Journal of Experimental Psychology: Copyright 1998 by the American Psychological Association, Inc.

Learning, Memory, and Cognition 0278-7393/98/S3.O0

1998, Vol. 24, No. 1,27-47

The Role of Attention During Encoding in Implicit

and Explicit Memory
Neil W. Mulligan
Illinois State University

In 5 experiments, participants read study words under conditions of divided or full attention.
Dividing attention reduced performance on the general knowledge test, a conceptual implicit
test of memory. Likewise, dividing attention reduced conceptual priming on the word-
association task, as well as on a matched explicit test, associate-cued recall. In contrast, even
very strong division of attention did not reduce perceptual priming on word-fragment
completion, although it did reduce recall on the matched explicit test of word-fragment-cued
recall. Finally, dividing attention reduced recall on the perceptual explicit tests of graphemic-
cued recall and graphemic recognition. The results indicate that perceptual implicit tests rely
minimally on attention-demanding encoding proceses relative to other types of memory tests.
The obtained pattern of dissociations is not readily accommodated by the transfer-appropriate-
processing (TAP) account of implicit and explicit memory. Potential extensions of the TAP
view are discussed.

Attention during encoding plays a critical role in later
memory performance. Dividing attention during encoding
reduces performance on many measures of memory, includ-
ing free and cued recall and recognition (e.g., Baddeley,
Lewis, Eldridge, & Thomson, 1984; Broadbent, 1958;
Cherry, 1953; Craik, Govoni, Naveh-Benjamin, & Ander-
son, 1996; Fisk & Schneider, 1984; Moray, 1959; Murdock,
1965; Norman, 1969b). The prevalence of these findings
encouraged earlier researchers to posit that attention is
required for the formation of durable memory traces (e.g.,
Broadbent, 1958; Cherry, 1953; MacKay, 1973; Moray,
1959; Norman, 1969a). A similar, central role for attention in
encoding has also been espoused in more recent work (e.g.,
Bentin, 1994; Cowan, 1988, 1995; Craik, 1989; Craik et al.,
1996; Fisk & Schneider, 1984).
The memory experiments that gave rise to this traditional
view generally used explicit (or direct) tests of memory. In
such tests, participants are directed to think back and try to
recall information about some specific (usually experimenter-
provided) event. Memory may be assessed with implicit (or
indirect) tests of memory, as well (Roediger & McDermott,
1993; Schacter, 1987). In these tests, memory for prior
events is inferred from the enhanced reprocessing of re-
cently experienced stimuli, a phenomenon known as repeti-
Neil W. Mulligan, Department of Psychology, Illinois State
University.
This research was supported in part by a university research
grant from Illinois State University. The assistance of Robert
Hitlan, Clayton Lovitt, Rick Sarles, and Molly McGuire is
gratefully acknowledged. Thanks to Theresa Blaxton for supplying
experimental materials and to Maria Stone, Marilyn Hartman,
Kavitha Srinivas, Steve Lindsay, and Leah Light for providing very
useful comments on a draft of this article.
Correspondence concerning this article should be addressed to
Neil W. Mulligan, who is now at the Department of Psychology,
Southern Methodist University, Dallas, Texas 75275. Electronic
mail may be sent via Internet to mulligan@mail.smu.edu.
tion priming. Research on implicit memory has garnered
great notice because repetition priming often differs from
performance on explicit tests in important ways. First, there
have been numerous reports of population dissociations
between implicit and explicit test performance, in which
participants from populations impaired on explicit memory
tests (e.g., people with amnesia, depression, or schizophre-
nia and older adults) show normal or near-normal levels of
repetition priming on implicit memory tests (e.g., Denny &
Hunt, 1992; Elliott & Greene, 1992; Light, 1991; Schwartz,
Rosse, & Deutsch, 1993; Shimamura, 1986, 1993). Second,
a number of experimental manipulations, such as levels of
processing, the read-generate manipulation, and study mo-
dality, also have produced dissociations between repetition
priming and performance on explicit tests (see Roediger &
McDermott, 1993, for a review).
Given the centrality of attention in theories of memory
encoding, it is of great interest that dividing attention during
encoding may also dissociate performance on implicit and
explicit tests. Several studies have reported that dividing
attention during encoding can reduce explicit test perfor-
mance while leaving repetition priming unaffected (Bentin,
Kutas, & Hillyard, 1995; Gabrieli et al., 1996; Isingrini,
Vazou, & Leroy, 1995; Jacoby, Woloshyn, & Kelley, 1989;
Kellogg, Newcombe, Kammer, & Schmitt, 1996; Mulligan
& Hartman, 1996; Parkin, Reid, & Russo, 1990; Parkin &
Russo, 1990; Russo & Parkin, 1993; M. E. Smith &
Oscar-Berman, 1990; Szymanski & MacLeod, 1996). Other
studies have demonstrated that retention of very poorly
attended stimuli is more likely to be detected with implicit
than explicit memory tests (Bornstein, Leone, & Galley,
1987; Eich, 1984; Hawley & Johnston, 1991; Jelicic, Bonke,
Wolters, & Phaf, 1992; MacKay, 1973; Mandler, Nakamura,
& Van Zandt, 1987; Merikle & Reingold, 1991; Seamon,
Marsh, & Brody, 1984). These results are so compelling that
some researchers have suggested that implicit test perfor-
mance largely reflects automatic encoding processes (Ben-

27
28 MULLIGAN
tin, 1994; Bentin et al., 1995; Besson, Fischler, Boaz, &
Raney, 1992; Graf & Mandler, 1984; Isingrini et al., 1995;
Jacoby, Toth, & Yonelinas, 1993; Jacoby et al., 1989; Jelicic
et al., 1992; Parkin et al., 1990; Parkin & Russo, 1990;
Shallice et al., 1994; Szymanski & MacLeod, 1996; see
Koriat & Feuerstein, 1976, for a related view). However,
recent research on attention and implicit memory (Gabrieli,
Stone, Shackleton, et al, 1996; Light & Prull, 1995b;
Mulligan, 1997; Mulligan & Hartman, 1996; Wood, Stadler,
& Cowan, in press) suggests that this view is insufficient. To
provide a framework for describing and summarizing the
relevant research, the transfer-appropriate-processing (TAP)
account of implicit and explicit memory (e.g., Blaxton,
1989,1992; Craik, Moscovitch, & McDowd, 1994; McDer-
mott & Roediger, 1996; Roediger, 1990; Roediger, Weldon,
& Challis, 1989; Weldon, Roediger, Beitel, & Johnston,
1995) is useful. I first describe this framework and then
review the relevant studies on attention and implicit memory.
The TAP framework is predicated on the notion that
performance on a memory test is enhanced to the extent that
encoding processes are reengaged at the time of retrieval
(Blaxton, 1989; Roediger, 1990; Roediger et al., 1989), an
assumption consistent with the encoding-specificity prin-
ciple (Tblving & Tomson, 1973). Consequently, the TAP
framework proposes that memory tests should be character-
ized primarily by the types of mental processes that they
engage rather than by test instructions (i.e., implicit vs.
explicit). Toward this end, the TAP framework differentiates
between two broad classes of cognitive processes: percep-
tual processes, defined as the analysis of perceptual or
surface-level features (but may also include other representa-
tions required for stimulus identification; e.g., Weldon,
1991), and conceptual processes, defined as the analysis of
meaning or semantic information.
The most widely used implicit memory tests involve the
identification or completion of degraded or ambiguous
perceptual cues, such as word fragments or stems, briefly
presented words, or fragmented pictures. The TAP frame-
work posits that these tests are primarily dependent on
perceptual processes during retrieval and, as a consequence,
are sensitive to variation in perceptual processes at encoding
but relatively insensitive to variation in conceptual pro-
cesses. In contrast, commonly used explicit tests (e.g., free
recall and recognition) are assumed to primarily use concep-
tual processses during retrieval (Craik et al., 1994; Roediger
et al., 1989). The TAP framework accounts for dissociations
between implicit and explicit tests in terms of the differences
between underlying processes rather than in terms of task
instructions. Specifically, because the most common implicit
and explicit memory tests rely differentially on conceptual
and perceptual processes, performance on these two types of
tests benefits from different types of encoding procedures
and is thus dissociable.
However, note that not all implicit tests are perceptual and
not all explicit tests are conceptual. With the appropriate
combinations of memory cues and task instructions, concep-
tual implicit tests and perceptual explicit tests may be
constructed. The TAP framework proposes that dissociations
occur between conceptual and perceptual tests, regardless of
their implicit-explicit status. This parsimonious approach
has had a good deal of success in accounting for and
predicting dissociations among memory tests (see Roediger
& McDermott, 1993, for a review).
Finally, a strength of this theory is that theorists need not
rely on intuition to categorize a test as perceptual or
conceptual. Rather, the TAP framework proposes a set of
criterial variables for operationally defining memory tests as
primarily conceptual or perceptual. A central criterion is
how test performance is affected by the read-generate study
manipulation. In this manipulation, some study words are
presented intact to be read, whereas others study words are
not physically presented but must be produced by the
participant from a conceptual cue (e.g., an antonym). In the
read condition, the perceptual characteristics of the study
word are processed to a greater degree than in the generate
condition, whereas in the generate condition, the conceptual
characteristics of the word are processed to a greater degree
(e.g., Jacoby, 1983b). In the TAP framework, memory tests
that exhibit the typical generation effect, in which the
generate condition leads to superior performance to that of
the read condition (e.g., Slamecka & Graf, 1978), are
assumed to reengage primarily conceptual processes and are
classified as conceptual memory tests. Memory tests exhibit-
ing the reverse generation effect, in which the read condition
leads to greater performance than the generate condition, are
assumed to reengage primarily perceptual processes and are
classified as perceptual memory tests (e.g., Blaxton, 1989;
Roediger, 1990; Roediger & McDermott, 1993; Roediger et
al., 1989). Other primary criterial variables are the study-
modality manipulation and the levels-of-processing manipu-
lation (e.g., Blaxton, 1989; Roediger, 1990; Roediger &
McDermott, 1993).1
With regard to the effects of divided attention, the TAP
distinction between perceptual and conceptual processes has
two implications. If, as is traditionally assumed, dividing
attention at study reduces the amount of semantic or
conceptual processing (e.g., Broadbent, 1971; Craik, 1983;
Craik &Byrd, 1982; Craik etal., 1996; Eagle &Ortof, 1967;
Koriat & Feuerstein, 1976; Norman, 1969a) but has little or
no effect on perceptual identification processes, then the
TAP framework predicts (a) that divided attention should
have little or no effect on perceptual tests and (b) that
divided attention should reduce performance on conceptual
tests.
With regard to perceptual implicit tasks, the prediction of
the TAP framework has been largely borne out. Provided the
manipulation of attention does not disrupt stimulus identifi-
cation, dividing attention leaves priming unaffected on the
perceptual tests of word-fragment completion (Mulligan &
1
Although perceptual implicit tests are typically unaffected by
the levels-of-processing manipulation (Roediger & McDermott,
1993), at least some of these tests are sensitive to this manipulation
under some conditions (Challis & Brodbeck, 1992; Thapar &
Greene, 1994; see Brown & Mitchell, 1994, for a review). The
exact significance of these results for the TAP framework is a
matter of current debate (e.g., Challis & Brodbeck, 1992; Thapar &
Greene, 1994).
 ATTENTION AND MEMORY
Hartman, 1996; Parkin et al., 1990), picture-fragment comple-
tion (Parkin & Russo, 1990; Russo & Parkin, 1993), lexical
decision (Kellogg et al., 1996; M. E. Smith & Oscar-
Berman, 1990; Szymanski & MacLeod, 1996), and percep-
tual identification (Gabrieli, Stone, Shackleton, et al., 1996;
see Mulligan & Hartman, 1996, for a review). These results
imply that perceptual priming has minimal reliance on
attentional resources (beyond those necessary for stimulus
identification; see Mulligan & Hartman, 1996, and Stone,
Ladd, Vaidya, & Gabrieli, 1996, for discussion). The only
exception is priming in word-stem completion, which ap-
pears to be reduced by divided attention (Gabrieh, Stone,
Shackleton, et al., 1996). However, the nature of this implicit
test is in dispute; some researchers argue that this test is
primarily perceptual, whereas others argue that it is primar-
ily conceptual (cf. Gabrieli et al., 1994; Keane, Gabrieli,
Fennema, Growdon, & Corkin, 1991; Roediger, Weldon,
Stadler, & Riegler, 1992).
Most of the studies on implicit memory and divided
attention have used perceptual priming tasks. The few
studies using conceptual implicit tests have produced mixed
results. Consistent with the prediction of the TAP frame-
work, Mulligan and Hartman (1996) recently reported that
dividing attention during encoding reduced later conceptual
priming on the category-exemplar production task, as well
as reducing recall on a comparable explicit test, category-
cued recall (see Gabrieli, Stone, Shackleton, et al., 1996;
Light & Prull, 1995b, for similar findings). However,
Isingrini et al. (1995), using a similar experimental design
and the same memory tests, found no effect of divided
attention on category-exemplar production, although cat-
egory-cued recall was reduced. Because the division-of-
attention task in Isingrini et al. appeared to be a weaker
manipulation of attention than that used by Mulligan and
Hartman and Gabrieli, Stone, Shackleton, et al. (1996) and
because Isingrini et al. (1995) reported a nonsignificant
reduction in priming for all of the divided-attention groups
(Isingrini et al. compared full vs. divided attention in each of
four age groups), Mulligan (1997) hypothesized that the
strength of the attention manipulation was important in
accounting for the divergent results. Using a memory-load
paradigm to divide attention over several levels, Mulligan
found that category-cued recall was reduced in all the
divided-attention conditions, whereas in the category-
exemplar production task, priming was reduced (in fact,
eliminated) only for the strongest division of attention but
not for milder divisions. This appears to resolve the discrep-
ancy between Mulligan and Hartman's results and those of
Isingrini et al. Conceptual priming in the category-exemplar
production task is affected by divided attention, but the
division of attention in Isingrini et al.'s study was too mild to
reveal this. In summary, conceptual priming in category-
exemplar production appears to be dependent on attention.
Experiments 1 and 2 tested the generality of the TAP
prediction that conceptual priming should be diminished by
divided attention by examining the effects of this manipula-
tion on the conceptual implicit tests of general knowledge
questions (Experiment 1) and word association (Experiment
2). The implications of the TAP framework derived earlier
29
focus this inquiry on the role of attention in postperceptual
aspects of memory encoding. Consequently, the attentional
manipulations used in these and subsequent experiments do
not prevent the identification of the study stimuli. Partici-
pants in divided-attention conditions were asked to identify
study words as well as carry out a secondary task. Further-
more, given the findings of Mulligan (1997), strong division
of attention seems important if one is to maximize the
probability of observing divided-attention effects on concep-
tual priming. Consequently, procedures were used that were
similar to those used by Mulligan and Hartman (1996,
Experiment 1) and Gabrieli, Stone, Shackleton, et al. (1996)
in uncovering effects of divided attention on conceptual
priming in the category-exemplar production test. Two
groups of participants were visually presented with a set of
words under either full- or divided-attention conditions. The
divided-attention task was a digit-monitoring task in which
participants were asked to monitor an aurally presented
series of digits for sequences of three odd numbers. This task
provides a rigorous division of attention, eliminating prim-
ing in the category-exemplar production task (Mulligan &
Hartman, 1996). The results of Mulligan (1997), which used
various memory loads to divide attention, are consistent
with this characterization. The digit-monitoring task in
Mulligan and Hartman reduced conceptual priming to the
same (approximately zero) level as the strongest divided-
attention condition in Mulligan (1997).
In Experiment 1, the memory test consisted of a set of
questions, presented as a general knowledge test. The
answers to some of the questions were presented in the study
list. Within the TAP framework, the general knowledge test
is considered conceptual because it is affected by conceptual
encoding manipulations, such as the generation manipula-
tion (Blaxton, 1989,1992) and levels of processing (Hamann,
1990), and is unaffected by perceptual encoding manipula-
tions, such as varying study modality (Challis et al., 1993).
The TAP framework implies that this implicit test should be
affected by divided attention during encoding.
Experiment 1
Method
Participants. Sixty-four undergraduates at Illinois State Univer-
sity participated in exchange for extra credit in psychology classes.
Design and materials. The experiment used a 2 X 2 design in
which study status (old vs. new) was manipulated within subjects
and attention at encoding (full vs. divided) was manipulated
between subjects.
The study and test materials were extracted from the set
developed by Blaxton (1989). One hundred general knowledge
questions and their single-word answers (e.g., "What is the fastest
animal on earth?" Cheetah) were chosen from this source. The 100
answers were randomly divided into two sets of 50. These critical
items were preceded by a primacy buffer and followed by a recency
buffer of 4 unrelated words each, yielding two study lists of 58
words. The general knowledge test consisted of the 100 questions
corresponding to the critical items (50 corresponding to studied, or
old, items and 50 to new items), plus 8 filler questions, also drawn
from Blaxton's materials. The resulting 108 questions were ran-
domly ordered.
30 MULLIGAN
The task used in the divided-attention condition was a digit-
monitoring task (as used in Mulligan & Hart man, 1996; see Craik,
1982; Jacoby, 1991). Participants monitored a tape-recorded series
of random digits, attempting to detect target sequences of three odd
digits in a row. The digits were played at the rate of 1 digit per 1.5 s.
There were 32 target sequences randomly distributed through the
series of 180 digits, subject to two constraints: (a) A minimum of 1
and a maximum of 5 numbers occurred between the end of one
target sequence and the beginning of the next, and (b) not more than
2 even digits occurred in sequence. The list of 180 digits was
recorded six times from beginning to end without pause between
repetitions.
Procedure. Participants were tested individually. The experi-
ment began with the study phase, in which each study word was
displayed in the center of a computer screen for 3 s. Participants
were instructed to read each word out loud and to try to remember
the words for a later, unspecified memory test. This was the sole
task for participants in the full-attention condition. Participants in
the divided-attention condition simultaneously performed the digit-
monitoring task. These participants were told to monitor the digits
for strings of three odd numbers in a row and to tap a pen on the
desk whenever they detected a target string. If a participant missed
a target sequence, the experimenter prompted the participant by
saying "Miss." The experimenter monitored and recorded errors in
reading the words and detecting target sequences of digits. Task
instructions emphasized that the digit-monitoring task and reading
and remembering the words were both equally important. Partici-
pants in the divided-attention condition were allowed approxi-
mately 30 s of practice on the digit-monitoring task before the start
of the study list.
A 3-min distractor task followed the study phase. The task was
used to disguise the relationship between the study session and the
subsequent memory test by making the memory test appear to be
one in a series of unrelated tasks. The distractor task was a city
names completion task. Participants were presented with a set of
index cards, on each of which was printed the first three letters of a
U.S. city (e.g., Bos— for Boston). The participants were asked to
complete each city word stem with the name of a U.S. city.
The general knowledge test followed the distractor task. Partici-
pants were presented with a test booklet consisting of the general
knowledge questions. Participants were instructed to try but were
not required to answer each question. Answers were to be written in
the space provided next to each question. Participants were given
15 s per question. Time was kept with an audiocassette tape that
emitted a beep every 15 s. No mention was made of the relationship
between this task and the studied words.
Results and Discussion
During the study task, participants in the divided-attention
condition correctly identified 97% of the study words and
detected 88% of the target digit strings. Thus, participants in
the divided-attention condition were able to identify the vast
majority of the study words while simultaneously identify-
ing almost all of the target sequences. However, the identifi-
cation of study words was not quite perfect in the divided-
attention condition. Consequently, the test data in this and all
subsequent experiments were analyzed in two ways, condi-
tionalized on correct identification at study and uncondition-
alized. In all cases, the two analyses led to identical conclus-
ions. Only the unconditionalized results are reported below.
The results of the general knowledge test are summarized
in Table 1. Mean proportions of correct answers are pre-
Table 1
Experiment 1: Proportions of Correct Responses in the
General Knowledge Test as a Function of Study Status and
Attention at Encoding
Study status
Attention
condition Old New Old - new
Full attention .30 .15 .15
Divided attention .25 .17 .08
sented as a function of attention at encoding and study
status. The results are straightforward: Dividing attention
reduced the amount of priming in the general knowledge
test. The proportions correct were submitted to a 2 X 2
analysis of variance (ANOVA), using study status (old vs.
new) as a within-subjects variable and attention at encoding
(full vs. divided) as a between-subjects variable. The
significance level for this and subsequent analyses was .05.
The analysis produced two significant effects. First, the main
effect of study status was significant, F(l, 62) = 60.25,
MSE = 0.0068, indicating significant overall priming.
Second, and more important for present purposes, the Study
Status X Attention interaction was significant, F(l, 62) =
5.17, MSE = 0.0068, indicating greater priming in the full-
than divided-attention condition. Follow-up t tests indicated
a significant effect of study status (i.e., priming) in both the
full- and divided-attention conditions, f(31) — 6.89 and
f(31) = 4.01, respectively.
The results of the present experiment were consistent with
Gabrieli, Stone, Shackleton, et al. (1996), Mulligan and
Hartman (1996), and Mulligan (1997), indicating that a
strong division of attention reduces conceptual priming in
both the category-exemplar production task and in the
general knowledge test. However, one potential problem
with using the general knowledge test as a measure of
conceptual priming is that it may be more susceptible to
explicit contamination than other conceptual implicit tests
(Vaidya, Gabrieli, Demb, Keane, & Wetzel, in press; Vaidya
et al., 1996). For example, amnesics are frequently impaired
on this test, which may be an indication of explicit contami-
nation (Blaxton, 1992, Experiment 1; Vaidya et al., in press).
In Experiment 2, the effects of divided attention were
assessed on the conceptual implicit memory test of word
association. This task is arguably a better measure of
implicit memory and one in which amnesics perform
comparably to nonamnesics (Graf, Shimamura, & Squire,
1985; Schacter, 1985). In addition, Experiment 2 used a
posttest questionnaire to determine whether participants
were aware of the relationship between the study episode
and the implicit test and the extent to which they had
deliberately tried to retrieve study items (Bowers & Schac-
ter, 1990; Light & Albertson, 1989; Mulligan & Hartman,
1996). As is shown, the results from the posttest question-
naire indicated that the word-association task was minimally
contaminated by explicit influences.
The word-association task has been used in prior research
to assess the effects of divided attention. Anooshian (1989)
concluded that dividing attention reduced conceptual prim-
ing in a variant of the traditional word-association task. In
 ATTENTION AND MEMORY
this study, participants were aurally presented with homo-
phones embedded in phrases biasing their less common
meaning. The phrases were presented either in focally
attended questions or on the unattended channel in a dichotic
listening task. Participants were later asked to free associate
to a set of old and new aurally presented homophones. The
interpretations of the homophones were inferred from the
free associations produced. Anooshian found that old homo-
phones were more frequently interpreted in line with their
less common meaning than were new homophones and that
this difference was larger for the focally attended than
unattended condition (i.e., greater conceptual priming in a
full-attention condition). However, in this study the implicit
word-association test followed an explicit test of homo-
phone recognition, which may have jeopardized the implicit
status of the word-association task.
An early study by Koriat and Feuerstein (1976), using the
traditional version of word association, suggests a different
conclusion. Koriat and Feuerstein presented participants
with an auditory study list either under intentional learning
instructions or under incidental learning instructions coupled
with a distractor task. This latter (divided attention) condi-
tion reduced performance on the explicit memory tests of
free recall and recognition but did not reduce priming on the
conceptual implicit test of word association. Thus, unlike
Gabrieli, Stone, Shackleton, et al. (1996) and Mulligan and
Hartman (1996), Koriat and Feuerstein found an apparent
dissociation between conceptual priming and conceptual
explicit memory. Specifically, Koriat and Feuerstein's re-
sults imply that conceptual priming may not rely on
attentional encoding processes.
There are several potential reasons for the divergent
results of Gabrieli, Stone, Shackleton, et al. (1996) and
Mulligan and Hartman (1996), on the one hand, and Koriat
and Feuerstein (1976), on the other. The most obvious has to
do with the differences between the word-association and
the category-exemplar production tests. The word-associa-
tion task typically makes use of highly associated items
(e.g., doctor-nurse), whereas the examples in category-
exemplar production are typically less strongly associated
with the category cues (in both Gabrieli, Stone, Shackleton,
et al., 1996, and Mulligan & Hartman, 1996, the studied
examples did not rank in the 10 most frequently produced
instances, according to the norms of Battig & Montague,
1969). It is possible that the use of very strong associates
eliminates the need for elaborative processing to produce
relatively accessible memory traces (Bentin et al., 1995;
Rabinowitz, Craik, & Ackerman, 1982).
A second difference among the studies is that they used
different tasks to divide attention. A digit-monitoring task
was used by Gabrieli, Stone, Shackleton, et al. (1996) and by
Mulligan and Hartman (1996), a memory-load task was used
by Mulligan (1997), and a digit symbol task was used by
Koriat and Feuerstein (1976). The results of Mulligan (1997)
underscore the importance of the strength of the divided-
attention manipulation. A memory test dependent on atten-
tion during encoding may not appear so if attention has been
weakly divided. It is difficult to gauge whether Koriat and
Feuerstein's manipulation was as strong a division of
31
attention as that used by Gabrieli et al. and by Mulligan and
Hartman.
A final consideration is that Koriat and Feuerstein (1976)
did not use parallel explicit and implicit memory tests.
Consequently, it is not possible to determine if the dissociat-
ing effects of divided attention are due to the manipulation of
test instructions (implicit vs. explicit) or to other differences
between the word-association task and the free-recall and
recognition tests (e.g., type of memory cue or response
demands). As demonstrated by Craik et al. (1996), different
explicit tests are differentially reliant on (i.e., sensitive to)
attention at encoding. To be more confident that divided
attention dissociates implicit and explicit test performance,
one should use a robust division of attention and demon-
strate its effects on the explicit test before one can confi-
dently interpret a null result on the parallel implicit test
(Mulligan & Hartman, 1996).
Further investigation of the word-association test is
important for two reasons. First, it is important to further test
the generality of the TAP prediction that conceptual priming
will be diminished by divided attention. Second, the Koriat
and Feuerstein (1976) results have been interpreted as
evidence that attention may not be required for intact
conceptual priming (e.g., Jacoby, Toth, & Yonelinas, 1993;
Jacoby, Woloshyn, & Kelley, 1989; Smith & Oscar-Berman,
1990). In Experiment 2, we addressed some of these issues
to determine if conceptual priming in the word-association
task is sensitive to a strong division of attention. We used
procedures similar to those used by Mulligan and Hartman
(1996, Experiment 1) and Gabrieli, Stone, Shackleton, et al.
(1996) in uncovering effects of divided attention on concep-
tual priming in the category-exemplar production test. These
same procedures were used in the present Experiment 1, in
which divided attention reduced priming in the general
knowledge test. Participants were visually presented with a
list of words under the same full- or divided-attention
conditions as in Experiment 1. As noted earlier, the digit-
monitoring task provides a strong division of attention,
eliminating priming in the category-exemplar production
task (Mulligan & Hartman, 1996).
Retention of the studied words was tested with the
implicit test of word association and with its explicit
counterpart, associate-cued recall. The cued-recall task was
identical to the word-association task in all ways except
instructions. Thus, it was possible to determine if the
divided-attention task was strong enough to reduce perfor-
mance on a parallel explicit test.
Experiment 2
Method
Participants. Sixty-four undergraduates at Illinois State Univer-
sity participated in exchange for extra credit in psychology classes.
Design and materials. The experiment used a 2 X 2 X 2 design
in which study status (old vs. new) was manipulated within
subjects and attention at encoding (full vs. divided) and memory
test (word association vs. associate-cued recall) were manipulated
between subjects.
Fifty critical cue-response word pairs were chosen from the
32 MULLIGAN
Shapiro and Palermo (1968) word-association norms. Each re-
sponse word was the primary response to its associated cue. The
normative associative probabilities of the word pairs varied from
.19 to .61, with a mean of .34. The study lists were composed of the
response word from each pair. The response words were randomly
divided into two groups of 25 critical words each. Eight additional
words were used as primacy and recency buffers, 4 preceding and 4
following the critical words, creating two study lists of 33 words
each.
Both the word-association test and the cued-recall test consisted
of the 50 critical cue words, 25 corresponding to studied words (the
old words) and 25 to nonstudied words (the new words). The
critical cue words were intermixed with 44 filler cues (also chosen
from Shapiro & Palermo, 1968), creating a test list of 94 cue words.
The divided-attention task was identical to that used in Experiment 1.
Procedure. The study and distractor phases were identical to
Experiment 1. After the distractor task, participants completed
either the word-association test or the cued-recall test. For the
word-association test, participants were told that this was a
normative word-association study and that we would use their
responses to develop materials for future research. The participants
were asked to read each cue word and write down the first word that
came to mind, working rapidly through the list. Participants
responded to each cue word in this task. No mention was made of
the relationship between this task and the studied words. After
finishing the word-association task, participants were given an
awareness questionnaire. The experimenter asked a series of
increasingly specific questions, to assess whether participants
noticed the relationship between the study task and the memory test
and the extent to which they had deliberately tried to retrieve study
items.
For the cued-recall task, participants were informed that this was
a memory test for the words that were read aloud earlier in the
experiment. These participants were told to use the cue words to try
to remember related studied words but that many of the cues were
not related to studied items. Participants were asked to go through
the cue words in order, writing down remembered words next to the
cue. As is standard in this task, participants were not required to
respond to each cue. Both the cued-recall and word-association
tasks were self-paced.
Results and Discussion
During the study task, participants in the divided-attention
condition correctly identified 99.5% of the study words and
detected 95% of the target digit strings. Mann-Whitney U
tests indicated that the percentages of study words identified
and target strings detected did not significantly differ as a
function of subsequent test condition (ps > .5), as would be
expected. Participants in the divided-attention condition
were able to identify virtually all study words while
simultaneously identifying the vast majority of the target
sequences.
Table 2 summarizes the results of the memory tests.
Proportions of critical items produced are presented as a
function of test type, attention at encoding, and study status.
It can be seen from the table that dividing attention produced
strong effects on associatet-cued recall and conceptual prim-
ing in the word-association task. Statistical analyses confirm
this summary. The data from the word-association and
cued-recall tests were submitted to separate 2 x 2 ANOVAs,
using study status (old vs. new) as a within-subjects variable
and attention at encoding (full vs. divided) as a between-
Table!
Experiment 2: Proportions of Responses in Word
Association and Associate-Cued Recall as a Function of
Study Status and Attention at Encoding
Study status
Memory test Old New Old - new
Word association
Full attention .35 .28 .07
Divided attention .28 .27 .01
Associate-cued recall
Full attention .31 .10 .21
Divided attention .17 .07 .10
subjects variable. For the word-association task, the analysis
revealed two significant effects: (a) a main effect of study
status, F(l,30) = 10.81, MSE = 0.0027, and (b) an
interaction between study status and attention, F(l, 30) =
5.25, MSE = 0.0027. Overall, significant priming was
obtained, but crucially, priming was greater in the full- than
divided-attention condition. Follow-up t tests indicated a
significant effect of study status (i.e., priming) in the
full-attention condition, r(15) = 4.23, but not in the divided-
attention condition (p > .5). Thus, priming was obtained in
the full-attention condition but not in the divided-attention
condition.
Analysis of the cued-recall data also yielded significant
effects of study status, F(l, 30) = 56.63, MSE = 0.0067, and
the Study Status X Attention interaction, F(l, 30) = 6.92,
MSE = 0.0067. These results indicated that recall of old
words exceeded the guessing rate for new words (i.e.,
participants showed above-chance retention) but that di-
vided attention reduced the difference between old and new
items. Therefore, divided attention also affected the explicit
version of the word-association task. In both the full-
attention condition and the divided-attention condition, t
tests indicated an effect of study status, t(l5) = 6.08 and
4.44, respectively. In contrast to the implicit version of this
test, there was evidence of retention in both the full- and
divided-attention conditions in the cued-recall test.
The current results indicate that divided attention has a
very similar effect on the implicit and explicit versions of the
word-association task. In fact, the present results are pre-
cisely in accord with Mulligan and Hartman's (1996,
Experiment 1) finding that conceptual priming in the cat-
egory-exemplar production task was also reduced to (essen-
tially) zero in an identical divided-attention condition. The
current results contrast with the findings of Koriat and
Feuerstein (1976) and imply that given a strong manipula-
tion of attention, conceptual priming in the word-association
task is greatly affected by divided attention.
Whenever parallel implicit and explicit tests are affected
in similar ways, there is always the potential concern that the
putative implicit test has been compromised by intentional
retrieval strategies. In the present case, it is possible that the
results are due to more effective use of intentional retrieval
in the full-attention condition of the word-association task.
One way to assess the extent to which intentional retrieval
 ATTENTION AND MEMORY
influences the results of a nominally implicit memory test is
with the use of posttest questionnaires (e.g., Bowers &
Schacter, 1990; Light &-Albertson, 1989; Mulligan &
Hartman, 1996). In the present study, none of the word-
association participants reported using intentional retrieval
at test. In fact, only 4 of the 16 full-attention participants and
2 of the 16 divided-attention participants reported that they
were aware that some of the words that they had produced
were from the study list (the proportion of aware participants
did not significantly differ across the full- and divided-
attention conditions: Fisher's exact test, p > .3, one-tailed).
Excluding the aware participants from the analysis does not
alter the results. The 12 unaware participants in the full-
attention condition produced 36% of the old items and 28%
of the new items, whereas the 14 unaware participants in the
divided-attention condition produced 26% of the old and
25% of the new items. The amount of priming was
significantly greater in the full- than divided-attention condi-
tion, F(l, 24) = 7.78, MSE = 0.0045. In addition, signifi-
cant priming was obtained in the full-attention condition,
t(\X) = 3.53, but not in the divided-attention condition
(P > -5).
Consistent with this assessment is the finding that divided
attention reduces conceptual priming to approximately zero
but leaves associate-cued recall substantially above zero.
(Mulligan & Hartman, 1996, Experiment 1, reported the
same result for category-exemplar production and category-
cued recall.) If performance on the word-association test
were badly contaminated by intentional retrieval, then
priming in the divided-attention condition would likely be
significant. This finding lends further credence to the
proposal that these implicit and explicit tests do not rely
(entirely, at least) on the same retrieval processes or memory
encodings.
As an aside, we note that the word-association task seems
to give rise to lower levels of intentional retrieval and test
awareness, as measured by the posttest questionnaire, than
category-exemplar production, which is a more widely used
conceptual priming test. In Mulligan and Hartman (1996,
Experiment 1), the proportions of test-aware participants
were 93% and 57% in the full- and divided-attention
conditions, respectively (the corresponding proportions
claiming intentional retrieval were 50% and 14%). For the
young adults in Light and Albertson's (1989) study (most
comparable with the full-attention condition here), 88%
reported being test aware whereas 34% reported an inten-
tional retrieval strategy. These figures compare with 25%
and 12.5% of the word-association participants claiming test
awareness in the full- and divided-attention conditions,
respectively (as noted above, 0% claimed an intentional
retrieval strategy). To the extent that the posttest question-
naire is a measure of the purity of an implicit test, these
results suggest that the word-association task may be a
superior measure of conceptual implicit memory, at least in
this regard.
In summary, the results of this experiment add to the
findings of Experiment 1 and the previously cited work
(Gabrieli, Stone, Shackleton, et al., 1996; Mulligan, 1997;
Mulligan & Hartman, 1996) indicating that dividing atten-
33
tion at study reduces conceptual priming in the word-
association task, as well as the category-exemplar produc-
tion and general knowledge tasks. The present results are
important in demonstrating that when the division of atten-
tion is strong enough to reduce performance on the explicit
test of associate-cued recall, the comparable implicit test of
word association is also affected. This is significant because
it is not known whether Koriat and Feuerstein's (1976)
divided-attention task, which had no measurable impact on
word association, would have affected a comparable cued-
recall test.
The effects of divided attention on priming are generally
consistent with the predictions of the TAP framework.
Recall that the TAP view predicts that for divisions of
attention that do not disrupt identification, perceptual prim-
ing will be unaffected but conceptual priming will be
reduced. Several studies indicate that this prediction holds
for perceptual priming on tests such as word-fragment
completion, picture-fragment completion, lexical decision,
and perceptual identification (Gabrieli, Stone, Shackleton, et
al., 1996; Kellogg et al., 1996; Mulligan & Hartman, 1996;
Parkin et al., 1990; Parkin & Russo, 1990; Russo & Parkin,
1993; M. E. Smith & Oscar-Berman, 1990; Szymanski &
MacLeod, 1996). In contrast, conceptual priming on tasks
such as category-exemplar production, word association,
and the general knowledge test is reduced by divided
attention. The results of Mulligan (1997) and (arguably) the
results of Experiment 2 underscore the importance of the
strength of the divided-attention manipulation, indicating
that weaker manipulations may not reduce conceptual
priming.
This last point raises the possibility that strong division of
attention may also affect perceptual priming tasks, even if
the attentional manipulation does not disrupt stimulus identi-
fication. To develop this possibility, first consider that some
divided-attention studies used dual-task manipulations in
which the identification of study items either was substan-
tially reduced or was not required (Bentin et al., 1995;
Hawley & Johnston, 1991; Light & Prull, 1995a; Weldon &
Jackson-Barrett, 1993). Generally, these studies indicate that
manipulations that reduce the likelihood of stimulus identifi-
cation also reduce the amount of perceptual priming (Haw-
ley & Johnston, 1991; but see Bentin et al., 1995, for an
important exception). These studies were excluded from our
earlier discussion because they may represent a qualitatively
different type of manipulation than divided-attention tasks
that disrupt only postperceptual encoding operations. For
example, reductions in perceptual priming observed in these
studies may result from disruptions in lexical access, as
argued by Weldon and Jackson-Barrett (1993, p. 523). Such
an account does not apply to the present results or to those
reviewed earlier, because overt identification of study words
(presumably requiring lexical access) was required in both
the full- and divided-attention conditions.
The lexical-access account and other accounts that at-
tribute decreased perceptual priming to truncated perceptual
analysis are in accord with the TAP framework. Specifically,
the TAP view posits that any manipulation disrupting
perceptual processes will reduce perceptual priming. The
34 MULLIGAN
alternative view is that divided-attention tasks that disrupt
stimulus identification represent especially strong divisions
of attention and that the resulting division of attention makes
a unique contribution to the reduction of perceptual priming
beyond the simple disruption of perceptual processing
(Hawley & Johnston, 1991).2 If so, a strong manipulation of
attention may reduce perceptual priming, even in tasks that
have heretofore been unaffected by divided attention and
even when the attentional manipulation does not prevent
stimulus identification. This contrasts with the TAP frame-
work, which views the strength of the attentional manipula-
tion as irrelevant, provided perceptual identification pro-
cesses are not disrupted.
Mulligan and Hartman (1996, Experiment 2), using study
procedures identical to Experiments 1 and 2, found that
dividing attention with the digit-monitoring task had no
effect on perceptual priming in word-fragment completion
but significantly reduced recall on the matched explicit test
of word-fragment-cued recall. In Experiment 3, I examine
the generality of this finding by using a short-term memory-
load paradigm to vary attention during encoding. In this
experiment, each study word was preceded by a short-term
attentional load consisting of zero (the full-attention condi-
tion), three, five, or seven digits and letters. Participants
were instructed to maintain the digits and letters in memory
until a RECALL signal was presented a few seconds later.
This technique has been used to control the division of
attention over several levels (e.g., Baddeley & Hitch, 1974;
Engle, Conway, Tuholslri, & Shisler, 1995; Logan, 1979;
Mulligan, 1997). The particular attentional loads were
chosen on the basis of prior research. Using this memory-
load paradigm, Mulligan (1997) found that an attentional
load of five not only reduced but eliminated conceptual
priming in the category-exemplar production task. In addi-
tion, these results indicated that the attentional load of five
produced comparable levels of priming and cued recall as
the digit-monitoring task used to divide attention in Mulli-
gan and Hartman (1996) and in the present Experiments 1
and 2 (see Mulligan, 1997, Table 1). Considering an atten-
tional load of five as a strong division of attention, I included
that level of load, as well as an even stronger division (i.e.,
the seven load).
Implicit memory for the study words was assessed with
the perceptual implicit test of word-fragment completion. As
noted above, this perceptual test has not been affected by
other (although possibly weaker) manipulations of attention.
In the event that the present manipulation did not affect this
test, it would be important to determine if the manipulation
affected a comparable explicit test. Consequently, a separate
group of participants were given a word-fragment cued-
recall test.
Experiment 3
Method
Participants. Fifty-six undergraduates at Illinois State Univer-
sity participated in exchange for extra credit in psychology classes.
Design and materials. The experiment used a 4 x 2 design in
which attentional load (zero, three, five, or seven) was manipulated
within subjects and memory test (word-fragment completion vs.
word-fragment-cued recall) was manipulated between subjects.
The critical study materials consisted of 96 relatively low-
frequency words (mean frequency of 13; KuCera & Francis, 1967),
from 5 to 10 letters in length. Two master study lists of 48 items
each were created by randomly dividing the study materials in half.
Within each master file, the words were randomly divided into four
groups (each consisting of 12 words), and each group was rotated
through each of the attentional load conditions. This produced four
study lists from each master list, for a total of eight study lists. The
words in each study list were randomly ordered. Twelve additional
(noncritical) words were also used; 4 were presented before the list,
as practice items, and 4 were placed at the beginning and end of
each list as primacy and recency buffers. Each study list was
presented to equal numbers of participants.
Each study item was preceded by an attentional load. Nonzero
loads were constructed by randomly selecting items from the set of
digits (1-9) and a set of letters (B, C, D, F, G, H, J, K, L) according
to the following rules: (a) Digits and letters occupied alternating
positions, with a digit in the first position, and (b) no repetition of
digits or letters occurred within a load. These materials and rules
were used in an attempt to minimize chunking of the loads.
The memory tests consisted of a total of 102 word fragments.
The fragments consisted of between 2 and 6 letters, with under-
scores in the place of missing letters. The fragments were
constructed to have only one legitimate completion. This was
verified with the spellcheck and dictionary facilities of the WordPer-
fect program (Version 5.1), which indicated that all fragments were
consistent with only one English word completion. In addition, the
experimental participants never completed a fragment with any
word other than the intended one. Of the 102 fragments, 96
corresponded to the critical items from the two master lists, 48 of
which corresponded to old items and 48 to new items. Fragments
from an additional 6 words were used as filler items. The fragments
were randomly ordered with the constraint that there be no more
than 2 items in a row of any single type (old, new, orfiller).The old
and new items were counterbalanced so that across participants,
each word appeared equally often in the old and new conditions.
Procedure. Participants were tested individually. The experi-
ment consisted of a study task and a memory test. The study task
was based on the procedures of Mulligan (1997). Each study trial
began with a ready prompt that was displayed for 500 ms. Next, the
digit-letter string constituting the attentional load (or a dash in the
zero-load condition) was presented for 3,250 ms. For the non-zero-
load trials, participants were instructed to read the digit-letter
string aloud and retain it in memory until the recall signal was
given. For the zero-load trials, participants were simply instructed
to say "blank" in response to the dash. Next, the study word was
presented for 3 s. Participants were instructed to read the word
aloud and try to remember it for a later, unspecified memory test.
Finally, either the word RECALL (in the non-zero-load conditions)
or the word BLANK (in the zero-load condition) appeared for 2,750
ms. Participants were instructed either to recall the digits and letters
(in the non-zero-load conditions) or to again say "blank" (in the
zero-load condition).
2
Consistent with this view is the finding that explicit test
performance may exceed chance following study conditions that
eliminated perceptual priming (e.g., Hawley & Johnston, 1991;
Weldon & Jackson-Barrett, 1993). This result may be problematic
for the lexical-access account and other accounts positing truncated
perceptual processing because above-chance explicit memory
implies that stimulus identification or lexical access occurred for at
least some study items.
 ATTENTION AND MEMORY
After the study task, participants were given either the word-
fragment-completion or word-fragment cued-recall test. For both
tests, participants were given a sheet of paper with the 102
fragments, which were numbered and arranged in columns. In the
word-fragment-completion task, participants were asked to try to
complete each fragment with an appropriate English word. Ex-
amples of a fragment and an appropriate completion were pro-
vided. No mention was made of the relationship between this task
and the study task. In the word-fragment cued-recall test, partici-
pants were informed that their memory was being tested. They
were also informed that some of the fragments corresponded to
words from the study list and some did not. The participants were
asked to try to recall words from the study list that completed the
fragments and, if they could not do so, to complete the fragment
with any appropriate word that came to mind. These were the same
guessing instructions used by Mulligan and Hartman (1996,
Experiment 2) in an attempt to equate response requirements across
the implicit and explicit tests. This renders the two memory tests
comparable in all ways except test instructions (i.e., implicit vs.
explicit).
For both memory tests, the participants were given 12 s for each
fragment. Time was kept with an audiotape that emitted a beep
every 12 s. Participants were instructed to use the entire 12 s on the
current fragment, not to work backward, and to move on to the next
fragment only after hearing the beep. The participants were given a
cover sheet to keep their place.
Thirty-two of the participants were given the word-fragment
completion test, and the other 24 were given the word-fragment
cued-recall test.
Results and Discussion
During the study task, the mean proportion of attentional
load items recalled was .99, .94, and .76 for the three-, five-,
and seven-load conditions, respectively. Because these data
deviated substantially from the normal distribution (espe-
cially in the three- and five-load conditions), a nonparamet-
ric test (Friedman's test of ranks) was used. This test
revealed significant differences across loads. Pairwise sign
tests indicated that all three load conditions significantly
differed from one another. In addition, Mann-Whitney U
tests performed within load conditions found no significant
effects of subsequent test condition (allps > .1). The mean
proportion of study words correctly identified was .99 and
did not vary as a function of attentional load (p > .5) or
subsequent test condition (p > .25). Thus, participants were
able to identify virtually all study words while simulta-
neously maintaining the attentional loads.
The results of the memory tests are presented in Figure 1.
Mean proportions of fragments completed are presented as a
function of test type and attentional load. As statistical
analyses confirm, attentional load had a significant impact
on performance in the explicit test of word-fragment-cued
recall but not in the comparable implicit test of word-
fragment completion.
To measure retention, the proportion of completed new
fragments was subtracted from the proportion of completed
old fragments in each of the load conditions. The resulting
data from the word-fragment cued recall and word-fragment
completion were submitted to separate ANOVAs, using
attentional load as a within-subjects variable. The effect of
attentional load was significant for cued recall, F(3, 69) =
35
3.62, MSE = 0.0258, but not for word-fragment completion
(p > .5). For the cued-recall test, pairwise comparisons of
the load conditions indicated significantly greater perfor-
mance in the zero- than the seven-load condition, r(23) =
2.51; in the three- than five-load condition, *(23) = 2.30; and
in the three- than seven-load condition, f(23) = 2.65. No
other differences were significant (all otherps > .15). Perfor-
mance in each of the load conditions was significantly
greater than for new fragments, all f(23)s > 3.37. For the
word-fragment condition, performance in each of the load
conditions was significantly greater than for new fragments;
that is, priming occurred in all load conditions, all rs(23) >
6.84.
Finding that dividing attention had a significant impact on
word-fragment cued recall but no measurable impact on
word-fragment completion is theoretically important. Be-
fore continuing, however, it is important to evaluate the
power of the latter statistical test. Power calculations were
performed (following Kirk, 1995, pp. 259-264) using a
to-be-detected effect size estimated from the cued-recall
data. The estimated effect size of attentional load on the cued
recall is / = 0.29, somewhat larger than Cohen's (1988)
definition of a "medium" effect size of/ = 0.25. With alpha
equal to .05 and a sample size of 32, the power of detecting
an effect of attentional load on word-fragment completion of
the same size as that obtained with cued recall was .84.
Because earlier research has demonstrated the importance
of the strength of attentional manipulations in the study of
attention and implicit memory (Mulligan, 1997), one motiva-
tion for this experiment was to provide a range of divided-
attention conditions, from mild to very strong, within one
experiment. A potential concern was that whereas it is
reasonable to claim that increasing loads represent increas-
ing division of attention, the present method did not
guarantee that the participants would maintain increasing
amounts of information in working memory. Consequently,
the possibility that nominally different load conditions do
not actually tax working memory to different extents must
be entertained. One way to evaluate this is to examine the
average number of items retained by participants in the
different load conditions by converting percentage of load
items recalled (given above) into the raw number of load
items recalled. The average numbers of load items recalled
significantly increased across load conditions: 2.97, 4.70,
and 5.32, in the three-, five-, and seven-load conditions,
respectively; Friedman's test of ranks: x2(2, N = 32) =
88.32. Not surprisingly, sign tests indicated that significantly
more items were maintained in either the seven- or five-load
condition than in the three-load condition (both Zs = 7.35).
More important for present purposes, significantly more
items were maintained in the seven- than five-load condition
(Z- 2.81). This is consistent with the assertion that the
seven-load condition represents a stronger division of atten-
tion than either the three- or five-load condition.
Another way to address this issue would be to limit the
analysis of the test data to only those participants showing
high levels of performance in the seven-load condition at
study. As a cutoff, I included only those participants that
recalled an average of more than five items in the seven-load
36 MULLIGAN

Word-Fragment Completion

.60-

.50

o
o
I .40

.30

.20
1+1 1 1

New
Attentlonal Load

B
Word-Fragment Cued Recall

.20
New
Attentional Load

Figure 1. Results of Experiment 3: Mean proportions of completions (and ± 1 standard error bars)
in (A) the word-fragment-completion task and (B) the word-fragment cued-recall test as a function of
attentional load at encoding.

condition. This guaranteed that the resulting set of partici-
pants maintained more information in working memory in
the seven- than the five-load condition. Fourteen of the 24
(58%) cued-recall participants and 18 of the 32 (56%)
word-fragment completion participants exceeded this cutoff.
On average, these participants recalled 2.99, 4.84, and 6.13
items, in the three-, five-, and seven-load conditions, respec-
tively. For each participant, the average number of load
items recalled increased in each successive load condition.
An analysis of the test data for this subset of participants
revealed the same pattern of results as that obtained with all
participants. Attentional load had a significant impact on the
 ATTENTION AND MEMORY
explicit memory test of word-fragment cued recall, F(3, 39) =
3.15, MSE = 0.0323, but not on the comparable implicit test of
word-fragment completion (p > .5).
Thus, even when attention is very strongly divided during
encoding, as in the five- and seven-load conditions of the
present experiment, later performance on the perceptual
priming task is unaffected. This result is consistent with the
results of Mulligan and Hartman (1996, Experiment 2), in
which attention was divided with the digit-monitoring task
used in the present Experiments 1 and 2. Of course, the
criticism could always be raised that an even more rigorous
division of attention could be implemented (say, a nine-load
condition), which would reduce later perceptual priming
without reducing study-word identification, a possibility
difficult to rule out. However, using the memory-load
paradigm, it is questionable whether larger load conditions
would actually have the effect of more rigorously dividing
attention. Preliminary results from our laboratory indicate
that for loads exceeding seven, retention of load items
decreases markedly, producing functional loads that may not
differ from the seven-load condition. Other methods of
dividing attention may circumvent this problem. In any
event, these results are consistent with the view that
perceptual priming has minimal requirements of attention-
demanding postperceptual encoding processes. In addition,
these results provide additional evidence that divided atten-
tion dissociates perceptual priming and explicit memory
performance, even when the implicit and explicit tests are
matched in terms of memory cues and response require-
ments (Mulligan & Hartraan, 1996).
The TAP framework has thus far been successful at
predicting and accounting for the effects of attention on
implicit memory. However, the TAP framework, as initially
proposed, was posited as a general theory of both implicit
and explicit memory performance (e.g., Blaxton, 1989;
Roediger, 1990; Roediger et al., 1989). This framework
emphasizes the importance of the types of processes en-
gaged by a memory test rather than the nature of the test
instructions (i.e., implicit vs. explicit). Consequently, disso-
ciations are expected when perceptual and conceptual tests
are compared, but tests within one of the categories (e.g.,
perceptual tests) are generally expected to be similarly
affected by experimental manipulations.3 For example, the
TAP framework implies that perceptual explicit tests should
react more like perceptual implicit tests than conceptual
explicit tests. This has been found with respect to several
experimental manipulations, including read-generate, study
modality, imagery versus no-imagery instructions, and typog-
raphy manipulations (e.g., Blaxton, 1989; see Roediger &
McDermott, 1993, for a review). A strong interpretation of
the TAP view (which appears consistent with the theory's
initial form) implies that because perceptual implicit tests
are unaffected by divided attention, perceptual explicit tests
should be similarly unaffected. Later versions of the TAP
framework present a somewhat more differentiated view,
proposing that explicit retrieval instructions may make a
memory test more reliant on prior conceptual processing
(e.g., Craik et al., 1994; Roediger et al., 1992; we return to
this issue in the General Discussion).
37
The final two experiments examine these issues. Experi-
ment 4 uses the graphemic-cued-recall test, an explicit
memory test that has been classified as perceptual (see
Blaxton, 1989; Roediger, 1990; Roediger et al., 1989).
Experiment 5 uses the graphemic recognition test, a new
memory test that appears to be primarily perceptual in nature
(Challis, Velichkovsky, & Craik, 1996). In these memory
tests, participants are presented with memory cues bearing
graphemic similarity to studied items (e.g., cheetohs as a cue
for the studied word cheetah). In the graphemic-cued-recall
test, participants are informed of this relationship and asked
to use the cues to try to recall studied words. In the
graphemic recognition test, participants are asked to recog-
nize words that have a similar appearance to studied words,
without having to recall the studied words.
The results of Experiment 3 and Mulligan and Hartman
(1996, Experiment 2) are relevant to Experiments 4 and 5. In
these studies, the effects of divided attention were assessed
on an explicit version of a perceptual implicit test. Specifi-
cally, word-fragment cues were coupled with explicit test
instructions. Dividing attention reduced recall on the result-
ing word-fragment cued-recall test. This may be taken as
evidence that perceptual explicit tests are negatively affected
by divided attention. However, it should be pointed out that
word-fragment cued recall has not been categorized as a
perceptual test according to the TAP criterial variables. Also,
as noted above, refinements to the TAP view suggest that
explicit retrieval instructions render memory tests sensitive
to prior conceptual processing (e.g., Craik et al., 1994;
Roediger et al., 1992). Consistent with this view, Roediger et
al. (1992) found that word-fragment cued recall was sensi-
tive to a levels-of-processing manipulation that left word-
fragment completion unaffected. A more complete evalua-
tion of the TAP view would use an explicit memory test
categorized as perceptual according to the TAP criteria.
Graphemic-cued recall has been so categorized because it
exhibits the reverse-generation effect (i.e., the read condition
leads to higher performance than the generate condition), the
purported signature of a perceptual memory test (Blaxton,
1989; Roediger, 1990). Consequently, Experiment 4 investi-
gates the effects of divided attention on this perceptual
explicit test. A strong interpretation of the TAP framework
implies that this test will be relatively insensitive to divided
attention.
However, it has recently been reported that the graphemic-
cued-recall test is affected by the traditional levels-of-
processing manipulation (Challis et al., 1996; Thapar &
3
This does not mean that the TAP view never predicts dissocia-
tions between perceptual tests. If two perceptual tests cue memory
by means of different modes or modalities (e.g., picture-fragment
completion vs. word-fragment completion), the TAP view predicts
that certain study manipulations (e.g., studying pictures vs. words)
will dissociate the tests (see Roediger, 1990, for discussion). In the
present discussion, I restrict consideration to perceptual tests that
cue memory in the same mode and modality (e.g., all test cues are
visually presented words or word fragments). Without further
refinement, the default prediction of the TAP view is that such tests
should be similarly affected by study manipulations.
38 MULLIGAN
Greene, 1994). This implies that despite being categorized
as perceptual on the basis of the read-generate manipula-
tion, this memory test may have a conceptual component.
Challis et al. developed a variant of this test, the graphemic
recognition test, that is not affected by levels of processing.
Although this test has not been categorized according to the
other TAP criterial variables, it is reasonable to assume that
it is a perceptual explicit test, minimally influenced by prior
conceptual processing (as indicated by its insensitivity to the
levels-of-processing manipulation). On this basis, the graphe-
mic recognition test is provisionally categorized as a percep-
tual explicit test. To the extent that this test does not rely on
conceptual retrieval processes, both versions of the TAP
framework lead to the expectation that the graphemic
recognition test will respond to experimental manipulations
in ways similar to perceptual implicit tests. Specifically, like
perceptual implicit tests, the graphemic recognition test
should have minimal reliance on attention at encoding. The
graphemic recognition test is examined in Experiment 5.
In contrast to the TAP view, the bulk of the attention and
memory literature suggests that the effects of divided
attention on explicit tests are pervasive and are to be
expected on the graphemic-cued-recall and graphemic recog-
nition tests. We begin by contrasting these alternative
implications in Experiment 4 using graphemic-cued recall.
For comparison, we include a matched conceptual explicit
test, semantic-cued recall (Blaxton, 1989), which is ex-
pected to be adversely affected by divided attention.4 In the
semantic-cued-recall test, memory for the same study items
is tested with semantic rather than graphemic cues.
Experiment 4
Method
Participants. Forty-four undergraduates at Illinois State Univer-
sity participated in exchange for extra credit in psychology classes.
Design and materials. The experiment used a 2 X 2 X 2 design
in which study status (old vs. new) was manipulated within
subjects and attention at encoding (full vs. divided) and memory
test (graphemic-cued recall vs. semantic-cued recall) were manipu-
lated between subjects.
The study materials and digit-monitoring task were the same as
those used in Experiment 1. For each of the critical study words, a
graphemic cue and a semantic cue were drawn from Blaxton's
(1989) materials. For example, the graphemic and semantic cues
for the study word cheetah were cheetahs and jaguar, respectively.
The graphemic- and semantic-cued-recall tests consisted of the
appropriate set of cues arranged in a random order on a sheet of
paper. Each cue was followed by an answer space. Of the 100 cues,
50 corresponded to old items, and 50 corresponded to new items.
Production of new items served as the estimate of guessing rate.
Across participants, cues were equally often in the old and new
conditions.
Procedure. The study and distractor phases were identical to
Experiments 1 and 2. After the distractor task, participants were
presented with either the graphemic- or semantic-cued-recall test.
Participants were asked to use the provided cues to recall words
from the study list. The relationship between the cues and the
to-be-recalled words was explained, and three examples from the
primacy and recency portions of the study list were provided.
Recall that the primacy and recency buffers consisted of filler items
that were not tested. Participants were informed that many but not
all of the cues were related to study items. Beyond this, no specific
guessing instructions were provided, and participants were not
required to respond to each cue. Participants were given 10 s per
cue. Time was kept with an audiotape that emitted a beep every 10
s. Participants were provided with a cover sheet to keep their place.
They were asked to use the entire 10 s on one cue and to move onto
the next cue at the beep. Twenty-four of the participants were given
the graphemic-cued-recall test, and the other 20 were given the
semantic-cued-recall test.
Results and Discussion
During the study task, participants in the divided-attention
condition correctly identified 97% of the study words and
detected 90% of the target digit strings. Mann-Whitney U
tests indicated that the percentages of study words identified
and target strings detected did not significantly differ as a
function of subsequent test condition (ps > .10), as would
be expected. As in the earlier experiments, participants in the
divided-attention condition were able to identify the vast
majority of the study words while simultaneously identify-
ing almost all of the target sequences.
Table 3 summarizes the results of the memory tests.
Proportions of critical items produced are presented as a
function of test type, attention at encoding, and study status.
The table demonstrates that dividing attention during acqui-
sition has strong effects on both the semantic- and graphemic-
cued-recall tests. The data from the memory tests were
submitted to separate 2 X 2 ANOVAs, using study status
(old vs. new) as a within-subjects variable and attention at
encoding (full vs. divided) as a between-subjects variable.
Beginning with the semantic-cued-recall test, all three
effects were significant: (a) a main effect of attention, F(l,
18) = 5.92, MSE = 0.0017; (b) a main effect of study status,
F(l, 18) = 32.86, MSE = 0.0022; and (c) an interaction
between study status and attention, F(l, 18) = 4.49, MSE =
0.0022. The results indicated that more critical items were
produced in the full- than divided-attention condition, that
recall of old words exceeded the guessing rate for new words
(i.e., participants showed above-chance retention), and that
the difference between old and new items (i.e., the measure
of retention) was greater in the full- than divided-attention
condition. Follow-up t tests indicated an effect of study
status both in the full-attention condition, f(9) = 4.85, and
the divided-attention condition, t(9) = 3.07, indicating
above-chance retention in both the full- and divided-
attention conditions. In summary, the results of the semantic-
cued-recall test indicated that dividing attention produced
the expected effect on this conceptual explicit memory test:
It reduced memory for old items.
The results of the graphemic-cued-recall test are identical
to the results of the semantic-cued-recall test. The analysis
4
The semantic-cued-recall test is similar to the associate-cued-
recall test of Experiment 2 in that both tests use semantic associates
as cues. Because the semantic-cued-recall test used in Experiment
4 was derived from Blaxton's (1989) materials, I refer to this test by
the same name used in Blaxton's article rather than calling it an
associate-cued-recall test.
 ATTENTION AND MEMORY
Table 3
Experiment 4: Proportions of Responses in
Graphemic- and Semantic-Cued Recall as a Function
of Study Status and Attention at Encoding
Study status
Memory test Old New Old — new
Graphemic-cued recall
Full attention .31 .05 .26
Divided attention .17 .05 .12
Semantic-cued recall
Full attention .14 .03 .11
Divided attention .07 .02 .05
yielded significant effects of attention, F(\y 22) - 6.18, MSE =
0.0093; study status, F(l, 22) = 73.40, MSE = 0.0057; and
the Study Status X Attention interaction, F(l, 22) = 9.51,
MSE = 0.0057. These results reveal the same pattern as
above. Specifically, more critical items were produced in the
full- than divided-attention condition, recall of old words
exceeded the guessing rate for new words, and the difference
between old and new items was greater in the full- than
divided-attention condition. As with the semantic-cued-
recall test, t tests indicate an effect of study status in both the
full- and divided-attention conditions, fs(ll) = 6.26 and
7.52, respectively, indicating above-chance retention in both
attention conditions.5 To summarize, the perceptual explicit
test of graphemic-cued recall suffered the same deleterious
effects of divided attention as semantic-cued recall, a
conceptual explicit test
A final and somewhat surprising aspect of the results in
Table 3, although of less importance for present purposes, is
that performance tended to be higher on the graphemic- than
the semantic-cued-recall test. To investigate, old-new differ-
ence scores were submitted to a 2 X 2 ANOVA, using test
(graphemic- vs. semantic-cued recall) and attention at
encoding as between-subjects variables. This analysis re-
vealed significant main effects of attention, F(l,40) =
13.81, MSE = 0.0082 (indicating, as above, that divided
attention reduced retention), and of test, F(l, 40) = 14.69,
MSE = 0.0082. The latter effect indicates greater recall of
studied items when graphemic cues were provided at test. In
addition, the graphemic test led to significantly greater
performance on new items (i.e., a higher guessing rate), F(l,
42) = 5.73,A/S£ = 0.0013.
Although surprising, there is precedence for the finding of
greater recall with the graphemic than semantic cues with
Blaxton's (1989) materials. We restrict consideration only to
those experiments and conditions that are most comparable
to the full-attention condition in the present experiment
(e.g., intentional learning instructions, visual study presenta-
tion, same type font for study words and test cues, and no
imagery, elaborative, or generative encoding instructions).
In Blaxton's (1989) Experiment 2 (visual study condition),
the mean old-new difference scores for the graphemic- and
semantic-cued-recall tests were .41 and .33, respectively. In
Blaxton's Experiment 3 (same type font at study and test,
averaged over elite-elite and italic-italic conditions, see pp.
39
665-666), the comparable means were .23 and .11, respec-
tively. These results are similar to the results in the present
full-attention condition, in which the respective means were
.26 and . 11 (see Table 3). Other relevant data come from the
no-context condition of Blaxton's (1992) Experiments 1 and
2. This condition was similar to the present experiment in
that participants read study words in isolation (as opposed to
generating study words or reading them in a semantic
context). In the no-context condition, the graphemic-cued-
recall test led to greater mean difference scores than the
semantic-cued-recall test for control participants (the exact
figures are not provided in the article, but estimates from
graphs of the data indicate the graphemic advantage was at
least .20). The same pattern of results was obtained for
temporal lobe epileptics (Blaxton, 1992). The only compa-
rable experiment producing conflicting results is the no-
context condition in Blaxton's (1989) Experiment 1, in
which performance on the semantic-cued-recall test was
somewhat higher than on the graphemic-cued-recall test.
There are quite likely several factors that differ across
these experiments that play a role in accounting for the
precise relationship between graphemic- and semantic-cued
recall (as well as accounting for the variability in overall
performance levels). For example, the graphemic and seman-
tic cues in the Blaxton (1989) materials, from which the
present materials were drawn, were not equated on such
variables as cue-target associative strength or search set
size, both factors that can affect recall (see Nelson, Schreiber,
& McEvoy, 1992, for a review). However, taken as a group,
these results indicate that for the Blaxton (1989) materials,
the graphemic cues may be more effective than the semantic
cues when nonelaborative-nongenerative encoding strate-
gies are used. The opposite appears to be true for elaborative
or generative encoding strategies (see Blaxton, 1989,1992).
To the extent that this summary accurately describes the
relationship between the efficacy of the graphemic and
semantic cues, it is in accord with general TAP tenets.
Specifically, to the extent that conceptual or elaborative
processes are invoked at encoding, tests using semantic cues
(which presumably reengage conceptual processes) should
have an advantage over tests using nonsemantic cues.
Although this aspect of the results may be in accord with
the TAP framework, this digression should not obscure the
main point of the present experiment: Perceptual as well as
conceptual explicit tests have at least some reliance on
attention at encoding. Contrary to the TAP expectation,
division of attention has similar effects on perceptual and
5
This pattern of t values occurred because the divided-attention
condition had lower variance than the full-attention condition,
despite producing a smaller mean effect of study status. Although
the difference in variances across the full- and divided-attention
conditions was not significant by Levene's test, it was nevertheless
prudent to reexamine the data with a nonparametric test. The
old-new difference scores were submitted to the Mann-Whitney V
test, resulting in a significant difference between attention condi-
tions (U = 26.0). Thus, the nonparametric test produced the same
conclusion as the parametric tests reported in the text: Divided
attentionreducesretentionin the graphemic-cued-recall test.
40 MULLIGAN
conceptual explicit tests, whereas it dissociates perceptual
implicit and explicit tests.
As noted earlier, a potential limitation of the present study
is that the graphemic-cued-recall test can be affected by the
levels-of-processing manipulation (Challis et al., 1996;
Thapar & Greene, 1994). Consequently, a stronger test of the
hypothesis that divided attention affects perceptual explicit
memory may be garnered by using an explicit test that is not
affected by levels of processing. Graphemic recognition is
such a test (Challis et al., 1996), and it is used in Experiment 5.
Experiment 5
Method
Participants. Sixty-four undergraduates at Illinois State Univer-
sity participated in exchange for extra credit in psychology classes.
Design and materials. The experiment used a 2 X 2 X 2 design
in which study status (old vs. new) and test-associate type
(graphemic vs. semantic) were manipulated within subjects and
attention at encoding (full vs. divided) was manipulated between
subjects.
The study materials and digit-monitoring task were the same as
those used in Experiment 4. The graphemic-recognition test was
modeled after Challis et al. (1996). The test lists were created from
the same set of graphemic and semantic cues used in Experiment 4.
Specifically, each test list consisted of all 100 graphemic associates
(50 corresponding to the old words and 50 to the new words) and
50 of the 100 semantic associates (25 corresponding to old and 25
to new words). Four test lists were created, so that both the
graphemic associates and the semantic associates were counterbal-
anced over study status (i.e., old vs. new). The test items were
randomly ordered on the test sheet. The participant's task was to
circle test items that looked like the words presented during the
study phase. For this test, the correct answers are the 50 graphemic
associates of the old words. All other circled test items are false
alarms. The graphemic associates of new words served as the
false-alarm rate for graphemic recognition. The semantic associates
were included to determine if participants adhered to instructions
and circled only graphemically similar and not semantkally similar
test items (Challis et al., 1996). If participants followed instruc-
tions, then false-alarm rates for the semantic associates of old
words would be no higher than the false-alarm rate for the semantic
associates of new words (the result obtained by Challis et al.,
1996).
Procedure. The study and distractor phases were identical to
Experiment 4. Following the distractor task, participants were
presented with the graphemic-recognition test. Following Challis et
al. (1996), participants were told that some of the words on the test
sheet looked like words that had been read in the earlier study
phase. Participants were told to focus on the appearance of the test
words and ignore their meaning, circling those test items that
looked like words from the study phase. Two examples from the
primacy and recency portions of the study list were provided. As in
the earlier experiments, the primacy and recency buffers consisted
of filler items that were not tested. Participants were instructed to
take their time and try to be as accurate as possible. The memory
test was unpaced.
Results and Discussion
During the study task, participants in the divided-attention
condition correctly identified 99% of the study words and
detected 86% of the target digit strings. As in the earlier
experiments, participants in the divided-attention condition
were able to identify the vast majority of the study words
while simultaneously identifying almost all of the target
sequences.
Table 4 summarizes the results of the memory test. Mean
proportions of "old" responses are presented as a function
of test-associate type, attention at encoding, and study
status. Focusing on the graphemic test items, Table 4
demonstrates that dividing attention during acquisition re-
duced graphemic recognition. The results for the semantic
test items indicate that participants were able to (and did)
follow the instructions to endorse only items graphemically
similar to studied words. A preliminary analysis confirmed
that the graphemic and semantic test items gave rise to
different patterns of results and should be analyzed sepa-
rately. In this analysis, proportions of circled test items were
submitted to a 2 X 2 X 2 ANOVA, using study status (old vs.
new) and test-associate type (graphemic vs. semantic) as
within-subjects variables and attention at encoding (full vs.
divided) as a between-subjects variable. The relevant signifi-
cant effects were (a) the Study Status X Test Associate
interaction, F(l, 62) = 56.42, MSE = 0.0060, indicating
that as expected, the difference between old and new words
was greater for graphemic than semantic test items and (b)
the three-way interaction, F(l, 62) = 4.34, MSE = 0.0060,
indicating that the enhanced old-new difference for graphe-
mic items was affected by attention at encoding. Although
there were other significant effects in this preliminary
analysis, these interactions indicated that the data would be
best analyzed separately for the graphemic and semantic test
items.
Data from the graphemic and semantic test items were
submitted to separate 2 X 2 ANOVAs, using study status as a
within-subjects variable and attention at encoding as a
between-subjects variable. For the semantic test items, none
of the effects were significant (all ps > .1), consistent with
the claim that participants were able to follow the instruc-
tions and did not differentially endorse test items that were
semantically similar to old items. The analysis of the
graphemic items revealed two significant effects, a main
effect of study status, F(l, 62) = 131.83, MSE = 0.0065,
and an interaction between study status and attention, F(l,
62) = 7.62, MSE = 0.0065. The results indicated that
recognition of items graphemically similar to old words (i.e.,
Table 4
Experiment 5: Proportions of "Old" Responses in the
Graphemic-Recognition Test as a Function of Study Status,
Test-Associate Type, and Attention at Encoding
Study status
Test-associate
type Old New Old - new
Graphemic
Full attention .33 .13 .20
Divided attention .28 .16 ' .12
Semantic
Full attention .12 .10 .02
Divided attention .15 .13 .02
 ATTENTION AND MEMORY
hits) exceeded the guessing rate for items graphemically
similar to new words (i.e., false alarms). This indicated that
participants showed above-chance retention. More impor-
tant, the difference between old and new items (i.e., the
measure of retention) was greater in the full- than divided-
attention condition. Follow-up f tests indicated an effect of
study status in both the full- and the divided-attention
conditions, f(31) = 11.24 and 5.64, respectively.
In summary, the results of the graphemic recognition test
converge with the results of Experiment 4, indicating that
dividing attention reduces memory for old items in percep-
tual explicit tests. Moreover, these results make it unlikely
that divided attention reduced graphemic-cued recall solely
because this test has some sensitivity to prior conceptual
processing (as indicated by its sensitivity to the levels-of-
processing manipulation). The graphemic recognition test,
which uses the same overt memory cues as graphemic-cued
recall but is unaffected by prior conceptual processing
(indicated by its lack of sensitivity to the levels-of-
processing manipulation), is nonetheless very affected by
divided attention.
General Discussion
This article reports five experiments on the role of
attention during encoding on later implicit and explicit
memory. Integrating the present findings with prior research
yields a generally coherent set of findings, summarized here
in terms of the TAP taxonomy. First, as described in the
introduction, attentional state at encoding has pervasive
effects on traditional tests of memory, such as free recall,
recognition, and cued recall, all conceptual explicit tests in
the TAP nomenclature (e.g., Baddeley et al., 1984; Craik et
al., 1996; Fisk & Schneider, 1984; Moray, 1959; Murdock,
1965; Norman, 1969b). Although not in great need of
replication, this finding was reproduced in Experiments 2
and 4, in which dividing attention reduced performance on
cued-recall tests. More important, Experiments 4 and 5
showed that performance on the perceptual explicit tests of
graphemic-cued recall and graphemic recognition are also
reduced by divided attention. Coupled with the finding that
word-fragment-cued recall (arguably a perceptual explicit
test; see Blaxton et al., 1996) is also affected by divided
attention (the present Experiment 3; Mulligan & Hartman,
1996), the pattern of results supports the more general
conclusion that performance on explicit tests, whether
perceptual or conceptual, is dependent on attention at
encoding.
Turning to implicit tests of memory, it appears that
perceptual priming is relatively unaffected by divided atten-
tion, provided that stimulus identification is not disrupted
(see discussion of Experiment 2). In Experiment 3, such
results were found for word-fragment completion, even
using a very strong division of attention (see Gabrieli, Stone,
Shackleton, et al., 1996; Mulligan & Hartman, 1996; Parkin
et al., 1990, for similar findings). Similar results have been
reported for perceptual identification (Gabrieli, Stone, Shack-
leton, et al., 1996), picture-fragment completion (Parkin &
Russo, 1990; Russo & Parkin, 1993), and lexical decision
41
making (Kellogg et al., 1996; M. E. Smith & Oscar-Berman,
1990; Szymanski & MacLeod, 1996). In contrast to percep-
tual implicit tests, priming in conceptual implicit tests
generally appears to be reliant on attention at encoding,
especially when attention is strongly divided (Mulligan,
1997). Division of attention can reduce conceptual priming
on tests such as category-exemplar production (Gabrieli,
Stone, Shackleton, et al., 1996; Mulligan, 1997; Mulligan &
Hartman, 1996), word association (Experiment 2), and
general knowledge questions (Experiment I). 6
Before considering the theoretical implications of these
results, I discuss two related methodological considerations
in examining the role of attention in implicit and explicit
memory. First, the results underscore the importance of
using matched implicit and explicit tests when intepreting
dissociations (Neely, 1989; Schacter, Bowers, & Booker,
1989). Before one can confidently assert mat the attentional
requirements of implicit memory differ from those of
explicit memory, it is important that dissociations be based
on comparable implicit and explicit tests, ideally tests that
differ only in task instructions. Second, the use of strong
manipulations of attention, coupled with multiple levels of
attention as in the memory-load paradigm of Experiment 3
(also see Mulligan, 1997), is quite useful because memory
tests differ in the extent to which they are sensitive to
attentional manipulations during encoding (e.g., Craik et al.,
1996).
The relevance of these considerations is made clear by
considering some of the details of the results of Experiment
3, in which the attentional load variable significantly re-
duced performance on word-fragment-cued recall but not on
word-fragment completion. Consider the zero- (i.e., full
attention) and three-load conditions. It is known from prior
research (Mulligan, 1997) that manipulation of attention
over this range reduces performance on the explicit test of
category-cued recall (it is reasonable to assume that other
explicit tests, such as free recall, would be likewise af-
fected). If Experiment 3 had consisted of only the word-
fragment-completion test and only the zero- and three-load
conditions, an inappropriate conclusion may have been
derived; to wit, division of attention dissociates implicit and
explicit tests of memory. That this conclusion is too strong
can be seen by considering the results of the matched
explicit test, word-fragment-cued recall. Over the same
memory loads, this explicit test is also unaffected by divided
attention. Over that range of the attentional manipulation,
there is no dissociation of explicit and implicit memory tests
that use word-fragment cues, even though other explicit
tests, such as category-cued recall, are affected. In fact, the
attention manipulation does dissociate word-fragment
completion and word-fragment-cued recall, but this is not
apparent unless the larger attention loads (i.e., stronger
divisions of attention) are considered.
6
This summary may be complicated by two recent conference
presentations indicating that the semantic verification task, a
priming task requiring a conceptual decision, is unaffected by
divided attention and levels of processing (Gabrieli, Stone, Vaidya,
et al., 1996; Light & Prull, 1995b).
42 MULLIGAN
This discussion is relevant to the conflicting results
concerning the word-association test. As noted earlier,
Koriat and Feuerstein (1976) found no effect of divided
attention on word association, whereas the present Experi-
ment 2 did. However, the Koriat and Feuerstein study did
not include a matched explicit test, and the strength of the
attentional manipulation was unclear. In the present Experi-
ment 2, a comparable explicit test was used (the associate-
cued recall), and the attentional manipulation was strong
enough to reduce performance on that explicit test. Under
these circumstances, priming in the word-association test
was reduced, essentially to zero. Although there may well be
other important differences between the discrepant studies",
the results of Experiment 2 strongly imply that priming in
the word-association task relies on attention, at least under
the present circumstances.
Next, I consider some of the theoretical implications of
the present results. As noted in the introduction, some
theorists have suggested that recollective memory (pre-
sumed to underlie explicit test performance) is crucially
dependent on attention at encoding, whereas nonrecollec-
tive, or fluency, aspects of memory do not rely (or have
minimal reliance) on attentional encoding (Bentin, 1994;
Bentin et al., 1995; Besson et al., 1992; Isingrini et al., 1995;
Jacoby, Toth, & Yonelinas, 1993; Jacoby, Woloshyn, &
Kelley, 1989; Jelicic et al., 1992; Parkin et al., 1990; Parkin
& Russo, 1990; Szymanski & MacLeod, 1996; see Shallice
et al., 1994, for a similar assumption). The present results,
coupled with the results of Gabrieli, Stone, Shackleton, et al.
(1996) and Mulligan and Hartman (1996), are inconsistent
with this view. The present results indicate that if prior
conceptual-elaborative processing influences the expression
of nonrecollective memory, as it does on conceptual implicit
tests, then attention at encoding is crucial for that expres-
sion.
A related issue is the attentional requirements for encod-
ing semantic information. What aspects of meaning, if any,
are encoded as a mere consequence of identifying a word
(i.e., automatically)? There is substantial evidence for auto-
matic semantic activation during the course of word identifi-
cation (e.g., Balota, 1983, 1990; Carr, 1992; Durante &
Hirshman, 1994; Hirshman & Durante, 1992; Neely, 1977,
1991), but the effects of such activation are typically
short-lived (e.g., Neely, 1977). Are there automatic semantic
encodings that lead to longer term retention? Rabinowitz et
al. (1982) and Bentin (1994; Bentin et al., 1995) argued that
core (or context-invariant) semantic features of words are
encoded automatically and impact long-term memory tests.
In contrast, elaborative encoding processes and the integra-
tion of words with the contexts in which they occur (both of
which also influence long-term retention) are hypothesized
to require attention (Bentin, 1994; Bentin et al., 1995;
Rabinowitz etal., 1982).
The results of this and previous studies (Gabrieli, Stone,
Shackleton, et al., 1996; Mulligan, 1997; Mulligan &
Hartman, 1996) are relevant to this viewpoint. To review,
divided attention reduced conceptual priming on three
implicit tests: category-exemplar production (Gabrieli, Stone,
Shackleton, et al., 1996; Mulligan, 1997; Mulligan &
Hartman, 1996), general knowledge questions (Experiment
1), and word association (Experiment 2). In two of these
(category-exemplar production and word association), prim-
ing was reduced essentially to zero. The encoding of what
might be called core semantic features would appear to be an
important determinant of performance in these tasks. It
could be argued that truly core semantic features do not
mediate priming in the category-exemplar production task,
in which the critical exemplars are typically not the most
common category examples from the category norms (Mul-
ligan & Hartman, 1996; Rappold & Hashtroudi, 1991).
However, core semantic features are presumably crucial for
priming in the general knowledge test, in which many of the
questions are essentially definitions of the to-be-produced
item, and in the word-association task, in which the critical
response term is the most common associate to the cue. In
these cases, prior processing of what must be considered
core semantic features is crucial. In all cases, this prior
processing appears to be importantly dependent on attention.
Passive identification of the study words does not produce
conceptual priming on these tasks either at all or at least at
much lower levels than in full-attention conditions. This
implies either (a) that contrary to Bentin (1994; Bentin et al.,
1995) and Rabinowitz et al. (1982), the encoding of core
semantic features requires attention, or (b) that the memorial
vestiges of automatic semantic processing during word
identification influence only some long-term memory tests.
In this last section, I discuss the ramifications of the
present results for the TAP framework. The TAP framework
has been successful in accounting for much of the data from
implicit and explicit research (see Roediger, 1990, and
Roediger & McDermott, 1993, for reviews). Despite this
success, a growing body of research, to which this study
adds, indicates that the TAP dichotomy between perceptual
and conceptual processing is not rich enough to account for
the data from implicit and explicit memory research, even,
unfortunately, in broad strokes. Because the TAP view posits
common perceptual processes underlying perceptual im-
plicit and explicit tests and common conceptual-elaborative
processes underlying conceptual implicit and explicit tests,
this view suggests that dissociations between memory tests
are more likely to fall along the perceptual-conceptual
dimension than the dimension of implicit-explicit test
instructions.
A first objection (noted by Roediger, 1990) has been that
population dissociations are far more likely to occur along
the latter dimension than the former (e.g., Cermak, Verfael-
lie, & Chase, 1995; Gabrieli, Stone, Shackleton, et al., 1996;
Light & Albertson, 1989; Schwartz et al., 1993; see Mosco-
vitch, Goshen-Gottstein, & Vriezen, 1994, and Moscovitch,
Vriezen, & Goshen-Gottstein, 1993, for reviews; although
see Blaxton, 1992, for application of the TAP framework to
population dissociations). For instance, amnesics, older
adults, and schizophrenics all show normal levels of concep-
tual priming on category-exemplar production but reduced
performance on the matched explicit tests of category-cued
recall, when compared with appropriate controls (Keane,
Gabrieli, Monti, Cantor, & Noland, 1993; Light & Albert-
son, 1989; Schwartz et al., 1993). Likewise, amnesics show
 ATTENTION AND MEMORY
normal conceptual priming on the word-association test but
reduced recall on the matched-associate-cued recall test
(Cermak et al., 1995). Finally, amnesics frequently show
normal levels of priming on perceptual implicit tests but not
on the perceptual explicit test of graphemic-cued recall
(Cermak etal., 1995).
The TAP framework has had greatest success in account-
ing for experimental dissociations in research with healthy,
normal participants (see Roediger & McDermott, 1993, for
review). Even in this arena, recently obtained dissociations
have failed to honor the perceptual-conceptual distinction in
much the same way as population dissociations. Weldon and
Coyote (1996) found a picture-superiority effect on concep-
tual explicit tests (category-cued recall and associate-cued
recall) but not on comparable implicit tests (category-
exemplar production and word association, respectively). A
similar dissociation was reported by Mulligan (1996) in an
investigation of the perceptual interference effect in memory.
Under certain circumstances, interfering with (but not
preventing) stimulus perception at study can increase perfor-
mance on later explicit memory tests. Mulligan (1996)
reported that perceptual interference enhanced recall on the
category-cued-recall test but had no effect on conceptual
priming in the category-exemplar production test. The
results of Weldon and Coyote (1996) and of Mulligan (1996)
are both examples of experimental dissociations between
conceptual implicit and explicit tests. McDermott and
Roediger (1996) recently reported that both exact and
conceptual repetition of study items influenced free recall
but not category-exemplar generation or category-cued
recall. Although not a dissociation of comparable implicit
and explicit tests, the results are a dissociation of different
types of conceptual memory tests. Similarly, Cabeza (1994)
reported a dissociation between two conceptual implicit
tests.
To summarize, there exist (a) population variables that
dissociate performance on conceptual implicit and explicit
tests, (b) population variables that dissociate performance on
perceptual implicit and explicit tests, (c) experimental
manipulations that dissociate conceptual implicit and ex-
plicit tests (i.e., the picture-superiority and perceptual-
interference effects), (d) other experimental manipulations
that dissociate conceptual explicit tests from one another
(i.e., exact and conceptual repetition), and (e) still other
experimental manipulations that dissociate conceptual im-
plicit tests from one another (see Cabeza, 1994). To this set
of results, may now be added a sixth: The experimental
manipulation of divided attention dissociates performance
on perceptual implicit and explicit tests.
As noted by several researchers (e.g., McDermott &
Roediger, 1996; Mulligan, 1996; Weldon & Coyote, 1996),
dissociations within the class of conceptual tests or within
the class of perceptual tests are problematic for the TAP
view and suggest that its continued viability requires further
refinement. These refinements may take any of several forms
(or, perhaps, multiple refinements may be necessary to
accommodate all the data). Here, we consider two refine-
43
ments to the TAP framework in terms of how they might
account for the effects of divided attention on implicit and
explicit tests of memory, as well as the other dissociations
listed above.
First, dissociations that appear problematic for the TAP
view may simply indicate that memory tests are not purely
conceptual or purely perceptual in nature (Blaxton, 1989;
Weldon & Coyote, 1996; Weldon et al., 1995). If so, then an
important challenge to the TAP framework is to establish a
principled characterization of the mixture of perceptual and
conceptual processes involved in a given memory test (see
Roediger & McDermott, 1993, for discussion).
Weldon et al. (1995; see also Craik et al., 1994; Roediger
et al., 1992) proposed a refinement to the TAP framework
along these lines by positing that conceptual processes play
a larger role in explicit than implicit tests of memory,
independent of the nature of the memory cues (e.g., Graf &
Mandler, 1984). If explicit test instructions encourage
conceptual retrieval processes, then this would account for
dissociations between perceptual implicit tests and compa-
rable explicit tests, such as the dissociation between word-
fragment completion and word-fragment-cued recall caused
by dividing attention at study. Specifically, if divided
attention reduces conceptual encoding processes, then word-
fragment completion, which engages perceptual retrieval
processes, would be unaffected by the variation in prior
conceptual processing. However, coupling the word-
fragment cues with explicit retrieval instructions renders a
test that engages both perceptual and conceptual retrieval
processes. Consequently, word-fragment-cued recall would
be sensitive to the divided attention manipulation (see
Mulligan & Hartman, 1996, for related discussion). This
version of the TAP framework may well account for other
dissociations between perceptual implicit and explicit tests
(e.g., Roediger et al., 1992; Weldon et al., 1995).
However, such a modification of the TAP framework
faces three problems. First, by itself, it does not account for
the observation that some perceptual explicit tests, like
graphemic-cued recall and word-fragment-cued recall, are
sensitive to the levels-of-processing manipulation, whereas
others, such as graphemic recognition, are not (Challis et al.,
1996; Roediger et al., 1992; Thapar & Greene, 1994).
Second, it does not seem to account for dissociations
between conceptual implicit and explicit tests. According to
the TAP framework, conceptual tests, either implicit or
explicit, are assumed to be sensitive to the same underlying
conceptual processes. The proviso that explicit test instruc-
tions engage a common set of conceptual processes more
fully does suggest that conceptual explicit tests may render
more sensitive measurement of prior conceptual processes
(see Mulligan, 1997, for discussion), but without further
specification, this view does not explain why a conceptual
implicit test may be completely insensitive to a manipulation
that has a marked effect on a comparable explicit test. A
complementary approach, suggested by Cabeza (1994), is to
articulate subclasses of conceptual processes and determine
both how encoding manipulations influence the subclasses
44 MULLIGAN
and how the subclasses map onto the various implicit and
explicit tests (this is similar to the component-process model
of Moscovitch, 1992; see also Moscovitch et al., 1993;
Vriezen, Moscovitch, & Bellos, 1995).
A third problem with this modification of the TAP
framework is that it renders the TAP criterial variables
superfluous. Recall that in the TAP framework, memory
tests are classified as primarily perceptual or conceptual
according to how they are influenced by a set of criterial
variables, such as the read-generate manipulation and
manipulations of study modality. Part of the power of the
TAP framework is that it allows predictions to be made on
the basis of the classification. For instance, a memory test
classified as perceptual is predicted to be unaffected by any
of a large number of conceptual encoding manipulations. If
we add the proviso that explicit test instructions encourage
conceptual retrieval processes, then the TAP framework
does not generate unambiguous predictions for explicit tests
designated as perceptual by the criterial variables.7
Another way to extend the TAP framework is based on the
overt episodic nature of explicit tests of memory. In
particular, it is commonly assumed that spatiotemporal
context plays a more important role in explicit than in
implicit tests (e.g., Bower, 1996; Humphreys, Bain, & Pike,
1989; Humphreys, Wiles, & Dennis, 1994; Jacoby &
Hollingshead, 1990; Light & LaVoie, 1993; Schwartz et al.,
1993). According to this view, conceptual implicit and
explicit tests are both sensitive to prior conceptual-
elaborative processing, but in addition, conceptual explicit
tests are also sensitive (or are at least more sensitive) to prior
encoding of study context (Mulligan, 1997). Likewise, this
view states that perceptual implicit and explicit tests may be
sensitive to the same prior perceptual processing, but in
addition, explicit tests are sensitive to prior encoding of
contextual information. Such a refinement parsimoniously
supplements the conceptual-perceptual processing distinc-
tion and provides an account for the effects of dividing
attention. There is good evidence that dividing attention
disrupts conceptual and contextual encoding (e.g., Craik,
1989; Rabinowitz et al., 1982; see Cowan, 1995, for a
review). To the extent that perceptual processes are unaf-
fected by divided attention, this account predicts that
dividing attention will disrupt tests that rely on contextual
encoding (i.e., perceptual and conceptual explicit tests), as
well as tests that are sensitive to conceptual processing (i.e.,
conceptual implicit or explicit tests).
A final point concerning this extension of the TAP
framework is that it suggests that implicit tests should be
relatively unaffected by manipulations of spatiotemporal
context. To my knowledge, there are only two published
studies on this topic, and they produced conflicting results.
Jacoby (1983a) reported that a change in environmental
context bad no effect on the implicit test of perceptual
identification. In contrast, S. M. Smith, Heath, and Vela
(1990) found that manipulating environmental context sub-
stantially reduced priming in the homophone-spelling task.
Additional research on this topic, including studies of the
influence of environmental context on conceptual implicit
tests, is required to more fully evaluate this extension of the
TAP framework.
7
One way around this problem is to expand the set of criterial
variables. This has two consequences for the utility of the TAP
framework. First, it increases the chances of conflicting outcomes
(i.e., some criterial variables indicate that the test is perceptual, and
others indicate that it is conceptual, as in the case of graphemic-
cued recall), thus decreasing the number of memory tests for which
unambiguous predictions may be made. Second, as the set of
criterial variables increases, the utility of classification decreases.
Specifically, if we must evaluate a memory test on increasingly
larger sets of variables before classification, this necessarily
reduces the number of variables about which the theory makes
predictions.
References
Anooshian, L. J. (1989). Effects of attentive encoding on analytic
and nonanalytic processing in implicit and explicit retrieval
tasks. Bulletin of the Psychonomic Society, 27, 5-8.
Baddeley, A., Lewis, V., Eldridge, M., & Thomson, N. (1984).
Attention and retrieval from long-term memory. Journal of
Experimental Psychology: General, 113, 518-540.
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H.
Bower (Ed.), The psychology of learning and motivation (pp.
47-89). New York: Academic Press.
Balota, D. A. (1983). Automatic semantic activation and episodic
memory encoding. Journal of Verbal Learning and Verbal
Behavior, 22, 88-104.
Balota, D. A. (1990). The role of meaning in word recognition, hi
D. A. Balota, G. B. Flores d'Arcais, & K. Rayner (Eds.),
Comprehension processes in reading (pp. 9-32). Hillsdale, NJ:
Erlbaum.
Battig, W. F, & Montague, W. E. (1969). Category norms of verbal
items in 56 categories: A replication and extension of the
Connecticut category norms. Journal of Experimental Psychol-
ogy Monographs, 80(3, Pt. 2).
Bentin, S. (1994). Semantic processing of ignored stimuli: The role
of attention in memory. In C. Umilta & M. Moscovitch (Eds.),
Attention & performance XV: Conscious and nonconscious
information processing (pp. 551-569). Cambridge, MA: MIT
Press.
Bentin, S., Kutas, M., & Hillyard, S. A. (1995). Semantic process-
ing and memory for attended and unattended words in dichotic
listening: Behavioral and electrophysiological evidence. Journal
of Experimental Psychology: Human Perception and Perfor-
mance, 21, 54—67.
Besson, M., Fischler, I., Boaz, T., & Raney, G. (1992). Effects of
automatic associative activation on explicit and implicit memory
tests. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 18, 89-105.
Blaxton, T. A. (1989). Investigating dissociations among memory
measures: Support for a transfer-appropriate processing frame-
work. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 15, 657-668.
Blaxton, T. A. (1992). Dissociations among memory measures in
memory-impaired subjects: Evidence for a processing account of
memory. Memory & Cognition, 20, 549-562.
Blaxton, T. A., Bookheimer, S. Y., Zeffiro, T. A., Figlozzi, C. M.,
Gaillard, W. D., & Theodore, W. H. (1996). Functional mapping
of human memory using PET: Comparisons of conceptual and
 ATTENTION AND MEMORY
perceptual tasks. Canadian Journal of Experimental Psychology,
50, 42-56.
Bornstein, R. F , Leone, D. R., & Galley, D. J. (1987). The
generalizability of subliminal mere exposure effects: Influences
of stimuli perceived without awareness. Journal of Personality
and Social Psychology, 53, 1070-1079.
Bower, G. H. (1996). Reactivating a reactivation theory of implicit
memory. Consciousness and Cognition, 5, 27-72.
Bowers, J. S., & Schacter, D. L. (1990). Implicit memory and test
awareness. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 16, 404—416.
Broadbent, D. E. (1958), Perception and communication. London:
Pergamon Press.
Broadbent, D. E. (1971). Decision and stress. San Diego, CA:
Academic Press.
Brown, A. S., & Mitchell, D. B. (1994). A reevaluation of semantic
versus nonsemantic processing in implicit memory. Memory &
Cognition, 22, 533-541.
Cabeza, R. (1994). A dissociation between two implicit conceptual
tests supports the distinction between types of conceptual
processing. Psychonomic Bulletin and Review, 1, 505-508.
Carr, T. H. (1992). Automaticity and cognitive anatomy: Is word
recognition automatic? American Journal of Psychology, 105,
201-238.
Cermak, L. S., Verfaellie, M., & Chase, K. A. (1995). Implicit and
explicit memory in amnesia: An analysis of data-driven and
conceptually driven processes. Neuropsyckology, 9, 281-290.
Challis, B. H., & Brodbeck, D. R. (1992). Level of processing
affects priming in word fragment completion. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition, 18,
595-607.
Challis, B. H., Chiu, C. -Y, Ken:, S. A., Law, J., Schneider, L.,
Yonelinas, A., & Tulving, E. (1993). Perceptual and conceptual
cueing in implicit and explicit retrieval. Memory' A 127-151.
Challis, B. H., Velichkovsky, B. M., & Craik, F. I. M. (1996).
Levels-of-processing effects on a variety of memory tasks: New
findings and theoretical implications. Consciousness and Cogni-
tion, 5, 142-164.
Cherry, E. C. (1953). Some experiments on the recognition of
speech with one and with two ears. Journal of the Acoustical
Society ofAmerica, 26, 554-559.
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences. Hillsdale, NJ: Erlbaum.
Cowan, N. (1988). Evolving conceptions of memory storage,
selective attention, and their mutual constraints within the
human information-processing system. Psychological Bulletin,
104, 163-191.
Cowan, N. (1995). Attention and memory: An integrated frame-
work. New York: Oxford University Press.
Craik, F. I. M. (1982). Selective changes in encoding as a function
of reduced processing capacity. In F. Klix, J. Hoffman, & E. van
der Meer (Eds.), Cognitive research in psychology (pp. 152—
161). Berlin, Germany: Springer-Verlag.
Craik, F I. M. (1983). On the transfer of information from
temporary to permanent memory. Philosophical Transactions of
the Royal Society of London, Series B, 302, 341-359.
Craik, F I. M. (1989). On the making of episodes. In H. L.
Roediger & F. I. M. Craik (Eds.), Varieties of memory and
consciousness: Essays in honour ofEndel Tulving (pp. 43-57).
Hillsdale, NJ: Erlbaum.
Craik, F. I. M., & Byrd, M. (1982). Aging and cognitive deficits:
The role of attentional resources. In F. I. M. Craik & S. E. Trehub
(Eds.), Aging and cognitive processes (pp. 191-211). New York:
Plenum Press.
Craik, F. I. M., Govoni, R., Naveh-Benjamin, M., & Anderson,
45
N. D. (1996). The effects of divided attention on encoding and
retrieval processes in human memory. Journal of Experimental
Psychology: General, 125, 159-180.
Craik, F. I. M., Moscovitch, M., & McDowd, J. M. (1994).
Contributions of surface and conceptual information to perfor-
mance on implicit and explicit tasks. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 20, 864—875.
Denny, E. B., & Hunt, R. R. (1992). Affective valence and memory
in depression: Dissociation of recall and fragment completion.
Journal ofAbnormal Psychology, 101, 575-580.
Durante, R., & Hirshman, E. (1994). Retrospective priming and
masked semantic priming: The interfering effects of prime
activation. Journal of Memory and Language, 33, 112-127.
Eagle, M., & Ortof, E. (1967). The effect of level of attention upon
"phonetic" recognition errors. Journal of Verbal Learning and
Verbal Behavior, 6, 226-231.
Eich, E. (1984). Memory for unattended events: Remembering
with and without awareness. Memory & Cognition, 12, 105-111.
Elliott, C. L., & Greene, R. L. (1992). Clinical depression and
implicit memory. Journal of Abnormal Psychology, 101, 572-
574.
Engle, R. W., Conway, A. R. A., Tuholski, S. W., & Shisler, R. J.
(1995). A resource account of inhibition. Psychological Science,
6, 122-125.
Fisk, A. D., & Schneider, W. (1984). Memory as a function of
attention, level of processing, and automatization. Journal of
Experimental Psychology: Learning, Memory, and Cognition,
10, 181-197.
Gabrieli, J. D. E., Keane, M. H., Stanger, B. Z., Kjelgaard, M. M.,
Corkin, S., & Growdon, J. H. (1994). Dissociations among
structural-perceptual, lexical-semantic, and event-fact memory
systems in Alzheimer, amnesic, and normal subjects. Cortex, 30,
75-103.
Gabrieli, J. D. E., Stone, M. V, Shackleton, K., Thompson-Schill,
S. L., Ladd, S. L., Vaidya, C. J., & Chan, A. (1996). Attention
and implicit memory for words. Manuscript submitted for
publication.
Gabrieli, J. D. E., Stone, M. V., Vaidya, C. J., Askari, N., Zabinski,
M. F, & Rabin, L. (1996, November). Neuropsychological and
behavioral evidence for the role of attention in implicit memory.
Paper presented at the 37th Annual Meeting of the Psychonomic
Society, Chicago.
Graf, P., & Mandler, G. (1984). Activation makes words more
accessible, but not necessarily more retrievable. Journal of
Verbal Learning and Verbal Behavior, 23, 553-568.
Graf, P., Shimamura, A., & Squire, L. (1985). Priming across
modalities and priming across category levels: Extending the
domain of preserved functions in amnesia. Journal of Experimen-
tal Psychology: Learning, Memory, and Cognition, 11, 386-396.
Hamann, S. B. (1990). Level-of-processing effects in conceptually
driven implicit tasks. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 16, 970-977.
Hawley, K. J., & Johnston, W. A. (1991). Long-term perceptual
memory for briefly exposed words as a function of awareness
and attention. Journal of Experimental Psychology: Human
Perception and Performance, 17, 807-815.
Hirshman, E., & Durante, R. (1992). Prime identification and
semantic priming. Journal of Experimental Psychology: Learn-
ing, Memory, and Cognition, 18, 255-265.
Humphreys, M. S., Bain, J. D., & Pike, R. (1989). Different ways to
cue a coherent memory system: A theory for episodic, semantic,
and procedural tasks. Psychological Review, 96, 208-233.
Humphreys, M. S., Wiles, J., & Dennis, S. (1994). Toward a theory
of human memory: Data structures and access processes.
Behavioral and Brain Sciences, 17, 655-692.
46 MULLIGAN
Isingrini, M., Vazou, E, & Leroy, P. (1995). Dissociation between
implicit and explicit memory tests: Effects of age and divided
attention on category exemplar generation and cued recall.
Memory & Cognition, 23, 462-467.
Jacoby, L. L. (1983a). Perceptual enhancement: Persistent effects
of an experience. Journal of Experimental Psychology: Learn-
ing, Memory, and Cognition, 9, 21-38.
Jacoby, L. L. (1983b). Remembering the data: Analyzing interac-
tive processes in reading. Journal of Verbal Learning and Verbal
Behavior, 22, 485-508.
Jacoby, L. L. (1991). Aprocess dissociation framework: Separating
automatic from intentional uses of memory. Journal of Memory
and Language, 30, 513-541.
Jacoby, L. L., & Hoi lings head, A. (1990). Toward a generate/
recognize model of performance on direct and indirect tests of
memory. Journal of Memory and Language, 29, 433-454.
Jacoby, L. L., Toth, J. P., & Yonelinas, A. P. (1993). Separating
conscious and unconscious influences of memory: Measuring
recollection. Journal of Experimental Psychology: General, 122,
1-16.
Jacoby, L. L., Woloshyn, V, & Kelley, C. (1989). Becoming
famous without being recognized: Unconscious influences of
memory produced by divided attention. Journal of Experimental
Psychology: General, 118, 115-125.
Jelicic, M., Bonke, B., Wolters, G., & Phaf, R. H. (1992). Implicit
memory for words presented during anaesthesia. European
Journal of Cognitive Psychology, 4, 71-80.
Keane, M. M., Gabrieli, J. D. E., Fennema, A. C , Growdon, J. H.,
& Corkin, S. (1991). Evidence for a dissociation between
perceptual and conceptual priming in Alzheimer's disease.
Behavioral Neuroscience, 105, 326-342.
Keane, M. M., Gabrieli, J. D. E., Monti, L. A., Cantor, J. M,, &
Noland, J. S. (1993). Amnesic patients show normal priming and
a normal depth of processing effect in a conceptually driven
implicit memory test. Society for Neuroscience Abstracts, 19,
1079.
Kellogg, R. T., Newcombe, C , Kammer, C , & Schmitt, K. (1996).
Attention and direct and indirect memory tasks with short- and
long-term probes. American Journal of Psychology, 109, 205-
217.
Kirk, R. E. (1995). Experimental design (3rd ed.). New York:
Brooks/Cole.
Koriat, A., & Feuerstein, N. (1976). The recovery of incidentally
acquired information. Acta Psychologica, 40, 463-474.
Kucera, H., & Francis, W. N. (1967). Computational analysis of
present-day American English. Providence, RI: Brown Univer-
sity Press.
Light, L. L. (1991). Memory and aging: Four hypotheses in search
of data. Annual Review of Psychology, 42, 333-376.
Light, L. L., & Albertson, S. A. (1989). Direct and indirect tests of
memory for category exemplars in young and older adults.
Psychology and Aging, 4, 487-492.
Light, L. L., & LaVoie, D. (1993). Direct and indirect measures of
memory in old age. In P. Graf & M. E. J. Masson (Eds.), Implicit
memory: New directions in cognition, development, and neuro-
psychology (pp. 207-230). Hillsdale, NJ: Erlbaum.
Light, L. L., & Prull, M. W. (1995a). Aging, divided attention, and
repetition priming. Swiss Journal of Psychology, 54, 87-101.
Light, L. L., & Prull, M. W. (1995b, November). Repetition
priming, divided attention and aging. Paper presented at the 36th
Annual Meeting of the Psychonomic Society, Los Angeles.
Logan, G. D. (1979). On the use of a concurrent memory load to
measure attention and automaticity. Journal of Experimental
Psychology: Human Perception and Performance, 5, 189-207.
MacKay, D. G. (1973). Aspects of the theory of comprehension,
memory and attention. Quarterly Journal of Experimental
Psychology, 25, 22-40.
Mandler, G., Nakamura, Y., & Van Zandt, B. J. S. (1987).
Nonspecific effects of exposure on stimuli that cannot be
recognized. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 13, 646-648.
McDennott, K. B., & Roediger, H. L. (1996). Exact and conceptual
repetition dissociate conceptual memory tests: Problems for
transfer appropriate processing theory. Canadian Journal of
Experimental Psychology, 50, 57-71.
Merikle, P., & Reingold, E. (1991). Comparing direct (explicit) and
indirect (implicit) measures to study unconscious memory.
Journal of Experimental Psychology: Learning, Memory, and
Cognition, 17, 224-233.
Moray, N. (1959). Attention in dichotic listening: Affective cues
and the influence of instructions. Quarterly Journal of Experimen-
tal Psychology, 11, 56-60.
Moscovitch, M. (1992). Memory and working-with-memory: A
component process model based on modules and central sys-
tems. Journal of Cognitive Neuroscience, 4, 257-267.
Moscovitch, M., Goshen-Gottstein, Y, & Vriezen, E. (1994).
Memory without conscious recollection: A tutorial review from a
neuropsychological perspective. In C. Umilta & M. Moscovitch
(Eds.), Attention and performance (pp. 619-660). Cambridge,
MA: MIT Press.
Moscovitch, M., Vriezen, E., & Goshen-Gottstein, Y. (1993).
Implicit memory in patients with focal lesions and degenerative
brain disorders. In F. Boiler & J. Grafman (Eds.), Handbook of
neuropsychology (pp. 133-173). Amsterdam: Elsevier.
Mulligan, N. W. (1996). The effects of perceptual interference at
encoding on implicit memory, explicit memory, and memory for
source. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 22, 1067-1087.
Mulligan, N. W. (1997). Attention and implicit memory: The
effects of varying attentional load on conceptual priming.
Memory & Cognition, 25, 11-17.
Mulligan, N. W., & Hartman, M. (1996). Divided attention and
indirect memory tests. Memory <fe Cognition, 24,453-465.
Murdock, B. B. (1965). Effects of subsidiary task on short-term
memory. British Journal of Psychology, 56, 413-419.
Neely, J. H. (1977). Semantic priming and retrieval from lexical
memory: Roles of inhibitionless spreading activation and limited-
capacity attention. Journal of Experimental Psychology: Gen-
eral, 106, 226-254.
Neely, J. H. (1989). Experimental dissociations and the episodic/
semantic memory distinction. In H. L. Roediger & F. I. M. Craik
(Eds.), Varieties of memory and consciousness: Essays in honour
ofEndel Tulving (pp. 229-270). Hillsdale, NJ: Erlbaum.
Neely, J. H. (1991). Semantic priming effects in visual word
recognition: A selective review of current findings and theories.
In D. Besner & G. W. Humphreys (Eds.), Basic processes in
reading—visual word recognition (pp. 264-336). Hillsdale, NJ:
Erlbaum.
Nelson, D. L., Schreiber, T. A., & McEvoy, C. L. (1992).
Processing implicit and explicit representations. Psychological
Review, 99, 322-348.
Norman, D. A. (1969a). Memory and attention: An introduction to
human information processing. New York: Wiley.
Norman, D. A. (1969b). Memory while shadowing. Quarterly
Journal of Experimental Psychology, 21, 85-93.
Parkin, A. J., Reid, T. K., & Russo, R. (1990). On the differential
nature of implicit and explicit memory. Memory <fe Cognition,
18,507-514.
Parkin, A. J., & Russo, R. (1990). Implicit and explicit memory and
 ATTENTION AND MEMORY
the automatic/effortful distinction. European Journal of Cogni-
tive Psychology, 2, 71-80.
Rabinowitz, J. C , Craik, E I. M., & Ackerman, B. P. (1982). A
processing resource account of age differences in recall. Cana-
dian Journal of Psychology, 36, 325-344.
Rappold, V. A., & Hashtroudi, S. (1991). Does organization
improve priming? Journal of Experimental Psychology: Learn-
ing, Memory, and Cognition, 17, 103-114.
Roediger, H. L., HI. (1990). Implicit memory: Retention without
remembering. American Psychologist, 45, 1043-1056.
Roediger, H. L., Ill, & McDermott, K. B. (1993). Implicit memory
in normal human subjects. In E Boiler & J. Grafman (Eds.),
Handbook of neuropsychology (pp. 63-131). Amsterdam:
Elsevier.
Roediger, H. L., ITI, Weldon, M. S., & Challis, B. H. (1989).
Explaining dissociations between implicit and explicit memory:
A processing account. In H. L. Roediger III & E I. M. Craik
(Eds.), Varieties of memory and consciousness: Essays in honour
ofEndel Tulving (pp. 3-41). Hillsdale, NJ: Erlbaum.
Roediger, H. L., m , Weldon, M. S., Stadler, M. A., & Riegler, G. H.
(1992). Direct comparison of two implicit memory tests: Word
fragment and word stem completion. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 18, 1251-1269.
Russo, R., & Parkin, A. J. (1993). Age differences in implicit
memory: More apparent than real. Memory & Cognition, 21,
73-80.
Schacter, D. L. (1985). Priming of old and new knowledge in
amnesic patients and normal subjects. Annals of the New York
Academy of Sciences: Vol. 444 (pp. 44—53). New York: New
York Academy of Sciences.
Schacter, D. L. (1987). Implicit memory: History and current
status. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 13, 501-518.
Schacter, D. L., Bowers, J., & Booker, J. (1989). Intention,
awareness, and implicit memory: The retrieval intentionality
criterion. In S. Lewandowsky, J. C. Dunn, & K, Kirsner (Eds.),
Implicit memory: Theoretical issues (pp. 47-65). Hillsdale, NJ:
Erlbaum,
Schwartz, B. L., Rosse, R. B., & Deutsch, S. I. (1993). Limits of the
processing view in accounting for dissociations among memory
measures in a clinical population. Memory & Cognition, 21,
63-72.
Seamon, J., Marsh, R., & Brody, N. (1984). Critical importance of
exposure duration for attentive discrimination of stimuli that are
not recognized. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 10, 465-469.
Shallice, T., Fletcher, P., Frith, C D., Grasby, P., Frackowiak,
R. S. J., & Dolan, R. J. (1994, April). Brain regions associated
with acquisition and retrieval of verbal episodic memory.
Nature, 368, 633-635.
Shapiro, S. I., & Palermo, D. S. (1968). An atlas of normative free
association data. Psychonomic Monograph Supplements, 2,
219-250.
Shimamura, A. P. (1986). Priming effects in amnesia: Evidence for
a dissociable memory function. Quarterly Journal of Experimen-
tal Psychology, 38A, 619-644.
Shimamura, A. P. (1993). Neuropsychological analyses of implicit
memory: History, methodology, and theoretical interpretations.
47
In P. Graf & M. E. J. Masson (Eds.), Implicit memory: New
directions in cognition, development and neuropsychology (pp.
265-285). Hillside, NJ: Erlbaum.
Slamecka, N. J., & Graf, P. (1978). The generation effect:
Delineation of a phenomenon. Journal of Experimental Psychol-
ogy: Human Learning and Memory, 4, 592-604.
Smith, M. E., & Oscar-Berman, M. (1990). Repetition priming of
words and pseudowords in divided attention and in amnesia.
Journal of Experimental Psychology: Learning, Memory, and
Cognition, 16, 1033-1042.
Smith, S. M., Heath, F. R., & Vela, E. (1990). Environmental
context-dependent homophone spelling. American Journal of
Psychology, 103, 229-242.
Stone, M., Ladd, S. L., Vaidya, C. J., & Gabrieli, J. D. E. (1996).
Word identification priming for ignored and attended words.
Manuscript submitted for publication.
Szymanski, K. E, & MacLeod, C. M. (1996). Manipulation of
attention at study affects an explicit but not an implicit test of
memory. Consciousness and Cognition, 5, 165-175.
Thapar, A., & Greene, R. L. (1994). Effects of level of processing
on implicit and explicit tasks. Journal of Experimental Psychol-
ogy: Learning, Memory, and Cognition, 20, 671-679.
Thriving, E., & Thomson, D. M. (1973). Encoding specificity and
retrieval processes in episodic memory. Psychological Review,
80, 352-373.
Vaidya, C. J., Gabrieli, J. D. E., Demb, J. B., Keane, M. M., &
Wetzel, L. C. (in press). Impaired priming on the general
knowledge task in amnesia. Neuropsychology.
Vaidya, C. J., Gabrieli, J. D. E., Keane, M. M., Monti, L. A.,
Gutierrez-Rivas, H., & Zarella, M. M. (1996). Evidence for
multiple mechanisms of conceptual priming on implicit memory
tests. Manuscript submitted for publication.
Vriezen, E., Moscovitch, M., & Bellos, S. A. (1995). Priming
effects in semantic classification tasks. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 21, 933-946.
Weldon, M. S. (1991). Mechanisms underlying priming on percep-
tual tests. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 17, 526-541.
Weldon, M. S., & Coyote, K. C. (1996). Failure to find the picture
superiority effect in implicit conceptual memory tests. Journal of
Experimental Psychology: Learning, Memory, and Cognition,
22, 670-686.
Weldon, M. S., & Jackson-Barrett, J. L. (1993). Why do pictures
produce priming on the word-fragment completion test? A study
of encoding and retrieval factors. Memory <fe Cognition, 21,
519-528.
Weldon, M. S., Roediger, H. L., m , Beitel, D. A., & Johnston, T. R.
(1995). Perceptual and conceptual processes in implicit and
explicit tests with picture fragment and word fragment cues.
Journal of Memory and Language, 34, 268-285.
Wood, N. L., Stadler, M. A., & Cowan, N. (in press). Is there
implicit memory without attention? A re-examination of task
demands in Eich's (1984) procedure. Memory & Cognition.
Received December 4,1996
Revision received March 19,1997
Accepted March 28,1997 •