=
=
=
—
=
=
_
=
*
~
—
—
_
‘
—
2

 

IIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

310515 3054

   

THESlS

  
  
   

”321352?

91'? ' {5: t ‘5. ‘ Pi‘ql i~,1“‘
"""" u
‘5! 1‘?
LEE? 3“ ”A :3,
I

I” "x.

I

o
l
I

I

This is to certify that the

thesis entitled
IN SEARCH OF lNTROSPECTION:
ATTENTIONAL SET AND VERBAL REPORTS ON EVALUATIVE JUDGMENTS.

presented by

ROBERT J. MacCOllN

has been accepted towards fulfillment
of the requirements for

M.A. Psychology

degree in

Major professor

 

 

 

Date Fri»: 255111.53

0-7539 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

RETURNING MATERIALS:
‘bV1ESIvj Place in book drop to

remove this checkout from
A::::;:::;_ your record. FINES wil]
be charged if book is
returned after the date
stamped below.

 

 

 

 

 

 

 

 

In Search of Introspection:

Attentional Set and Verbal Reports on Evaluative Judgments
By

Robert J. MacCoun

A'THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1983

(L
I :1

"6

K’)

ABSTRACT

in Search of Introspection:

Attentional Set and Verbal Reports on Evaluative Judgments

BY

Robert J. MaoCoun

Nisbett and Hilson (1977) have argued that people have no access to
cognitive processes. and that verbal reports are generated by implicit
causal theories rather than introspection. However. recent social
psychological and cognitive theory suggests that people may be able to
accurately report on causal stimuli when they process information
"mindfully." In order to test this hypothesis. subjects were provided
with trait descriptions of thirty-two target persons and were asked to
evaluate them. One third of the subjects were instructed to "think out
loud" during the task. one third were instructed to carefully attend to
the task. and the remaining third were given no special instructions.
Verbal reports were provided either immediately or one week later. An
additional group of yoked observers provided verbal reports after
examining raters' responses to the stimuli. Subjects achieved a
significant degree of accuracy. However. these reports did not

significantly differ across conditions.

ACKNOWLEDGMENTS

First and foremost. I am eternally grateful to my committee members.
Bill Crane. Norb Kerr. Larry Hesse'. and Tom Carr. for accompanying me in
my exploration.of a topic described most aptly (by Herb) as a "snake
pit." Their tolerance for ambiguity and their firm but supportive
prodding served as a reliable induction of "mindfulness." Perhaps I
should have used them as my experimental manipulation.

Thanks to Lonnie Supniek and Bob Grossman of Kalamazoo College for
introducing me to the tradeoff between empirieism and phenomenology. and
especially for not allowing me to resolve it prematurely (in either
direction).

Kudos and a tip of the lid to Jack Condon and Rob Hymes for their
calm intervention during numerous episodes of computer panic.

Finally. for Dad. Tassia. Timmer and Kimmer. The Hymo. El Jaybo. The
Bee Man. The Duman. The Human Cannonball. The Harv. Rutzie. Cap'n Jack.
LePro and Noeilla. the Geiss and Greco families. and the lovely Beverly
Watson (Rule Brittanial) -- all of whom provided direct pep talks and
distraction when needed -~ and for all my magnificent family and friends

(the distinction becomes blurred). I offer my ancestral motto as a

blessing:

NE GROW GREEN

11

TABLE OF CONTENTS
PAGE
List of Tables . . . . . . . . . . . . . . . . . . . . . .
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . 1
Nisbett's challenge to introspectionism . . . . . . . . , 2

Theoretical and meta-theoretical
critiques of Nisbett and Wilson . . . . . . . . . . . . M

Methodological critiques of Nisbett and Wilson . . . . . . 5
Recent empirical research on verbal reports . . . . . . . 7

Ericsson and Simon's (1980)
information-processing model . . . . . . . . . . . . . .10

Implications of Ericsson and Simon‘s
model for the social realm . . . . . . . . . . . . . . .13

Statement of hypotheses . . . . . . . . . . . . . . . . . 18

II. METHOD . . . . . . . . . . . . . . . . . . . . . . . . . . 22

III. RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . 28
IV. DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . 32

Some speculation regarding introspective access . . . . . 35

List of References . . . . . . . . . . . . . . . . . . . . "O

111

1.

2.

3.

LIST OF TABLES
Experimental Design and Cell Sizes . . . .

Analysis of Variance: Evaluations by
Attentional Set and Stimulus Traits . . .

Analysis of Variance: Mean Accuracy
Correlations by Attentional Set.
Actor/Observer Status. and Sex of Subject

Mean Accuracy Correlations by
Attentional Set and Actor/Observer Status

1v

”6

N7

N8

CHAPTER 1

INTRODUCTION

Robert Zajonc (1980) has recently observed that social psychology
began invoking cognitive theories long before the information-processing
revolution which overtook experimental psychology in the mid-1960's.
Festinger's theory of "cognitive dissonance" (1957). Hovland's notion of
"counterargumentation" (eg. Hovland. Janis. & Kelley. 1953). and Heider's
original work on attribution theory (1957) all serve as reminders that
social-psychology has been rather cognitive all along. These theories.
although Ph9fl0m6n010gically rich and vivid. lack the detail and rigor
which the information-processing theorists have since developed.
Consequently. there has been a growing trend towards the adoption of
explicit information-processing models of "social cognition" (eg.
Higgins. Herman. & Zanna. 1981).

With this increase in theoretical sophistication and articulation.
however. has come a critical posture towards some of the overly
simplistic assumptions implicit in traditional cognitive social
psychology. For example. Langer (1978) has questioned the image of people
as "intuitive scientists" (eg. Kelley. 1967) searching for causes and
explanations in order to control their social environment: she argues
that. on the contrary. most social behavior is governed by "scripts"
(Abelson. 1976) or pre-established sequences of appropriate situational
behavior. and that we are therefore "mindless" a great deal of the time.
Similarly. Taylor and Fiske (1978) have reviewed a growing body of
research suggesting that rather than thoughtfully reflecting on social
judgments. we typically respond "off the top of our head." Perhaps the

strongest challenges have come from Richard E. Risbett and his colleagues

1

(Nisbett a Wilson. 1977: Nisbett & Bellows. 1977: Wilson and Nisbett.
1978: Hisbett & Ross. 1980). who argue that we are not able to
introspectively monitor higher-order cognitive processes. and from Daryl
Dem (196?. 1972). who argues that cognitive constructs such as attitudes
may not even exist until requested.

These critiques contribute much to the development of a more
sophisticated understanding of cognitive processes and their impact on
social behavior. However. we must avoid "throwing out the baby with the
bath-water": i.e.. eliminating the rich phenomenological texture which
has characterized our traditional social-psychological theories in an
attempt to gain theoretical and methodological rigor. A more fruitful
approach would be to determine the necessary and sufficient conditions in
which phenomenology plays a role in social cognition (cf. Lieberman.
1979: Vallacher a Wegner. in press). The present thesis is an attempt to
elucidate the conditions under which we can accurately introspect and
report on higher-order social-cognitive processes. Recent literature on
verbal reports on cognitive processes will be reviewed. and then research
will be presented testing the hypothesis that accurate verbal reports can

be obtained when subjects form social judgments "mindfully."

' e e o n ros eat 0 8
By 1977. a recurring anomaly in the results of research on cognitive
dissonance (e.g.. Aronson & Mills. 1959) and attribution theory (e.g..
Hisbett & Valins. 1971) threatened to undermine a central assumption of
both paradigms. the assumption that cognitive events were mediating the
observed changes in behavior. This anomaly was the inability of
researchers to obtain spontaneous verbal reports from subjects on these

hypothesized cognitive events. verbal reports which the theories'

predictions. Faced with this dilemna. Richard E. Nisbett. an active
researcher in attribution theory (e.g.. Jones & Nisbett. 1972). and his
colleague Timothy D. Wilson proposed three counterintuitive hypotheses
which would resolve the anomaly. First. people have little or no direct
access to higher-order cognitive processes. Not only can we seldom report
on the existence of critical stimuli affecting our cognitive processes:
we are often not aware that such processes even took place. Second. when
people do report on cognitive processes. their reports are not based upon
a recollection of cognitive events but rather upon implicit. a priori
theories about the causal factors most likely to have influenced their
responses. In doing so. people may adopt a "representativeness" heuristic
(Tversky & Kahneman. 1974). selecting as causal these factors which "seem
similar" to the response. Third. when verbal reports do correctly
correspond to behavioral changes. these reports are not due to
introspection. but are simply "good guesses"; correct matches between a
priori theories and actual cognitive events. Thus. consistent with Bem's
(1972) self-perception theory. Nisbett and Wilson deny actors access to
any information which could not be made available to observers given
sufficient observation.

In addition to a review of literature considered consistent with
these propositions. Nisbett and Wilson (1977) presented the results of
several experiments they performed to specifically test them. A brief
summary of these experiments will suffice at present. First. they
demonstrated numerous failures of subjects to report the effects on their
judgments of such influential factors as (1) semantic cuing effects. (2)
position effects. (3) "anchoring" effects. and (fl) "halo" effects.

Second. they demonstrated erroneous reporting by subjects of various

11

non-influential factors. such as (1) the emotional impact of literary
passages. (2) the effects of distraction on reaction to a film. and (3)
the effects of reassurance on willingness to take electric shocks.
Similar results have been obtained in studies by Nisbett and Bellows

(1977). and Wilson. Hull. and Johnson (1981).

‘leo ; . 2.. ie .- ... r c; ou‘: . r 826 an« :-

One theoretical weakness with Nisbett and Wilson's thesis is that
their reasoning can be interpreted tautologically. For example. Smith and
Miller (1978) have suggested that since Nisbett and Wilson regard both
correct and incorrect reports as supporting their theory. it cannot be
falsified: i.e.. "representativeness could come to be defined at least in
part by the fact that self-reports are correct" (p. 356). Similarly.
Cotton (1980) has warned that a subscription to Nisbett and Wilson's
thesis may lead researchers to ignore subjects' verbal reports when they
don't support the hypothesis being tested. even when the hypothesis is
wrong. These flaws are not fatal to the thesis. but they do suggest the
need for strict a priori operationalism in its application.

A metatheoretical problem in Nisbett and Wilson's argument is raised
by their use of the word "process." Note that they do not explicitly deny
the possibility of introspective access to cognitive "products" or
"content": only cognitive "processes." Smith and Miller (1978) regard
this distinction as the creation of another tautology: if a report is
accurate. it reports "content." but if it is not. it fails to report
"process." "Content" and "process" can thus be defined by whether or not
the report was accurate. An exact test of Nisbett and Wilson is difficult
because no acceptable definitions of product and process exist: As White

(1980) has pointed out. every process can be infinitely broken down into

smaller processes. At an extreme level. social judgments have their
origin in the activities of sub-atomic particles. so in a sense. the
level of "product" is what we define it (of. Hempel. 1966). Ultimately.
"process" is a theoretical construct and by definition cannot be
observed: we can watch an apple fall but we can never "see gravity"

(Mandler. 1975).

Such semantic hair-splitting on both sides of the argument indicates
the need for a more adequate definition of the question examined by
Nisbett and Wilson's experiments. The question is clarified by Nisbett
and Ross (1980. p. 218): "Is the actor aware that particular responses
are functionally related to particular stimuli or situational factors.
however many or complex the'links?" They further suggest that the
question is only of interest for social psychologists when the functional
relationships are "typical of the sorts of 'causes' and 'effects' about

which we assume ourselves to be knowledgeable" (Ibid.).

Metnogglggicgl Cgitigues of Nisbett and Wilson

Several authors (Smith A Miller. 1978: Ericsson & Simon. 1980:
Wright & Rip. 1981) have questioned the appropriateness of the
between-subjects design used by Nisbett and Wilson (1977) to test their
propositions. It is argued that it is unreasonable to expect subjects to
be able to perceive the systematic effects of a variable upon their
behavior when they are only exposed to one level of that variable. First
of all. between-subjects designs are typically intended to obscure the
variables of interest from the subject in a social-psychological
experiment. (Consider the elaborate subterfuge of a double-blind
experiment using a placebo. for example.) Moreover. even the researcher

can only determine the systematic impact of a between—subjects variable

6

by comparing the different levels of that variable. not by examining a

single level.

\

Nisbett and Ross (1980) argue. however. that the within-subjects
design has the disadvantage of making public the additional cues it
provides. so that it is not clear whether the subject is reporting
introspective knowledge. or knowledge gained from perceiving his or her
own behavior. Further. they claim that the between-subjects design more
closely approximates normal day-to-day conditions: "We know a warm
Belgian psychology instructor or a cold Belgian psychology instructor.
not two physically identical Belgian psychology instructors. one of whom
is warm and one of whom is cold...life itself has a between design" (p.
220).

In a certain technical sense. Hisbett and Ross are correct: we will
never be exposed to two different levels of a single variable at two
different points in time with all other variables held constant. But if
we are to be that stringent in our criteria for experimentation. it is
unlikely that any experiment can qualify. A recognition of this fact is
implicit in the common acceptance of the within-subjects design as a
research tool: as long as such artifacts as carry-over effects are
accounted for. most social psychologists (and journals) recognize their
validity.

It has also been suggested that Nisbett and Wilson did not
adequately address certain measurement issues inherent in their program
of research (Smith L Miller. 1978: White. 1980: Wright & Rip. 1981). For
example. they did not control for the possible effects of social
desirability. evaluation apprehension. and demand characteristics upon

subjects' reports. Subjects should be able to comfortably retrieve and

report on cognitive events; further. if task demands are too great.
subjects may simply report a "public" theory because to do so is less
taxing. Smith and Miller (1978) and White (1980) have claimed that
Nisbett and Wilson did not always ask subjects the exact questions their
analyses assumed. and that an adequate test must ensure that subjects are
perceiving and answering the question intended by the experimenter.
Finally. many authors have criticized the procedure used by Nisbett and
his colleagues for analyzing their data (Smith & Miller. 1978: White.
1980: Wright A Rip. 1981). For example. Nisbett and Bellows (1977)
performed analyses of variance (ANOVA's) on mean judgment ratings and
reports across subjects. a procedure which treats individual differences
as error variance. thus ignoring the crucial question of how aware a
subject is of how the stimuli influenced his or her own judgment. Indeed.
information- processing pioneer Herbert Simon (1976) has claimed that the
ANOVA is a "useless" procedure for testing process models of cognition.
To illustrate this point. Smith and Miller obtained Nisbett and Bellow's
(1977) data and reanalyzed it. comparing each subject's ratings with
his/her own reports. They obtained small but significant correlations
such that subjects who reported a given item as influential tended to be

the same subjects whose judgment ratings actually exceeded the mean on

that item.

Recent c be e orts

Several recent studies have been conducted adopting designs and
procedures that overcome many of the aforementioned method- ological
problems (Weiss S Brown. reported in Nisbett & Ross. 1980: Wright & Rip.
1981: White. 1980). White (1980) presented subjects with a series of

thirty-two person-perception tasks in which subjects were asked to rate

their liking for a fictitious person on a 9-point scale after being
exposed to four pieces of information about him. Subjects were then asked
to rate the effect of each item of information on their judgment. Each
item of information represented one level of a four-level factor: each
level of a factor was presented with each level of each other factor
twice across tasks. but within subjects. Accuracy scores were computed by
correlating the within-subject mean actual effects for a factor with the
reports by each subject on the perceived influence of that factor. so
that accuracy scores were individualized. not aggregated.

In addition. two between-subjects conditions were run: an
experimental condition in which subjects completed each report
immediately following the judgment to which it refered. and an "actors'
control" condition. in which subjects completed their reports on the
factors one week after they completed their liking judgments. This latter
condition was included to investigate the possible impact of a priori
public theories on the reports after the subjects' memory of the judgment
process had presumably decayed. White found that the experimental group
was significantly more accurate on three of the five traits: furthermore.
he demonstrated that an analysis across subjects. like that of Nisbett et
al (1977. etc.). eliminated the effects.

Wright and Rip (1981) had 140 high school juniors and their parents
rank 32 hypothetical colleges given information on five factors. One
factor had four levels. and the other four had three levels. creating a u
x 3 x 3 x 3 x 3 design. a subset of which was presented within-subjects.
After ranking. each family member rated the "desirability" of each level
of each factor on a 7—point scale. first based upon their own previous

ranking judgments. and then again for their estimation of the ratings of

the other two family members. The estimation ratings for family members
were presumed to serve as a good "actor versus observer" test because (1)
the observers had also completed the judgment task at the same time. and
(2) the observers had a history of observing the actor. Wright and Rip
found that self-reports were significantly more accurate than
observer-reports. accounting for “0% more variance. They did find a wide
range of individual differences in accuracy. however. and a second
experiment was performed to attempt to explain some of this "error"
variance. Subjects performed a similar judgment task. but after a
manipulation which lead some subjects to heavily utilize some factors
that subjects otherwise tended to ignore. After the judgment task.
subjects were to again report the perceived influence of each level of
each factor: in this experiment. however. half of the subjects were given
an "impression-management" set in which certain factors were implied to
be more normatively important than others. The remaining subjects were
given an "accurate retrieval" set. Wright and Rip found that subjects who
had heavily utilized "non-normative" factors tended to be less accurate
than subjects whose judgment process was consistent with the "normative"
set. Subjects in the "accurate retrieval" set condition achieved accuracy
not attributable to impression-management. and achieved significantly
greater accuracy than observers did.

These studies by White (1980) and Wright and Rip (1981) would appear
to indicate that with the use of more appropriate methodology. Nisbett
and Wilson's thesis is simply wrong. Evidence from Weiss and Brown
(reported in Nisbett & Ross. 1980). and Wilson. Laser. and Stone (1982).
however. indicates that such a conclusion would be premature. Weiss and

Brown had female subjects report the quality of their mood and the state

10

of several factors which might conceivably influence it (eg. day of the
week. the weather. etc.) every day for two months. At the end of this two
month period. they completed a final questionnaire assessing their
perceptions of the relative impact of the different factors on their
mood. Weiss and Brown performed a multiple regression on the data in
order to obtain objective weights of the impact of each factor on the
subjects' mood. They compared these weights to the "subjective" weights
derived from the post-experimental questionnaire and found large
discrepancies. Furthermore. a group of "observers" yielded almost
identical subjective weights. One flaw in this procedure was that they
asked subjects for causal reports in a correlational design. However.
Wilson. et a1. (1982) recently replicated their results by using a
similar procedure but soliticiting reports on "predictors" of mood. Both
studies used a within-subjects design with individualized. not
aggregated. analyses and supported Wisbett and Wilson's (1977) thesis.
contrary to White (1980) and Wright and Rip (1981).

The reason for this discrepancy is not readily apparent. but a

review of a recent information-processing model by Ericsson and Simon

(1980) may help to clarify it.

as 's 1 n r at o - ocessin Model
Ericsson and Simon's model is basically consistent with previous
information-processing models (eg. Newell & Simon. 1972) but it is
specifically oriented to addressing the issue of verbal reports on
cognitive events. While a thorough review of the model is beyond the
scope of this thesis. its central tenets can be briefly summarized.
First. information that has recently been processed is kept in short-term

memory (STM). where it is available for further processing. According to

ll

Ericsson and Simon (see also Payne. Braunstein. & Carroll. 1978: Webber &
Rich. 1981). when asked to "think aloud." the subject can externalize
this information. They refer to this procedure as "concurrent" reporting.
However. STM has only limited capacity. and as new information enters
STM. previously stored information is either transferred to long-term
memory (LTM). or is not stored at all. Therefore. after passage of time.
information can only be reported if it has been stored in LTM and then
retrieved to STM for the report. a process Ericsson and Simon call
"retrospective" reporting.

Ericsson and Simon further distinguish three different levels of
verbalization. Level 1 verbalization is direct: information is reproduced
in the form it appeared during processing. According to Ericsson and
Simon (1980) and Payne. Braunstein and Carroll (1978). this form of
verbalization should not alter the information or the cognitive process.
Level 2 verbalization requires an intermediary phase of recoding
information which was processed in a nonverbal form into verbal code: for
example. attempting to describe what a recently-handled pomegranate felt
like. Level 2 verbalization has two drawbacks: it slows reporting down.
and. as is typically the case when recoding takes place. information is
often lost. On the other hand. Ericsson and Simon claim that task
performance should not be changed in either structure or course. Finally.
level 3 verbalization requires further intermediary processing: scanning.
filtering. inference. or generation. One example of level 3 verbalization
would be to "describe only those thoughts during task performance that
are of an injunctive form." Another example might be an attempt to
"verbalize those thoughts during task performance that seem task

irrelevant. and do so in a high. squeaky voice." Level 3 verbalization

12

may alter the ongoing cognitive processes. and. as a result. the task
performance as well. This is consistent with attribution research (e.g..
Enzle & Shopflocher. 1978) suggesting that direct attributional
questions. such as requests for motives or reasons for behavior. may
change or even create create cognitive processes that might not otherwise
occur. The drawbacks of both level 2 and level 3 processing will be
accentuated by tasks that create a "cognitive strain" for subjects in
that the verbalization process makes it difficult to perform the task
(Ericsson & Simon. 1980: White. 1980).

Because of the limitations on level 2 and level 3 verbalization.
Ericsson and Simon (1980) distinguish two important types of query:
directed vs. undirected probing. To insure that concurrent reporting does
not alter the ongoing cognitive processes. Ericsson and Simon recommend
undirected probing: for example. "think out loud" as opposed to "tell me
what criteria you are using in your decision process." etc. However. for
retrospective reporting. they suggest that more direct probing (eg. "did
you use any subgoals? If so. what were they?") may prompt the retrieval
of relevant information from LTM: without such prompts. the subject may
base the report on information currently available (i.e. the outcome
behavior) as opposed to information previously available during
performance. Of course. probe decisions must ultimately be dictated by
the unique features of the task. and Ericsson and Simon provide a good
framework for such decisions.

Having articulated their model for verbal reports. Ericsson and
Simon (1980) reviewed the work of Nisbett and Wilson (1977). and found
that their results were generally consistent with the model. The

inability of subjects to report on cognitive events during judgment tasks

13

can be explained by Nisbett and Wilson's inappropriate choice of query
procedures: a complete reliance upon retrospective probes. most of which

were "misdirected." For example:

With questions like "I noticed that you took more
shock than average. Why do you suppose you did?
(Nisbett A Wilson. 1977. P. 237)." it is not even
clear to us. nor probably to the subjects. that
memory for the cognitive processes should be the
information source for the answer. If subjects
can generate their answers without consulting
their memories of the cognitive process...this
might often be more efficient than retrieving
information from memory. (Ericsson a Simon.

1980. p. 2A6).

Ericsson and Simon's model can also shed some light on the results
of Weiss and Brown (In Nisbett A Ross. 1980). and of Wilson. et al.
(1982). Recall that those authors used a within-subjects design and
individualized analyses. and found that subjects were unable to report
the effects of various potentially influential factors on their mood.
However. since their estimates were obtained at the end of a two-month
period. it is clear that these reports were extremely retrospective. and
hence. following Ericsson and Simon's model. would require a great deal
of information "embedded" in LTM. if stored at all. On the other hand.
White (1980) found that subjects were more accurate when they reported on
the perceived influence of factors immediately following each stimulus

presentation than when they gave their reports a week later.

Im ' e Soc a calm.

Ericsson and Simon's model. together with the critical input
provided by Smith and Miller (1978). White (1980). and Wright and Rip
(1981) would appear to provide an excellent set of guidelines for
recognizing the conditions under which veridical data on cognitive events

may be obtained. Furthermore. it appears to resolve the counterintuitive

14

predicament created by Nisbett and Wilson (1977) -- but only in the
laboratory. Nisbett's point. that "a subject can be shown to ignore or
deny the influence of some factor that was necessary to produce his
response in a given situation" (Nisbett A Ross. 1980. p. 218). has
nevertheless been made. Ericsson and Simon (1980) provide us with
conditions under which this demonstration will occur. and also when it
will not necessarily occur. but their model is geared toward the
researcher and subject in the laboratory. The studies by Weiss and Brown
(in Nisbett & Ross. 1980) and Wilson. et al. (1982) might have yielded
different results had they adopted concurrent probing techniques. but
they force us to recognize that we might be just as inaccurate as an
outside observer when asked to indicate the potent factors influencing
our mood over the past two months. In the hectic. mundane. vibrant.
neurotic. sexy. ambiguous social world that we "researchers" and our
"subjects" live in. what constitutes a "concurrent report" or a
"retrospective report"? Without the prodding and probing of an
experimenter. when do we attempt to monitor cognitive events? When do we
retrieve information on social judgment processes? When do we attempt to
ascertain the impact of situational factors upon our behavior?

Some clues to these questions may be provided by research on the
"mindfulness/mindlessness" distinction by Ellen Langer (Langer. 1978:
Langer. Blank. and Chanowitz. 1978: Chanowitz and Langer. 1980. etc.).
and by recent research on "information search" by Kruglanski (1980).
Pyszczynski and Greenberg (1981). Swann. Stephenson. and Pittman (1981).
and Wong and Weiner (1981). Langer (1978) argues that well-learned
sequences of social behavior become automated and form what Abelson

(1976: Shank & Abelson. 1977) has termed "scripts" (of. Neisser. 1976.

15

pp. 89-93). Scripts then allow us to engage in relatively complex
behavior. such as driving a car. typing a letter. ordering food in a
restaurant. or even holding a conversation. while remaining "mindless."
i.e.. while not actively attending to the situation or our behavior.

An early study by Langer. Blank. and Chanowitz (1978) demonstrates
this phenomenon while also suggesting factors that can instigate
"mindfulness" in day-to-day existence. Experimental confederates
approached students waiting to use a copying machine at a university
library. and asked to make either five (small favor) or twenty (big
favor) copies. Langer et a1. created three different reasons for,the
request by varying its semantic content. One third of the subjects
received a "real information" request: "Excuse me. I have five (twenty)
pages. May I use the copy machine because I'm in a rush." According to
Langer (1978). this request was "script-congruent" because people in our
culture have learned to expect to hear an explanation accompanying a
request for a favor. A "placebic information" request was presented to
another third of the subjects: "Excuse me. I have five (twenty) pages.
May I use the Xerox machine because I have to make copies?" This request
was designed so as to be redundant with the control request ("Excuse me.
I have five (twenty) pages. May I use the Xerox machine?") given to the
final third of the subjects. but structurally congruent with the "real
information" request. and therefore. structurally congruent with the
script.

It was hypothesized that a person who processed the request
mindlessly would not notice that the "reason" for the placebic request
was not really a justifiable reason at all. and would therefore follow

the script and grant the favor. 0n the other hand. a person who mindfully

16

processed the script might detect the discrepancy. and subsequently be
less likely to comply with it. Langer. et al. found that when the request
was small. there was significantly less compliance in the control
condition (60!) than in either the placebic or real information
conditions (93% and 941. respectively). However. when the request was
large. compliance in the control and placebic conditions was identical
(241) and significantly less than in the real information condition.
Langer argues that we process mindfully when a situation is novel or when
it is costly not to do so. But when a situation is highly familiar and
has a good fit to a well-learned script. we will process "mindlessly"
because it requires less effort.

Langer's thesis also receives indirect support from recent research
on information search in social situations. It has been demonstrated that
individuals will search for social information when unexpected events
occur (Pyszczyski A Greenberg. 1981). when frustrated (Wong & Weiner.
1981). and when their experiences are incongruent with their belief
systems (Ibid.). This line of theorizing owes much to D. E. Berlyne
(e.g.. 1960. 196A). who theorized that the acquisition and processing of
information are motivated by novelty. "conceptual conflict." and
uncertainty.

One link between the mindfulness/mindlessness distinction and
Ericsson and Simon's (1980) information-processing model is provided in
their discussion of verbal reports on "automatic and controlled
processing." a distinction introduced by Shiffrin and Schneider (1977:
Schneider & Shiffrin. 1977). Automatic processing occurs when a
well-learned sequence of elements in LTM is activated without utilizing

STM capacity. Controlled processing. however. requires attention and STM

17

capacity. This distinction is analogous to the difference between a
computer program compiled in machine language and an interactive program
run from a terminal. According to Schneider and Shiffrin (1977):

When an automatic sequence is initiated. its nodes are

activated and hence the associative information enters

STM. This fact does not. however. mean that the various

elements of the automatic process must be available to

consciousness or recallable at a later time...unless
attention is directed to the process when it occurs

or unless the sequence includes an automatic attention-
calling response. than the information may be immediately
lost from STM. and the subject may be quite unaware
that the process took place. (p. 156).
Ericsson and Simon (1980) view this as consistent with their model:
information not present in STM simply is not accessible for reporting.

In a sense. Langer's mindfulness/mindlessness distinction may simply
be a higher-order analogue of the controlled/ automatic processing
distinction. Script-processing is basically automatic and does not
require the demands on STM that controlled. mindful processing does.
While it is likely that the greater complexity of higher-order processes
responsible for social behavior requires more "attention-calling
responses" to coordinate subprocesses (cf. Carr. 1979). the net result
may be similar: during mindless processing in social situations.
information may drift in and out of STM in such a brief period that we
are not able to retrieve it moments later if we didn't "catch it the
first time."

If Langer is correct in her claim that "people may process only a
minimal amount of information to get them through the day" (1978. p. 50).
and if the above reasoning is correct. than the results of the Weiss and
Brown (in Nisbett A Ross. 1980) and Wilson. et al. (1982) studies make

sense: the subjects were most likely processing mindlessly a great deal

of the time that potent variables were influencing their mood. and

18

therefore did not store that information in LTH where it would be

available for future reporting. (If you. the scholarly reader. stop this
moment and ask yourself "What did I just read? Am I taking full breaths?
Am I teasing any muscles? Am I daydreaming?". perhaps you will discover

that you have been mindlessly processing this thesis!)

By asking subjects to "think out loud." Ericsson and Simon's (1980)
concurrent probe technique may serve as a direct instigation of
mindfulness in the laboratory. It allows subjects to report on
information while it is still stored in STM. and therefore bypasses the
difficulties inherent in retrospective reporting. but it nevertheless
creates an intrusion than makes it an unlikely tool for research on
social interaction. and bears little or no resemblence to any real life
introspective behavior that may be taking place. However. by instigating
mindful processing. it may increase the likelihood that subjects will

attend to and store causal information for later retrospective reporting.

Statement Qfi Hypotheses

The present thesis suggests that the experience of concurrent
reporting will induce mindful processing in subjects. thus enabling them
to introspectively monitor their evaluative judgment processes and hence
yield more accurate verbal reports. Furthermore. it is predicted that
subjects will also mindfully process information if they are forewarned
that they will have to do so. Note that Nisbett and his colleagues have
neglected to provide such a warning for their subjects. Indirect support
for this prediction is provided by a recent study by Kraut and Lewis
(1982). These authors forwarned all their subjects that they would be
asked to report on factors that might influence their judgments. They

found a reliable correlation between actors' subjective influence and

19

actual influence measures (r : .u2). and a weak but reliable correlation
after partialling out naive observers' subjective expected influence
ratings (r = .15). However. since they did not vary this intervention
across subjects. their study can't provide conclusive evidence for its
impact. Therefore. the present study will test the following hypotheses:
I. Overall. subjects will provide better than
chance accuracy in reporting on the influence
of causal factors on their evaluative judgments.
II. Actors who provide concurrent reports will
provide more accurate verbal reports than actors
in a control condition.
III. Actors who are instructed to attend to the
process of reaching their evaluative judgments
and forewarned that they will be asked to report
on this process (i.e.. "Mindful" actors) will
provide more accurate verbal reports than actors
in a control condition.

In addition to examining the effect of attentional set on subjects'
verbal report accuracy. the present study will attempt to demonstrate
that actors have introspective access to their judgment formation
processes. and that they therefore are able to provide verbal reports
with accuracy above and beyond that which can be accounted for by
implicit causal theories (cf. Nisbett A Wilson. 1977) and self-perception
effects (Dem. 196?). In order to do so. two types of "observers" will be
used. First of all. in an attempt to rule out the role of implicit lay
theories. half of the actors will be asked to provide retrospective
verbal reports a week after they make their evaluative judgments. This
procedure was adopted by White (1980). and is based on the assumption
that any information obtained by introspection will not be retained for
an entire week. Since subjects will be randomly assigned to conditions.

there is no reason to expect a preponderance of "correct" implicit lay

theories in either the immediate or delayed conditions. Therefore. any

20

increase in the accuracy of the immediate retrospective reports will rule
(out the role of these "explanatory systems." suggesting the following
hypothesis:
IV. Actors who provide verbal reports immediately
following the evaluative judgments will be more
accurate than actors who provide verbal reports
a week or more later.

Finally. an additional "observer" condition will be included in
order to examine an alternative. "self-perception." explanation for
Hypotheses I through IV. This explanation states that actors may be able
to provide accurate reports. without any recourse to introspection or
implicit causal theories. by simply observing the covariation of their
evaluative judments with the stimuli in a repeated-measures design. In an
attempt to rule out this possibility. one-third of the subjects will be
randomly assigned to a "yoked observer" condition. These subjects will
read the same instructions that were presented to the actor to whom they
are yoked. examine the same stimuli in the same order. and examine the
actor's judgment for each stimulus presentation. Finally. they will be
asked to report on the relative influence of the stimulus factors upon
the actor. Since these observers will observe the same covariation data
as the actor. any increase in accuracy by the actor will provide evidence
for the additional impact of introspection. However. observers who are
yoked to actors in the concurrent report condition will also hear a tape
recording of the actor's "think out loud" data. To the extent that these
tapes convey the actor's introspection regarding the formation of
evaluative judgments. these yoked observers should also improve in
accuracy somewhat. Therefore. the final hypothesis is that:

V. Actors in the "mindful" condition will provide more

accurate verbal reports than those of their yoked
observers.

21

In summary. these hypotheses call for a 3 (Attentional Set: Concurrent
Report vs. Mindful vs. Control) X 3 (Actor/Observer Status:

Actor/Immediate vs. Actor/Delay vs. Yoked Observer) factorial design.

CHAPTER 2

METHOD
Suhlggfia_ann_ng§ign, A 3 (Attentional Set: Concurrent vs. Mindful vs.

Control) X 3 (Actor/Observer Status: Actor/Immediate vs. Actor/Delay vs.
Yoked Observer) X 2 (Sex of subject) factorial design was employed. One
hundred and forty six participants. 65 males and 81 females. were
recruited as subjects from Introductory Psychology classes and received
extra course credit for their participation. Cell sizes ranged from 5 to
11. with a median cell size of 9 subjects. (See Table 1).
Stimglgs mgtecigl . The materials and task were similar to those used by
White (1980). Subjects were presented with thirty-two brief descriptions
of fictitious individuals. Each description contained five pieces of
information. representing one level each of the following five
dichotomous factors: gender. intelligence. attractiveness.
extraversion-introversion. and liberalism-conservatism. Thus. each
subject was exposed to every possible combination of the five factors.
The order in which these factors were presented was counterbalanced
across the descriptions to prevent possible order effects from
influencing subjects' responses. For example:

Joan tends to be conservative in her political attitudes.

She is usually introverted. She is considered to be
unattractive. She has a high IQ.

Doug has an average IQ. He is considered to be attractive.
He tends to be liberal in his political attitudes. He is
usually extraverted.

After reading each description. subjects were asked to indicate how much
they believed they would like the target person. based solely upon the

information given. Subjects made their responses on the following scale:

22

23

: 1 : 2 : 3 : : 5 : 6 : 7 :_§_: 9 :
Would like Would neither Would like a
very little like nor dislike great deal

Proceguge. Subjects were scheduled individually for half-hour sessions.

Subjects in the Actor/Delay conditions were also scheduled for an
additional half-hour session one week after the initial session. Each
subject was seated in front of a tape recorder with a built-in'microphone

on a table. The experimenter then placed a folder in front of the subject

and said:

.This folder contains materials you will use

during the experiment. including your instructions

for the task you will be asked to perform today.

I will be in the next room during the experiment.

so if you have any questions or would like me

to sign your extra-credit card. please come and get me.
You may now open the folder and begin reading.

All subjects then opened the folder and read the following cover sheet:

Thank you for participating in today's research.

We are interested in how people form impressions

of others given very little information. We

believe that an understanding of the impression-
formation process may eventually lead to an opportunity
for smoother relationships between friends. spouses.
families. neighbors. co-workers. and perhaps even
nations. Although today's research is just one small
step towards such an understanding. we nevertheless
believe that it is important and we appreciate your
help. However. if you should decide that you are
not interested in participating. simply give your
experimental credit card to-the experimenter

and he or she will sign it for you. If you are
willing to participate. please turn the page and
read and sign the departmental consent form. When
you have done so. move on to the page which

follows the consent form for further instructions.

After reading and signing the consent form. subjects read the
remaining instructions in the folder. which differed for each

experimental condition. Subjects in the Actor/Immediate and Actor/Delay

conditions read the following:

24

Today we will ask you to read thirty-two brief
descriptions of individuals approximately

eighteen to twenty-two years of age. After

reading each description. you will be asked to rate
how much you believe you would like this person. given
the information available and nothing else. There

is no right or wrong answer to this decision:

it is a strictly personal and subjective judgment.

and research suggests that such ratings vary from
individual to individual. Please do not spend more than
30 seconds on each judgment: you will probably

find that this is more than enough time. When you
have completed all 32 ratings. please inform your
experimenter and he or she will give you another
brief task to complete.

You will notice a tape recorder and a stack of index
acards on the desk. When you have completed reading these
instructions. please press the "on" button on the

tape recorder. Next. notice that there is a number
written on the back of each card. Please read the

number of the first card out loud into the microphone

and then flip over the card and read the description
which is written on it.

The remaining instructions were different for each experimental condition

and are described below.

Level of attention manipulations. Subjects in the Concurrent reporting

condition than read the following instructions:

As you examine each description. please "think
out loud." verbalizing into the microphone the way
in which you use the information in each description

to create your "liking" judgment. Later. you will
be asked to explain this process.

Subjects in the Mindful condition read the following instructions:

As you examine each description. please pay careful

attention to the way in which you use the information
in each description to create your "liking" judgment.
Later. you will be asked to explain this process.

Subjects in the Control condition were given no special attentional
instructions. All subjects than read the following:

Then. circle the number which best characterizes
your judgment on the appropriate rating scale

on the following page. and also read the number
you have chosen out loud into the microphone. Then

25

repeat the process with the next card until you
have made all 32 "liking" judgments. When you are
done. or if you have any questions. please inform

your experimenter. who will be waiting outside in the
hall.

Actor-observer manipulations and retrospective repggtg. After completing
the 32 evaluative ratings. subjects in the Actor/ Immediate conditions
were given a questionnaire asking them to consider each of the five
dichotomous factors that appeared in the stimuli and "rate how
influential you believe that characteristic was in contributing to your
'liking' judgment" using the following scale:

I :_l_:_g_:_§_5_fl_:_§_3_§_5_1_3_§_3_2_3

Not at all Somewhat ' Extremely
influential influential influential

Subjects in the Actor/Delay conditions were asked to return a week
later to "fill out a brief questionnaire." These subjects were than
scheduled for an additional half— hour session for additional extra
course credit seven to nine days later. Those that attended the second
session then filled out the same retrospective report questionnaire that
subjects in the Actor/Immediate conditions were given. Subjects in the
Actor/Delay conditions who did not attend the second session were dropped
from the study. .

The procedure for subjects in the Observer conditions differed
somewhat from that detailed above. After reading the cover sheet and
reading and signing the consent form. these subjects read the following

instructions:

Recently. we asked a student volunteer to read a
series of brief descriptions of other people

and then decide how much he or she liked them.

based only upon the limited information given.

Today. we would like you to read the same instructions
that the student read (on the next page). read the
same descriptions of other people. and then see

26

the "liking" ratings given by that student. Later.
we will ask you to fill out a brief questionnaire.

First. we'd like you to read the instructions on the next
page. These are the instructions that the student read
prior to making the judgments. We only ask you to

read them so that you will understand the student's
task: YOU WILL NOT BE ASKED TO FOLLOW THE SAME
INSTRUCTIONS YOURSELF. YOUR instructions are on the
page which follows. Please read the next page. make

sure that you understand the task that the student was
asked to perform. and then read YOUR instructions on

the page.which follows. If anything is unclear or
confusing. feel free to ask your experimenter questions.

Observer subjects were yoked to subjects in the Actor/Immediate

conditions. Thus. they then read the same instructions. either Concurrent

report. Mindful. or Control. that the Actor encountered. Then they were

told:

Now that you have read the student's instructions.
you will listen to the tape recording of the student
reading and rating the cards. Note that the index
cards described in the instructions are sitting

next to you on the table. These have been arranged

in the same order that they were in when the student
read them. '

When you have completed reading these instructions.
please turn the switch for the tape recorder to the
"ON" position. You will then hear the student.

When the student reads the number on the back of card.
you may flip that card over and read it yourself.
After you have read the card. you may listen to the
student and read his or her "liking" rating which

is circled on the completed response form on the desk.
Please DO NOT flip over the next card until you hear
the student read its number out.

You may find yourself forming "liking" judgments for

the cards. However. YOUR task is simply to observe

the student forming his or her own judgments. When you
have heard the student rate all 32 cards. please turn off
the tape recorder and your experimenter will give you

a brief questionnaire to fill out.

Subjects in the Observer conditions then listened to the tape and

read the 32 descriptions and accompanying ratings of the Actor to whom

27

they were yoked. After they had completed this task. they were given a
questionnaire that asked them to consider each of the five dichotomous
factors and "rate how influential you believe that characteristic was in
contributing to the 'liking' judgments you heard and read today" using
the same scale used for retrospective reporting by subjects in the
Actor/Immediate and Actor/Delay conditions.

Ancilliary dependent measgres. After completing the retrospective report
scales. all subjects were asked (on the next page of the questionnaire)
to "rank the five characteristics in the order which you believe they
were influential" in forming the "liking" judgments. Next. every subject
was asked to complete the 23-item "Self-Consciousness Scale" (Fenigstein.
Scheier. and Buss. 1975.) Subjects were then asked if they believed that
they usually knew why they liked someone. and to speculate on the purpose
of the research they had just participated in. Finally. subjects in the
Actor/Immediate and Actor/Delay conditions were asked to list the kind of
thoughts that went throught their mind during the task. to report on the
degree of difficulty they encountered in attempting to monitor their
judgment formation. and to report whether or not the presence of the tape

recorder made them feel self-conscious.

CHAPTER 3

RESULTS

Evaluative judgments. Mean "liking" judgments were computed for each
subject and analyzed in a 2 (Sex of Subject) X 3 (Attentional Set) X 3
(Actor/Observer Status) between—subjects analysis of variance. No

significant effects were found. The grand mean "liking" judgment was 5.35

(n 1A6). falling just above the midpoint on the 9-Point scale.

In order to assess the causal impact of the stimulus traits. the
revaluative judgments were subjected to a 2 (Gender) X 2 (Intelligence) X-
2 (Attractiveness) X 2 (Extraversion/Introversion) X 2 (Liberal/
Conservative) X 3 (Attentional Set) repeated measures analysis of
variance. with subjects (nested within Attentional Set) as the
replication factor. The five stimulus traits were all dichotomous.
within-subjects variables. making this analysis extremely sensitive.
Because of the enormous number of possible interaction sources. and
because the verbal report probes in this study did not allow subjects to
report the influence of such interactions. only main effects were
considered in this analysis. These effects and their accompanying error
terms are presented in Table 2.

As in the previous analysis. no differences in "liking" judgments
were found as a function of attentional set. suggesting that neither the
concurrent report procedure nor the "mindfulness" instructions interfered
with the evaluation process. Three of the five stimulus traits had a
significant causal impact on the evaluations. thus meeting the
prerequisite for assessing subjects' awareness of causal influences.
First. subjects preferred stimulus persons of high intelligence (M =

5.62) to those of average intelligence (M : 5.u0). F(1.135) = 23.72.

28

29

p < .0005. Second. they preferred attractive persons (M = 6.09) to
unattractive persons (M = “.93). F(1.135) = 202.63. p < .0005. Finally.
extraverts (M : 6.00) were preferred to introverts (M : 5.03). F(1.135) =
113.76. p < .0005.

Accuracy conceigtiggg. Following Kraut and Lewis (1982). accuracy scores
were created in the following fashion: First. point-biserial correlations
were computed between each dichotomous trait and the 32 "liking"
judgments. within subjects. The absolute value of each of these five
correlations was used as a measure of the actual influence of that trait
on the subject's evaluations. These five "objective influence" measures
were then correlated with the five respective subjective influence
ratings provided by the subject during the retrospective report. Thus.
each subject had an "accuracy correlation" representing their ability to
detect the causal impact of the traits. These correlations were then
treated as a random dependent variable in a 2 (Sex of Subject) X 3
(Attentional Set) X 3 (Actor/Observer Status) between-subjects omnibus
analysis of variance. presented in Table 3. No significant effects were
detected.

Cell and marginal mean accuracy scores are presented for the
Attentional Set and Actor/Observer Status variables in Table A. Although
no planned comparisons between cell means were reliable. several
interesting trends are present. First. subjects were able to provide
better greater accuracy than expected by chance. with a total mean
accuracy correlation of .35 (p < .005). accounting for over twelve per
cent of the variance. and supporting Hypothesis I. Consistent with
Hypothesis III. actors in the "mindful" condition were more accurate than

actors in the control condition. a difference of thirteen points.

o Hypothesis
e than actors in the

contrary t
less accurat

However.

been somewhat
1y have actually

onal task m.

pear to have
While the dir

suggesting the
pective report
icted by Hypot
port for eith

condition up
1 condition.

th the retros

t this additi
ection

ing process.
heses IV and V.
or hypothesis.

centre
the

interfered wi
e marginal means is as pred
re slight. pr

nal analyses

of th
oviding no sup

were conducte
ese included

d using alternative

differences a
analyses of the

Several additio
s of the accur
laticn betwee

n ratings fo
nce ratings.

acy scores. Th

n the absolut

e effect sizes (the

formulation
ach dichotomous

difference between the mea r each level of e
the correlation between the

) and the subjective influe

trait
fluence absolute point—biserial correlations and the

objective in
ctive influenc
s and subjecti

e rankings. 3
plicated

ve influence

as analyses re

e differences

rankings. The

and no reliabl

subje

effect size
the pattern of accuracy scores in Table A.
were detected.
ses. Scores were computed for each subject for the three
in. Scheier. and

cale (Fenigste

in a 2 (Sex 0

0 8? an

f the Self—Co
f Subject) X 3

sub—scales o nsciousness S
Buss. 1975). Thea
I Set) X 3 (Ac
icant differen

ach analyzed
ysis of

e scores were e
tor/Observer
ces were foun

(Attentiona
d for the Public and

No signif
n effect for

F‘20128) 3

variance.
f—Consciousness sub-scales. However. a mai
y sub-scale.

Private Sel
Anxiet

tional Set wa
On a thirty-

Atten

“.67“. p < .05. concurrent subjects reported

anxiety (M - "mindful" subjects
87). Accuracy

eatest social
ct) X 3

the gr
(M 3 ‘8055)e

lations were t

(Sex of Subje

corre herefore subj

3!

(Attentional Set) X 3 (Actor/ Observer Status) analysis of variance using
Social Anxiety as a covariate. The only reliable source was the
covariate. F(1.127) 8 6.87. p 8 .01. An additional independent variable.
Private Self-Consciousness. was constructed by performing a median split
on the Private Self-Consciousness sub-scale; however an analysis of
variance of the accuracy correlations yielded no significant effects for
this variable.

Actors were asked to rate how difficult it was for them to be aware
of how they formed their "liking" judgments. Interestingly. a 2 (Sex of
Subject) X 3 (Attentional Set) X 2 (Actor: Immediate vs. Delay) analysis
of variance for this question revealed a main effect for Attentional Set.
F(2.88) s 6.39. p < .05. Actors in the concurrent condition reported
experiencing the greatest difficulty (H 3 “.41). followed by "mindful"
actors (M : 5.06). and finally. control actors (H z 5.65).

Actors were also asked whether the presence of the microphone made
them feel self-conscious. and whether or not they believed that they
usually knew why they liked someone. Analyses of variance conducted for
each variable yielded no significant differences between conditions. The
mean rating for the latter question was 2.11 (where 1-a "I always know
why I like someone"). suggesting that subjects have relatively strong

implicit causal theories for this task.

CHAPTER M
DISCUSSION

The present study was an attempt to demonstrate that actors are
capable of introspectively monitoring social jUdgments. and can provide
verbal reports on causal influences with a degree of accuracy greater
than can be accounted for by self-perception effects or by simply
reporting an implicit causal theory. It was not successful in this
attempt. However. it did support Hypothesis I. demonstrating that
subjects are able to provide verbal reports of significant accuracy.
Furthermore. it replicated evidence reviewed by Ericsson and Simon
(1980). demonstrating that the technique of concurrent reporting does not
interfere with judgment processes for verbal tasks.

While Hypothesis III was not reliably supported. the pattern of
results nevertheless is consistent with the prediction that actors will
provide more accurate verbal reports when instructed to pay careful
attention to the judgment process and to anticipate reporting on the
process retrospectively. However. the opposite trend is indicated for
actors who provide concurrent reports during the judgment phase (i.e.
"think out loud"). although this pattern is not statistically
significant. It appears that this procedure may interfere with the
retrospective reporting process. Actors who provided concurrent reports
displayed greater "social anxiety" than actors who did not. It is
.possible that the concurrent report procedure induces sufficient arousal
to constrict attention. and thereby reduce the accuracy of the verbal
report (of. flasher A lacks. 1979). This possibility is also supported by
the fact that these subjects also reported the greatest difficulty in

attending to their judgment formation process. An alternative possibility

32

34

two. Assuming that actors in the delay condition were not likely to
retain information obtained from introspection or the perception of
covariation of five dichotomous variables for a full week. they were left

with only their implicit causal theory. If these assumptions are true. it

appears that these implicit theories were relatively accurate in
accounting for this task. In addition. the perception of covariation
appears to have played an additional role in this study; observers who
heard the concurrent and especially the "mindful" instructions tended to
be more accurate. a pattern that would not be predicted by the implicit
theory explanation.

However. it is not possible to indicate how many of these three
sources enabled the actors to provide significantly accurate immediate
retrospective reports. It may be that while actors did indeed introspect
on the judgment formation process. it did not allow accuracy greater than
that possible using the other two sources. An additional possibility is
that observers and actors in the delay condition were able to generate
their verbal reports by introspectively monitoring their own reactions to
the stimuli. To the extent that these stimuli exerted a similar influence
on most of this relatively homogenous sample of subjects. such a process
might yield similar reports for actors and observers.

Future research along this vein would do well to select stimuli with
known counterintuitive causal effects. to discourage "correct" implicit
theories. Moreover. great care should be exercised in constructing
introspective probe~. More sensitivity can be gained by asking subjects
to report on each level of a stimulus factor independently. Finally.
multiple measures of both the evaluative judgments and the retrospective

reports should be obtained. in order to increase the reliability of these

35

variables. Attenuation due to poor reliabilty may obscure important
trends in the data.

Although beyond the scope of the current study. data analyses did
indicate a number of reliable interaction among stimulus traits; for
example. the main effect for attractiveness was more polarized for female
targets than for male targets. while the pattern was reversed for
intelligence. As with other research in this area. the probes adopted in
the present study only allowed subjects to report on main effects. thus
reducing the subjects' ability to articulate everything that they
perceive. In fact. research in this area has tended to assume a general
linear process model. which may be quite misleading if subjects form
their judgments by using some sort of prototypenmatching process (e.g..
"this target sounds like a librarian-type: this one sounds like a dumb
blonde." etc.). If this is the case. however. it is not clear what form
adequate probes might take. One possibility might be to use open-ended
probes. and then to develop a coding scheme using independent coders. A
more efficient variation of this approach might be to collect concurrent
and/or open-ended retro-spective reports in a pilot session. and then
use this data to develop a battery of forced-choice probes for use in the
experiment itself. Unfortunately. any forced-choice procedure seems to
run the risk of providing an insufficient variety of choices to

accomodate every response available.

Introspective Schemata. The present thesis has argued that subjects may
typically form evaluative judgments in a relatively automatic.
"mindless." fashion. and that as a result. they may not attend to the

formation process and their verbal reports may not reach their potential

36

accuracy. This argument presupposscs that subjects do not possess or
utilize an automatic process. or "schema" (cf. Neisser. 1976) for
monitoring such cognitive processes. While I am not aware of any research
explicitly examining this issue. my hunch is that this presupposition is
probably accurate for most people. if not all of us. A related argument
has recently been offered by Sabini and Silver (1981). They argue that
even if we have access to such processes. we must understand them if we
are to provide accurate reports on them: "Here access. without criteria
of starting and inclusion and knowledge of how to apply them. does not
ensure an accurate report" (p. 178). An indirect argument for the
evolutionary improbability of "introspective causal schemata" has been
offered by Skinner (1966. p. 1207):

Behavior may have advantages which have played no

role in its selection. The converse is also true.

Events which follow behavior but are not necessarily

produced by it may have a selective effect. All

current characteristics of an organism do not

necessarily contribute to its survival and procreation.

yet they are all nevertheless "selected." Useless

structures with associated useless functions are

as inevitable as superstitious behavior. Both
become more likely as organisms become more

sensitive to contingencies.

However. the argument that intropective schemata do not naturally
"evolve" does not rule out the possibility that they can be learned and
"calibrated" against independent. objective scientific evidence. One
strategy for future research might be to train a group of subjects to
introspect accurately by providing immediate feedback. Such subjects
might eventually be able to provide accurate reports while their
attentional capacity is fully occupied with other demands. Of course.

such a research program is highly vulnerable to all the methodological

pitfalls suffered by the early Structuralist tradition: demand

37

characteristics. expectancy effects. and alteration of the ongoing
cognitive process. Eliminating these problems is by no means a trivial
undertaking. but it does not appear to be impossible. and the potential
benefits may be great. especially as a means of training important

decision-makers such as Judges. generals. and doctors.

iiera c c o e s c act 0 act on- e

Current theorizing regarding cognitive processes (e.g.. Miller.
Galanter. & Pribram. 1960: Powers. 1980: Shallice. 1978: etc.) suggests
that the control of action is organized hierarchically: complex behaviors
are the cumulative product of a series of subordinate. molecular acts.
This hierarchical structure appears to conform to a principle of
"executive ignorance“ in which subordinate levels of control operate
automatically for well-learned behaviors. as discussed earlier.

Recently. Vallacher and Wagner (in press) have developed a model of
action-identification that appears at least roughly isomorphic to the
above action-control model. "Action-identification" refers to an
individual's phenomenological report of what s/he is doing at any given
point in time. According to Vallacher and Wagner. at any moment. an
individual has a "prepotent identity." or definition of what they are
doing (e.g.. "l'm writing my thesis"). However. if requested to do so.
individuals are able to provide a multitude of additional identifications
for the action. and are able to reliably assign these to superordinatc
("I'm trying to please my thesis committee") or subordinate ("I'm peeking
at the keys of this word processor") levels of a hierarchy. They
postulate three basic principles of such hierarchies:

1. An action is maintained in terms of its
prepotent identity.

38

2. When an action can be identified at both a higher
and a lower level. there will be a tendency for
the higher level identity to become prepotent.

3. When an action cannot be maintained in terms of its
prepotent identity. there will be a tendency for
a lower level identity to become prepotent.

Vallacher and Wegner's approach is valuable because it focuses on
people's phenomenological experience of a task. To the extent that the
action-identification hierarchy and the action-control hierarchy
correspond for a given domain. this approach leads to an important
implication for research on verbal reports: individuals may not be able
to accurate report on a stimulus that influences their behavior at a
subordinate level. For example. it is possible that Nisbett and Wilson's
subjects identified their task at a relatively high level (e.g.. "I want
to appear like I'm a confident and intelligent subject") and were
therefore less likely to notice details regarding the stimulus materials.

Locke and Pennington (1982) offer a related argument. They offer a
taxonomy in which "causes" are either internal or external. They consider
"reasons" to be a special type of internal cause. and further distinguish
between "situational reasons" and "psychological reasons." They argue
that we have privileged access to our reasons. "by definition." and that
to the extent that our behavior is driven by our reasons. actors will be
capable of providing accurate verbal reports.

These approaches provide a potentially rich source for future
research on verbal reports. One strategy would be to assess subjects'
action-identities and "reasons" for acting. A careful task analysis might

then lead to criteria for predicting which subjects would be most likely

to yield accurate reports. An additional strategy might be to disrupt the

39

performance of a subset of subjects. This may induce a sudden state of
"mindfulness" that would enhance their awareness of the impact of causal

stimuli upon their behavior.

LIST OF REFERENCES

40

LIST OF REFERENCES

Abelson. R. P. Script processing in attitude formation and
decision making. In J. 3. Carroll & J. W. Payne (Eds.).

Milan and social Psbaliec- Hillsdale. N. J.:

Lawrence Erlbaum Associates. 1976.

Aronson. E.. A Mills. J. The effect of severity of initiation

on liking for a group. Joggflal 91 Abgogmgl agg Social
Psychology. 1959. £2.177-181.

Bem. D. J. Self-perception: An alternative explanation of

cognitive dissonance phenomena. figygnglggigal fleligu.
1967. 15, 183-200.

Bem. D. J. Self-perception theory. In L. Berkowitz (Ed.).

Advance; lg experimental social psychology (Vol. 6).
New York: Academic Press. 197?.

Berlyne. D. E. Conflict. arogsgl. and gynigusifiy. Toronto:
McGraw-Hill Book Company. Inc.. 1960.

Berlyne. D. E. Uncertainty and conflict: A point of contact
between information-processing theory and behavior theory
concepts. In R. J. C. Harper et al. (Eds.). Ing,29gn1112

processes: eadin s. Englewood Cliffs. New Jersey:
Prentice-Hall. Inc.. 196“.

Carr. T. H. Consciousness in models of human information
processing: primary memory. executive control and input
regulation. In G. Underwood e R. Stevens (Eds.). Agassi;

.2; consc ousness (y_l. l: Egycnologicgl lssues). London:
Academic Press. 1979.

Chanowitz. B.. & Langer. E. Knowing more (or less) than you
can show: Understanding control through the mindlessness-
mindfulness distinction. In J. Garber & M. E. P. Seligman

(Eds.). Human helplgggnggg: theory. gesearch. and applic-
ation . New York: Academic Press. 19 0.

Enzle. M. E.. A Shopflocher. D. Instigation of attribution
processes by attributional questions._Per§9nality 22g

222L633 telesales! agile—tie. 1978. .94 59‘1-599-

Ericsson. K. A.. & Simon. H. A. Verbal reports as data.

figycholgglcgl m. 1980. 31. 215-251.

Fenigstein. A.. Scheier. H. F.. & Buss. A. H. Public and
private self-consciousness: Assessment and theory.

dgygngl‘gf Consulting and Clinical Psychology. 1975.
£1. 522-527.

1:1

Festinger. L.'A,tneocy.gf goggltiye gissgnange. Stanford.
California: Stanford University Press. 1957.

Hasher. L.. e Zacks. R. T. Automatic and effortful processes

in memory. Journal.2£ gyggglmgnggl Egycnology:
E22233. 1979. 1.9.8. 356-388.

Hempel. C. G. Philosophy.gf natural science. In E. 5 M.
Beardsley (Eds.). Foundations of Philosophy series.

Englewood Cliffs. New Jersey: Prentice-Hall. Inc.. 1966.

Higgins. E. T.. Herman. C. P. . : Zanna. M. P. (Eds.).
Soclgl cognltlgg: IRE Ontarlo symposium (!_l. l).

Hillsdale. N. J.: Lawrence Erlbaum Associates. 1981.

Hovland. C. I.. Janis. I. L.. e Kelley. H. H. Communication

32g persuasiog. New Haven. Conneticut: Yale University
Press. 1953.

Jones. E. 5.. & Nisbett. R. E. The actor and the observer:
Divergent perceptions of the causes of behavior. In

E. E. Jones. et al (Eds.). Attributlgp: pegcelylng the cggses
2i anaylgr. Morristown. N. J.: General Learning Press. 1972.

Kelley. H. H. Attribution theory in social psychology. In

D. Levine (Ed.). Nebzgsgg gympgglym gg,motivation
(flgl. 1;). Lincoln. Nebraska: University of Nebraska
Press. 1967.

Kraut. R. 8.. & Lewis. S. H. Person perception and self-
awareness: Knowledge of influences on one's own

judgments. Journal of Personality and Social Psychology.

Kruglanski. A. W. Lay epistemo-logic -- Process and contents:

Another look at attribution theory. Psychological Review.
1980. 21. 70-87.

Langer. E. J. Rethinking the role of thought in social inter-
action. In J. H. Harvey. W. J. Ickes. R. F. Kidd (Eds.).
MWmWWe c (lei-II)-

Potomac. Hd.: Lawrence Erlbaum Associates. 1978.

Langer. E. J.. Blank. A.. & Chanowitz. B. The mindlessness of

ostensibly thoughtful action. lgflnflgl‘gf Begsoggllty.gng
Soclgl Psychology. 1978.‘1§. 635-692.

Lieberman. D. A. Behaviorism and the mind: A (limited) call

for a return to introspection. Amgglgan Egycnologlst.
1979. 33. 319-333.

Locke. D.. A Pennington. D. Reasons and other causes: Their

role in attribution processes. ngLnal.gfl,EQL§gnaliyy
ang Soclal Psychology. 1982. 5;, 212-223.

42

Handler. 6. 51nd gag emotlog. New York: Wiley. 1975.

Miller. G. A.. Galanter. E. H.. 8 Pribram. K. Plans and

EDS structyge 9f ehavior. New York: Holt. Rinehart.
and Winston. 1960.

Neisser. U. Cognitiog sad reality. San Fransisco: W. H. Freeman
and Co.. 1976.

Nisbett. R. H.. 8 Bellows. N. Verbal reports about causal influences
on social judgments: Private access versus public theories.

Jggnngl 9f Pegsonallty and Social Psychology. 19770.22!

Nisbett. R. 8.. 8 Ross. L. Human inference: Strategies and

ghggyggmlngg'gf sgclgl judgment. Englewood Cli fs. NTJ.:
Prentice-Hall. Inc.. 1980.

Nisbett. R. E.. & Valins. S. Perceivlng 223 causes 9! Qge's
gun benaylgc. New York: General Learning Press. 1972.

Nisbett. R. E.. 8 Wilson. T. D. Telling more than we can know:

Verbal reports on mental processes. nggnglgglggl
Review. 1977. £3. 231-259.

Payne. J. W.. Braunstein. H. L.. 8 Carroll. J. S.
Exploring predecisional behavior: An alternative approach
to decision research. chanizationgl e vior 33g Human
Performance. 1978. gg. 17-NU.

Powers. W. T. A systems approach to consciousness. In J. M.

Davidson 8 R. J. Davidson (Eds.). In: RSXQDQQLQLQBY 2!

consciousness. New York: Plenum Press. 1980.

Pyszcynski. T. A.. & Greenberg. J. The role of disconfirmed
expectancies in the instigation of attributional

processing. ngnn l g£,Personality agg_Social Psychology.
19819 £19,. 31-38.

Sabini. J.. 8 Silver. N. Introspection and causal accounts.

Journal.gf Personality 229 Social Psychology. 1981.
319,. 171-179.

Schneider. W.. e Shiffrin. R. M. Controlled and automatic
human information processing: I. Detection. search. and

attention. Psychologlcgl Review. 1977. eg. 1-66.

Shallice. T. The dominant action-system: An information-
processing approach to consciousness. In K. 8. Pope
L J. L. Singer (Eds.). ID? stream-2! consciousness.
New York: Plenum Press. 1978. ‘7‘

43

Shiffrin. R. H.. 8 Schneider. W. Controlled and automatic
human information processing: II. Perceptual learning.
automatic attending. and a general theory. Psychglgglcgl
heylew. 1977. 81. 127-190.

Simon. H. A. Discussion: Cognition and social behavior. In
J. S. Carroll 8 J. W. Payne (Eds.). Cognltloh ghg social
pghgylgr. Hillsdale. N. J.: Lawrence Erlbaum Associates.
1976. pp. 253-267.

Smith. 8. R.. 8 Miller. F. D. Limits on perception of
cognitive processes: A reply to Nisbett and Wilson.

Psychologlcal Review. 1978. §§. 355-362.

Swann. W. B.. Stephenson. B.. 8 Pittman. T. S. Curiosity and
control: On the determinants of the search for social
knowledge. Journal pf Personality 229 Social Psychology.
1981. 39. 635-642. '7

 

Taylor. S. B.. 8 Fiske. S. T. Salience. attention. and
attribution: Top of the head phenomena. In L. Berkowitz

(Ed.). vances lh,experlmehpal social psychology (V_l. 11).

New York: Academic Press. 1978.

Tversky. A.. 8 Kahneman. D. Judgment under uncertainty:
Heuristics and biases. Spighgg. 197M. 1&5. 1129-1131.

Vallacher. R.. 8 Wegner. D._§_theory pf action identification.
Hillsdale. N. J.: Lawrence Erlbaum Associates. in press.

Weber. R.. 8 Rich. H. Concurrent and retrospective reports on
cognitive processes. Paper presented at the Midwestern

 

ngghglggiggl A§§9018£109 Cohvention. Detroit. Michigan.
May. 1981.

White. P. Limitations on verbal reports of internal events:
A refutation of Nisbett and Wilson and of Ram. ngghglpglgal

Wilson. T. 0.. Hull. J. 0.. 8 Johnson. J. Awareness and self-
perception: Verbal reports on internal states. qurnal‘gg
Personality 229 Social Psychology. 1981. 39. 53-71.

Wilson. T. D.. Laser. P. S.. 8 Stone. J. I. Judging the predictors
of one's own mood: Accuracy and the use of shared theories.

219.11.120.21 21' W MWv 1932' l§0 537-555.

Wilson. T. D.. 8 Nisbett. R. E. The accuracy of verbal reports
about the effects of stimuli on evaluations and behavior.
Social Psychology. 1978.'fll, 118-131.

Wong. P. T. P.. 8 Weiner. B. When people ask "why" questions.
and the heuristics of attributional search.

9.!
Personalihy‘ghg’SOcial Psychology. 1981. 39. 650-663.

4’.

Wright. P.. 8 Rip. P. D. Retrospective reports on the causes
of decisions. quhnal‘gg Personality and Social Psychology.

Zajonc. R. Cognition and social cognition: A historical perSpective.

In L. Festinger (Ed.). Bephpspectlgh§.gh pggial psychology.
New York: Oxford University Press. 1980. 180-209.

TABLES

45

TABLE 1

‘Experimental Design and Cell Sizes

ATTENTIONAL SET

CONCURRENT MINDFUL CONTROL
sex or SUBJECT: N ' r M F M F
ACTOR/IMMEDIATE 7 1O 8 11 7 1o
ACTOR/DELAY 7 10 10 8 6 9

YOKED OBSERVER 5 7 8 8 7 8

46

TABLE 2

Analysis of Variance: Evaluations by Attentional Set

and Stimulus Traits

SOURCE
Attentional Set
Gender
Intelligence
Attractiveness
Extraversion
Political Ideology
Subjects
Subjects x Gender

Subjects

’8

Intelligence
Subjects x Attractiveness
Subjects x Extraversion

Subjects

Ideology

SUM OF SQUARES

8.51
2.00
55.26
1u81.7u
1021.60
3H.80
228u.uo
185.98
319.48
987.19
1212.31

1297.90

DF

2

135
135
135
135
135
135

MEAN SQUARE
”.26
2.00

55.26
1481.74
1021.60

3U.80

228H.N0

F RATIO
.25
1.H5
23.72 *
202.63 '
113.76 *
3.62

47

TABLE 3

Analysis of Variance: Mean Accuracy Correlations by
Attentional Set. Actor/Observer Status. and Sex of Subject

SOURCE SUM OF SQUARES DF MEAN SQUARE F RATIO
Sex of Subject .831 1 .831 2.30
Attentional Set .482 2 .281 61.29
Actor/Observer Status .089 2 .025 0.13
Sex x Attentional Set .572 2 .286 1.53
Sex x Actor/Observer Status .203 2 .101 0.51
Attention x Actor/Observer .863 A .116 0.62
Sex x Attention x .

Actor/Observer .935 A .239 1.25
Explained 3.137 17 .185 0.99
Residual 23.97 128 .187

Total 27.11 1H5 .187

TABLE 9

Mean Accuracy Correlations by Attentional Set
and Actor/Observer Status

ACTOR/IMMEDIATE

ACTOR/DELAY

YOKED OBSERVER

CONCURRENT

.28

(17)

.37
(17)

.32

(12)

.32 *

(”6)

ATTENTIONAL SET

MINDFUL

.97 *

(19)

.33
(18)

.98 *

(16)

.N3 is!

(53)

NOTE: Cell sizes appear in parentheses.

* p < .05
** p < .01

*’* p < .005

CONTROL

.29 *

(R7)

.37 ***
<53)

.33 "
(50)

.3" *

(N3)

.35 {I}

(196)

HICHIGRN STnTE UNIV. LIBRQRIES
ll11|"1|11111111111"1111111111111H11|
31293105153054