Ease of Retrieval as Information - Another Look at the Availability Heuristic

Abstract

Experienced ease of recall was found to qualify the implications of recalled content. Ss who had to recall 12 examples of assertive (unassertive) behaviors, which was difficult, rated themselves as less assertive (less unassertive) than subjects who had to recall 6 examples, which was easy. In fact, Ss reported higher assertiveness after recalling 12 unassertive rather than 12 assertive behaviors. Thus, self-assessments only reflected the implications of recalled content if recall was easy. The impact of ease of recall was eliminated when its informational value was discredited by a misattribution manipulation. The informative functions of subjective experiences are discussed.

Full text

Journal of Personality and Social Psychology
1991,
Vol.
61,
No.
2,195-202
Copyright
1991
by the American Psychological Association, Inc.
0022-3514/91/J3.00
Ease of Retrieval as Information: Another Look at the Availability Heuristic
Norbert Schwarz
Zemrum fur Umfragen, Methoden und Analysen
ZUMA, Mannheim, Federal Republic of Germany
Herbert Bless
Universitat Mannheim
Mannheim, Federal Republic of Germany
Fritz Strack
Max Planck Institut fur psychologische Forschung
Munich, Federal Republic of Germany
Gisela Klumpp, Helga Rittenauer-Schatka,
and Annette Simons
Universitat Heidelberg, Heidelberg, Federal Republic of
Germany
Experienced ease of recall was found to qualify the implications of recalled content.
Ss who
had to
recall
12
examples of assertive (unassertive) behaviors, which
was
difficult, rated themselves
as
less
assertive (less unassertive) than subjects who had to recall 6 examples, which was
easy.
In fact, Ss
reported higher assertiveness after recalling
12
unassertive rather than
12
assertive
behaviors.
Thus,
self-assessments
only
reflected the implications of recalled content if recall
was
easy.
The impact of
ease of recall was eliminated when its informational value was discredited by a misattribution
manipulation. The informative functions of subjective experiences are discussed.
One of the most widely shared assumptions in decision mak-
ing as well as in social judgment research holds that people
estimate the frequency of an
event,
or the likelihood of its occur-
rence,
"by the
ease
with which instances or associations come to
mind" (Tversky
&
Kahneman,
1973,
p,
208).
Since Tversky and
Kahneman introduced this
availability
heuristic,
it has stimu-
lated a tremendous amount of research in social cognition (see
Sherman
&Corty,
1984;
Strack,
1985,
for
reviews).
However, the
classic studies on the issue are surprisingly ambiguous regard-
ing
the underlying
process.
For example, in the most frequently
cited study (Tversky & Kahneman, 1973, Experiment 8), sub-
jects were read two lists of names, one presenting 19 famous
men and 20 less famous women and the other presenting 19
famous women and 20 less famous men. When asked, subjects
reported that there were more men than women in the
first
list
but more women than men in the second list, even though the
opposite
was
true
(by
a difference of
1).
Presumably, the famous
names were easier to recall than the nonfamous ones, resulting
in an overestimate. In fact, subjects were able to recall about
50%
more of
the
famous than of the nonfamous names. It re-
mains unclear, however, what drives the overestimate: Were
subjects' judgments based on the phenomenal experience of
the ease or difficulty with which they could bring the famous
The reported research is based in part on the diploma theses of
Gisela Klumpp, Annette Simons (Experiment 1), and Helga Ritte-
nauer-Schatka (Experiment 2), conducted at the Universitat Heidel-
berg under the direction of Norbert
Schwarz,
Herbert Bless, and Fritz
Strack. The studies
were
supported
by
Grants
Schw278/2
and Str264/2
from the Deutsche Forschungsgemeinschaft to Norbert Schwarz and
Fritz Strack and Grant Schw278/5 to Norbert
Schwarz,
Herbert Bless,
and Gerd Bohner.
We thank Leslie Clark, Tory Higgins, Bob Wyer, and three anony-
mous reviewers for helpful comments on a draft of
the
article.
Correspondence concerning
this
article should
be
addressed
to
Nor-
bert Schwarz, Zentrum fur Umfragen, Methoden und Analysen,
ZUMA, P.O. Box 122155,
D-6800
Mannheim, Federal Republic of
Germany.
and nonfamous names to mind, as Tversky and Kahneman's
interpretation suggests? Or were their judgments based on the
content of their recall, with famous names being overrepre-
sented in the recalled sample?
In a related study (Tversky
&
Kahneman,
1973,
Experiment
3),
subjects were found to overestimate the number of words
that began with the letter
r
but to underestimate the number
of
words that had r as the third letter. Similarly, Gabrielcik and
Fazio (1984) observed that exposing subjects to subliminally
presented words containing the letter t increased subjects' esti-
mates of the frequency of t words. Again, these findings may
reflect either that subjects could generate more words begin-
ning with an
r,
or including a / if primed or that they relied on
the
ease
with which relevant exemplars could be called to mind.
Similar ambiguities apply to other studies (see Sherman &
Corty, 1984; Strack, 1985; Taylor, 1982, for reviews). Typically,
the manipulations that are introduced to increase the subjec-
tively experienced ease of recall are also likely to affect the
amount of subjects' recall.
As a
result,
it
is
difficult to evaluate if
the obtained estimates of frequency, likelihood, or typicality
are based on subjects' subjective experiences or on a biased
sample of recalled information. As Taylor (1982) noted, the
latter possibility would render the availability heuristic rather
trivial—after all, "one's judgments are always based on what
comes to mind"
(p.
199).
In the present article, we report three studies that were de-
signed to disentangle the impact of content of recall and of the
subjective experience of ease or difficulty that may accompany
recall. In all studies, we introduced conditions under which the
implications of experienced ease of recall were opposite to the
implications of the content of recall per se. In addition, we
manipulated the perceived diagnosticity of the experienced
ease of recall, using misattribution manipulations (Experiment
3).
We first introduce the basic logic and the findings of our
experiments and subsequently discuss the informational func-
tions of experienced ease of recall in the context of a more
general conceptualization of the informational functions of
subjective experiences
(Schwarz,
1990;
Schwarz
&
Clore,
1988).
195

196 SCHWARZ ET AL.
Experiment
1:
If It Is So Difficult to Recall,
It Cannot Be Typical
Suppose that people are asked to report a certain number of
examples of particularly assertive, or of particularly unasser-
tive,
behaviors that they have recently engaged in. Presumably,
reporting these behaviors would increase their cognitive accessi-
bility in memory, making it more likely that these behaviors
come to mind when the people are later asked to evaluate their
own assertiveness.
As
a result, one should find that people who
had to report assertive behaviors report higher assertiveness
than people who had to report unassertive behaviors, reflecting
that the previously reported examples resulted in a biased sam-
ple of relevant behaviors. As long as people consider only the
content of what they recall, the more examples they have to
report, the more pronounced this effect should be. Such con-
tent-based predictions may be derived from numerous models
of self-related judgment
(e.g.,
Bern, 1972; Wyer
&
Srull, 1989).
Suppose, however, that people not
only
rely on what
comes
to
mind but also pay attention to the subjective experiences that
accompany the recall process. If
so,
the subjective experience
that it is very difficult to recall examples of one's own assertive
behaviors may imply that one cannot be that assertive after all,
or thinking of examples would not be that difficult. To the
extent that the experienced difficulty of recall increases with
the number of examples that are to be reported, ease of recall
and content of recall would lead to different conclusions:
Whereas the content of the recalled examples would suggest
that one is very assertive (or very unassertive), the difficulty
experienced in recalling these examples would suggest that
they cannot be frequent and typical. Hence, one may conclude
that one is probably not as assertive (or unassertive) as the re-
called behaviors would seem to imply. Accordingly, the experi-
enced difficulty of recall may qualify the implications of re-
called content.
In Experiment
1,
we tested this prediction by asking subjects
to describe either 6 or 12 examples of very assertive or very
unassertive behaviors in which they had
engaged.
Pretests indi-
cated that most subjects could easily generate 8 or 9 behaviors
but found it very difficult to generate more than 10. Subse-
quently, subjects were asked to rate their own assertiveness
along several items.
If subjects base assessments of their own assertiveness solely
on the relevant behavior that
comes
to mind, subjects who have
to report examples of assertive behavior should rate themselves
as more assertive than subjects who have to report examples of
unassertive behavior. Moreover, the more examples subjects
have to report, the more pronounced the impact of recalled
content should
be.
Thus,
a
content-based judgment process will
predict additive effects of type of example and number of exam-
ples requested. If subjects consider the content of their recall in
the light of the ease or difficulty with which they can generate
the requested examples, however, these additive effects should
not be obtained. Rather, the impact of recalled content should
be less pronounced, the more examples subjects have to report
because the difficulty of doing
so
should imply that the recalled
examples are not very frequent and typical. Hence, subjects
should rate themselves as less assertive (or unassertive) after
recalling
12
rather than 6 examples, indicating that the implica-
tions of recalled content are qualified by the ease or difficulty
with which this content could be brought to mind.
Method
Forty female students at a German university participated in a 2
(examples
of assertive
vs.
unassertive behaviors)
X 2 (6
vs.
12
examples)
factorial between-subjects experiment. Subjects were randomly as-
signed to conditions and tested in groups of
4.
All subjects were informed that
the
study
was
concerned with devel-
oping role-playing scenarios that could be used in future relaxation
training. To help with the development of these scenarios, they were
asked to describe either 6 or 12 examples of situations in which they
"behaved
very
assertively and felt at
ease"
or
of situations
in which
they
"behaved unassertively and felt insecure." Pretests indicated that gen-
erating
6
examples
was
experienced
as an
easy task,
whereas
generating
12
examples was difficult. Subjects reported their examples on answer
sheets that provided three lines for each example.
After completion of this task, subjects were asked to answer some
general
questions,
purportedly designed
to explore
students'
interest
in
participating in a relaxation-training program. These questions asked
subjects to evaluate their assertiveness, their feelings of insecurity, and
their feelings of anxiety along 10-point scales. Given the high internal
consistency of
these ratings
(Cronbach's alpha
=
.76),
the mean of these
ratings was used as the major dependent variable; higher values indi-
cated higher assertiveness.
In addition, subjects rated how difficult it was to generate the re-
quested number of examples on a scale ranging from
not
at
all difficult
(1)
to very
difficult
(10),
thus providing
a
direct measure of experienced
ease of recall.
Results
Manipulation
check.
Analyses of subjects' reported ease of
retrieval indicated that subjects found it easier to report
6
{M=
5.1) rather than
12
(M= 7.2) examples, F(l, 36) = 4.2, p < .05.
No other effect emerged
(F <
1).
Mean
differences.
As shown in Table
1,
subjects who had to
describe examples of assertive behaviors rated themselves as
more assertive after describing
6
(M
=
6.3) rather than
12
(M=
5.2) examples. Conversely, subjects who had to describe exam-
ples of unassertive behaviors rated themselves as less assertive
after describing 6 (M = 5.2) rather than
12
(M= 6.2) examples.
This crossover pattern was reflected in a marginally significant
Valence X Number of Examples Requested interaction, F(l,
36) =
3.40, p
<
.071,
whereas neither of the main effects reached
significance (F< I).1
These
findings
indicate that subjects did consider the experi-
enced ease of recall in evaluating their own assertiveness. In
fact, subjects rated themselves
as more
assertive after describing
examples of assertive rather than unassertive behaviors only if
the recall task
was
easy When the recall task
was
difficult, their
self-rating was opposite to the implications of recalled content,
despite the fact that more examples had been recalled. This
pattern of findings could not be accounted for on the basis of
recalled content per se.2 Rather, it reflected that the implica-
tions of recalled content were qualified by the ease with which
the respective content could be brought to mind.
1
The statistical weakness of
this
interaction
is
of little concern be-
cause this pattern replicates consistently in subsequent experiments.
2 It
was
conceivable,
however, that the representativeness of the re-
called examples decreased with increasing number of
examples
re-
quested. This issue was addressed in Experiment
2,
which ruled out
this possibility,
as
described below.

EASE OF RETRIEVAL 197
Table
1
Ratings ofAssertiveness
as a
Function
of
Valence
and
Number of
Recalled Behaviors
Type of behavior
No.
recalled examples Assertive Unassertive
6
12
6.3
5.2
5.2
6.2
Note,
n = 9
or
10
per condition. Mean
score
of
three
questions
is
given;
possible range is
1
to
10;
higher values reflect higher assertiveness.
Correlational
analyses.
This
conclusion
is
further supported
by correlational analyses. Specifically, the more difficult sub-
jects who had
to
recall assertive behaviors
found
the recall task,
the lower the assertiveness they reported, r(20)
=
-. 35,
p
= .
12.
In contrast, the more difficult subjects who had to report exam-
ples of unassertive behaviors found the task, the higher the
assertiveness they reported, r(20) = .66, p
<
.002.
Both correla-
tions differ reliably from one another (z = 3.38, p < .001).
Discussion
In summary, the present findings suggest that the content of
recall affected self-judgments in the direction of the valence of
the recalled behaviors only if the
recall
process itself was experi-
enced as easy If the recall process elicited experiences of diffi-
culty, on the other hand, the content of recall affected self-judg-
ments in a direction opposite to the implications of the recalled
behaviors. Hence, we may conclude that the phenomenal expe-
rience of
ease
or difficulty of recall may qualify the implications
of what comes to mind, even to the extent that the inferences
drawn are opposite in valence to the implications of recalled
content.
Experiment 2: An Extended Replication
Given the marginal significance of the interaction obtained
in Experiment 1, a replication of this finding would be wel-
come. Experiment 2 was designed to provide this replication
and extended the previous study
by
manipulating the perceived
diagnosticity of experienced ease of retrieval. Subjects were
again asked to report 6 or
12
examples of assertive or unasser-
tive behavior. However, some subjects were informed that most
participants of a previous study had found it easy to complete
this task, whereas other subjects were told that most previous
participants found the task difficult. We expected that subjects
who were told that their subjective experience of
ease
or diffi-
culty of recall
was
shared
by
most other subjects would
be
likely
to attribute their experience to characteristics of the task. If
so,
they might perceive their subjective experience of ease or diffi-
culty as being less diagnostic and might therefore be less likely
to consider it when making
a
self-judgment (cf. Kelley,
1967).
In
contrast, subjects who were told that their own experience of
ease or difficulty contradicted the typical experience of similar
others might be more likely to perceive their apparently unique
phenomenal experience as reflecting the frequency of the re-
spective behaviors in their own repertoire. If
so,
they would be
more likely to rely on their subjective experience of ease or
difficulty of retrieval in making self-judgments.
Method
One hundred
fifty-eight
students
(113
women and
45 men)
of a West
German teachers' college were asked to report either
6
or
12
examples
of assertive
or
unassertive behavior and
were
informed either that most
previous participants had found it easy or that most previous partici-
pants had found it difficult to complete this task. The latter manipula-
tion should increase the perceived diagnosticity of ease or difficulty of
retrieval under conditions in which subjects' own experience deviates
from the alleged experience of previous participants, that
is,
in which
subjects find the task easy (6 examples) but are told others found it
difficult or in which subjects find the task difficult (12 examples) but
are told others found it easy. Conversely, this information should de-
crease perceived diagnosticity under conditions in which subjects' ex-
perience coincides with the alleged experience of most other partici-
pants.
In combination, these manipulations constituted a 2 (examples of
assertive
vs.
unassertive behaviors) x 2 (6
vs.
12
examples)
X
2 (low vs.
high diagnosticity of ease of
recall)
factorial between-subjects design.
Subjects were randomly assigned to conditions, and all subjects were
tested in one session in a lecture hall setting.
The cover story and procedure used closely followed the procedure
described in Experiment 1, except that the study was said to be con-
cerned with the development of assertiveness (rather than relaxation)
training. After completing the requested number of
examples,
subjects
responded
to a
short questionnaire that purportedly assessed students*
interest in participating in assertiveness
training.
Specifically, subjects
rated their own assertiveness, their desire to be more assertive, and
their interest in assertiveness training
along
9-point
bipolar
scales.
The
mean score of these variables (Cronbach's alpha = .72)
was
used as the
dependent variable; high values indicated high assertiveness.
Results
Manipulation
check.
Analysis of variance again indicated
that subjects experienced recalling 6 examples as easier (M
=
4.8) than recalling
12
examples (M= 7.4), F(i, 142) = 5.2, p <
.02.
No other effect reached significance.
Mean
differences.
Analyses of variance including subjects'
sex revealed a main effect of sex, F(\, 142) = 7.15, p < .01,
indicating that men reported higher assertiveness than did
women. However, sex of subjects did not interact with any of
the experimental variables
(all
F<
I)
and
was
therefore ignored
in the following analyses.
As shown in Table 2, the present study replicated the pre-
vious
findings.
Specifically, subjects who had to describe exam-
ples of assertive behaviors rated themselves as more asser-
Table 2
Ratings ofAssertiveness
as a
Function
of
Valence
and
Number
of
Recalled Behaviors
and
Others'
Ease of Recall
No.
recalled
examples
6
12
Assert
5.1
4.7
Diagnosticity of
ease
of
recall
Low
Unassert
5.0
5.2
Assert
5.2
4.7
High
Unassert
4.5
5.4
Note. n=
18
to
20
per
condition.
Mean
score
of
fivequestions
is
given;
possible range is
1
to 9; higher values reflect higher assertiveness. As-
sert = assertive; Unassert = unassertive.

198 SCHWARZ ET AL.
tive after describing
6
examples (M= 5.2) rather than
12
exam-
ples (M= 4.7). Conversely, subjects who had to describe exam-
ples of unassertive behavior rated themselves as less assertive
after describing 6 examples (A/= 4.7) rather than
12
examples
{M= 5.3). This crossover pattern was reflected in a significant
Valence X Number of Examples Requested interaction, F(l,
142) = 6.35, p < .02. As in Experiment 1, no main effects of
valence or number of examples emerged (F
<
1).
Contrary to predictions, however, the manipulation of the
perceived diagnosticity of the experienced ease of recall had no
significant impact on subjects' self-assessment. Although sepa-
rate analyses revealed
a
significant Valence
X
Number of Exam-
ples interaction under high-diagnosticity conditions, F(l,
76)
=
5.98, p <
.03,
but not under low-diagnosticity conditions (F<
1),
the triple interaction failed to reach significance (F<
1).
We
return to this issue below.
Correlational analyses. The interpretation that subjects'
self-
assessments of assertiveness were mediated by the subjective
experience of ease of retrieval is again supported by correla-
tional analyses. Specifically, subjects who had to report exam-
ples of unassertive behavior reported higher assertiveness the
more difficult they found the task (r
=
.32,
p
<
.002).
In
contrast,
subjects who had to report examples of assertive behavior re-
ported lower assertiveness the more difficult they found the
task
(r = —.
12,
p =.
15),
and both correlations differ significantly
from one another
(z
= 2.77, p < .003).
Self-perception. A possible alternative account of the ob-
tained findings of this and the previous experiment required
additional analyses.
To
the extent that recalling many examples
is difficult, the more examples subjects are to report, the more
the representativeness of the recalled examples may decrease.
Thus,
subjects who have to recall 12 examples may eventually
include examples that are
less
extreme
to
complete their
task.
If
so,
a content-based judgmental process may produce a similar
pattern of
findings,
reflecting that the inclusion of less extreme
examples may dilute the impact of more extreme ones (cf. Nis-
bett, Zukier, & Lemley, 1981).
To
test this possibility, the extremity of the last two examples
provided by
5
randomly selected subjects from each condition
of Experiment
2 was
rated
by two
independent judges on
a
scale
ranging from
very unassertive (1)
to very assertive
(11).
Interrater
reliability was high
(r =
.92), and examples of assertive behav-
iors (M
=
9.6) differed reliably from examples of unassertive
behaviors (M
=
2.8), F(l, 32) = 663.0, p <
.001.
More impor-
tant, planned comparisons indicated that the last 2 examples
provided by subjects who had to report 12 examples were, if
anything, better exemplars of the requested type of behavior
than the last
2
examples provided by subjects who had to recall
only
6
examples:
(Ms =
9.8 and
9.4),
F<
1,
for
12
and
6
assertive
behaviors, respectively;
(Ms
= 2.3 and 3.3), F(\, 32) = 4.26, p <
.05,
for
12
and 6 unassertive behaviors, respectively.
Thus,
although differences in the quality of the recalled be-
haviors did emerge in the unassertive examples conditions,
these differences
were
opposite to those found in subjects' judg-
ments, lending no support to the hypothesis that
subjects*
judg-
ments were mediated by differential content rather than by the
subjective experience of ease of recall.
Discussion
In summary, Experiment 2 replicated the previously ob-
tained interaction, indicating that the implications of recalled
content were qualified
by
the
ease
with which this content could
be brought to mind. In contrast to our expectations, however,
informing subjects that most previous participants found the
task easy or difficult, respectively, did not result in a significant
triple interaction, although separate analyses under each diag-
nosticity condition provided some support for our reasoning.
In retrospect, the standard that we introduced might
have
been
less
relevant for subjects' self-judgments than
we
had assumed
a
priori: Whereas the diagnosticity manipulation referred to
others' experience, the requested judgment was not
a
compara-
tive
one.
Individuals may
use
their own feelings and aspirations,
rather than the behavior of
others,
in evaluating how assertive
they are. If they can easily recall situations in which they be-
haved unassertively, for example, this may be bothersome no
matter if others can do so just as easily or not. Moreover, in-
forming subjects about others' experienced ease or difficulty of
recall may have focused their attention even more on their own
phenomenal experience, thus increasing its impact. Accord-
ingly, a different manipulation of the diagnosticity of ease of
retrieval is used in Experiment 3.
At the same time, the failure to obtain a pronounced impact
of the alleged typical recall performance renders a possible
variation of the ease-of-recall account less plausible. Specifi-
cally,
one might
argue
that asking subjects to report
12
examples
of
a
given class of behaviors may convey that most people are
probably able to do
so.
If
so,
the experienced difficulty in meet-
ing this expectation might imply that one has
less
of the respec-
tive trait than many other people. Accordingly, the obtained
findings would reflect the experience of ease or difficulty in
meeting a certain standard, rather than the experience of ease
or difficulty of
recall
per
se.
If
so,
however, one would expect a
pronounced impact of explicit information about the ease or
difficulty with which others can perform the recall task. That
this manipulation showed little effect renders an ease-of-meet-
ing-a-standard account less compelling.
Experiment 3: Misattributing the Ease of Recall
Reflecting the assumption that others' performance may not
be germane to the informational value of subjective experi-
ences,
as
discussed
above,
a
different manipulation of perceived
diagnosticity of ease of retrieval was used in Experiment
3.
As
previous research on the informational functions of another
subjective experience—namely mood—indicated, the diagnos-
ticity of subjective states can be manipulated by misattribution
manipulations. For
example,
attributing one's feelings correctly
(Schwarz
&
Clore,
1983,
Experiment
1)
or incorrectly (Schwarz
&
Clore,
1983,
Experiment
2;
Schwarz,
Servay,
&
Kumpf,
1985)
to a transient source that is irrelevant to the judgment at hand
was found to eliminate the impact of affective states on subse-
quent judgments (see Schwarz, 1987, 1990; Schwarz & Clore,
1988,
for reviews). Similarly, misattributing one's arousal to an
irrelevant source (e.g., Zanna & Cooper, 1974; Zillman &
Bryant, 1974) was found to eliminate the effects of arousal
states (see Zanna
&
Cooper, 1976; Zillman, 1978, for reviews).
Following these lines of research, we introduced a misattri-
bution manipulation in Experiment 3. Specifically, we in-
formed subjects that the study was concerned with the impact
of different types of music on the recall of autobiographical
experiences. All subjects were exposed, by means of head-

EASE OF RETRIEVAL 199
phones, to a piece of meditation music. Some subjects were
informed that this music facilitated the recall of situations in
which one behaved assertively and felt at ease, whereas others
were informed that it facilitated the recall of situations in which
one behaved unassertively and felt insecure. After these in-
structions, subjects had to describe either 6 or 12 examples of
assertive or unassertive behaviors, replicating the previous ex-
periments.
Following Kelley's
(1972)
reasoning about augmentation and
discounting effects,
we
expected the alleged
side
effects of
expo-
sure to meditation music to moderate the impact of subjects'
experienced ease of recall on their self-assessment of assertive-
ness.
Most important, subjects whose subjective experience
contradicted the alleged
side
effect of the music would consider
this experience particularly diagnostic and would use it in their
self-assessment. Accordingly, their self-rating would only re-
flect the content of recall if recall was easy, and not if it was
difficult, replicating the previously obtained findings.
Predictions are somewhat more complex, however, for sub-
jects who can attribute their subjective experience to the music,
thus rendering it nondiagnostic. On the one hand, these sub-
jects may turn to the content of their recall to evaluate their
assertiveness, given that the informational
value
of their subjec-
tive experience has been discredited. If
so,
they should report
higher assertiveness after recalling
12
rather than
6
examples of
assertive behaviors and lower assertiveness after recalling 12
rather than 6 examples of unassertive behaviors, despite the
difficulty that they experienced in doing
so.
Thus, within each
valence condition, discrediting the informative value of experi-
enced ease or difficulty of recall should result in a data pattern
that is opposite to the pattern observed in the previous experi-
ments.
On the other hand, discrediting the informational value of
experienced ease of retrieval may lead to self-assessments that
are opposite to the implications of recalled content under some
conditions. Much as self-assessments were found to deviate
from the content of recall when recall
was
difficult, they may be
expected to deviate from the content of recall if recall is easy,
but for the wrong reasons. Specifically, subjects who can easily
recall
6
examples of assertive
(or
u nassertive) behavior but
attrib-
ute the experienced ease to the alleged impact of the music may
discount their subjective experience. This may render the typi-
cality and frequency of the recalled behaviors
dubious,
resulting
in lower ratings of assertiveness (or unassertiveness, respec-
tively).
If
so,
subjects who had to report
6
examples of assertive
behavior may actually rate themselves
as less
assertive than sub-
jects who had to recall 6 examples of unassertive behaviors
under
these
conditions.
A
similar prediction cannot
be
derived,
however, for subjects who have to recall 12 examples under
low-diagnosticity conditions. In this case, the misattribution
manipulation implies that the impact of the music may render
recall difficult, suggesting that the content of recall is informa-
tive despite the experienced difficulty. Accordingly, these sub-
jects'
self-ratings should reflect content, and no reversal be-
tween valence conditions should be obtained.
In summary, discrediting the informative value of experi-
enced ease of recall may result in
a
data pattern that
is
opposite
to the pattern observed in the previous experiments, either
within or between the valence conditions, whereas the previ-
ously obtained pattern should replicate when the informational
value of subjects' phenomenal experience is not called into
ques-
tion.
Method
Seventy-eight female students of a West German university partici-
pated in a
2
(examples of assertive
vs.
unassertive behaviors)
X
2
(6
vs.
12 examples) x 2 (high vs. low diagnosticity of experienced ease of
recall) factorial between-subjects experiment. Subjects were randomly
assigned to conditions and received all instructions from cassette
players by means of headphones.
Subjects were told that the study was part of a research program
concerned with the impact of
music
on the recall of autobiographical
memories of different emotional
valence.
The present study
was
said to
investigate the recall of a specific type of situation, namely situations
in which one felt insecure or felt assertive and at ease, respectively.
Accordingly, subjects
were
asked to describe either
6
or
12
examples of
situations in which they "behaved
assertively
and felt at ease" or of
situations in which they "behaved unassertively and felt
insecure"
while
listening to meditation music, presented by means of headphones.
To manipulate the perceived diagnosticity of experienced ease of
recall, subjects were told that the music was known to facilitate the
recall of autobiographical memories that pertain to experiences char-
acterized by assertiveness or by insecurity, respectively. If the music is
said to facilitate the respective recall task, this alleged side effect
should result in low diagnosticity of the experienced ease of retrieval
associated with recalling
6 examples
but high diagnosticity of
the
diffi-
culty associated with recalling I2 examples. Conversely, if the music is
said to facilitate recall of experiences that are opposite in valence to
those the person
is
asked to
recall,
this
should imply high diagnosticity
of easy recall but low diagnosticity of difficult recall.
To draw attention to the alleged impact of
the
music, subjects were
exposed to the music for 30 s and were asked to rate the music along
several items before they began to work on the recall
task.
In addition,
it
was
emphasized that the music might only develop its impact after a
certain amount of
exposure,
and subjects were told that the time of
exposure was varied in the present study. This information seemed
necessary because the subjective experience of difficulty of recall does
only emerge after a certain number of examples are reported.
To
draw
attention to this information, subjects received a note that said the
following:
You will be exposed to the music
( ) for the first four minutes,
( ) for the first eight minutes,
( ) for the full duration of
the
recall task.
For all subjects, the full duration option was checked by the experi-
menter.
Following this information, subjects provided descriptions of the
requested behavioral examples while listening to the music. Subse-
quently, they rated their own assertiveness, as well their desire to be
more assertive, along
9-point
scales. The mean score of
these
ratings
was used as the major dependent variable, with higher values indicat-
ing
higher
assertiveness.
In addition,
a
rating of the experienced
ease of
retrieval was obtained on a scale ranging from
easy (1)
to
difficult
(9).
Results and Discussion
Manipulation
check.
As in the previous studies, subjects re-
ported that recalling
6
examples (M= 5.9)
was
easier than recal-
ling 12 examples {M
=
6.7), although this difference did not
reach significance,
F{\,
70) = 1.76,
ns.
No
other effects emerged.
Mean
differences.
Subjects' self-ratings of assertiveness are

200
SCHWARZ
ET AL.
Table
3
Ratings of Assertiveness as a Function of
Valence,
Number of
Recalled
Behaviors,
and Diagnosticity of Ease of Recall
No.
recalled
examples
6
12
Assert
3.9
4.8
Diagnosticity of ease of
recall
Low
Unassert
5.1
4.0
Assert
5.6
4.5
High
Unassert
3.1
4.4
Note,
n is
9
or
10
per
condition.
Mean
score
of
two
questions
is
given;
possible range is
1
to
9;
higher values reflect higher assertiveness. As-
sert = assertive; Unassert = unassertive.
shown in Table 3, as a function of the experimental variables.
Analysis of variance again indicated a significant Valence X
Number of Examples Requested interaction, F(\, 70) = 4.09,
p < .04. However, this interaction was qualified by a significant
triple interaction, F(l, 70) = 9.75, p <
.001,
involving level of
diagnosticity of experienced ease of retrieval.
Diagnoses of this triple interaction revealed that the previ-
ously obtained Valence x Number of Examples Requested in-
teraction was restricted to conditions in which the alleged side
effects of the music did not discredit the diagnosticity of experi-
enced ease or difficulty of
retrieval,
F{i, 70) = 5.97, p
<
.02, for
the simple interaction under high diagnosticity. Specifically,
subjects who had to recall assertive behaviors reported higher
assertiveness after describing 6 examples (M
=
5.6) rather than
12 examples {M= 4.5), r(70) = 1.55, p < .06, one-tailed. Con-
versely, subjects who had to recall unassertive behaviors rated
themselves as less assertive after describing 6 examples (M -
3.1) rather than 12 examples (M
=
4.4), ?(70) = 1.91, p < .03,
one-tailed. This pattern replicates the key findings of Experi-
ments
1
and 2, further illustrating the robustness of the effect,
although no crossover was obtained in the present study.
This was not so, however, under low-diagnosticity condi-
tions,
in which the alleged side effects of the music rendered the
experienced ease or difficulty of retrieval uninformative. In that
case, comparisons within each valence condition suggest that
subjects tended to rely on the content of recall, rather than on
the ease with which that content came to mind. Specifically,
subjects who had to report examples of assertive behaviors re-
ported higher assertiveness after recalling 12 (M
=
4.8) rather
than 6 (M
=
3.9) examples, f(70) = 1.27, p < .10, one-tailed.
Similarly, subjects who had to recall examples of unassertive
behaviors reported being less assertive after recalling 12 (M =
4.0) rather than 6 (M = 5.11) examples, f(70) = 1.52, p < .06,
one-tailed. This impact of content of recall is reflected in a
simple interaction that is opposite in direction to the one ob-
tained under high diagnosticity conditions, F( l, 70) = 3.88,
p < .06.
At the same
time,
diagnosis of this simple interaction
by
com-
parisons between both valence conditions indicates that sub-
jects who had to recall 6 examples of assertive behaviors rated
themselves as less assertive (M
=
3.9) than subjects who had to
recall 6 examples of unassertive behaviors (A/= 5.1), /(70) =
1.59, p < .06, one-tailed. This suggests that subjects discounted
the experienced ease of retrieval, resulting in a reversal between
both valence conditions. Finally, such a reversal was not ob-
tained when subjects had to recall
12
examples (M= 4.8 and 4,0
for assertive and unassertive examples, respectively), as sug-
gested by the reasoning outlined in the introduction to this
study.
In summary, the previously obtained Valence X Number of
Examples interaction only emerged when the informational
value of subjects' subjective experience of ease or difficulty of
recall was not discredited. When the alleged side effects of the
music did discredit the implications of subjects' subjective expe-
rience, subjects did not rely on their subjective experience.
Rather, subjects who had to recall
6
examples of assertive behav-
iors reported lower assertiveness than subjects who had to recall
6 examples of unassertive behavior, presumably reflecting that
they discounted the experienced ease of retrieval. This latter
finding renders comparisons within each valence condition
under low diagnosticity somewhat ambiguous, which would
otherwise suggest that subjects in the low diagnosticity condi-
tions relied primarily on the content of recall, as reflected in the
more pronounced impact of recalling 12 rather than 6 exam-
ples.
Correlational
analyses.
As in the previous experiments, sub-
jects who had to recall examples of unassertive behaviors re-
ported higher assertiveness the more difficult they found the
recall task
(r =
.23, p
=
.
15),
provided that the diagnosticity of
ease of recall was not called into question. When the diagnosti-
city of ease of recall was discredited, on the other hand, both
variables were uncorrelated
(r =
.05), as would be expected on
the basis of the present theorizing, although the difference be-
tween correlations did not reach significance (z = .60). Finally,
no significant correlation of reported ease of recall and asser-
tiveness emerged under either high
(r
=
.03) or low (r = —.02)
diagnosticity conditions for subjects who had to report exam-
ples of assertive behavior, in contrast to the previous experi-
ments. Subjects might have taken the induced expectations into
account in making their ratings, thus obscuring the correla-
tional relationships obtained in the previous experiments.
General Discussion
Ease
of
Recall
as
Information
In combination, the reported findings provide consistent
support for the assumption that the implications of recalled
content may be qualified by the ease or difficulty with which
that content can be brought to mind. Although this assumption
has enjoyed great popularity since Tversky and Kahneman's
(1973) introduction of the availability heuristic, it had not been
adequately tested in previous studies. Most important, these
studies were open to the alternative interpretation that manipu-
lations that affected the ease of recall could just as well affect
the content of recall. Accordingly, the obtained findings could
either be attributed to biased recall, reflecting what has been
called the accessibility bias (cf. Iyengar, 1990) or to the accom-
panying subjective experience of greater ease of retrieval, re-
flecting the operation of the availability heuristic (Tversky &
Kahneman, 1973), as Taylor (1982) noted.
By constructing conditions under which the implications of

EASE OF RETRIEVAL 201
what was recalled contradicted the implications of the subjec-
tively experienced ease or difficulty with which it came to
mind,
we
could disentangle the impact of both sources of infor-
mation. In three experiments, subjects attributed themselves
higher assertiveness after recalling
6
rather than
12
examples of
assertive behavior and lower assertiveness after recalling 6
rather than 12 examples of unassertive behavior. If their judg-
ment was solely based on the content of what they recalled,
their self-attributions should have been more extreme the more
examples they recalled—in particular because content analyses
of the reported examples (presented as part of Experiment 2)
provided no evidence that the larger number of examples re-
quested decreased their representativeness. Accordingly, a
judgmental process that
is
based on recalled content
as the
only
. source of information cannot account for the observed results.
Rather, the present findings indicate that people paid atten-
tion to the subjective experience of ease or difficulty of recall in
drawing inferences from recalled content. Apparently, our sub-
jects concluded that they can't be that assertive (or unassertive)
if it
is so
difficult to recall the requested number of examples. In
line with that assumption, their ratings of ease of recall were
positively correlated with their self-assessment of assertiveness
if they had to report examples of assertive behavior but were
negatively correlated if they had to report examples of unasser-
tive
behavior.
Most important,
however,
discrediting
the
experi-
enced ease of recall by misattribution manipulations (Experi-
ment 3) reversed the otherwise obtained pattern of
findings.
In
this case, subjects reported higher assertiveness after recalling
12
rather than
6
examples of assertive behavior and lower asser-
tiveness after recalling
12
rather than
6
examples of unassertive
behavior.
In summary, we conclude that people not only consider what
they recall in making a judgment but also use the ease or diffi-
culty with which that content comes to mind as an additional
source of information. Most notably, they only rely on the con-
tent of their recall if
its
implications are not called
into
question
by the difficulty that they experience in bringing the relevant
material to mind. In Experiments
1
and 2, the difficulty of
recalling
12
examples qualified the conclusions drawn from the
content of recall to such a degree that the obtained judgments
were, in fact, opposite to the implications of recalled content.
Thus,
the present studies extend previous research on the
avail-
ability heuristic by drawing attention to the other end of the
ease-of-recall continuum: Our findings suggest that difficulty
in recall may decrease judgments of frequency, probability, or
typicality, much as ease of
recall
has been assumed to increase
these judgments.
Informative Functions of Subjective Experiences
In addition to bearing on the operation of the availability
heuristic, the present studies extend previous research on the
informative functions of subjective experiences. This research
has primarily been concerned with the informational value of
affective
states (see
Schwarz,
1987,1990;
Schwarz
&
Clore,
1988,
for reviews) and has demonstrated that people may use their
perceived affective state as a source of information, according
to a "how do I feel about it" heuristic (e.g., Schwarz & Clore,
1983;
Schwarz et al., 1985; Schwarz, Strack, Kommer, &
Wagner, 1987). In doing so, people (misinterpret their pre-ex-
isting affective state as a reaction to the object of judgment,
resulting in more favorable evaluations under elated than under
depressed moods, unless the diagnosticity of their feelings for
the judgment at hand is called into question. Accordingly, peo-
ple were found to rely on their feelings at the time of judgment
only under conditions in which they could assume those feel-
ings to reflect their affective reaction to the object of judgment.
As Clore and Parrott (in press) noted, it is informative to
apply this logic to the operation of the availability heuristic. In
principle, the availability heuristic reflects the correct insight
that it is easier to recall frequent rather than rare events. What
renders
this
heuristic error prone
is
that the experienced
ease of
retrieval may reflect the impact of variables other than fre-
quency, such as the event's salience or vividness (cf. Nisbett &
Ross,
1980).
Hence, we may conclude that inappropriate appli-
cations of the availability heuristic reflect a process of misattri-
bution: People rely on their subjective experience of ease of
retrieval to the extent that they (mis)attribute it to frequency of
occurrence, rather than to the impact of other variables.
The same logic may be extended to other subjective experi-
ences,
such as feelings of familiarity. For example, Jacoby and
Dallas
(1981;
see also Jacoby & Kelley, 1987; Jacoby, Kelley,
Brown,
&
Jasechko,
1989)
observed in
a
recognition experiment
that subjects could accurately identify
rare
words that had previ-
ously been shown to them but provided numerous false alarms
in response to common words. As Clore and Parrot (in press)
noted, subjects "apparently misattributed to recency of expo-
sure the sense of familiarity that actually came from frequency
of exposure."
In fact, the current analysis may be extended
to the
operation
of priming phenomena in general (cf. Clore
&
Parrott, in press).
This research (see Higgins & Bargh, 1987; Martin & Clark,
1990,
for
reviews)
indicates that exposure to
a
concept increases
the likelihood that this, rather than another, concept is subse-
quently used in interpreting ambiguous information. However,
this effect
is
only obtained i f subjects
are
not
aware
of the poten-
tial impact of the priming episode. Specifically, correlational
analyses reported by Lombardi, Higgins, and Bargh (1987) in-
dicated that priming effects were limited to subjects who were
not able to consciously recall the primed concepts. In a more
direct experimental test, Strack, Schwarz, Bless, Kiibler, and
Wanke (1990) observed that priming effects were not obtained
when subjects were subtly reminded of the priming episode.
These
findings
suggest that priming effects may only emerge if
people misattribute the thoughts that come to mind to the im-
pact of the stimulus information. If they, correctly, attribute the
emerging thoughts to the impact of the priming procedure, on
the other hand, priming effects seem unlikely to be obtained.
As this discussion indicates, social cognition research may
benefit from paying closer attention to phenomenal experi-
ences,
a theme that has only recently captured researchers' at-
tention (see Bargh, 1989; Jacoby & Kelley, 1987, for reviews).
Although the default option is probably to use the content that
comes to mind without further qualification, judgmental pro-
cesses may also involve the use of subjective experiences as an
additional source of data, which
may
qualify the implications of
thought content. In using these data—be they affective states,
the ease of retrieval, a sense of familiarity, or an emerging train

202 SCHWARZ ET AL.
of thoughts—people may pay attention to their possible causes
to determine their informational value. Most important, they
will only
use
these subjective experiences
as a basis
of judgment
if they can (misjattribute them to the impact of the object of
judgment. Whereas much remains to be learned about the in-
formative functions of different subjective
experiences,
the pres-
ent discussion suggests that the general logic developed in re-
search on the informative functions of affective
states (see
Clore
& Parrott, in
press;
Schwarz,
1990;
Schwarz
&
Clore,
1988)
may
provide a fruitful heuristic framework for their conceptualiza-
tion.
References
Bargh, J. A. (1989). Conditional automaticity: Varieties of automatic
influence in social perception and cognition. In
J.
S.
Uleman
&
J.
A.
Bargh (Eds.),
Unintended thought
(pp. 3-51). New York: Guilford
Press.
Bern, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Ad-
vances in experimental social psychology (Vol. 6, pp.
1-62).
San
Diego,
CA:
Academic Press.
Clore, G. L., & Parrott,
W.
G. (in press). Moods and their vicissitudes:
Thoughts and feelings as information. In J. Forgas (Ed.), Emotion
and
social
judgment.
Elmsford, NY: Pergamon Press.
Gabrielcik, A., & Fazio, R. H. (1984). Priming and frequency estima-
tion:
A
strict test of the availability heuristic.
Personality
and Social
Psychology
Bulletin,
10,
85-89.
Higgins, E. T, & Bargh, J. A. (1987). Social cognition and social per-
ception. Annual
Review
of
Psychology,
38, 369-425.
Iyengar,
S,
(1990).
The accessibility
bias in
politics:
Television
news
and
public opinion.
International Journal
of
Public Opinion
Research,
2,
1-15.
Jacoby,
L. L.,
&
Dallas,
M.
(1981).
On the relationship between autobio-
graphical memory and perceptual learning. Journal of Experimental
Psychology:
General,
110,
306-340.
Jacoby, L. L., & Kelley, C. M.
(1987).
Unconscious influences of mem-
ory for a prior event.
Personality and Social Psychology
Bulletin,
13,
314-336.
Jacoby, L. L., Kelley, C. M., Brown, J.,
&
Jasechko,
J.
(1989).
Becoming
famous overnight: Limits on the ability to avoid unconscious influ-
ences of the past.
Journal
of
Personality
and
Social
Psychology,
56,
326-338.
Kelley, H. H. (1967). Attribution theory in social psychology. In D.
Levine (Ed),
Nebraska
symposium on
motivation
(Vol.
15,
pp. 192-
238).
Lincoln: University of Nebraska Press.
Kelley, H. H. (1972). Causal schemata and the
attribution
process.
Morristown, NJ: General Learning Press.
Lombardi, W. J, Higgins, E. T, & Bargh, J. A. (1987). The role of
consciousness in priming effects on categorization: Assimilation
versus
contrast
as
a function of awareness of the priming
task.
Person-
ality
and
Social Psychology
Bulletin,
13,
411-429.
Martin, L. L., & Clark, L. F. (1990). Social cognition: Exploring the
mental processes involved in human social interaction. In M. W
Eysenck (Ed.),
Cognitive
psychology:
An
international review
(pp.
265-310). New
York:
Wiley.
Nisbett, R. E.,&Ross,
L.
(1980).
Human
inference:
Strategies
and
short-
comings
of
social
judgment.
Englewood Cliffs, NJ: Prentice-Hall.
Nisbett, R. E., Zukier, H., & Lemley, R. (1981). The dilution effect:
Nondiagnostic information weakens the impact of diagnostic infor-
mation.
Cognitive
Psychology,
13,
248-277.
Schwarz, N. (1987). Stimmung als
Information:
Vntersuchungen
zum
Einflufi von Stimmungen auf die
Bewertung
des eigenen Lebens
[Mood as information]. West Berlin, Federal Republic of Germany:
Springer-Verlag.
Schwarz, N. (1990). Feelings as information: Informational and moti-
vational functions of affective states. In E. T. Higgins & R. Sorren-
tino (Eds.),
Handbook
of
motivation
and
cognition:
Foundations
of
social behavior
(Vol.
2, pp. 527-561). New
York:
Guilford Press.
Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judg-
ments of well-being: Informative and directive functions of affective
states.
Journal
of
Personality and Social
Psychology,
45,
513-523.
Schwarz,
N.,
&
Clore, G. L.
(1988).
How do I feel about
it?
Informative
functions of affective
states.
In K. Fiedler
&
J. Forgas (Eds.),
Affect,
cognition,
and
social behavior
(pp. 25-43). Toronto, Ontario, Can-
ada: Hogrefe International.
Schwarz,
N.,
Servay,
W,
&
Kumpf,
M.
(1985).
Attribution of arousal as
a mediator of the effectiveness of fear-arousing communications.
Journal
of
Applied Social
Psychology,
15,
74-84.
Schwarz, N., Strack, F., Kommer, D., & VSfcgner, D (1987). Soccer,
rooms, and the quality of your life: Mood effects of judgments of
satisfaction with life in general and
with
specific life-domains.
Euro-
pean
Journal
of
Social
Psychology,
17,
69-79.
Sherman,
S.
J., &Corty,
E.
(1984).
Cognitive heuristics. InR.
S.
Wyer&
T.
K. Srull(Eds.),
Handbook of'social cognition
(Vol.
1,
pp.
189-286).
Hillsdale, NJ: Erlbaum.
Strack, F. (1985). Urteilsheuristiken [Judgmental heuristics]. In D.
Frey & M. Irle (Eds.),
Theorien
der
Sozialpsychologie
(Vol. 3, pp.
239-267). Bern, Switzerland: Huber.
Strack, F., Schwarz, N., Bless, H., Ktibler, A., & Wanke, M. (1990).
Remember the priming
episode:
A
test
of Jacoby s
bifocational model
of
memory.
Manuscript submitted for publication.
Taylor,
S.
E.
(1982).
The availability bias in social perception and inter-
action. In D. Kahneman, P. Slovic, & A. Tversky (Eds.),
Judgment
under
uncertainty:
Heuristics
and
biases
(pp.
190-200). Cambridge,
England: Cambridge University Press.
Tversky,
A.,
&
Kahneman,
D.
(1973).
Availability:
A
heuristic for
judg-
ing frequency and probability.
Cognitive
Psychology,
5,
lOl-l^l.
Wyer, R. S., & Srull, T. K. (1989).
Memory
and
cognition
in its
social
context.
Hillsdale, NJ: Erlbaum.
Zanna, M.
P.,
&
Cooper,
J.
(1974).
Dissonance and the
pill:
An attribu-
tion approach to studying the arousal properties of dissonance.
Journal
of
Personality
and
Social
Psychology,
29, 703-709.
Zanna, M. P., & Cooper, J. (1976). Dissonance and the attribution
process. In
J.
Harvey, W Ickes,
&
R. E Kidd
(Eds.),
New directions in
attribution research
(Vol. 1, pp. 199-217). Hillsdale, NJ: Erlbaum.
Zillman, D. (1978). Attribution and misattribution of excitatory reac-
tions.
In
J.
H. Harvey, W I.
Ickes,
&
R.
F Kidd
(Eds.),
New directions
in attribution research
(Vol.
2, pp. 335-368). Hillsdale,
NJ:
Erlbaum.
Zillman, D., & Bryant, J. (1974). Effect of residual excitation on the
emotional
response
to provocation and delayed aggressive behavior.
Journal
of
Personality
and
Social
Psychology,
30,
782-791.
Received August
8,1990
Revision received March
6,1991
Accepted March 21,1991 •