
a slower process comes into play in which more
information is recovered and evaluated. This was
the case for the explicit memory task used in our
experiments wherein participants had to reflect
on the retrieved information to determine
whether it was adequate to support a recollec-
tion decision. In these cases, decisions based on
recollection are slower than those based on
familiarity, the reverse of what occurs when
merely deciding whether an item is old or new
or when task demands are fairly minimal (e.g.,
dual-process model; Boldini, Russo, & Avons,
2004; Toth, 1996; Yonelinas & Jacoby, 1994).
Earlier studies, typically those that employ the
PDP model (Jacoby, 1991; Toth et al., 1994),
describe recollection as a slow and effortful
process much like the second stage of the two-
stage model. Thus our model is not inconsistent
with such theories of recollection. While earlier
models help disentangle the contributions of the
recollection and familiarity when one is engaged
in high-demand decisions (e.g., stage two), our
model helps to disentangle processes involved
when the additional strategic and monitoring
processes associated with such tasks either are
not implicated or come into play before such
processes take effect (see also discussions in
MacLeod, 2008, and in studies by Gardiner &
Richardson-Klavehn and their colleagues).
On a final note, like the strength models our
proposal does not explain adequately the pattern
of results associated with K responses in Experi-
ment 1a. It would seem that priming associated
with K responses is based on a different system that
is not influenced by the same memory strength that
determines explicit memory performance. In
short, the classic systems dissociation between
explicit and implicit memory, at least as assessed
by priming, may apply only when comparing
priming with familiarity. Such a proposal would
be consistent with Schacter et al.’s (2004, 2007)
distinction between two types of priming*one
that is mediated by the MTL and can be influenced
by processes that also underlie explicit memory
(Ostergaard & Jerningan, 1993), and one that is
mediated by structures outside the MTL that may
be unique to implicit memory (Jacoby, 1991;
Moscovitch et al., 1993; Toth et al., 1994).
Manuscript received 30 June 2009
Manuscript accepted 28 May 2010
First published online 17 August 2010
REFERENCES
Aggleton, J. P., & Brown, M. W. (1999). Episodic
memory, amnesia, and the hippocampal-anterior
thalamic axis. Behavioral and Brain Sciences,22,
425444.
Balota, D. A., Cortese, M. J., Sergent-Marshall,
S. D., Spieler, D. H., & Yap, M. J. (2004). Visual
word recognition of single-syllable words. Journal of
Experimental Psychology: General,133, 283316.
Berry, C. J., Shanks, D. R., & Henson, R. N. A. (2008).
A unitary signal-detection model of implicit and
explicit memory. Trends in Cognitive Sciences,12,
367373.
Boldini, A., Russo, R., & Avons, S. E. (2004). One
process is not enough! A speedaccuracy tradeoff
study of recognition memory. Psychonomic Bulletin,
11, 353361.
Bowers, J., & Schacter, D. L. (1993). Priming of novel
information in amnesic patients: Issues and data. In
P. Graf & M. Masson (Eds.), Implicit memory: New
directions in cognition, development, and neuropsy-
chology (pp. 303326). Hillsdale, NJ: Lawrence
Erlbaum Associates Inc.
Butler, L. T., & Berry, D. C. (2001). Implicit memory:
Intention and awareness revisited. Trends in Cogni-
tive Sciences,5(5), 192197.
Cabeza, R., Ciaramelli, E., Olson, I. R., & Moscovitch,
M. (2008). The parietal cortex and episodic memory:
An attentional account. Nature Reviews Neu-
roscience,9, 613625.
Cermak, L. S. (1993). Automatic versus controlled
processing and the implicit task performance of
amnesic patients. In P. Graf & M. Masson (Eds.),
Implicit memory: New directions in cognition, devel-
opment, and neuropsychology (pp. 287301). Hills-
dale, NJ: Lawrence Erlbaum Associates Inc.
Ciaramelli, E., Grady, C. L., & Moscovitch, M. (2008).
Top-down and bottom-up attention to memory: A
hypothesis (AtoM) on the role of the posterior
parietal cortex in memory retrieval. Neuropsycho-
logia,46, 18281851.
Coltheart, M. (1981). The MRC Psycholinguistic Da-
tabase. Quarterly Journal of Experimental Psychol-
ogy,33A, 497505.
Dewhurst, S. A., & Conway, M. (1994). Pictures,
images, and recollective experience. Journal of
Experimental Psychology: Learning, Memory and
Cognition,20, 10881098.
Dewhurst, S. A., Holmes, S. J., Brandt, K. R., & Dean,
G. M. (2006). Measuring the speed of the conscious
components of recognition memory: Remembering
RECOLLECTIVE PERFORMANCE ADVANTAGES 695
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
is faster than knowing. Consciousness and Cogni-
tion: An International Journal,15, 147162.
Diana, R. A., Reder, L. M., Arndt, J., & Park, H.
(2006). Models of recognition: A review of argu-
ments in favor of a dual-process account. Psycho-
nomic Bulletin & Review,13,121.
Diana, R. A., Yonelinas, A. P., & Ranganath, C. (2007).
Imaging recollection and familiarity in the medial
temporal lobe: A three-component model. Trends in
Cognitive Sciences,11, 379386.
Dunn, J. C. (2004). Rememberknow: A matter of
confidence. Psychological Review,111(2), 524542.
Dunn, J. C. (2008). The dimensionality of the re-
memberknow task: A state-trace analysis. Psycho-
logical Review,115(2), 426446.
Eichenbaum, H., Yonelinas, A. P., & Ranganath, C.
(2007). The medial temporal lobe and recognition
memory. Annual Review of Neuroscience,30, 123
152.
Gabrieli, J. D., Keane, M. M., Stanger, B. Z., Kjelgaard,
M. M., Corkin, S., & Growdon, J. H. (1994).
Dissociations among structuralperceptual, lexical
semantic, and eventfact memory systems in Alz-
heimer, amnesic, and normal subjects. Cortex,30,
75103.
Gardiner, J. M. (1988). Functional aspects of recollec-
tive experience. Memory & Cognition,16, 309313.
Gardiner, J. M., Gregg, V. H., & Karayianni, I. (2006).
Recognition memory and awareness: Occurrence of
perceptual effects in remembering or in knowing
depends on conscious resources at encoding, but not
at retrieval. Memory & Cognition,32, 227239.
Gardiner, J. M., Richardson-Klavehn, A., Ramponi, C.,
& Brooks, B. M. (2001). Involuntary level-of-proces-
sing effects in perceptual and conceptual priming. In
M. Naveh-Benjamin, M. Moscovitch, & H. L.
Roediger III (Eds.), Perspectives on human memory
and cognitive aging: Essays in honor of Fergus Craik.
Philadelphia: Psychology Press.
Goshen-Gottstein, Y., & Moscovitch, M. (1995). Re-
petition priming for newly formed associations:
Perceptual and conceptual influences. Journal of
Experimental Psychology: Learning, Memory, and
Cognition,21, 12211248.
Goshen-Gottstein, Y., Moscovitch, M., & Melo, B.
(2000). Intact implicit memory for newly formed
verbal associations in amnesic patients following
single study trials. Neuropsychology,14(4), 570578.
Graf, P., & Schacter, D. L. (1985). Implicit and explicit
memory for new associations in normal and amnesic
subjects. Journal of Experimental Psychology:
Learning, Memory and Cognition,11, 501518.
Hintzman, D. L., & Hartry, A. L. (1990). Item effects in
recognition and fragment completion: Contingency
relations vary for different sub-sets of words. Journal
of Experimental Psychology: Learning Memory, and
Cognition,16, 955969.
Horton, K. D. (1989). Word stem completion norms.
Unpublished manuscript, Wilfrid Laurier Univer-
sity, Waterloo, Ontario, Canada.
Horton, K. D., Wilson, D. E., & Evans, M. (2001).
Measuring automatic retrieval. Journal of Experi-
mental Psychology: Learning Memory and Cogni-
tion,27, 958966.
Jacoby, L. L. (1991). A process dissociation framework:
Separating automatic from intentional uses of mem-
ory. Journal of Memory and Language,30, 513541.
Jacoby, L.L., & Dallas, M (1981). On the relationship
between autobiographical memory and perceptual
learning. Journal of Experimental Psychology: Gen-
eral,3, 306340.
Kinoshita, S. (2001). The role in involuntary aware
memory in the implicit stem and fragment comple-
tion tasks: A selective review. Psychonomic Bulletin
& Review,8,5869.
Kinoshita, S., & Wayland, S. V. (1993). Effects of
surface features on word-fragment completion in
amnesic subjects. American Journal of Psychology,
106,6780.
Kirchhoff, B. A., Wagner, A. D., Maril, A., & Stern,
C. E. (2000). Prefrontal-temporal circuitry for epi-
sodic encoding and subsequent memory. The Journal
of Neuroscience,20, 61736180.
Kucera, H., & Francis, W. (1967). Computational
analysis of present-day American-English. Provi-
dence, RI: Brown University Press.
MacLeod, C. M. (2008). Implicit memory tests:
Techniques for reducing conscious intrusion. In J.
Dunlosky & R. A. Bjork (Eds.), Handbook of
metamemory and memory (pp. 245263). New
York: Psychology Press.
MacLeod, C. M., & Masson, M. E. J. (2000). Repetition
priming in speeded word reading: Contributions of
perceptual and conceptual processing episodes.
Journal of Memory and Language,42, 208228.
Masson, M. E. J., & MacLeod, C. M. (1997). Episodic
enhancement of processing fluency. In D.L. Medin
(Ed.), The psychology of learning and motiviation,
Vol. 37 (pp. 155210). San Diego, CA: Academic
Press.
Mickes, L., Wais, P. E., & Wixted, J. T. (2009).
Recollection is a continuous process: Implications
for dual-process theories of recognition memory.
Psychological Science,20(4), 509515.
Moscovitch, M. (1992). Memory and working with
memory: A component process model based on
modules and central systems. Journal of Cognitive
Neuroscience,4, 257267.
Moscovitch, M. (2008). The hippocampus as a ‘‘stupid’’
domain-specific module: Implications for theories of
recent and remote memory, and of imagination.
Canadian Journal of Experimental Psychology,62,
6279.
Moscovitch, M., Vriezen, E., & Goshen-Gottstein, Y.
(1993). Implicit tests of memory in patients with
focal lesions or degenerative brain disorders. In H.
Spinnler & F. Boller (Eds.), Handbook of neurop-
sychology, Volume 8 (F. Boller & J. Grafman,
Volume Editors, pp. 133173). Amsterdam: Elsevier.
Musen, G., & Squire, L.R. (1992). Nonverbal priming in
amnesia. Memory & Cognition,20, 441448.
Ostergaard, A. L., & Jernigan, T. L. (1993). Are word
priming and explicit memory mediated by different
brain structures? In P. Graf & M. E. J. Masson
(Eds.), Implicit memory: New directions in cognition,
development and neuropsychology (pp. 327349).
Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
696 SHELDON AND MOSCOVITCH
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
Rastle, K., Harrington, J., & Coltheart, M. (2002).
358,534 nonwords: The ARC Nonword Database.
Quarterly Journal of Experimental Psychology,55A,
13391362.
Richardson-Klavehn, A., & Gardiner, J. M. (1995).
Retrieval volition and memorial awareness in stem
completion: An empirical analysis. Psychological
Research,57, 166178.
Richardson-Klavehn, A., & Gardiner, J. M. (1996).
Cross-modality priming in stem completion reflects
conscious memory, but not voluntary memory.
Psychonomic Bulletin & Review,3, 238244.
Richardson-Klavehn, A., & Gardiner, J. M. (1998).
Depth-of processing effects on priming in stem
completion: Tests of the voluntary contamination,
conceptual processing, and lexical processing hy-
potheses. Journal of Experimental Psychology:
Learning, Memory, Cognition,24, 593609.
Roediger, H. L., & McDermott, K. B. (1993). Implicit
memory in normal human subjects. In H. Spinnler &
F. Boller (Eds.), Handbook of neuropsychology,
Volume 8 (F. Boller & J. Grafman, Volume Editors,
pp. 133173). Amsterdam: Elsevier.
Rotello, C. M., & Macmillan, N. A. (2006). Remember
know models as decision strategies in two experi-
mental paradigms. Journal of Memory and Lan-
guage, 55(4), 479494.
Rotello, C. M., Macmillan, N. A., & Reeder, J. A.
(2004). Sum-difference theory of remembering and
knowing: A two-dimensional signal detection
model. Psychological Review,111, 588616.
Rugg, M. D., Mark, R. E., Walla, P., Schloerscheidt,
A. M., Birch, C. S., & Allen, K. (1998). Dissociation
of the neural correlates of implicit and explicit
memory. Nature,392, 595598.
Schacter, D. L. (1987). Implicit memory: History and
current status. Journal of Experimental Psychology:
Learning, Memory, and Cognition,13, 501518.
Schacter, D. L., Church, B. A., & Bolton, E. (1995).
Implicit memory in amnesic patients: Impairment of
voice-specific priming. Psychological Science,6,
2025.
Schacter, D. L., Dobbins, I. G., & Schnyer, D. M.
(2004). Specificity of priming: A cognitive neu-
roscience perspective. Nature Reviews Neuroscience,
5, 853862.
Schacter, D. L., Wig, G. S., & Stevens, W. D. (2007).
Reductions in cortical activity during priming.
Current Opinion in Neurobiology,17, 171176.
Shimamura, A. P. (1985). Problems with the finding of
stochastic independence as evidence for multiple
memory systems. Bulletin of the Psychonomic So-
ciety,23, 506508.
Shimamura, A. P., Salmon, D. P., Squire, L. R., &
Butters, N. (1987). Memory dysfunction and word
priming in dementia and amnesia. Behavioral Neu-
roscience,101, 347351.
Squire, L. R. (2004). Memory systems of the brain: A
brief history and current perspective. Neurobiology
of Learning and Memory,82, 171177.
Squire, L. R., Shimamura, A. R., & Graf, P. (1987).
Strength and duration of priming effects in normal
subjects and amnesic patients. Neuropsychologia,25,
195210.
Stark, S. M., Gordon, B., & Stark, C. E. (2008). Does
the presence of priming hinder subsequent recogni-
tion or recall performance? Memory,16, 157173.
Toth, J. P. (1996). Conceptual automaticity in recog-
nition memory: Levels-of-processing effects on
familiarity. Canadian Journal of Experimental Psy-
chology,50, 123138.
Toth, J. P., Reingold, E. M., & Jacoby, L. L. (1994).
Toward a redefinition of implicit memory: Process
dissociations following elaborative processing and
self-generation. Journal of Experimental Psychol-
ogy: Learning, Memory and Cognition,20, 290303.
Tulving, E. (1983). Elements of episodic memory.
Oxford, UK: Clarendon Press.
Tulving, E. (1985). Memory and consciousness. Cana-
dian Psychology,26,112.
Turk-Browne, N. B., Yi, D. J., & Chun, M. M. (2006).
Linking implicit and explicit memory: Common
encoding factors and shared representations. Neu-
ron,49, 917927.
Wagner, A. D., Gabrieli, J. D. E., & Verfaellie, M.
(1997). Dissociations between familiarity processes
in explicit-recognition and implicit-perceptual mem-
ory. Journal of Experimental Psychology: Learning,
Memory and Cognition,23(2), 305323.
Wagner, A. D., Maril, A., & Schacter, D. L. (2000).
Interactions between forms of memory: When
priming hinders new episodic learning. Journal of
Cognitive Neuroscience,12,5260.
Westmacott, R., Black, S. E., Freedman, M., & Mos-
covitch, M. (2004). The contribution of autobiogra-
phical significance to semantic memory: Evidence
from Alzheimer’s disease, semantic dementia, and
amnesia. Neuropsychologia,42,2548.
Westmacott, R., & Moscovitch, M. (2002). Temporally
graded semantic memory loss in amnesia and
semantic dementia: Further evidence for opposite
gradients. Cognitive Neuropsychology,19, 135163.
Westmacott, R., & Moscovitch, M. (2003). The con-
tribution of autobiographical significance to seman-
tic memory. Memory & Cognition,31, 761774.
Wixted, J. T. (2007). Dual-process theory and signal-
detection theory of recognition memory. Psycholo-
gical Review,114, 152176.
Wixted, J. T., & Stretch, V. (2004). In defense of the
signal detection interpretation of remember/know
judgments. Psychonomic Bulletin and Review,11(4),
616641.
Yang, J., Weng, X., Guan, L., Kuang, P., Zhang, M.,
Sun, W., et al. (2003). Involvement of the medial
temporal lobe in priming for new associations.
Neuropsychologia,41, 818829.
Yonelinas, A. P. (2002). The nature of recollection and
familiarity: A review of 30 years of research. Journal
of Memory and Language,46(3), 441517.
Yonelinas, A. P., & Jacoby, L. L. (1994). Dissociations
of processing in recognition memory: Effects of
interference and of response speed. Canadian
Journal of Experimental Psychology,48, 516535.
RECOLLECTIVE PERFORMANCE ADVANTAGES 697
Downloaded By: [Canadian Research Knowledge Network] At: 14:18 25 April 2011
... Greater strength tends to translate into faster identification RTs. Similar experimental findings have been obtained when word-stem completion and lexical decision tasks were used to measure priming (Sheldon & Moscovitch, 2010). ...
... However, if the association between priming and recognition is driven by fluency, then this does not explain why an association persisted in Experiment 3A and 3B when the influence of fluency was precluded. That is, the priming effect was still greater for studied items that were recognized than those that were not recognized (see also Ostergaard, 1998;Sheldon & Moscovitch, 2010, for a similar findings). It is also not clear how a fluency attribution account would explain why the priming-source association persisted in Experiments 3A and 3B. ...
... We did not model item effects directly, so cannot rule out that such item effects contributed to the association we observed here. However, Sheldon and Moscovitch (2010) showed that the association they observed between recollection and priming (using remember-know and lexical decision tasks respectively) could not be attributed to item effects. This suggests that the association here between priming and source memory is also unlikely to arise from item effects alone. ...
Article
Full-text available
We present new behavioral data and modeling that links priming, recognition, and source memory. In four experiments, we found that the magnitude of the priming effect, as measured with identification response time in a gradual clarification task, was (1) greater for studied items receiving correct source decisions than incorrect source decisions, and (2) increased as confidence in the source decision increased. Building on the framework for modeling recognition and priming proposed by Berry, Shanks, Speekenbrink, and Henson (2012), we developed a single-system model in which source memory decisions are driven by the same memory strength signal as recognition and priming. We formally compare the model against a multiple-systems model, in which the (implicit) memory signal driving priming is distinct to the (explicit) one driving recognition and source memory. The single-system model reproduces the qualitative patterns of the association between source memory and priming better than the multiple-systems model. Comparison of the quantitative fits was not as clear-cut, however: the single-system model tended to fit better in Experiments 1 and 2, but not in Experiments 3A and 3B, where the observed association between priming and recognition was weaker. Our investigation is an initial attempt at linking priming, recognition, and source memory in the same modelling framework, and provides a basis for further exploration and refinement.
... Table 1 showcases studies from high-impact journals including Nature and Journal of Experimental Psychology: General, attesting to versatility of the miss-CR contrast method. Researchers have used this approach in conjunction with diverse methodologies, from electroencephalogram (EEG) assessments (Addante, 2015;Addante et al., 2023) to behavioral response time measures (Sheldon & Moscovitch, 2010). The studies also employ a wide range of stimuli including words (Woollams et al., 2008), faces (Lehmann et al., 2004), and line drawings (Kark et al., 2016(Kark et al., , 2020. ...
Article
Full-text available
A widely adopted approach in research on unconscious perception and cognition involves contrasting behavioral or neural responses to stimuli that have been presented to participants (e.g., old items in a memory test) against those that have not (e.g., new items), and which participants do not discriminate in their conscious reports. We demonstrate that such contrasts do not license inferences about unconscious processing, for two reasons. One is Kelley’s Paradox, a statistical phenomenon caused by regression to the mean. In the inevitable presence of measurement error, true awareness of the contrasted stimuli is not equal. The second is a consequence, within the framework of Signal Detection Theory, of unequal skewness in the strengths of target and nontarget items. The fallacious reasoning that underlies the employment of this contrast methodology is illustrated through both computational simulations and formal analysis, and its prevalence is documented in a narrative literature review. Additionally, a recognition memory experiment is reported which tests and confirms a prediction of our analysis of the contrast methodology and corroborates the susceptibility of this method to artifacts attributable to Kelley’s Paradox and strength skewness. This work challenges the validity of conclusions drawn from this popular analytic approach.
... One kind that has flown under the radar of Tulvingian theories concerns implicit-i.e., nonconscious-retrieval of episodic information. In a noteworthy study, Sheldon and Moscovitch (2010) employed the remember-know procedure to examine the relation between recollection and performance on two implicit memory tasks: lexical decision and word stem completion. They found that remembered words were associated with greater priming effects than were words given a "known" rating (or studied but non-recognized words). ...
Article
The Galilean explanatory style is characterized by the search for the underlying structure of phenomena, the positing of "deep" explanatory principles, and a view of the relation between theory and data, on which the search for "crucial data" is of primary importance. In this paper, I trace the dynamics of adopting the Galilean style, focusing on the science of episodic memory. I argue that memory systems, such as episodic and semantic memory, were posited as underlying competences producing the observable phenomena of memory. Considered in idealized isolation from other systems, episodic memory was taken to underlay the ability of individuals to remember events from their personal past. Yet, in reality, memory systems regularly interact, standing in many-to-many relations to actual memory tasks and experiences. Upon this backdrop, I explore a puzzle about the increasing prominence of the notion of autonoetic consciousness in Tulving's theory of episodic memory. I argue that, contrary to widespread belief, the prominence is not best explained by the purported essential link between autonoetic consciousness and episodic memory. Rather, it is explained by the fact that autonoetic consciousness, hypothesized to uniquely accompany episodic retrieval, was considered a source of crucial data, predictable only from theories positing a functionally distinct episodic memory system. However, with the emergence of a new generation of theories, positing wider memory systems for remembering and imagination, the question of the relation between episodic memory and autonoetic consciousness has been reopened. This creates a pressing need for de-idealization, triggering a new search for crucial data.
... First, there are a number of reports of "selfless" memories, not accompanied by autonoetic experiences, in both clinical and extra-clinical contexts (Klein and Nichols, 2012;Gentry, 2021;Millière and Newen, 2022). Second, episodic information has been shown to be "implicitly" retrievable for a variety of cognitive tasks (Sheldon and Moscovitch, 2010;Wimmer and Shohamy, 2012). Third, there is neuropsychological evidence for selective impairments of episodic simulation and self-related processing (Arzy et al., 2009;Andelman et al., 2010). ...
Article
Full-text available
This paper offers a modeling account of episodic representation. I argue that the episodic system constructs mental models: representations that preserve the spatiotemporal structure of represented domains. In prototypical cases, these domains are events: occurrences taken by subjects to have characteristic structures, dynamics and relatively determinate beginnings and ends. Due to their simplicity and manipulability, mental event models can be used in a variety of cognitive contexts: in remembering the personal past, but also in future-oriented and counterfactual imagination. As structural representations, they allow surrogative reasoning, supporting inferences about their constituents which can be used in reasoning about the represented events.
... However, there is behavioural evidence (Sheldon & Moscovitch, 2010), and neural evidence (Waldhauser, et al, 2016), that context-specific memory reactivation, termed "ecphory" (Tulving, 1983;Waldhauser, et al), is a rapid unconscious response to incoming sensory information whose output is not consciously apprehended but can contribute to performance on a variety of tasks (Moscovitch, 2008). Subsequently, slower processes make the recovered content available to consciousness as episodic memory. ...
Article
Full-text available
The term "mind" is used here to refer to all of the mechanisms in the brain that generate one’s responses including thinking and experiencing. These mechanisms may be innate, or they may be learned but based on innate abilities. Most of these mechanisms respond in a rapid but flexible way to situations that one confronts frequently; some respond to occasional threats or opportunities; some are temporary mechanisms for responding to expected situations, or for controlling intended behavioural sequences; and some are capable of more generalised problem-solving. All of these mechanisms are unconscious; there is no evidence of conscious processing, and much evidence—based on psychological research, evolutionary principles, and theoretical considerations—confirming that no mental processing occurs in consciousness. Consciousness is a changing array of information in various forms; such as sights, sounds, and felt experiences, but consciousness is adaptive, and this can only be because conscious information enhances its possessor’s responses in some situations. The mechanisms of mind generally operate with unconscious information, but they sometimes benefit from access to conscious information, and some mechanisms may have evolved to function solely with conscious information or use only conscious information in some situations.
Article
According to traditional automation theory, an automatic process should be "purely unconscious", independent of limited attention resources, and executed at the same time as other processing tasks without interference. Implicit memory is considered to be a tool that provides unconscious and automatic cognitive processes and that is not be affected by any type of attention resource. Memory includes two important links: coding and retrieval. Coding is mainly responsible for the preliminary processing of information and the generation of memory traces. Retrieval promotes or suppresses the connection of these memory representations after coding. Previous studies have mostly discussed the effect of interference on implicit memory from the point of view of coding and considered that implicit memory tests are immune to coding interference. However, there are doubts about whether the interference in the retrieval stage will affect the implicit memory, and more importantly, it is not known whether this difference is caused by the different types of memory tests. Because of the intersection between the types of implicit memory tests, this study involves four experiments. The effects of retrieval interference on an identification-perceptual implicit test, an identification-conceptual implicit test, a production-perception implicit test and a production-concept implicit test were investigated. In this study, two new production tasks were designed. The learning-test paradigm was used to explore the relationship between retrieval interference and different types of implicit memory by setting up digital interference tasks at the same time in the retrieval stage of the test. The results showed that (1) under the condition of no interference, both the lexical judgement task (identification-perceptual test) and semantic classification task (identification-conceptual) display a significant priming effect, and under the condition of interference, the priming effect of the two types of tasks disappears. (2) Under the condition of no interference, both the production lexical judgement task (production-perceptual test) and the production semantic classification task (production-conceptual test) show obvious priming effects; however, under interference conditions, the two kinds of production judgements still have obvious priming effects. Because of the priming effect for both interference conditions, we performed repeated 2 (with or without interference) × 2 (processing level) analysis of variance tests for the priming amounts (reaction time and ACs) of the two production experiments to determine whether interference would influence the priming effect. The results showed that the priming amounts of the two experiments under interference conditions were significantly lower than those under noninterference conditions. Therefore, compared with those for noninterference conditions, the priming effects of the two implicit identification tests disappeared under retrieval interference. Although the priming effects of the two implicit production memory tasks significantly decreased, there was still a significant priming effect. In conclusion, the retrieval processing of different types of implicit memory tests is affected by interference, and unconscious memory retrieval processing is not completely automated processing but is also regulated by attention resources. Identified implicit memory is more easily affected by retrieval interference than productive implicit memory. Under retrieval interference, there is a separation of the identified implicit memory and production implicit memory, and identified priming is more easily affected by retrieval interference than is productive priming.
Article
Memory systems consolidation is often conceived as the linear, time-dependent, neurobiological shift of memory from hippocampal-cortical to cortico-cortical dependency. We argue that contrary to this unidirectional view of memory reorganization, information about events may be retained in multiple forms (e.g., event-specific sensory-near episodic memory, event-specific gist information, event-general schematic information, or abstract semantic memory). These representations can all form at the time of the event and may continue to coexist for long durations. Their relative strength, composition, and dominance of expression change with time and experience, with task demands, and through their dynamic interaction with one another. These different psychological mnemonic representations depend on distinct functional and structural neurobiological substrates such that there is a neural-psychological representation correspondence (NPRC) among them. We discuss how the dynamics of psychological memory representations are reflected in multiple levels of neurobiological markers and their interactions. By this view, there are only variations of synaptic consolidation and memory dynamics without assuming a distinct systems consolidation process.
Article
Conventional wisdom indicated that implicit memory was a kind of automatic processing. However, recent studies found that implicit memory was more vulnerable to be distracted than explicit memory in some conditions. Therefore, it is not clear on the processing mechanisms of implicit memory. In order to resolve this confusion, we can use variables that have different effects on the automated processing and controlled processing. Previous studies have found that cognitive fatigue would have a strong impact on the controlled processing, while its impact on automated processing was less. However, few studies have explored the effect of cognitive fatigue on implicit memory. Therefore, are effects of cognitive fatigue different on implicit memory and explicit memory? In this study, effects of cognitive fatigue on memory were assessed by drawing a comparison between the participants with cognitive fatigue condition and those without. We used lexical decision task and lexical recognition task to test implicit memory and explicit memory respectively. Thirty participants were recruited in the experiment. All of them participated in experiments on two occasions, two days apart at least, once in fatigue condition and once in resting condition. They carried out pre-test before formal test to ensure the validity of the experimental results. The pre-test setting was the same as the formal test. In fatigue condition, participants carried out pre-test first. After that, they carried out the fatigue task containing 320 calculation questions. Then, they carried out formal test. In the resting condition, participants rested between the pre-test and the formal test. Participants were instructed to make response to appropriate items by pressing the keyboard. The Reaction Time and Accuracy data in retrieval phase were recorded in order to assess implicit memory and explicit memory. Results showed that implicit memory performance declined under fatigue condition, but the explicit memory performance was not. In the explicit memory performance, the results of present study were consistent with previous results. Cognitive fatigue did not affect explicit memory [F(1, 29)=.119, p > .05]. However, implicit memory was affected by cognitive fatigue [F(1, 23)=9.816, p < .01, ηp2=.253]. The performance of implicit memory significantly decreased after the fatigue task. It reflected that implicit memory was more vulnerable to cognitive fatigue than explicit memory. In conclusion, results from the present study revealed that effects of cognitive fatigue on implicit memory and explicit memory were different. Implicit memory was more sensitive to cognitive fatigue than explicit memory. The results of present study were clearly inconsistent with the traditional views. Conventional wisdom thought that implicit memory was a kind of automatic processing, which could be immune to effects of cognitive fatigue. However, in the present study, we found inverse results in which implicit memory retrieval was easier to be impacted than explicit memory by cognitive fatigue. These results revealed that although implicit memory was an automated processing, it might own different features from other automated processing. This study firstly used cognitive fatigue to explore the difference between implicit memory and explicit memory. We found implicit memory is more vulnerable to cognitive fatigue than explicit memory. The present study provides support for further research on the processing mechanism of implicit memory.
Article
The distinction between episodic and semantic memory was first proposed in 1972 by Endel Tulving and is still of central importance in cognitive neuroscience. However, data obtained over the past 30 years or so support the idea that the frontiers between perception and knowledge and between episodic and semantic memory are not as clear cut as previously thought, prompting a rethink of the episodic-semantic distinction. Here, we review recent research on episodic and semantic memory, highlighting similarities between the two systems. Taken together, current behavioral, neuropsychological, and neuroimaging data are compatible with the idea that episodic and semantic memory are inextricably intertwined, yet retain a measure of distinctiveness, despite the fact that their neural correlates demonstrate considerable overlap.
Article
Full-text available
Abstract A neuropsychological model of memory is proposed that incorporates Fodor's (1983) idea of modules and central systems. The model has four essential components: (1) a non-frontal neocortical component that consists of perceptual (and perhaps interpretative semantic) modules that mediate performance on item-specific, implicit tests of memory, (2) a modular medial temporal/hippocampal component that mediates encoding, storage, and retrieval on explicit, episodic tests of memory that are associative/cue dependent, (3) a central system, frontal-lobe component that mediates performance on explicit tests that are strategic and on procedural tests that are rule-bound, and (4) a basal ganglia component that mediates performance on sensorimotor, procedural tests of memory. The usefulness of the modular/central system construct is explored and evidence from studies of normal, amnesic, agnosic, and demented people is provided to support the model.
Article
This study was addressed to the question of whether the medial temporal lobe (MTL) plays a critical role in implicit memory for new associations. Priming for new associations was examined in two different tasks in 18 patients with focal lesions all involving the MTL. In Experiment 1, following a study phase for pairs of unrelated words, subjects performed a perceptual identification task on old, recombined, and new pairs of words presented at brief exposure durations. In contrast to control subjects, and despite a normal level of item priming, the patients failed to show superior identification of the old pairs relative to the recombined pairs, the measure of associative priming. In Experiment 2, subjects engaged in speeded naming of the print color for previously studied words presented in the original color or in a different old color, and for unstudied words. Again, in contrast to control subjects and despite a normal level of item facilitation on color naming reaction time (RT), the patients failed to show priming for recently experienced new associations between words and colors. Explicit recognition memory by the patients was abnormal in both experiments. This study records an absence of priming for new associations, in two different tasks in which the nature of the stimuli was considerably different, in a large group of patients with lesions in the MTL. Although some previous research has reported significant associative priming in other tasks for patients with MTL lesions, the present results suggest that this region is critical for forming new associations of the types assessed here.
Article
A comparison of incidental and intentional stem-completion tests confirmed that cross-modality priming occurs when performance conforms completely to the retrieval intentionality criterion, indicating involuntary-not voluntary-retrieval in the incidental test. However, an on-line measure of awareness in the incidental test, and a process-dissociation analysis of the intentional test, indicated only within-modality, but not cross-modality, transfer of involuntary retrieval that is unaccompanied by memorial awareness. These results imply that conscious memory should not be equated with voluntary retrieval, and unconscious memory should not be equated with involuntary retrieval, because involuntary retrieval can be accompanied by memorial awareness.
Article
This paper describes a computerised database of psycholinguistic information. Semantic, syntactic, phonological and orthographic information about some or all of the 98,538 words in the database is accessible, by using a specially-written and very simple programming language. Word-association data are also included in the database. Some examples are given of the use of the database for selection of stimuli to be used in psycholinguistic experimentation or linguistic research. © 1981, The Experimental Psychology Society. All rights reserved.
Article
Five experiments investigated repetition priming on an indirect speeded word reading (naming) test, a task intended to circumvent conscious recollection. Reading a word or generating it from a semantic cue (either a phrase or an antonym) produced reliable priming of similar magnitude on this indirect test of memory. Efforts to encourage conscious recollection elevated response latencies in speeded reading and improved performance on a direct test of recognition memory, without creating a difference in the amount of priming observed in the Read and Generate conditions. We also found more priming for visually than for auditorily studied words, consistent with the standard pattern for indirect tests assumed to be data-driven. Speeded word reading provides a good measure of repetition priming because the fully exposed target word recruits both perceptual and conceptual aspects of the initial interpretive encoding episode.