Blog 2. Choice A.
Memory Process. Levels of Processing.
Your Name:___________________________________________
1. Give full reference of the article in the APA format.
2. Complete the experiment and report your results:
a. Letter Recognition _______% correct
b. Rhyme Recognition _______%correct
c. Synonym Recognition _______% correct
d. Lure Recognition _______% correct
**Attach actual results to the printout copy (just the cumulative results table)
3. Questions to answer:
a. In this demonstration, how are you asked to evaluate words to induce a shallow level of processing? How are you asked to evaluate words to induce a deep level of processing? What are “lures” in the context of this demonstration?
b. What is incidental learning? How do researchers typically study incidental learning?
c. What other methods, other than the one used in this demonstration, could be used to induce a deep level of processing for a given word?
4. Create glossary for 4 terms from the original study article and at the end give reference to where your definitions are taken from.
Example:
Automaticity is the ability to do things without occupying the mind with the low level details required. It is usually the result of learning, repetition, and practice.
Reference:
Psychology Wikia (n.d.). Retrieved on June 1, 2015 from http://psychology.wikia.com/wiki/Automaticity
a.
b.
c.
d.
5. Search for a media file that you think is relevant to this study. Describe the media, justify your choice of relevance, and provide full APA reference of the media file. For correct referencing refer to APA materials or the www.apa.org website.
The origins of levels-of-processing effects
in a conceptual test:
Evidence for automatic influences of memory
from the process-dissociation procedure
DAFNA BERGERBEST and YONATAN GOSHEN-GOTTSTEIN
Tel-Aviv University, Ramat Aviv, Israel
In three experiments, we explored automatic influences of memory in a conceptual memory task, as affected
by a levels-of-processing (LoP) manipulation. We also explored the origins of the LoP effect by
examining whether the effect emerged only when participants in the shallow condition truncated the
perceptual processing (the lexical-processing hypothesis) or even when the entire word was encoded
in this condition (the conceptual-processing hypothesis). Using the process-dissociation procedure
and an implicit association-generation task, we found that the deep encoding condition yielded higher
estimates of automatic influences than the shallow condition. In support of the conceptual processing
hypothesis, the LoP effect was found even when the shallow task did not lead to truncated processing
of the lexical units. We suggest that encoding for meaning is a prerequisite for automatic processing on
conceptual tests of memory.
AUTOMATIC PROCESSES IN CONCEPTUAL TESTS 1253
Evidence for the contribution of automatic influences
of memory to conceptual tests comes primarily from the
behavioral pattern of amnesic patients (e.g., exemplar
generation: Graf, Shimamura, & Squire, 1985; Keane,
Gabrieli, Monti, Cantor, & Noland, 1993; Keane et al.,
1997; but see Blaxton, 1992; association generation: Carlesimo,
1994; Vaidya, Gabrieli, Keane, & Monti, 1995)
and from task dissociationsin studies that satisfied Schacter,
Bowers, and Booker’s (1989) retrieval-intentionality
criterion (Culp & Rajaram, 1999; Goshen-Gottstein &
Kempinsky, 2001; McDermott & Roediger, 1996; Mulligan,
1997; Vaidya & Gabrieli, 2000; Vaidya et al., 1997;
Weldon & Coyote, 1996; Zeelenberg, Pecher, Shiffrin, &
Raaijmakers, in press; Zeelenberg, Shiffrin, & Raaijmakers,
1999).
In this article, we wished to provide convergingevidence
for the idea that automatic processes mediate performance
on conceptual tests by applying the PD procedure to the
association-generation task. The PD procedure (Jacoby,
1991) is a technique that yields estimates of controlled
and of automaticinfluences of memory by comparing two
test conditions. In the inclusion condition, controlled and
automatic influences work in concert to facilitate performance.
In the present study, participants were provided
with test cues (e.g., MILK) and were asked to generate an
associated studied word (e.g., COOKIES) or, failing to do
so, to produce the first association that comes to mind.
Hence, studied words were produced either because they
were consciously recollected or because they came to
mind automatically.
In contrast,in the exclusion condition,the two influences
of memory work in opposition to each other—controlled
influenceswork to minimize particular responses, whereas
automaticinfluenceswork to promotethese same responses.
In the present study, participantswere asked to provide associated
words that were not presented earlier. In this condition,
automatic influences of memory promoted respondingwith
studiedwords, whereas controlledinfluences
minimized responding with these words.
Performance in the exclusion condition underestimates
the contribution of automatic influences of memory because
some of the words that were automatically retrieved
may have also been consciously recollected and were,
therefore, excluded (Jacoby, 1991). To correct for this underestimation,
Jacoby translated performance in the two
test conditions into a set of simple equations.
Specifically, assuming that consciously controlled and
automatic influences are independent (for a sample of debates
on this issue, see Bodner, Masson, & Caldwell,
2000;Curran & Hintzman,1997;Hintzman& Curran, 1997;
Jacoby, Begg, & Toth, 1997; Jacoby & Shrout, 1997; Jacoby,
Yonelinas, & Jennings, 1996; Jones, 1987; Joordens
& Merikle, 1993; Reingold & Toth, 1996; RichardsonKlavehn,
Gardiner, & Java, 1996), the probability that a
studied word will be reported in the inclusion (I) condition
can be estimated by the probability of controlled recollection
of the item (C) plus the probability of the word
automatically (A) coming to mind when controlled recollection
fails [A (1 2 C )]; that is, I = C 1 A(1 2 C ). For
the exclusion condition (E), a studied word will be reported
only if it is not consciously recollected yet comes
to mind automatically;that is,
E = A(1 2 C).
The probability of controlled recollection can be estimated
as the difference in the probability of responding
with studied words in the inclusion and exclusion conditions;
that is, C = I 2 E. Automatic influences may then
be computed as A = E / (1 2 C). Jacoby (1991) noted that
the estimate of automatic influences (A) reflects a contribution
of both automatic influences (M) and the baseline
probability (B) of providing a word without having seeing
it in the study phase. Jacoby (1991) assumed that M and B
are additive (A = M 1 B) and took as evidence for automatic
influences an estimate of A being higher than baseline
(but see Wainwright & Reingold, 1996).
The PD Procedure and the Locus of
Levels-of-Processing (LoP) Effects
Because the PD procedure can separate the contribution
of automatic and consciously controlled processes to
performance, it provides an opportunity to identify the
locus of different experimental effects. In particular, using
the PD procedure, one can ask whether an experimental
effect is mediated by automatic processes or whether this
effect is merely a by-product of consciously controlled
contamination. To illustrate, Toth et al. (1994, Experiment
1) asked why directing participants’ attention to the
meaning of studied words during encoding (i.e., the deepencoding
condition)led to better performance than did directing
their attentionto the perceptual form of these same
words (i.e., the shallow-encoding condition) even on implicit
perceptual tests (Challis & Brodbeck, 1992; Thapar
& Greene, 1994; for a review, see Brown & Mitchell,1994).
These LoP effects suggest that encoding of the meaning of
words can affect automatic influences on implicit tests
much as it affects controlled influences on explicit tests
(e.g., Craik & Lockhart, 1972).
By applyingthe PD procedure,Toth et al. (1994) showed
that the LoP manipulation affected the estimate of controlled
influences but that it did not affect the estimate of
automaticinfluences. Therefore, the effects of LoP on perceptual
implicittests, which could falsely be interpreted as
genuine automatic effects, are more likely the by-product
of contamination by conscious processes (but see Bodner
et al., 2000).
In the present study, we extended the Toth et al. (1994)
design to explore conceptualtests. Thus, we asked whether
the LoP effects that are often reported in implicit conceptual
tests are only a by-product of consciouscontamination
or whether these effects also reflect a genuine contribution
of automatic influences of memory. We hypothesized that,
LoP effects on conceptual,in contrast with perceptual,tests
of memory reflect, at least in part, a genuine contribution
of automatic influences. Hence, we predicted that an LoP
effect would be found even on the automatic estimates.
1254 BERGERBEST AND GOSHEN-GOTTSTEIN
Our prediction was based on at least 3 reasons: First, the
hypothesisthat LoP effects on conceptual tests of memory
represent a genuine automatic influence is in harmony
with the finding that dividing attention during encoding
resulted in lower estimates of automatic influences relative
to a full-attention condition in the category-exemplar
generation task. The magnitude of the estimate of automatic
influences decreased when attention was divided at
encoding (Schmitter-Edgecombe, 1999; Experiment 2),
ostensibly because dividing attention at encoding reduced
the amount of conceptual processing (Mulligan & Hartman,
1996).
Second, unlike LoP effects on perceptual tests, LoP effects
on measures of conceptual priming are quite robust.
Such effects were reported on the association-generation
task (e.g.,Weldon & Coyote, 1996; but see Nelson, Benett,
& Xu, 1997, Experiment 3), the category-exemplar generation
task (e.g., McDermott & Roediger, 1996; Mulligan,
Guyer, & Beland, 1999; Srinivas & Roediger, 1990;
Vaidya et al., 1997;Weldon & Coyote, 1996), the generalknowledge
question task (e.g., Hamann, 1990; Thapar &
Greene, 1994), the sentence-completion task (GoshenGottstein
& Peres, 1998), and in amnesic patients (Keane
et al., 1997).
Finally, according to the transfer-appropriate processing
(TAP) framework (e.g., Roediger et al., 1989), recapitulation
of conceptual processing enhances performance on
conceptualtasks. BecausetheLoP manipulationis presumed
to affect processing of meaning (but see RichardsonKlavehn
& Gardiner, 1998) ostensibly leading to differences
in conceptual processing, we expected more recapitulation
of conceptual processing in the deep-encoding
condition than in the shallow condition.
To summarize, the primary goal of this study was to
seek evidence for automatic influences of memory on a
conceptual task, and, if we found such evidence, we further
wished to test whether the LoP manipulation would
affect these automatic influences. To this end, we manipulated
LoP during encoding and then applied the PD procedure
to the conceptual association-generation task (Experiment
2). To anticipate our results, we found an
automatic contributionto the LoP effect. For sake of comparison,
we also tested the effect of the LoP manipulation
on the implicit association-generation task (Experiment
1). Finally, Experiment 3 was designed to better
comprehend the nature of the automatic contribution to
the LoP effect.
EXPERIMENT 1
The purpose of Experiment 1 was to replicate the LoP
effect in the implicit test of association-generation using
LoP instructions identical to those used by Weldon and
Coyote (1996) in their association-generationexperiment.
Method
Participants. Twenty-four introductory psychology students participated
in the experiment to fulfill course requirements. All had
normal or corrected-to-normal vision.
Design and Materials. Encoding condition (deep, shallow, or
unstudied) was manipulated within subjects. In a pilot study, 120 associatively
related cue–target word pairs were selected in the following
manner. A preliminary list of 360 cue words was presented
to 53 participants who were asked to write down, for each cue, the
first association that came to mind (Bergerbest & Goshen-Gottstein,
1999b). The 120 word pairs were selected so that the cue word would
elicit the target word by approximately 30% of the participants.
These 120 pairs were presented in the test list.
The experimental condition of the cues was defined by the status
of their respective targets. To this end, the 120 word pairs were randomly
divided into three lists of 40 pairs, each to be allocated to one
of the encoding conditions. For each session, targets from two of the
lists were included in the study list, corresponding to either the deep
or the shallow encoding condition, and targets from the third list were
not studied. The three lists were counterbalanced so that each participant
was presented with an equal number of targets in the three
encoding conditions and that, across participants, each target would
be allocated an equal number of times to each of the three encoding
conditions. Two buffer words were added at the beginning and two
at the end of the study list, producing a list of 84 words. Ten additional
words, which did not match any of the target words and were
not associated with any of the cue words, were used for practicing
the two encoding tasks.
Procedure. Individually tested participants were told that they
would be shown a list of words and were asked to make one judgment
per word by pressing a key on the computer keyboard. For each
word in the deep-encoding condition, participants were required to
rate the pleasantness of the meaning on a 5-point scale. In the shallowencoding
condition, they were required to count the number of vowels
per word.
Participants were informed that the two tasks would be presented
in a random order and that task instructions would appear above each
of the presented words. No more than four consecutive words appeared
with the same instruction.
Each participant received 10 practice trials. Every trial began with
a warning signal in the middle of a Macintosh computer screen for
0.5 sec. The warning signal was replaced by the target word and task
instructions determining the relevant task (“pleasantness,” “vowels”).
Task instructions appeared 2 cm above the target word. Target
words were presented for 3 sec with the next trial beginning 1 sec later.
The 120 test cues were presented in random order, one cue at a
time, and participants were asked to generate for each cue the first
association that came to mind. The study and test phases lasted 6
and 30 min, respectively. Both study targets and test cues were presented
in Gilboa font (8 mm high, 6 mm wide).
After test, participants were asked whether they had noticed any
connection between the first and the second phases of the experiment
and whether they had tried to intentionally recall studied words
to provide associations on the test phase.
Results and Discussion
For each participant, the proportion of target responses
that were generated in each encoding condition was comTable
1
Experiment 1: Proportions (Ps) and Standard Errors
of Target Words Generated in the Three Encoding Conditions
and Priming Scores, Computed by Subtracting Performance
in the Unstudied Conditions From Performance in the
Deep and Shallow Encoding Conditions
Encoding Conditions
Deep Shallow Unstudied
P SE P SE P SE
Proportion of targets .37 .03 .31 .02 .28 .02
Priming .09 .02 .03 .02
AUTOMATIC PROCESSES IN CONCEPTUAL TESTS 1255
puted. Priming scores were calculated, for both the deep
and the shallow encodingconditions,by subtractingthe proportion
of unstudied targets from that of studied targets.
Table 1 presents the proportion of target words under each
encoding conditionand the corresponding priming scores.
Examination of participants’ responses revealed that
targets were most likely to be generated in the deepencodingcondition,lesslikelyto
be generatedin the shallowencoding
condition, and least likely to be generated in the
unstudied condition. A one-way analysis of variance
(ANOVA) revealed that, indeed, the effect of encoding
condition was significant [F(2,22) = 24.92, MSe = .17,
p , .0001]. Post-hoc analysis (Levin, Serlin, & Seaman,
1994) revealed that there was significant priming in both
the deep-encoding condition [t(23) = 4.61, p = .0001] and
the shallow condition [t(23) = 1.84, p = .039, one tailed].
Most importantly, the LoP effect was significant [t(23) =
2.96, p = .007]. The results replicate Weldon and Coyote’s
(1996) finding of an LoP effect on the implicit association-generationtask
using the identical set of instructions.
There are two reasons to believe that automatic influences
of memory contributed to this LoP effect. First, responses
to the posttest questionnaire revealed that not
even a single participant reported trying to intentionally
recall studied words to provide associations in the test
phase. Therefore, all the findings of this experiment represent
responses of participants who, according to their
subjective reports, did not try to explicitly recollect studied
items. Second, and more important, even when analyzing
the responses of participants who were unaware
during the test phase that they had generated previously
presented associations (i.e., Bowers & Schacter, 1990), we
still found a significant LoP effect [t(10) = 2.08, p = .032,
one tailed].
Still, because the test phase immediately followed the
study phase, and because it included a high percent (66%)
of cues that were associated to studied words, the possibility
of conscious contamination cannot be ruled out.
Therefore, in Experiment 2, we applied the PD procedure
to the association-generation task to eliminate the possibility
of conscious contamination.
EXPERIMENT 2
The purpose of Experiment 2 was to use the PD procedure
to provide converging evidence for the idea that automatic
processes mediate performance in the associationgeneration
task and that an LoP effect can be found on
these processes. During study, participants studied a list of
words under deep- or under shallow-encoding conditions.
During test, participants were provided with cues and
were asked to produce associations,under either inclusion
or exclusion instructions.
Method
Participants. Sixty introductory psychology students, all with
normal or corrected-to-normal vision, took part to fulfill course requirements.
None had participated in Experiment 1. Because exclusion
scores of zero (no target generation) result in an underestimation
of the automatic contribution to performance [because A = E/
(1 2 C)] (e.g., Jacoby, 1996), an additional 7 participants were
tested to replace participants with exclusion scores of zero, as advocated
by Jacoby and colleagues. Replacement of the participants
did not affect the pattern of results.
Design and Materials. Encoding condition (deep, shallow, unstudied)
and test condition (inclusion, exclusion) were manipulated
within subjects. The materials were identical to those used in Experiment
1.
Procedure. The study phase was identical to that of Experiment
1. During test, which lasted approximately 40 min, participants
were told that their memory would be tested for the words that they
had studied. They were informed that they were to see a list of words,
one word at a time, accompanied by two types of retrieval instructions.
When presented with the instruction “old” (i.e., inclusion),
they were to say a studied word that was associated with the cue
word. If they could not recollect a studied word, they were to say the
first associated word that came to mind. If the instruction “new” appeared
(i.e., exclusion), participants were required to say aloud the
first associated word that came to mind but to exclude words they
recollected as having appeared in the study phase, replacing them
with another word that came to mind.
To ensure that participants understood the instructions, they practiced
the two tasks. Then, the 120 cue words were presented in a different
random order for each participant. Half of the randomly chosen
cues, corresponding to target words from each of the three encoding
conditions (i.e., 20 cues), appeared in the inclusion condition and the
remaining half appeared in the exclusion condition. Across participants,
each cue was presented equally often in the two test conditions.
The “old” and “new” instructions appeared in random order with
no more than four consecutive cues appearing with the same task instruction.
Random presentation of the two test conditions increased
the likelihood that controlled recollection was the same for the inclusion
and exclusion conditions (see Jacoby, 1998).
Results and Discussion
Table 2 presents the proportions of target words generated
under each experimental condition and the estimates
derived from the PD procedure equations. In this and the
subsequent experiment, only the analysis of the estimates
of controlled and automatic influences will be reported
because it is statistically redundant to analyze performance
in the inclusion and exclusion conditionsas well as
the pattern of the derived estimates (see Jacoby, 1996).
Examination of participants’ performance revealed that
in the inclusion condition,participantswere more likely to
Table 2
Experiment 2: Proportions (Ps) and Standard Errors of Target
Words Generated as a Function of the Three Encoding
Conditions and the Two Test Conditions, and Estimates
of Controlled and Automatic Processes
Encoding Conditions
Deep Shallow Unstudied
P SE P SE P SE
Test Conditions
Inclusion .58 .02 .33 .02 .23 .01
Exclusion .19 .01 .26 .01 .27 .01
Estimates
Controlled .38 .03 .07 .02
Automatic .32 .02 .28 .01 .25 .01
Automatic-Baseline .07 .02 .03 .01
1256 BERGERBEST AND GOSHEN-GOTTSTEIN
generate targets following deep than following shallow
encoding, and that in both encoding conditions more target
words were generated than in the unstudied condition.
In contrast, in the exclusion condition, participants were
more likely to generate unstudied than studied words and
were more likely to generate studied words after shallow
encoding than after deep encoding.
Analysis of the PD estimates revealed that deep encoding
produced more controlled retrieval than did shallow
encoding [t(59) = 7.31, p , .0001]. More important, the
estimate of automatic influences was higher than baseline
performance, in both deep- [t(59) = 3.39, p = .001] and
shallow- [t(59) = 1.99, p = .025, one tailed] encoding conditions.
Thus, evidence for automatic influences of memory
was found with the PD procedure when applied to a
conceptual test of memory. Finally, the estimates of automatic
influences revealed that deep encoding produced
significantly higher automaticinfluencesthan did shallow
encoding [t(59) = 1.68, p = .049, one tailed]. This finding
of an LoP effect on the estimates of automatic influences
converges with the finding of a LoP effect on the implicit
test in Experiment 1.
Next, we compared baseline performance in the inclusion
and exclusion conditions to examine whether there
was evidence for a change in response strategies across
the two tasks. This analysis revealed that the baseline in
the inclusion condition was significantly lower than the
baseline in the exclusion condition [t(59) = 2.74, p =
.008]. This difference may have been the result of participants’
tendency, during the inclusion condition, to provide
a studied word even if it was not related to the cue.1
Thus, participants may have reported studied words even
when responding to a baseline cue. This would have reduced
the likelihood of generating the appropriate (unstudied)
target for the baseline cues. In contrast, in the exclusion
condition, participants may have tried to avoid
providing studied words, and so may have been more
likely than in the inclusion conditionto report the unstudied
associates as response to the relevant cues. Indeed, we
found that associates that were intended to serve as cues
to generate unstudied targets were used to generate studied
targetsmore often in the inclusioncondition(M = 0.068,
SE = 0.008) than in the exclusion condition (M = 0.011,
SE = 0.002) [t(59) = 7.12, p , .001].
Toth et al. (1994) suggested that higher baselines in the
exclusion, relative to the inclusion, condition were a signature
for a generate-recognize strategy (but see Bodner
et al., 2000, who questioned the validity of this signature
by demonstratingthat it did not emerge even when test instructions
explicitly guided participantsto use a generaterecognize
strategy). Although our baselines showed an
opposite pattern, we still wished to correct for the differences
in baselines by estimating the conscious and automatic
influences. To do so, we used Wainwright and Reingold’s
(1996) correction equations.
Wainwright and Reingold’s corrections for differences
between baselines. Several attempts have been
made to derive measures of automaticinfluences of memory
that correct for unequal performance in baseline conditions.
Of all the proposals, Wainwright and Reingold’s
(1996) remains the most comprehensive. These authors
presented seven different models for correcting unequal
performance in the baseline conditionsand were careful to
articulate the underlying assumption behind each model.
We now describe the estimates that were derived from the
three models that assume independence between controlled
and automatic processes. In the General Discussion,
we argue that if this assumption is incorrect, and participants
used a generate-recognize strategy, then our
findings of an LoP effect would only be enhanced.
The three independence models differ with regard to
their assumptions of the relation between a guessing parameter
(which is estimated by performance in the baseline
conditions) and the controlled and automatic influences
(see equations and underlying assumptions of the
three models in the Appendix).Reanalysis of the results of
Experiment 2 following the three guessing models is presented
in Table 3.
The estimates of automatic influences derived from the
HITS 2 FA model yield negative values for the automatic
influences. Because the true automatic contribution must
be either nil (A = 0) or positive, this model obviously underestimates
the automatic contribution (probably more
in the deep condition than in the shallow condition; see
General Discussion). Because the magnitude of this underestimation
is unknown, this model turns out to be uninformative.
In contrast, the results of the other two models,
the independent-guessing model (which “yields
corrected estimates that are numerically identical to the
Buchner, Erdfelder, & Vaterrodt-Plunnecke[1995] model”;
Wainwright & Reingold, 1996, p. 241) and the additive
model, yield results that are both reasonable and consistent
with each other. Both the independent-guessing
model [t(59) = 2.77, p , .01] and the additive model
[t(59) = 3.05, p , .01] provide support for automatic influences
of memory under deep, but not under shallow
Table 3
Experiment 2: Estimates (Es) and Standard Errors
of Controlled and Automatic Processes According
to the Different Correction Methods Described
by Wainwright and Reingold (1996)
Encoding Conditions
Deep Shallow
Correction Method E SE E SE
Hits 2 FA
Controlled .43 .04 .11 .03
Automatic 2.16 .15 2.02 .02
Independent guessing
Controlled .39 .04 .10 .03
Automatic .08 .03 .01 .02
Additive
Controlled .39 .04 .09 .03
Automatic .06 .02 .01 .01
Note—All three models assume independence between controlled and
automatic influences of memory.
AUTOMATIC PROCESSES IN CONCEPTUAL TESTS 1257
(ts , 1), encoding, and a significant LoP effect for the automatic
influences of memory [t(59) = 1.78, p = .039, one
tailed, and t(59) = 1.84, p = .035, one tailed, respectively].
To summarize, evidence for automatic influences of
memory and for an LoP effect on these influences was
found both with Jacoby’s (1991) original PD equations
and when the corrected estimates were derived from
Wainwright and Reingold’s (1996) models.
EXPERIMENT 3
To increase the probability that controlled and automatic
processes are independent, as required by the PD
equations, Jacoby and colleagues have recently suggested
using a different version of exclusion instructions (see Jacoby,
1998; also see Bodner et al., 2000). One goal of this
experiment was to extend our findings to the new version
of the instructions.2 Thus, we asked participants in the exclusion
condition to use each test cue as a cue to recall a
studied associated word and only then to replace the studied
word with another, unstudied,associatively-related word.
An even more important goal of this experiment was to
critically examine our interpretation of the LoP effect in
Experiment 2. We interpreted this effect as stemming
from the reduced encoding-of-meaningof the words in the
shallow condition (vowel counting) relative to the deep
condition ( pleasantness task), which lessened the automatic
influences of memory for these words. However, the
LoP effect could be interpreted as arising from either of
two sources.
First, as we have suggested so far, participants in the
shallow condition did not elaborately encode the meaning
of the words (henceforth, the “conceptual-processing hypothesis”;
Richardson-Klavehn & Gardiner, 1998), and
the LoP effect may have resulted from the difference in
the amount of encoding-for-meaning that words underwent
in the shallow and deep conditions. Second, in the
shallow condition,participantsmay not have even encoded
the lexical/perceptual units with which they were presented
(the “lexical-processing hypothesis”; RichardsonKlavehn
& Gardiner, 1998). That is, participants in the
shallow conditionmay have set themselves to restrict perceptual
processing by, for example, only checking for
vowels, thereby truncating the perceptual analysis of the
stimuli (Richardson-Klavehn & Gardiner, 1998; Thapar &
Greene, 1994; for an example of truncated processing in
nonverbal stimuli, see Goshen-Gottstein & Ganel, 2000).
If this is so, then the Experiment 2 LoP effect may have
been mediated by the difference between words for which
the lexical units were accessed (deep condition) relative to
words for which the lexical units were not accessed (shallow
condition), rather than from a difference in the encodingfor-meaning.
According to the conceptual-processing hypothesis,
LoP effects are the product of differential encoding for
meaning. Hence, if a shallow-encoding task was used in
which the lexical units would be processed in their entirety
but the encoding-for-meaning would remain minimal,
then LoP effects should still be observed. According to the
lexical processing hypothesis, however, the source of LoP
effects in Experiments 1 and 2 was the truncated lexical
processing of words in the shallow condition. Therefore,
if a shallow-encodingtask were used in which the lexical
units would be processed in their entirety, the LoP effect
would be eliminated.
Critically, in the posttest interview of Experiment 1,
several participants remarked that as they were counting
the vowels they did not always notice the actual word with
which they were presented. Therefore, our finding of an
LoP effect on the automatic estimates might have been the
product of truncated lexical processing of words in the
shallow condition rather than of reduced processing-formeaning
of these words.
With regard to perceptual implicit tests, several studies
have attempted to distinguish between the conceptualprocessing
and the lexical-processing hypotheses for LoP
effects. For example, using the stem-completion task,
Richardson-Klavehnand Gardiner (1998; see also Challis,
Velichkovsky, & Craik, 1996) found a LoP effect when
performance in the semantic condition (i.e., rating pleasantness
of meaning)was compared with that of a graphemic
condition (i.e., counting enclosed spaces in letters), which
ostensibly leads to truncated lexical processing. However,
an LoP effect was not found when performance in the semantic
condition was compared with that of a phonemic
condition (i.e., counting syllables), which ostensibly allows
for processing of the entire lexical unit but with minimal
processing of the words’ meaning.
These results support the lexical processing hypothesis,
according to which direct participants’ attentionto the letter
level, as in counting enclosed spaces, can reduce the
lexical processing of the studied words and hence produce
the observed LoP effect. Although the lexical processing
hypothesis seems to account for LoP effects in at least
some perceptual tests, it may not account for LoP effects
in conceptual tests, where encoding-for-meaning should
play a more critical role.
In Experiment 3, we distinguished between the two hypotheses
by replacingthe vowel-countingtask used in Experiment
2 with a syllable-counting task. We chose the
syllable-counting task because it seems to demand less
processing of the words’ meaning than that required in the
pleasantness-ratingtask. Yet, as suggested by the results in
the phonemic condition of Richardson-Klavehn and Gardiner
(1998), encoding in this task seems to be of the entire
lexical unit. Will a LoP effect emerge despite the processing
of the entire lexical unit under shallow encoding?
Method
Forty-eight introductory psychology students, none of whom had
participated in the previous experiments, participated in the experiment
to fulfill course requirements. Five additional participants were
run to replace participants with exclusion scores of zero. As in Experiment
2, this did not affect the pattern of results.
The method was identical to that of Experiment 2 except that, in
the shallow-encoding condition, participants were asked to count the
number of syllables per word. Also, during the exclusion trials, par-
1258 BERGERBEST AND GOSHEN-GOTTSTEIN
ticipants were asked to use each test cue as a cue to recall a studied
associated word and only then to replace the studied word with an
unstudied associatively related word. Participants were told to provide
the first association that came to mind if they could not recollect
a related studied word.
Results and Discussion
The results were scored as in Experiment 2 and are presented
in Table 4. The general pattern of results was similar
to that in Experiment 2. However, in contrast to Experiment
2, the difference between baselines was not
significant [t(47) = 1.60, p . .1].
As in Experiment 2, deep encoding resulted in higher
consciously controlled estimates than did shallow encoding
[t(47) = 12.15, p , .0001]. More importantly, the estimate
of automatic influences were higher than baseline
performance following deep encoding [t(47) = 2.61, p =
.012]. Thus, Experiment 3 provides a replication of the
Experiment 2 finding of automatic influences of memory
followingthe deep-encoding condition on the associationgeneration
task. The estimate of automatic influences following
shallow encoding did not differ significantly from
baseline [t(47) , 1]. The interpretation of this null effect
is discussed in the General Discussion. For now, suffice it
to say that the absence of automaticinfluences of memory
following shallow encoding was consistent with the absence
found using Wainwright and Reingold’s (1996)
equations, which corrected for unequal baselines (Experiment
2).
Most importantly, an examination of the LoP effect revealed
a significantly larger estimate of automatic influences
in the deep-encoding condition than in the shallowencoding
condition [t(47) = 1.93, p = .03, one tailed]. This
LoP effect replicates that of Experiment 2 where the
pleasantness-rating task was contrasted with the vowelcounting
task. Because the syllable-countingtask used in
our experiment presumably does not induce truncatedlexical
processing, the ensuing LoP effect supports the notion
that this effect was based on reduced processing-ofmeaning
in the shallow condition rather than on the
absence of lexical processing.
Correcting for differences between baselines. The
baselines in Experiment 3 did not differ significantly. Still,
as in Experiment 2, we reanalyzed the results accordingto
the three guessing-correction models that were suggested
by Wainwright and Reingold (1996); the results are presented
in Table 5.
As in Experiment 2, the HITS 2 FA model yielded negative
values for the estimates of automatic influences and
was therefore rejected. In contrast, the two other models,
the independent-guessing model and the additive model,
yielded results that are both reasonable and consistent
with each other. Both the independent-guessing model
[t(47) = 1.67, p = .05, one tailed] and the additive model
[t(47) = 2.11, p = .04] provided support for automatic influences
of memory in the deep-encoding condition but
not the shallow condition (ts , 1). Moreover, in both models,
the LoP effect for automatic influences was significant
[independent-guessing model, t(47) = 1.80, p = .039,
one tailed; additive model, t(47) = 1.89, p = .033, one
tailed].
Thus, as in Experiment 2, evidence for automatic influences
of memory and for an LoP effect on these influences
was found both with Jacoby’s original PD equations
and when the corrected estimates were derived from
Wainwright and Reingold’s (1996) models.
GENERAL DISCUSSION
In the present study, we addressed three questions. First,
do automaticinfluences of memory affect performance on
the association-generationtask? Second,if they do, can such
processes be affected by an LoP manipulation? Third, are
the LoP effects, if found, due to differences in lexical processing
between the shallow and deep conditions or due to
differencesin the relativeamountof processing-of-meaning?
In Experiment 1, we found a conceptual priming effect,
which was sensitive to LoP. This effect was found even
when one considers only the performance of test-unaware
participants. Using the Experiment 1 stimuli, we applied
the PD procedure to the association-generationtask in Experiment
2 and, in reply to the first question, found evidence
for automatic influences of memory.
Table 4
Experiment 3: Proportions (Ps) and Standard Errors
of Target Words Generated as a Function of the Three
Encoding Conditions and the Two Test Conditions,
and Estimates of Controlled and Automatic Processes
Encoding Conditions
Deep Shallow Unstudied
P SE P SE P SE
Test Conditions
Inclusion .63 .02 .35 .02 .24 .02
Exclusion .18 .02 .25 .01 .27 .02
Estimates
Controlled .45 .03 .09 .02
Automatic .32 .02 .27 .01 .25 .01
Automatic-Baseline .07 .02 .02 .01
Table 5
Experiment 3: Estimates (Es) and Standard Errors
of Controlled and Automatic Processes According
to the Different Correction Methods Described
by Wainwright and Reingold (1996)
Encoding Conditions
Deep Shallow
Correction Method E SE E SE
Hits 2 FA
Controlled .48 .04 .13 .03
Automatic 2.30 .23 2.10 .04
Independent guessing
Controlled .45 .03 .11 .03
Automatic .05 .03 2.01 .02
Additive
Controlled .45 .03 .10 .03
Automatic .05 .02 .01 .02
Note—All three models assume independence between controlled and
automatic influences of memory.
AUTOMATIC PROCESSES IN CONCEPTUAL TESTS 1259
In response to the second question, we found that the
automatic influences were greater following deep than
following shallow encoding. Thus, results of the PD procedure
converged with those of conceptual priming. In the
domain of perceptual tests, Toth et al.’s (1994) suggested
that LoP effects were a by-product of contamination by
consciouslycontrolled processes. In contrast, in our study,
when a conceptual test was used, the LoP effects showed
up both on the measure of conceptual priming and on the
estimates of automatic influences of memory as derived
from the PD procedure. Therefore, the LoP effect that was
found in conceptual priming seems to be not merely a byproduct
of conscious contamination.
The finding of converging evidence of conceptual priming
with the PD estimates leads to the conclusion that,
at least for the association-generation task, the measures
of conceptual priming may not have been subject to much
contamination by controlled processes (also see McBride
& Shoudel, in press; Nelson et al., 1997; SchmitterEdgecombe,
1999). The suggestion that conscious contamination
was not responsible for the LoP effects in Experiment
1 conforms with the finding that LoP effects
were found with both test-aware and test-unaware participants
(see Bowers & Schacter, 1990). This seems most
reasonable for tests such as association generation, which
neither place any constraints on the answers generated by
participants nor require any “correct answers.” Participants
are least likely to be motivated to try to recollect words
from an earlier study phase to assist their performance in
this task as opposed to other conceptual tasks. Providing
the first association that automatically comes to mind
seems to demand very little—and may indeed be performed
without—conscious control.
The Lexical- Versus the
Conceptual-Processing Hypothesis
Our third question, whether the LoP effect resulted
from differences in lexical processing or from differences
in conceptual processing, was addressed in Experiment 3.
To this end, we replaced the vowel-counting (shallow)
task, which ostensiblyleads to truncatedlexical processing,
with the syllable-countingtask, which allows the processing
of entire lexical units. As predicted by the conceptualprocessing
hypothesis, LoP effects were still found on the
estimates of automatic influences of memory derived from
the PD procedure (Experiment 3). Thus, although words
in both the shallow- and the deep-encoding tasks were
encoded in their entirety, the additional processing-ofmeaning
that words underwent in the deep-encoding condition
produced higher levels of automatic influences of
memory.
Our results establish that adding conceptual information
above and beyond the activation of the lexical unit enhances
automatic influences. This contrasts with the finding
of Richardson-Klavehn and Gardiner (1998), who
found that once the entire lexical unit was accessed, priming
did not benefit from additionalprocessing-of-meaning.
The different results are easily explained by considering
the different requirements that are made by the memory
tests in the two studies. In Richardson-Klavehn and Gardiner’s
(1998) stem-completion study, a perceptual test
was used, in which successful performance presumably
depends on the recapitulation of perceptual processing
(Roediger et al., 1989;Masson & MacLeod, 2002). Therefore,
because materials presumably underwent similar
perceptual processing under the pleasantness-rating (semantic)
task and under the syllable-counting (phonemic)
task, it should not be surprising that the magnitude of perceptual
priming in these tasks was similar.
In the present study, in contrast, a conceptual test was
used, one in which successful performance presumably
depends on the recapitulation of conceptual processing
that studied material undergoes during encoding (e.g.,
Roediger et al., 1989). Because materials undergo more
conceptual processing under the pleasantness-rating task
than under the syllable-countingtask, it makes sense that
the automaticinfluences were larger in the deep-encoding
condition than in the shallow-encoding condition.
Furthermore, the absence of conceptual processing in
the shallow condition did not enable a contribution of automatic
influences (see estimates of these influencesin the
independent-guessingmodel and additive model shown in
Table 3). Even in the syllable-counting task (Experiment
3, Table 5), when participants perceptually encoded
the entire words (as demonstrated by Richardson-Klavehn
& Gardiner, 1998), no evidence for an automatic contribution
was found. This establishes that a minimal amount
of encoding-for-meaning is required if conceptual priming
is to be found. If this amount is not available (i.e., shallow
encoding), priming cannot emerge.
The only other published report of the PD procedure
appliedto the association-generationtask is that of Nelson
et al. (1997). Although these authors found support for automatic
influences of memory, they did not find evidence
that the LoP manipulation affected either performance in
the implicittest or the estimates of automaticinfluences as
derived from the PD procedure. It is unclear what accounts
for the different results. A comparison of associationgeneration
studies that have shown an LoP effect (Vaidya
et al., 1997, Experiment 3; Weldon & Coyote, 1996, Experiment
5; our study) or failed to show one (Nelson et al.,
1997, Experiment 3; Vaidya et al., 1997, Experiment 2)
suggests that certain variables, which seem to play a role
in finding LoP effects in other studies (e.g., whether LoP
was manipulated between blocks or in random order: Thapar
& Greene, 1994; but see Mulligan et al., 1999; and the
strength of cue-to-target associations: Vaidya et al., 1997)
cannot account for the discrepant findings between our
study and that of Nelson et al. For example, the strength of
cue-to-target associations was very similar in our study
and in Nelson et al.’s study, yet LoP affected performance
only in our study. It also seems that the discrepant findings
cannot be resolved by attributingthem to different encoding
tasks in the two studies because the encoding tasks of
Experiments 1 and 2 are very similar to Nelson et al.’s encoding
tasks.
We argue that our data are more reliable because the application
of the PD procedure depends on obtaining intra-
1260 BERGERBEST AND GOSHEN-GOTTSTEIN
experimental baselines under both inclusion and exclusion
conditions (e.g., Jacoby, 1998; Toth et al., 1994).
Nelson et al. (1997), however, acquired baselines from
extra-experimental participants. This does not allow for a
comparison of the automatic estimates with an intraexperimental
baseline measure, does not allow for a comparison
of baseline performance in the inclusion and exclusion
conditionsto assess the validity of the independence
assumption, and provides no means to correct for baseline
differences. Moreover, the results of Nelson et al. describe
an impossible-to-prove null effect and might be the result
of insufficient statistical power.
Does the Conclusion That LoP Affects
Automatic Influences of Memory Depend on the
Validity of the Independence Assumption?
The estimation of automaticinfluences in our study depends
on the independence assumption. Does the conclusion
that LoP affects automatic influences also depend on
this assumption? We suggest that even if participants in
our study used a generate-recognize strategy instead of a
direct-retrieval strategy,3 the conclusion derived from our
findings is still valid.
Several signatures for the use of a generate-recognize
strategy have been proposed. These include reduced baseline
performance in the exclusion condition as compared
to the inclusion condition (e.g., Jacoby, 1998; Jacoby,
Toth, & Yonelinas, 1993; Toth et al., 1994), as well as
paradoxical results on the automatic estimates, such as reduced
estimates of automatic influences following deep
than shallow encoding or automatic estimates that are significantly
below baseline (e.g., Bodner et al., 2000; Curran
& Hintzman, 1995; Jacoby, 1998; Russo, Cullis, &
Parkin, 1998; Toth et al., 1994).
In our study, none of the proposed signatures for a
generate-recognize strategy were found. However, Bodner
et al. (2000) demonstratedthat the generate-recognize signature
of lower exclusion baseline may be absent even
when a “generate-recognize”strategy is used. Thus, the absence
of a “traditional” signature for a generate-recognize
strategy in our study cannot rule out the possibility that
participants did, nevertheless, use such a strategy. Moreover,
in Experiment 2, participants were biased in the inclusion
condition to report studied words even when responding
to a baseline cue, thereby reducing baseline
inclusion performance. This bias may have concealed elevated
inclusion baseline performance—the signature of
a generate-recognize strategy—which may, conceivably,
have emerged if the bias were eliminated.
Nevertheless, Bodner et al. (2000) have suggested that
if a generate-recognize strategy is used by participants, in
which target words come to mind automatically and are
then submitted to a recognition check, then items that
were encoded for meaning (e.g., deep-encoding condition)
would be recognized most easily and would, therefore,
most likely be withheld on exclusion trials. Such encoding
conditions would lead to artificially low exclusion
performance and to an underestimation of automatic influences
(for similar ideas, see Curran & Hintzman, 1995;
Jacoby, 1998; Russo et al., 1998).We argue, therefore, that
even if a generate-recognize strategy had been used in our
study, it would have resulted in an underestimation of the
true automatic influences, particularly following deep encoding.
According to this logic, the real LoP effects on automatic
influences are, most probably, even larger than the
reported results.
Bodner et al.’s (2000) proposal was supported by comparing
stem-completion performance following a study
condition in which participants read a word (the “read”
condition) with performance following a study condition
in which participants read a word and then generated an
associate (the “associate” condition).Because the perceptual
information was equated in the two conditions (i.e.,
participants in both conditions read the word), the associate
condition should lead to estimates of automatic influences
that were at least as large as those in the read
condition.Yet, not only were the estimates of automaticinfluences
in the associate conditionlower than those found
in the read condition, they were even lower than baseline
(for similar results, see Curran & Hintzman, 1995; Jacoby,
1998; Richardson-Klavehn & Gardiner, 1998; also see
Russo & Andrade, 1995; Russo et al., 1998).
Pursuing this logic, even if participants in our study did
use a generate-recognize strategy, it could be argued following
Bodner et al. (2000) that the PD equations would
lead to a more pronounced underestimation of automatic
influencesin the deep-encodingconditionthaninthe shallowencoding
condition. Therefore, the LoP effects reported in
our study on the automatic influences might underestimate
the true LoP effects. In sum, our study has established
that automatic influences of memory on the conceptual
association-generation task are dependent on the earlier
encoding of items, with attention directed to the meaning
of words during study increasing the likelihood that these
words will later come to mind automatically.
REFERENCES
Bergerbest,D., & Goshen-Gottstein, Y. (1999a, November). Applying
the process-dissociation procedure to conceptual tests of memory.
Paper presented at the annual meeting of the Psychonomic Society,
Los Angeles.
Bergerbest, D., & Goshen-Gottst ein, Y. (1999b). Measures of
Hebrew-word association norms and their implications to tests of
memory. Psychologia, 8, 83-96.
Blaxton, T. A. (1992). Dissociations among memory measures in
memory-impaired subjects: Evidence for a processing account of
memory. Memory & Cognition, 20, 549-562.
Bodner,G. E., Masson, M. E., & Caldwell, J. I. (2000). Evidence for
a generate-recognize model of episodic influences on word-stem completion.
Journal of Experimental Psychology: Learning, Memory, &
Cognition, 26, 267-293.
Bowers, J. S., & Schacter, D. L. (1990). Implicit memory and test
awareness. Journal of Experimental Psychology: Learning, Memory, &
Cognition, 16, 404-416.
Brown, A. S., & Mitchell,D. B. (1994).A reevaluation of semantic versus
nonsemantic processing in implicit memory. Memory & Cognition,
22, 533-541.
Buchner,A., Erdfelder,E., & Vaterrodt-Plunnecke,B. (1995).Toward
unbiased measurement of conscious and unconscious memory
AUTOMATIC PROCESSES IN CONCEPTUAL TESTS 1261
processes within the process-dissociation framework. Journal of Experimental
Psychology: General, 124, 137-160.
Carlesimo,G. A. (1994). Perceptual and conceptual primingin amnesic
and alcoholic patients. Neuropsychologia, 32, 903-921.
Challis, B. H., & Brodbeck, D. R. (1992). Level of processing affects
priming in word fragment completion. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 18, 595-607.
Challis, B. H., Velichkovsky, B. M., & Craik, F. I. M. (1996). Levelsof-processing
effects on a variety of memory tasks: New findings and
theoretical implications. Consciousness & Cognition, 5, 142-164.
Craik, F. I. M., & Lockhart,R. S.(1972).Levels of processing: A framework
for memory research. Journal of Verbal Learning & Verbal Behavior,
11, 671-684.
Culp, C., & Rajaram, S. (1999, November). Dividing attention at encoding
and retrieval differentially affects conceptual priming. Paper presented
at the annual meeting of the Psychonomic Society, Los Angeles.
Curran, T., & Hintzman, D. L. (1995). Violations of the independence
assumption in process dissociation. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 21, 531-547.
Curran, T., & Hintzman, D. L. (1997). Consequences and causes of
correlations in process dissociation. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 23, 496-548.
Goshen-Gottstein, Y., & Ganel,T. (2000).Repetition priming for familiar
and unfamiliar faces in a sex-judgment task: Evidence for a
common route for the processing of sex and identity. Journal of ExperimentalPsychology:
Learning,Memory, & Cognition, 26, 1198-1214.
Goshen-Gottstein, Y., & Kempinsky, H. (2001). Probing memory
with conceptual cues at multiple retention intervals: A comparison of
forgetting rates on implicit and explicit tests. Psychonomic Bulletin &
Review, 8, 139-146.
Goshen-Gottstein,Y., & Peres, N.(1998,November).Conceptuallydriven
implicit tests of memory produce cross-language priming effects.
Paper presented at the annual meeting of the Psychonomic Society.
Graf, P., & Schacter, D. L. (1985). Implicit and explicit memory for
new associations in normal and amnesic patients. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 11, 501-518.
Graf, P., Shimamura, A. P., & Squire, L. R. (1985). Priming across
modalities and across category levels: Extending the domain of preserved
function in amnesia. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 11, 385-395.
Hamann, S. B. (1990). Level of processing effects in conceptually driven
implicit tasks. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 16, 970-977.
Hintzman, D. L., & Curran, T. (1997). More than one way to violate
independence: Reply to Jacoby and Shrout (1997). Journal of Experimental
Psychology: Learning, Memory, & Cognition, 23, 511-513.
Jacoby, L. L. (1991). A process dissociation framework: Separating automatic
from intentional uses of memory. Journal of Memory & Language,
30, 513-541.
Jacoby,L. L. (1996). Dissociating automatic and consciously controlled
effects of study/test compatibility. Journal of Memory & Language,
35, 32-52.
Jacoby, L. L. (1998). Invariance in automatic influences of memory: Toward
a user’s guide for the process-dissociation procedure. Journal of
Experimental Psychology: Learning, Memory, & Cognition, 24, 3-26.
Jacoby, L. L., Begg, I. M., & Toth, J. P. (1997). In defense of functional
independence: Violations of assumptions underlying the
process-dissociation procedure? Journal of Experimental Psychology:
Learning, Memory, & Cognition, 23, 484-495.
Jacoby, L. L., & Shrout, P. E. (1997). Toward a psychometric analysis
of violations of the independence assumption in process dissociation.
Journal of Experimental Psychology: Learning,Memory, & Cognition,
23, 505-510.
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, 139-154.
Jacoby, L. L., Yonelinas, A. P., & Jennings, J. (1996).The relation between
conscious and unconscious (automatic) influences: A declaration
of independence. In J. Cohen & J. W. Schooler (Eds.), Scientific
approaches to the question of consciousness (pp. 13-47). Mahwah,
NJ: Erlbaum.
Jones, G. V. (1987). Independence and exclusivity among psychological
processes: Implications for the structure of recall. Psychological
Review, 94, 229-235.
Joordens, S., & Merikle, P. M. (1993). Independence or redundancy?
Two models of conscious and unconscious influences. Journal of Experimental
Psychology: General, 122, 462-467.
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 task. Society for Neuroscience Abstracts, 19, 1079.
Keane, M. M., Gabrieli, J. D. E., Monti, L. A., Fleischman, D. A.,
Cantor, J. M., & Noland, J. S. (1997). Intact and impaired conceptual
memory processes in amnesia. Neuropsychology, 11, 59-69.
Levin, J. R., Serlin, C., & Seaman, A. (1994). A controlled, powerful
multiple-comparison strategy for several situations. Psychological
Bulletin, 115, 153-159.
Masson, M. E. J., & MacLeod, C. M. (2002). Covert operations: Orthographic
recoding as a basis for repetition priming in word identification.
Journal of Experimental Psychology: Learning, Memory, &
Cognition, 28, 858-871.
McBride, D. M., & Shoudel, H. (in press). Conceptual processing effects
on automatic memory. Memory & Cognition.
McDermott, K. B., & Roediger, H. L., III (1996). Exact and conceptual
repetition dissociate conceptual memory tests: Problems for
transfer appropriate processing theory. Canadian Journal of Experimental
Psychology, 50, 57-71.
Moscovitch, M., Vriezen, E. R., & Goshen-Gottstein, Y. (1993).
Implicit tests of memory in patients with focal lesions or degenerative
brain disorders. In F. Boller & J. Grafman (Eds.), Handbook of neuropsychology
(Vol.8, pp. 133-173). Amsterdam: Elsevier.
Mulligan, N. W. (1997). Attention and implicit memory tests: The effects
of varying attentional load on conceptual priming. Memory &
Cognition, 25, 11-17.
Mulligan, N. W., Guyer, P. S., & Beland, A. (1999). The effects of
levels-of-processing and organization on conceptual implicit memory
in the category exemplar production test. Memory & Cognition, 27,
633-647.
Mulligan, N. W., & Hartman, M. (1996). Divided attention and indirect
memory tests. Memory & Cognition, 24, 453-465.
Nelson, D. L., Benett, D. J., & Xu, J. (1997). Recollective and automatic
uses of memory. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 23, 872-885.
Reingold, E. M., & Goshen-Gottstein, Y. (1996a). Automatic retrieval
of new associations under shallow encoding conditions. Consciousness
& Cognition, 5, 117-130.
Reingold, E. M., & Goshen-Gottstein, Y. (1996b). Separating consciously
controlled and automatic influences in memory for new associations.
Journal of Experimental Psychology: Learning, Memory,
& Cognition, 22, 397-406.
Reingold, E. M., & Toth, J. P. (1996). Process dissociations versus task
dissociations: A controversy in progress. In G. Underwood (Ed.), Implicit
cognition (pp. 159-202). Oxford: Oxford University Press.
Richardson-Klavehn, A., & Gardiner, J. M. (1998). Depth-ofprocessing
effects on primingin stem completion: Tests of the voluntarycontamination,
conceptual-processing, and lexical-processing hypotheses.
Journal of Experimental Psychology: Learning,Memory, &
Cognition, 24, 593-609.
Richardson-Klavehn, A., Gardiner, J. M., & Java, R. I. (1996).
Memory: Task dissociations, process dissociations, and dissociations
of consciousness. In G. Underwood (Ed.), Implicit cognition (pp. 85-
158). Oxford: Oxford University Press.
Roediger, H. L., III (1990). Implicit memory: Retention without remembering.
American Psychologist, 45, 1043-1056.
Roediger,H. L., III, & McDermott, K. B. (1993). Implicit memory in
normal human subjects. In F. Boller & J. Grafman (Eds.), Handbook
of neuropsychology (Vol.8, pp. 63-131). Amsterdam: Elsevier.
Roediger, H. L., III, Weldon, M. S., & Challis, B. H. (1989). Explaining
dissociations between implicit and explicit measures of retention:
A processing account. In H. L. Roediger III & F. I. M. Craik
(Eds.), Varieties of memory and consciousness: Essays in honour of
Endel Tulving (pp. 3-14). Hillsdale, NJ: Erlbaum.
1262 BERGERBEST AND GOSHEN-GOTTSTEIN
Russo, R., & Andrade, J. (1995). Directed forgetting effect in wordfragment
completion: An application of the process-dissociation procedure.
Quarterly Journal of Experimental Psychology, 48A, 405-
423.
Russo, R., Cullis, A. M., & Parkin, A. J. (1998). Consequences of violating
the assumption of independence in the process dissociation
procedure: A word fragment completion study. Memory & Cognition,
26, 617-632.
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.
Schmitter-Edgecombe,M. (1999). Effects of divided attention on perceptual
and conceptual memory tests: An analysis using a processdissociation
approach. Memory & Cognition, 27, 512-525.
Srinivas, K., & Roediger,H. L., III(1990). Classifying implicit memory
tests: Category association and anagram solution. Journal of
Memory & Language, 29, 389-412.
Thapar, A., & Greene, R. L. (1994). Effects of level of processing on
implicit and explicit tasks. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 20, 671-679.
Toth, J. P., Reingold, E. M., & Jacoby, L. L. (1994). Toward a redefinition
of implicit memory: Process dissociations following elaborative
processing and self generation. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 20, 290-303.
Vaidya, C. J., & Gabrieli, J. D. E. (2000). Picture superiority in conceptual
memory: Dissociative effects of encoding and retrieval tasks.
Memory & Cognition, 28, 1165-1172.
Vaidya, C. J., Gabrieli, J. D. E., Keane, M. M., & Monti, L. A.
(1995). Perceptual and conceptual memory processes in global amnesia.
Neuropsychology, 9, 580-591.
Vaidya, C. J., Gabrieli, J. D. E., Keane, M. M., Monti, L. A.,
Gutierrez-Rivas, H., & Zarella,M. M. (1997). Evidence for multiple
mechanisms of conceptual priming on implicit memory tests.
Journal of Experimental Psychology: Learning,Memory, & Cognition,
23, 1324-1343.
Wainwright, M. J., & Reingold, E. M. (1996). Response bias correction
in the process dissociation procedure: Approaches, assumptions,
and evaluation. Consciousness & Cognition, 5, 232-254.
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.
Zeelenberg,R., Pecher,D., Shiffrin, R. M., & Raaijmakers, J. G. W.
(in press). Semantic context effects and priming in word association.
Psychonomic Bulletin & Review.
Zeelenberg, R., Shiffrin, R. M., & Raaijmakers, J. G. W. (1999).
Priming in a free association task as a function of association directionality.
Memory & Cognition, 27, 956-961.
NOTES
1. We thank Michael Masson for suggesting this possibility to us.
2. In the exclusion condition of Experiment 2, participants were asked
to write down for each cue word the first association that came to mind,
but to replace associations they recollected as words that appeared in the
study phase with another association. It could be argued that these instructions
directed participants to use a generate-recognize strategy,
thereby violating the independence assumption. Note, however, that
lower performance in the inclusion relative to the exclusion baseline conditionin
Experiment 2 was in the oppositedirection to that taken as a signature
(e.g., Toth et al., 1994) for a generate-recognize strategy in the exclusion
condition (but see Bodner et al., 2000).
3. The relation between controlled and automatic influences of memory
may actually be one of exclusivity, in which items either can be consciously
recollected or can come to mind automatically, but not both (for
support of the exclusivity assumption, see Richardson-Klavehn et al.,
1996). If so, performance in the exclusion condition would be the estimate
of automatic influences (e.g., Reingold & Toth, 1996). This would
lead to the difficult-to-embrace conclusion that, in our study, automatic
influences following deep encoding were lower than those following
shallow encoding.
APPENDIX A
Model Equations for the PD Procedure, Assuming Independence
Between Controlled and Automatic Influences of Memory, by Correction Method
Correction Method Model Equations
Hits 2 FA I = C 1 U – C * U 1 Gi
(Conscious and guessing–exclusivity E=U – C * U 1 Ge
Automatic and guessing–exclusivity) C = I – E – d
U = (E – Be)/(1 – C)
Independent guessing I = C 1U – U * C 1(1 – C) * (1 – U) * Gi
(Conscious and guessing–independence E = U – U * C 1 (1 – C) * (1 – U) * Ge
Automatic and guessing–independence) C = [(I – r * E) 1 (r – 1)]/r
U = [1/(1 – Be)] * {[E/(1 – C)] – Be}
Additive I = C 1 (U1 Gi
) – C * (U 1 Gi
)
(Conscious and guessing–independence E = (U1 Ge) – C * (U 1 Ge)
Automatic and guessing–exclusivity) C = (I – E – d)/(1 2 d)
U = [E /(1 – C)] – Be
Note—I, proportion of “old” responses in inclusion;C, estimate of consciousinfluences; E, proportion
of “old” responses in exclusion;U, estimate of unconsciousinfluences;Gi
, probability of
guessing in inclusion; Bi
, base rate from inclusion; Ge, probability of guessing in exclusion; Be,
base rate from exclusion;r = (1 – Bi)/(1 – Be); d = Bi – Be. Adapted from Wainwright and Reingold
(1996).
(Manuscript received February 16, 2000;
revision accepted for publication July 29, 2002.)
