INTERPERSONAL RELATIONS AND GROUPPROCESSES

Naturalistic Social Cognition: Methodology, Assessment, and Validation

William Ickes, Eric Robertson, William Tooke, and Gary TengUniversity of Texas at Arlington

The authors described a research paradigm for the study of naturalistic social cognition (i.e., thestudy of the thoughts and feelings of individuals engaged in spontaneous, unstructured dyadic inter-actions). This paradigm was created by incorporating the thought-listing technique developed byBrock (1967), Greenwald (1968), and Cacioppo and Petty (1981) into the unstructured dyadic inter-action paradigm of Ickes (1982, 1983). Data from the first "expanded dyadic interaction paradigm"study provided evidence for the interrater reliability and the construct validity (i.e., face and contentvalidity, concurrent validity, divergent and convergent validity) of the thought and feeling measuresobtained by this procedure. The degree of subjects' behavioral involvement in their interactions wasrelated to a number of thought-feeling indexes (e.g., total number of entries, percentage of positivepartner entries), and its relations with the percentages of positive, neutral, and negative entries werefurther moderated by internal correspondence (Brickman, 1978) and private self-consciousness(Fenigstein, Scheier, & Buss, 1975). In addition, some interesting parallels in the behavioral andthought-feeling correlates of gender were noted. These findings suggest that the expanded dyadic-interaction paradigm will be particularly useful in exploring the relations between naturalistic socialcognition and naturalistic social behavior.

Few problems in the history of psychology as a science haveproved to be as intractable as the problem of studying people'sthoughts and feelings as they naturally occur. As Griffin (1984)noted, the frustration resulting from the "confusing and contra-dictory results" of the 19th-century introspectionist approachled most experimental psychologists to abandon the study ofhuman consciousness altogether. Although the typical responsewas simply to ignore the phenomena of thoughts and feelingsbecause of the difficulties inherent in studying them, many psy-chologists went so far as to "virtually deny their existence or atleast their accessibility to scientific analysis" (p. 436).

The difficulties of developing appropriate research paradigmsfor the study of thoughts and feelings are certainly real enough,but one might expect that researchers in cognitive psychologywould have made great strides toward surmounting thesedifficulties over the past several decades. According to Griffin(1984), however, this is not the case. He contended that "anyserious attention to conscious thoughts or subjective feelings"is "conspicuously absent from most of contemporary cognitivepsychology," and went on to criticize cognitive psychology for

The authors would like to thank Tim Baltisberger; Guadalupe Bar-ron, Scott Blackwood, Stanley Gaines, Diane Grim, and Chuck Over-street for their assistance in coding the data. We would also like to thankCarol Marangoni for her comments during the planning stages of theresearch, Ira Bernstein, Calvin Garbin, and David Kenny for their ad-vice regarding the data analyses, and Charles Carver for his endorse-ment of our hypotheses regarding the Self-Consciousness Scale factors.

Correspondence concerning this article should be addressed to Wil-liam Ickes, Department of Psychology, University of Texas at Arlington,Arlington, Texas 76019.

its naive, erroneous belief that conscious experience is "nothingbut" information processing. "Information-processing isdoubtless a necessary condition for mental experience, but is itsufficient? Human minds do more than process information;they think and feel. We experience beliefs, desires, fears, expec-tations, and many other subjective mental states" (p. 457).

Similar criticisms have arisen even within the more circum-scribed domain of cognitive social psychology—a domain inwhich one might expect to find a special concern with people'sthoughts and feelings as they naturally occur in everyday socialexperience. For example, when the eminent cognitive psycholo-gist Ulric Neisser was invited to comment on the papers pre-sented at a major research symposium on "Social Knowing,"his criticisms focused on (a) "the extraordinarily narrow per-spective from which scholars in this field approach their work,"and (b) Neisser's suspicion that the symposium speakers' "pri-mary interest was not in the phenomena themselves but in thetheoretical issues and concepts of social psychology." Regardingthis second point, he said:

They see little need to investigate "social knowing" as it actuallyoccurs in the world, or even to read what others have discoveredabout it, because (I think) they are not very interested in "socialknowing" anyway . . . Productive scientific activity, like percep-tual activity, involves exploration as well as observation andthought; what must be explored are the phenomena as they actuallyoccur [italics added]. (1980, p. 603)

Neisser's (1980) sentiments are clearly in accord with thoselater echoed by Griffin (1984)—that after more than a centuryof research activity, most psychologists still have not come togrips with the fundamental methodological problems of study-ing people's thoughts and feelings as they naturally occur.

Journal of Personality and Social Psychology, 1986, Vol. 51, No. 1,66-82Copyright 1986 by the American Psychological Association, Inc. 0022-3514/86/100.75

66

NATURALISTIC SOCIAL COGNITION 67

Not all contemporary psychologists have chosen to ignore

these problems, however. At least a few researchers since Wundt

and Freud (1900/1939) have attempted to confront them di-

rectly, and the history of such attempts is chronicled in an in-

sightful and informative review by Cacioppo and Petty (1981).

The techniques used by some of these researchers include a va-

riety of "mechanical assessment procedures" such as button-

pushing (e.g., Hovland, Lumsdaine, & Sheffield, 1949), "sig-

naled stopping" (e.g., Carter, Ruggels, Jackson, & Hefner,

1973), dial-turning (e.g., Peterman, 1940), electrophysiological

recording (e.g., Cacioppo & Petty, 1979; Schwartz, 1975;

Schwartz, Fair, Salt, Mandel, & Klerman, 1976), and reaction

time measures (e.g., Geller & Shaver, 1976; Lingle & Ostrom,

1979; Rogers, Kuiper, & Kirker, 1977).

Alternative techniques employed by other researchers in-

clude a variety of "spoken and written assessment procedures"

(e.g., Craighead, Kimball, & Rehak, 1979; Hurlburt, 1979,

1980; Kagan, 1977; Klinger, 1978; Meichenbaum, 1977;

Wright, 1974). Of these nonmechanical techniques for assessing

cognitive responses, one promising approach appears to be the

thought-listing technique "developed by Brock and Greenwald

at Ohio State University in the late 1960s (Brock, 1967; Green-

wald, 1968)" (Cacioppo & Petty, 1981, p. 315). This technique,

which was later refined and elaborated by Petty and Cacioppo

(e.g., Cacioppo & Petty, 1979, 1981; Petty & Cacioppo, 1977,

1979) requires subjects to list their thoughts in response to stim-

ulus material presented by the researcher. The resulting

thoughts are content coded by the subjects or by independent

judges, or both, into various dimensionally denned categories,

and the indexes representing the combined or aggregated re-

sponses within each category are then analyzed as data.

Evidence from the studies reviewed by Cacioppo & Petty

(1981) led them to conclude that the thought-listing procedure

(a) "has provided a reliable and valid measure of cognitive re-

sponses"; (b) is generally nonreactive "because it does not affect

the responses to the task under investigation"; (c) is apparently

"sensitive to environmental manipulations and to individual

differences"; and (d) "appears to tap thoughts that mediate

affective responses rather than post hoc rationalizations for

these responses" (p. 337).

To date, there are at least three published reports of attempts

to study subjects' spontaneous thoughts about social interac-

tions that are either (a) anticipated but not actually experienced

by the subject (Cacioppo, Glass, & Merluzzi, 1979), or (b) pas-

sively witnessed by the subject (Davison, Robins, & Johnson,

1983; Harvey, Yarkin, Lightner, & Town, 1980). In the Caci-

oppo et al. (1979) study, high and low socially anxious male

subjects were asked to list their thoughts following a 3-min in-

terval in which they were anticipating a discussion of campus

issues with an unfamiliar female undergraduate. In the Davison

et al. (1983) and Harvey et al. (1980) studies, subjects were

asked to list their thoughts in response to either an audiotape or

a videotape of an interaction in which they did not actually

participate.

The results of all three investigations revealed that indexes of

the content of the subjects' listed thoughts were meaningfully

related to (a) the subjects' personality traits (social anxiety in

the Cacioppo et al., 1979, study, fear of negative evaluation in

the Davison et al., 1983, study) and/or (b) experimental manip-

ulations imposed by the researchers (amount of social-evalua-

tive stress in the Davison et al. study, cognitive set and severity

of outcome in the Harvey et al., 1980, studies). These findings

suggest that the thought-listing technique might profitably be

applied to study the cognitive responses of subjects involved in

actual, as opposed to imagined or simulated, social interactions.

In the present article, we describe how the thought-listing

technique can be incorporated into the dyadic interaction para-

digm of Ickes (1982, 1983) to permit the study of naturalistic

social cognition, that is, the study of the thoughts and feelings

experienced by individuals while they are engaged in spontane-

ous, unstructured dyadic interaction.

Ickes' dyadic interaction paradigm provides a context for the

thought-listing procedure that has many inherent advantages.

These advantages include the flexibility of the paradigm as a

tool for studying the effects on dyadic interaction of a wide

range of cognitive (e.g., Ickes, Patterson, Rajecki, & Tanford,

1982), dispositional (e.g., Ickes & Barnes, 1977, 1978), situa-

tional (e.g., Ickes, 1984), and sociodemographic variables (e.g.,

Ickes & Turner, 1983). Other advantages include the unobtru-

sive audio- and video-recording of spontaneous, naturalistic in-

teraction behavior that is relatively free of task demands or

other traditional sources of bias. The audio- and videotape data,

which are collected in a controlled observational setting, yield

a wide range of objective behavioral measures (most of which

are aggregated, multiple-act measures) that can be analyzed at

both the individual and dyad levels of analysis. (For detailed

reviews and discussions of the paradigm, see Ickes, 1982,1983.)

Our current studies indicate that the addition of the thought-

listing technique brings an entirely new dimension to the dyadic

interaction paradigm. Whereas the original paradigm succeeds

in capturing much of the overt interaction, (i.e., the partici-

pants' manifest behaviors), the expanded paradigm appears to

succeed in capturing much of the covert interaction (i.e., the

participants' latent thoughts and feelings) as well. In addition,

the expanded paradigm makes it possible for researchers to

study the links between naturalistic social cognition and natu-

ralistic social behavior in ways that we are only beginning to

explore in our research.

We had two major goals in this first report of research using

the expanded dyadic interaction paradigm. Our first goal was

to demonstrate that the expanded dyadic interaction paradigm

can be used to obtain reliable and valid measures of the actual

thoughts and feelings that individuals experience in unstruc-

tured dyadic interactions. Our second goal was to examine and

explore the strongest and most consistent patterns of correla-

tion between the various behavioral and thought-feeling mea-

sures used in this research. An important corollary of this sec-

ond goal was to assess the paradigm's utility in revealing empir-

ically the ways in which social behavior and social cognition are

related in unstructured, dyadic interactions.

Method

Subjects and Design

The subjects were 31 female and 29 male undergraduates enrolled inintroductory psychology classes at the University of Texas at Arlington.

Virtually all of them participated in a general pretesting at the begin-ning of the semester, in which they completed the Self-ConsciousnessScale (Fenigstein, Scheier, & Buss, 1975) and a modified version ofBrickman's (1978) Internal-External Correspondence Scale. (The few

68 ICKES, ROBERTSON, TOOKE, AND TENG

subjects who did not complete these measures during the pretestingwere asked to complete them at the end of the session in which theywere run.)

Unlike the subjects in previous studies in this program of research,the subjects in this study were not preselected on the basis of relevantpersonality or sociodemographic variables (cf. Ickes, 1984; Ickes &Barnes, 1977, 1978; Ickes, Schermer, & Steeno, 1979; [ekes & Turner,1983; Rajecki, Ickes, & Tanford, 1981). Instead, they volunteered toparticipate through the usual sign-up sheet procedure, which was modi-fied so that two separate sign-up sheets were used to schedule the sub-jects who would be paired as dyad members within each session. Theseparate sign-up sheets, which not only bore the names of different ex-periments and different experimenters but also directed subjects to re-port to different waiting areas, were used to help ensure that (a) twofriends or previously acquainted individuals would not sign up for thesame session, and (b) each scheduled pair of subjects would be unlikelyto meet and interact before their session began.

This sign-up procedure resulted in a relatively heterogeneous sampleof subjects, whose assignment of dyad partners was essentially random.The heterogeneity of the dyads is indicated by the fact that virtually allcombinations of three gender compositions (male-male, female-fe-male, and female-male) by four ethnic compositions (white-black,white-chicano, white-oriental, and white-white) were represented. Be-cause of the small ns resulting from these Ethnicity X Gender composi-tions, it was not feasible to analyze the data using either ethnic composi-tion or sex composition as between-dyads independent variables. Thediversity of the dyad compositions is stili useful, however, in contribut-ing to the generalizability (i.e., external validity) of the findings obtainedin this research.

Setting and Equipment

The setting for the study was the suite of laboratory rooms depictedin Figure 1. The room at the left of the figure was an observation room

furnished with a long couch (1), a coffee table with an FM wireless mi-crophone concealed within its base (2), some bookcases (6 and 7), amedium-sized table near the door (5), and a small table in a corner ofthe room (3) that supported a slide projector oriented to project slidesonto a screen (4) in the opposite corner, Across the hallway from theobservation room was the camera room (15), which appeared on casualinspection to be a darkened storage area filled with stacked cardboardboxes and other materials. In actuality, one of the boxes concealed acolor video camera with a zoom lens. The camera was oriented to focusthrough the open doorways of both the observation and the camerarooms on the area of the couch and coffee table.

Adjacent and to the right of the observation room was the controlroom. It contained a microcomputer workstation (8 and 9) that in-cluded an Apple II microcomputer with disk drives and a printer. Alongthe right wall of the control room was the experimenter's control station.Here the experimenter could sit in front of a single table (10) housinga microphone and intercom system (not depicted) and two identicalvideocassette recorders. The two VCRs were connected by a Y-adapterto a cable that passed from the control room over the ceiling of thehallway to connect to the camera in the camera room. Behind the tablecontaining the VCRs were two adjacent tables that supported identical25" color TV monitors (11 and 12) oriented to face through one-waymirrors into identical test cubicles (13 and 14). The test cubicles wereeach equipped with an intercom speaker, a remote start/pause switchconnected to the VCR and TV monitor system unique to each cubicle,and a supply of "Thought and Feeling Coding Forms" (to be describedsubsequently).

Procedure

Following the directions on their respective sign-up sheets, the twosubjects scheduled for each session reported to different waiting areaswithin the psychology building. These areas were physically isolated

70 13

\ M

Figure 1. Schematic diagram of the laboratory setting.

NATURALISTIC SOCIAL COGNITION 69

from each other but were on the same floor as the suite of research

rooms just described.

Collection of the videotape data. After activating and checking the

video camera, the experimenter collected the two subjects from their

respective waiting areas, led them down the hall into the observation

room, and asked them to leave their belongings on the table (5) and to

take a seat on the couch (1). The experimenter then explained that the

first part of the study involved filling out copies of a questionnaire. How-

ever, when he attempted to get copies of the questionnaire from a boxon one of the bookcases (6), he "discovered" that only one copy of the

questionnaire was left. Stating that he needed to get additional copiesand would return "in a minute or two," the experimenter left the room.

Once out of sight, he consulted a digital watch to time the 5-min interval

in which the subjects were covertly audio- and videotaped.

At the end of this period, the experimenter returned, announced that

the study was half over, and questioned the subjects for possible suspi-

cion of the videotaping before proceeding further. (No subject evi-

denced such suspicion in this study.) Following the test for suspicion,the experimenter conducted a partial debriefing in which he explained

that the first part of the study had been designed "to examine the behav-

ior of two people in an initial interaction." He then explained to the

subjects how the videotape of their interaction had been made, assured

them that any data coded from the tapes would be used for statistical

purposes only, and asked them to sign a release form giving their consent

for the tapes to be used in this way. (All subjects agreed to sign the re-lease.)

Collection of the thought and feeling data. The experimenter then

explained that the second part of the study was designed to complement

the first part by permitting the two subjects to create independent, writ-

ten records of the specific thoughts and feelings they each rememberedhaving experienced during the time they spent waiting in the observa-

tion room. Prior to explaining this procedure in more detail, the experi-menter asked the subjects to accompany him down the hallway, where

each was then seated in one of the cubicles (13 and 14) that adjoined

the control room. (The subjects were given no physical or visual access

to the control room itself during any phase of the procedure.)From his station in the control room (10), the experimenter switched

on the intercom system and asked the subjects to read silently a page

of printed instructions while he read the same instructions aloud. Theinstructions read as follows:

Each of you will independently view a videotape of the 5-minuteperiod you were left alone together in the other room. You will eachbe viewing a separate, but identical, copy of the videotape. On thetable in front of you is a stack of printed forms and a remote start/pause control for the videotape player.

Once the tape has started to play, I would like you to stop—orpause—the tape each time you come to a point where you remem-ber experiencing a particular thought or feeling during the 5-min-ute period in which you and the other person were left alone to-gether.

In the upper-left corner of the TV screen, a tinier will appearwhile the tape is playing. Each time you stop the tape at a pointyou remember experiencing a particular thought or feeling, writethe time shown on the screen in the column labeled "TIME" onthe form in front of you. Then, in the column labeled"THOUGHT OR FEELING," briefly write what that thought orfeeling was in the box provided. PLEASE USE COMPLETE SEN-TENCES WHEN WRITING DOWN YOUR THOUGHTS ANDFEELINGS. Next, in the column labeled "T/F," put a "T" if whatyou have written constitutes (by your own definition) a thought, oran "F" if it constitutes a feeling. Use your own, personal definitionsof "thought" and "feeling" here.1 Finally, at the top of the sheetwrite in today's date and then circle the letter, "M" or "F," thatrepresents your sex.

Please be candid and truthful in writing down your thoughts andfeelings. Your responses will be completely confidential; no one but

the experimenter will see them, and they will be used for statisticalpurposes only. DO NOT PUT YOUR NAME ON THE FORMSto ensure that these data will remain anonymous. Use as manyforms as you need, but place only one thought or feeling ineach box.

Begin at the point on the videotape where I (the experimenter)leave the room. Write down each of the thoughts and feelings youremember having had, and stop writing them down only at the mo-ment on the tape when I return to the room. When you have fin-ished, I will collect your completed forms and then we will start thelast portion of the experiment, which will take approximately fiveminutes to complete.

I WILL NOW START THE VIDEOTAPE.

After reading these instructions aloud, the experimenter concluded

by reminding the subjects to record only those thoughts and feelings

they remembered having had during the interaction. He then activated

simultaneously both subjects' VCRs, in all cases starting the videotape

at a point several seconds before the experimenter had left the observa-

tion room. The two subjects then completed their thought-and-feeling

records according to the instructions given.Both our own observations and the reports of the subjects themselves

indicated that none of the subjects had any problems with any aspects

of the thought-and-feeling recording procedure that would require dis-

carding their data. In fact, several subjects spontaneously volunteered

the observation that seeing the events on the videotape made it much

easier than they had expected to recall their specific thoughts and feel-

ings during the interaction. Clearly, an important advantage of usingthe videotape to cue subjects' recall of their thoughts and feelings is that

it reproduces as completely as possible the total context in which these

subjective, often "episodic" events were coded into memory (see God-

den & Baddeley, 1975; Smith, Glenberg, & Bjork, 1978; and Tulving,1983, on context dependency in human memory).

Collection of the final self-report measures. In the final part of the

study, the subjects remained in their respective cubicles long enough to

complete a brief posttest questionnaire designed to assess (a) the degree

of discomfort each subject experienced during the 5-min period they

waited together, and (b) the degree to which each subject liked his or

her partner. When each subject had completed the questionnaire and

returned it to the experimenter who was waiting outside in the hall, the

subject was then debriefed more fully, was asked not to discuss the studywith potential future subjects, and was then thanked and released.

Dependent Measures

Behavioral measures. Two types of behavioral data—static behaviors

and dynamic behaviors—were coded from the videotapes by two inde-

pendent judges who were kept blind with respect to the subjects' scores

on the personality measures. The static behaviors, which either oc-

curred only once or did not vary much across time, were represented

by measures of interpersonal distance (shoulder-to-shoulder), body ori-

entation, and openness of body posture. The interrater reliabilities for

these three static behavior measures were .98, .70, and .87, respectively,

based on the entire sample. The more temporally variable dynamic be-

haviors included the total frequency and duration of verbalizations (i.e.,

speaking turns), directed gazes, mutual gazes, expressive gestures, and

expressions of positive effect. Also included were measures of the num-

1 After debating for some time the problem of how to get our judges

to reliably differentiate thoughts from feelings when content coding the

entries, we finally decided to let the subjects make this decision them-

selves for each of the entries they listed (cf. Cacioppo & Petty, 1981, pp.

324-325). No subjects so far (out of about 150 run in our first two

studies using the expanded dyadic interaction paradigm) have reported

having had any difficulty distinguishing their thoughts from their feel-

ings.

70 ICKES, ROBERTSON, TOOKE, AND TENG

Table 1

Direct and Meta-Perspective Perceptions of Self, Partner,

Otker(s), and Environment

Direct perspectives Meta perspectives

S>S S's perception of S S > P > S S's perception of Fsperception of S

S>P S's perception of P S > P > P S's perception of P'sperception of P

S>O S's perception of O S>P>O S's perception of P'sperception of O

S>E S's perception of E S > P > E S's perception of P'sperception of E

Note. S = subject, P = partner, O = other person(s), E = environmentalobject, event, or circumstance.

ber of conversation sequences initiated and the number of verbal rein-forcers provided by each dyad member during the course of the interac-tion. The interrater reliabilities of these dynamic behavior measures

ranged from .89 to .99, with an average r of .94. (For the criteria usedto operationally define these behaviors, see Ickes & Turner, 1983, pp.214-215.)

Self-report measures. Two types of self-report data were obtained—

the dyad members' thought-and-feeling data and their responses to theitems on the posttest questionnaire. The thought-and-feeling data werecoded by two independent judges (not those coding the behavioral data)

who were also kept blind with respect to the subjects' personality scores.These judges recorded (a) the total number of entries (thoughts + feel-

ings) for each subject and (b) the emotional valence of the entry (+, 0,or -). They also recorded, in terms suggested by Laing, Phillipson, andLee (1966), (c) whether the entry represented a direct perspective (sub-ject's own perspective) or a meta-perspective (subject's adoption of his

or her partner's perspective), and (d) the target or object of the perception

expressed by the entry (self, partner, other person(s), environmental ob-ject, event or circumstance; see Table 1).

Our emotional valence dimension is essentially equivalent to the po-

larity dimension used in previous thought-listing studies of attitude

change. Likewise, our target or object dimension is similar to the target

dimension used in earlier research, although the specific target or object

categories we employ differ markedly from those of attitude change re-searchers (cf. Cacioppo & Petty, 1981, 319-324). The direct versus

meta-perception distinction has its origin in the work of Laing et al.(1966; see also Goffman, 1969), and to our knowledge has not pre-viously been applied to the classification of cognitive responses gener-

ated by means of the thought-listing technique.Clearly, the categories denned in Table 1 are not the only ones possi-

ble. Subjects might, for example, adopt the experimenter's or a room-mate's perspective or even the "group perspective" of the interactingdyad (Wegner & Giuliano, 1982) in considering the various targets orobjects of perception. In practice, however, we have found that the Table

1 categories are sufficient to categorize more than 98% of the thought-feeling entries, with the frequency of the remaining entries being too

low to j ustify using them as data in any event.The entire thought-feeling data set was based on 439 entries (283

thoughts, 156 feelings). The judges' rates of agreement in coding these

entries were (a) 100% for number of entries, (b) 97% for valence of entry,(c) 98% for direct- versus meta-perspective, and (d) 94% for the targetor object of the perception expressed by the entry. The cases in whichthe judges originally disagreed were later resolved through reexamina-

tion and discussion.The coded thought-and-feeling data for the subjects within each dyad

were then entered into the laboratory microcomputer as the input to asoftware program called COLLECT YOUR THOUGHTS. COLLECT YOURTHOUGHTS is an Applesoft Basic program written specifically for this

program of research by the first author. It produces as output various

summary indexes that represent simple aggregations (i.e., counts) of thefrequency of occurrence of entries in specific thought-feeling catego-ries. These summary indexes (and the corresponding percentage mea-

sures derived from them) were used as the primary dependent measures

in analyses reported in the Results section of this article, where they aredescribed in more detail.

The self-report measures on the posttest questionnaire included (a)a three-item measure of the degree of discomfort that dyad membersexperienced in their interactions, and (b) a four-item measure of thedyad members' liking for their partners (i.e., the Interpersonal Judg-

ment Scale or US; Byrne, 1971). For this sample, the internal consis-tency of the three-item measure of discomfort ("How self-conscious didyou feel when you were with the other person?" "To what degree did

the interaction seem awkward, forced, and strained to you?" and "Towhat degree did the interaction seem smooth, natural, and relaxed toyou?") was .76. The internal consistency of the four-item IJS was .84.All items were rated on 11-point scales ranging from not at all (0) to

extremely (10).

Results and Discussion

The logic of the procedures commonly used to establish the

construct validity of ability or trait measures was extended by

analogy to establish the validity of our method for recording

and coding the thoughts and feelings that dyad members experi-

ence in their initial, unstructured interactions. The results re-

ported here concern the (a) face and content validity, (b) concur-

rent validity, and (c) convergent/discriminant validity of the

thought-feeling indexes.

Face and Content Validity

The data represented by the written thoughts and feelings of

the subjects in this study appeared by inspection to satisfy the

criteria of face and content validity. None of the recorded

thought-feeling entries were facetious, frivolous, or in any other

way indicative that the subjects failed to take their task seriously.

Concurrent Validity

To provide evidence for the concurrent validity of our

thought-feeling assessment, we designated certain thought-

feeling measures as those one might expect to evidence concur-

rent validity in terms of their correlation with our posttest ques-

tionnaire indexes of discomfort during the interaction and lik-

ing for partner (i.e., the IJS). More precisely, we expected that

the number and percentage of negative self entries (direct per-

spective) would correlate positively with the three-item mea-

sure of discomfort during the interaction, whereas the number

and percentage of positive partner entries would correlate posi-

tively with the four-item IJS index of liking for partner.

These tests of the concurrent validity of the thought-feeling

measures can be regarded as somewhat stringent in that not all

direct, negative self entries imply personal discomfort in inter-

action (i.e., they can imply low self-esteem, fatigue, illness, and

other physical and/or psychological states, as well). Similarly,

not all positive partner entries imply liking for one's partner

(i.e., they can imply respect in the absence of liking). In addi-

tion, these tests were stringent in one other respect. To further

explore the discriminant validity of the thought-feeling assess-

ment, we also examined the correlations between (a) discomfort

NATURALISTIC SOCIAL COGNITION 71

during the interaction and the number and percentage of nega-

tive partner entries, and (b) liking for partner and the number

and percentage of positive self entries. These measures were

chosen to provide, in each case, at least one discriminant valid-

ity comparison that varied the most relevant "target" of the

thought-feeling index (self vs. partner) without varying the in-

dex's affective tone (positive or negative).

The correlations were computed at both the individual and

dyad levels of analysis, consistent with the recommendations of

Kenny and LaVoie (1985). At the individual level, each subject

is treated as a single case, and the subjects' pairing into dyads

is disregarded. At the dyad level, each dyad is treated as a single

case, and the data take the form of the summed or averaged

scores of the two dyad members for each of the variables being

correlated. In addition to the individual and dyad correlations,

a third type of correlation—the adjusted individual-level

correlation—was computed. It represents the residual individ-

ual difference covariance that remains when the variance for

the dyad-level correlation is "partialed out" of the variance for

the unadjusted individual-level correlation (for details, see

Kenny & LaVoie, 1985).

As the data in the upper portion of Table 2 reveal, self-re-

ported discomfort was indeed correlated with measures of di-

rect, negative self entries. Its correlation with the number of

negative self entries was primarily evident at the individual

level, whereas its correlation with the percentage of negative self

entries was primarily evident at the dyad level.

Each of these significant correlations may be viewed as sup-

porting our prediction that the dyad members' self-reported

discomfort would bear a positive relation to the number and

percentage of direct, negative self entries they reported. How-

ever, the level at which a given correlation is observed further

qualifies its meaning and interpretation (cf. Kenny & LaVoie,

1985). In general, a significant dyad-level correlation indicates

that the dyad members' responses are not statistically indepen-

dent of each other, but instead reflect mutual, interaction-based

influence that caused their responses either to (a) converge and

become more similar (positive dyad-level correlations) or (b)

diverge and become less similar (negative dyad-level corre-

lations).

In the present case, the data pattern suggests that their shared

interaction experience led the dyad members to become similar

in the proportion, but not necessarily in the number, of negative

self entries they reported. None of the correlations between self-

reported discomfort and the measures of direct, negative part-

ner entries was significant, a fact that tends to bolster our confi-

dence in the discriminant validity of the thought-feeling mea-

sures.

As the data in the lower portion of Table 2 reveal, self-re-

ported liking for one's interaction partner was correlated, as

expected, with measures of direct, positive partner entries.

These effects were significant at both the individual and dyad

levels. The dyad-level correlations document the nonindepen-

dence of the dyad members' responses. They indicate that dyad

members tended to report similar numbers and percentages of

positive thoughts and feelings about each other and similar rat-

ings of liking for each other as a consequence of their shared

interaction experience. The adjusted individual-level corre-

lations further reveal, however, that individual differences were

preserved in spite of this mutual influence. In other words, indi-

Table 2

Concurrent and Discriminant Thought-Feeling Correlates of

Self-Reported Discomfort and Liking for Partner

Thought-feeling measure UNADJ DYAD ADJ

Self-reported discomfort

# negative self entries (D)% negative self entries (D)# negative partner entries (D)% negative partner entries (D)

.23*

.28*

.13

.12

.20

.42**

.18

.24

.29

.10

.05-.06

Liking for partner

# positive partner entries% positive partner entries# positive self entries% positive self entries

.49****

.47****-.02-.11

.57****

.46**

.09-.02

.31*.48***

-.18-.25

Note. UNADJ = unadjusted individual-level correlations, df = 58;DYAD = dyad-level correlations, df= 28; ADJ = adjusted individual-level correlations, dj'= 29. D = direct perspective.*p < .05. **p < .01. ***p< .005. ****;> < .001.

viduals who reported the largest number and percentage of pos-

itive partner entries also tended to report the greatest amount

of liking for their partners. None of the correlations between

liking for one's partner and the measures of direct, positive self

entries was significant, again supporting the discriminant valid-

ity of the thought-feeling measures.

In summary, the patterns of relation between specified

thought-feeling measures and the indexes of self-reported dis-

comfort and liking for one's interaction partner revealed consis-

tent support for both the concurrent and the discriminant va-

lidity of the thought-feeling assessment.

Convergent/Discriminant Validity

To provide further evidence for the convergent/discriminant

validity of the thought-feeling assessment, we extended the

logic of assessing the predictive validity of personality traits. If

specific thought-feeling measures can be predicted to relate to

specific personality traits, then evidence for such relations

should support the convergent and—by virtue of the patterns

of such correspondence—the discriminant validity of the

thought-feeling measures at the same time that it supports the

predictive validity of the trait measures themselves.

Following this reasoning, we predicted that (a) measures of

thoughts and feelings about self from the subjects' own direct

perspective should relate to their scores on the private serf-con-

sciousness factor of Fenigstein et al.'s (1975) Self-Conscious-

ness Scale, (b) measures of thoughts and feelings about self

from their partners' perspective (i.e., meta-perspective self en-

tries) should relate to subjects' scores on the public self-con-

sciousness factor, and (c) measures of negative thoughts and

feelings about self (summed across both perspectives—direct

and meta) should relate to their scores on the social anxiety

factor. We further predicted that (d) measures of the predomi-

nance of feelings in the subjects' reports should relate to their

level of internal correspondence, as measured by the first au-

72 ICKES, ROBERTSON, TOOKE, AND TENG

Table 3

Thought-Feeling Correlates of Private Self-Consciousness,

Social Anxiety, and Internal Correspondence

Thought-feeling measure UNADJ DYAD ADJ

Private self-consciousness

*> self entries (D) .21* .18 .23% self entries (D) .31" .46** .16# neutral self entries (D) .34*** .36* .31*% neutral self entries (D) .35*** .39* .31*

Social anxiety

# negative self entries% negative self entries

# negative self thoughts

% negative self thoughts* negative self feelings% negative self feelings

.14

.25*

.29*44****

-'l8°b

-.06b

.17

.28

.31*

.40*-.16b

-.02.

.11

.22

.28.

.49***-.19b

-.10b

Internal correspondence

# feelings .14.27*

-.10.19

.39*

.37*

Note. UNADJ = unadjusted individual-level correlations, df = 58;DYAD = dyad-level correlations, df= 28; ADJ = adjusted individual-level correlations, df= 29. D = direct perspective. Correlations for thenegative self thought measures having an a subscript diner significantly,p < .05, from those for the corresponding negative self feeling measureshaving a b subscript.*p < .05. **p < .01. ***p < .005. ****p < .001.

thor's modification of Brickman's (1978) Internal-External

Correspondence Scale.2

The first three of these predictions derived not only from the

theoretical definitions of the three Self-Consciousness Scale fac-

tors (cf. Buss, 1980; Carver & Scheier, 1981; Fenigstein et al.,

1975; Scheier & Carver, 1983), but also from the results of rele-

vant empirical research (e.g., Turner, Scheier, Carver, & Ickes,

1978). The fourth prediction derived from Brickman's (1978)

conception of internally correspondent individuals as those

whose subjective reality is dominated by feelings. In order to

score as internally correspondent on Brickman's Internal-Ex-

ternal Correspondence Scale, individuals must consistently

choose behaviors correspondent to their feelings (i.e., as op-

posed to desired external consequences) across a set of forced-

choice scale items.

Private self-consciousness. As predicted, private self-con-

sciousness scores were positively correlated with measures of

thoughts and feelings about the self from the subjects' own di-

rect perspective (see Table 3, upper part). In general, the pattern

of results for the set of conceptually and empirically overlapping

measures of direct, self entries reveals that their correlations

with private self-consciousness were primarily at the dyad level.

According to Kenny and LaVoie's (1985) analysis, these sig-

nificant dyad-level correlations would suggest that the dyad

members reported similar numbers and percentages of direct,

self-relevant thoughts and feelings as a consequence of their

shared interaction experience. Despite this mutual influence at

the dyad level, however, the presence of significant residual co-

variance for the last two thought-feeling measures indicates

that the rank ordering of individuals still tended to be pre-

served. In other words, the persons who ranked highest in pri-

vate self-consciousness still tended to report the greatest num-

ber and percentage of direct-perspective self entries.

In addition to testing the validity prediction just described,

we tested the ancillary hypothesis that the total number of

thoughts and feelings recalled by the subjects would vary ac-

cording to their level of private self-consciousness. This hypoth-

esis was not supported, inasmuch as all of the correlations be-

tween subjects' private self-consciousness scores and the num-

ber of thoughts, feelings, and total entries they reported were

nonsignificant and in the range of -.02 to .08. In other words,

private self-consciousness was not associated with reporting a

greater number and percentage of all thoughts and feelings, but

only of those thoughts and feelings whose object was the self.

This outcome suggests that the disposition of private self-con-

sciousness may not facilitate subjects' introspective access to all

of their subjective experience, but only to those aspects for

which relevant self-schema are available. Although such a con-

clusion is speculative and in need of further confirmation, it is

consistent with the position taken by many researchers who

have studied the role of self-schema in human cognitive pro-

cesses (e.g., Bargh, 1982; Greenwald, 1981; Hastie, 1981; Hull

& Levy, 1979; Kuiper, 1981; Markus, 1977).

Public self-consciousness. The predicted correlations were

not found between subjects' public self-consciousness scores

and the number and frequency of their meta-perspective

thoughts and feelings about themselves (i.e., self entries reflect-

ing the subjects' adoption of their partners' perspective). The

relevant correlations were all nonsignificant, both at the indi-

vidual level (adjusted and unadjusted) and at the dyad level.

Social anxiety. As the data in the middle part of Table 3

reveal, the predicted correlation between subjects' social anxi-

ety scores and the measures of their negative thoughts and feel-

ings about themselves was significant only for the percentage of

such entries, r = .25, p < .05, and not for the number, r = .14,

ns. In an attempt to further clarify this relation, we broke down

the negative self entry measures into indexes of negative self

thoughts and negative self feelings.

We found—contrary to our intuitive expectations—that so-

cial anxiety was clearly related to the number and percentage of

negative self thoughts but was not related to the number and

percentage of negative self feelings. This pattern of results was

observed across all three types of correlation (unadjusted indi-

vidual level, adjusted individual level, and dyad level), indicat-

ing that the covariation at the individual level was still reliable

even though the shared interaction experience of the dyad

members led them to report similar numbers and percentages

of negative self entries. Furthermore, in both the unadjusted

and the adjusted individual-level data, the correlations for the

though! measures were significantly different from those for the

correspondingyee/;«# measures (designated by the subscripts in

Table 3).

A simple interpretation of this last outcome is that social anx-

iety is a cognitive, rather than an affective, disposition. A more

2 To the best of our knowledge, Brickman did not publish the original

version of his Internal-External Correspondence Scale prior to hisdeath in 1982. We lack the space to reproduce either the original or themodified version here, but both versions can be obtained on request

from the nrst author.

NATURALISTIC SOCIAL COGNITION 73

complicated alternative explanation is that social anxiety ismanifested as a "self-attributed" emotion, that is, one that isexperienced whenever a change in one's level of generalizedarousal is "explained" in terms of the available cognitions thatcould plausibly account for such a change (e.g., Bern, 1972; Pat-terson, 1976; Schachter, 1964; Schachter & Singer, 1962). Thissecond explanation is clearly less parsimonious than the firstand leads to theoretical complications that cannot be consid-ered here. We are intrigued, however, by the possibility that the

expanded dyadic interaction paradigm—through its assess-ment of naturally occurring, subject-differentiated thoughtsand feelings—can be used to provide more definitive tests ofcontemporary theories of emotion such as Schachter and Sing-er's (1962) cognition-arousal theory.

Internal-external correspondence. According to Brickman's(1978) conceptualization, feelings should predominate in themental life of individuals as a function of the degree to which

they report themselves to be internally versus externally corre-spondent. As the data in the lower part of Table 3 indicate, thisprediction received clearcut support in terms of the adjustedindividual-level correlations. In general, the individuals havingthe highest internal correspondence scores also tended to reportthe greatest number and percentage of feelings, as opposed tothoughts.

Another aspect of Brickman's (1978) conceptualization ofinternal-external correspondence was tested, that is, his predic-tion that "females will emphasize internal correspondence"whereas "males will emphasize external correspondence" (p.19). Consistent with this prediction, the average internal corre-

spondence score of the women in the present sample was sig-nificantly greater, M - 16.8, than that of the men, M = 14.6,((58) = 2.57,p<,02.

In summary, the data offered moderate to strong support forthe convergent/discriminant validity of the thought-feelingmeasures theoretically related to social anxiety, internal corre-spondence, and private self-consciousness. However, they pro-vided no such support in the case of public self-consciousness.Because a lack of support could be due to the poor predictivevalidity of the relevant trait measure rather than to the invalid-ity of the corresponding thought-feeling measure(s), the pres-ence of such support is clearly more informative than its ab-

Exploratory Findings

Behavioral correlates of specific thought-feeling indexes.Certain patterns of correlation between the various thought-feeling indexes and the various behavioral measures were virtu-ally self-evident in their strength and consistency. The first twoof these patterns, both represented in Table 4, revealed that thedynamic behaviors most clearly indicative of interactional in-volvement were significantly correlated with (a) the total num-ber of thoughts and feelings that subjects reported and (b) thepercentage of those thoughts and feelings that were positive andthat focused on the interaction partner as their target or object.

The first data pattern revealed a positive relation between theamount of behavioral involvement subjects displayed and theamount of concomitant mental activity they remembered expe-riencing during these initial interactions. The positive corre-lations at the dyad level indicate that their shared interaction

experience led the dyad members to display similar levels ofinvolvement through the various interaction behaviors and toreport similar numbers of thoughts and feelings as well. In addi-tion, the significant adjusted individual-level correlations revealthat the dyad members who looked most often at their partnersand provided them with the largest number of verbal reinforcersalso tended to report the greatest number of thoughts and feel-ings. Because these two specific behaviors (looking at one's part-ner and encouraging him or her to continue talking) are the

ones most prototypic of the listener role in social interaction(Duncan & Fiske, 1977), the adjusted correlational data clearlysuggest that "good listeners" are the individuals most likely torecall a large number of the thoughts and feelings they experi-ence in their social encounters.

The second data pattern revealed a positive relation betweenthe subjects' interactional involvement and the percentage oftheir total entries that expressed positive thoughts or feelingsabout their interaction partner (%POS PARTNER). The positivecorrelations at the dyad level indicate that their shared interac-tion experience led the dyad members to display similar levelsof interactional involvement and to report similar percentagesof positive partner entries. The significant adjusted individual-level correlations further reveal that the dyad members whotalked least often but who looked longest and smiled most oftenat their interaction partners tended to report the largest percent-age of positive partner entries. Apparently, the "silent worship"phenomenon—displaying one's positive reactions to anothermore through nonverbal than through verbal expression—hasbeen captured in these results.

Because causal direction cannot be specified in the correla-tional data reported in Table 4, these findings lend themselvesto either or both of the following interpretations:

1. Increasing interactional involvement with one's partnernot only stimulates the partner to reciprocate this behavior butalso generates an increasingly larger number of interactionevents that can serve as the stimuli for both persons' thoughtsand feelings, particularly positive thoughts and feelings about

one's partner.2. As the number of thoughts and feelings that the dyad

members experience increases, the likelihood increases thatsome of these thoughts and feelings will serve as the stimuli foractions that represent increased involvement with each other(talking, looking, smiling, etc.), with positive thoughts and feel-ings about one's partner being particularly likely to lead to this

outcome.We believe that both of these processes occur in interaction

and that both interpretations are therefore valid in the limitedsense that each is incomplete without the other. In other words,we believe that private mental experience and public interac-tion behavior are intertwined through reciprocal causal re-lations that encompass both of the processes just described.

The same type of interpretation can be offered for the thirdpattern of correlations that was evident in the data (see Table 5).These data reveal that interactional involvement was positivelycorrelated with the percentage of positive thoughts and feelings(%POS) the dyad members reported, and negatively correlatedwith the percentage of negative thought-feeling entries(%NEG). The effects were generally strongest at the dyad level,revealing interaction-based similarity in the dyad members' re-sponses. The adjusted individual-level correlations further re-

74 ICKES, ROBERTSON, TOOKE, AND TENG

Table 4Behavioral Correlates of the Total Number of Thought-Feeling Entries and the Percentage of Positive Partner Entries

Behavioral measure

Total no. of entries Positive partner entries (%)

UNADJ DYAD ADJ UNADJ DYAD ADJ

Verbalizations (frequency)Verbalizations (duration)Directed gazes (frequency)Directed gazes (duration)Positive affect (frequency)Positive affect (duration)Gestures (frequency)Gestures (duration)Verbal reinforcers (frequency)Mutual gazes (frequency)Mutual gazes (duration)

.33"

.33"

.43"*'

.37"*

.37"*

.20

.39*"

.36*"44****

.40*

.50***

.49***

.47***

.45**

.34*

.48***

.45**

.42*

.49***

.39*

.13-.13

.33*

.07

.15-.01

.20

.18

.52"*

.30*

.40*

.40*

.50*

.32*

.28*

.34*

.25*

.31*

.41*** .46**** .47*"** _52...

.33*

.43**» _53«,

.40*' .34*

.58"**

.46"

-.46***.20.05.44**.31*.00

-.18-.11

.23

Note. UNADJ = unadjusted individual-level correlations, df =correlations, df= 29.*p<.05. "p<.01.***p<.005. **"p<.001.

58; DYAD = dyad-level correlations, df= 28; ADJ = adjusted individual-level

vealed that the subjects' predominant affective state was most

clearly expressed at the individual level by the frequency of

partner-directed gazes.

The findings for the %POS thought-feeling index appear to

be a more general form of those reported in Table 4 for the

%POS PARTNER index, encompassing positive thoughts and feel-

ings about a variety of objects or targets that are not limited to

one's interaction partner (e.g., self, third-party others, environ-

mental objects or events). As in the earlier case, a high percent-

age of positive thoughts and feelings could be either the cause

or the consequence of a high level of interactional involvement,

with both relations possible if causal reciprocity is assumed.

By the same token, the pattern of negative correlations for the

%NEG index might also reflect reciprocal causal relations.

That is, a high percentage of negative thoughts and feelings

could be either the cause or the consequence of a low level of

interactional involvement.

To further explore these findings, we tested the possibility that

personal moderating variables such as those described by Sny-

der and Ickes (1985, pp. 896-904) may have regulated the de-

gree of correspondence between the subjects' predominant

affective states and their level of interactional involvement.

These ancillary analyses focused on two personal moderating

variables: private self-consciousness and internal correspon-

dence.

For subjects at the high and low levels of each of these moder-

ating variables, we recomputed the correlations between the

most reliable, highly aggregated, multiple-act measures of their

predominant affective states (i.e., the %POS, %NEU, and

%NEG thought-feeling indexes) and the behavioral measures

of their interactional involvement. Because the results of these

analyses revealed that most of the significant moderating vari-

able effects occurred only at the individual level of analysis (un-

adjusted correlations), only the unadjusted individual-level

effects are reported here.

Tests of independent moderating variable effects. Private self-

Table 5

Behavioral Correlates of the Percentages of Positive and Negative Thought-Feeling Entries

Positive entries (%)

Behavioral measure

Verbalizations (frequency)Verbalizations (duration)Directed gazes (frequency)Directed gazes (duration)Positive affect (frequency)Positive affect (duration)Gestures (frequency)Gestures (duration)Verbal reinforcers (frequency)Mutual gazes (frequency)Mutual gazes (duration)

UNADJ

.44****

.41****

.49**"

.60*"*

.41*"*

.40**"

.46"**

.38*"

.42**"

DYAD

.54"*

.55*"*

.54***

.70"**

.45"

.61"**

.56****

.45"

.48"*

.68****

.64****

ADJ

-.10-.08

.33*

.20

.23

.00

.17

.22

.23

Negative entries (%)

UNADJ

-.28*-.18-.32"-.34***-.19-.23*-.24*-.18-.25*

DYAD

-.34*-.27-.29-.41*-.22-.36*-.29-.20-.30-.39*-.40*

ADJ

.04

.12-.58*"*-.11-.09

.01-.10-.14-.10

Note. UNADJ = unadjusted individual-level correlations, df= 58; DYAD = dyad-level correlations, df-

correlations, df= 29.*p< .05. **p< .01. "*p < .005. *"*p< .001.

' 28; ADJ = adjusted individual-level

NATURALISTIC SOCIAL COGNITION 75

INTERNAL CORRESPONDENCEAS A MODERATOR VARIABLE

VARIANCE (r*| ACCOUNTED FOR BY PERCENTAGE Of POSITIVE ENTRIES60

HIGHINT-CORRESP

LOWINT-CORRESP

PRIVATE SELF-CONSCIOUSNESSAS A MODERATOR VARIABLE

VARIANCE Ir'l ACCOUNTED FOR BY PERCENTAGE OF POSITIVE ENTRIES60

HIGHPRIVATE S-C

LOWPRIVATE S-C

SPK TUflNS GAZES SMILES GESTURES VERB RFS

INTERACTION BEHAVIORS

SP« TURNS GAZES SMILES GESTURES VERB RFS

INTERACTION BEHAVIORS

VARIANCE (r*) ACCOUNTED FOR BY PERCENTAGE OF NEUTRAL ENTRIES

60

VARIANCE lr j) ACCOUNTED FOR BY PERCENTAGE OF NEUTRAL ENTRIES

HIGHINT-CORRESP

LOWINT-CORHESP

V/////A

SPK TURNS GAZES SMILES GESTURES VERB RFS

INTERACTION BEHAVIORS

HIGHPRIVATE S-C

LOWPRIVATE S-C

-.34 -34

SPK TURNS GAZES SMILES GESTURES VERB RFS

INTERACTION BEHAVIORS

VARIANCE {rz) ACCOUNTED FOR BY PERCENTAGE OF NEGATIVE ENTRIES60

HIGHINT-COHHESP

LOWINT-CORRESP

-.42

SPK TURNS GAZES SMILES GESTURES VERB NFS

VARIANCE (r2) ACCOUNTED FOR BY PERCENTAGE OF NEGATIVE ENTRIES

60

HIGHPRIVATE S-C

LOWPRIVATE S-C

SPK TURNS GAZES

INTERACTION BEHAVIORS INTERACTION BEHAVIORS

Figure 2. Independent moderating variable effects of internal correspondence and private self-conscious-ness. (SPK TURNS = frequency of verbalizations; GAZES = frequency of directed gazes; SMILES =frequency of expressions of positive affect; GESTURES = frequency of expressive gestures; and VERBRFS = frequency of verbal reinforcers.)

76 ICKES, ROBERTSON, TOOKE, AND TENO

consciousness was not correlated in this sample with either in-

ternal correspondence or gender (both rs = .02, m), whereas

internal correspondence was significantly related to gender (r =

.32, df= 58, p < .01) in the manner described earlier (women

tend to score higher than men).

To determine the independent effects of private self-con-

sciousness and internal correspondence as personal moderating

variables, we split the distributions of scores for each variable

at the median to create high and low groups. We then examined

the correlations within each group between the relevant

thought-feeling indexes and the nonredundant (i.e., frequency

only) measures of the five individual-level behaviors (i.e., as op-

posed to dyadic behaviors such as mutual gaze) implicated in

the Table 5 results. By partialing out the effect of gender in each

case, we were able to compare the moderating effects specific

to internal correspondence with those specific to private self-

consciousness. These effects are depicted in Figure 2.3

An examination of the data in Figure 2 suggests at least four

observations worth noting. First, both internal correspondence

and private self-consciousness appear to moderate the strength

of the correlations between the percentage of positive thoughts

and feelings that the subjects reported (%POS index) and behav-

ioral measures of their interactional involvement. Second, at

the high levels of both moderating variables, the correlations

between the %POS thought-feeling index and the behavioral

measures are quite substantial, with r2s accounting for 35%-

50% of the variance in behavior. Third, internal correspondence

appears to moderate these effects at least as strongly as does

private self-consciousness.

Fourth, despite their similarity in moderating the relation be-

tween the %POS index and interactional involvement, internal

correspondence and private self-consciousness appear to

differentially moderate the relation between the %NEU and

%NEG indexes and interactional involvement. Specifically, in-

ternal correspondence systematically moderates the relation of

%NEG, but not that of %NEU, with interactional involvement.

In contrast, private self-consciousness systematically moderates

the relation of %NEU, but not that of %NEG, with interac-

tional involvement.

Although these features of the data in Figure 2 are evident by

visual inspection, we felt it was desirable to provide estimates

of their statistical significance as well. Toward this end, we sub-

jected the measures of the five nonredundant, individual inter-

action behaviors to a principal-components factor analysis in

order to first determine whether the behavioral data could be

further reduced. The algorithm extracted a single factor having

an eigenvalue of 3.69 and accounting for so much of the total

factor variance (74%) that the solution could not be rotated to

extract additional factors from the residual. Each subject's

score on this factor was then computed as a global, unitary mea-

sure of his or her level of interactional involvement, and this

measure was used in a reanalysis of the independent moderating

effects of internal correspondence and private self-conscious-

ness.

The results of this reanalysis, represented by the data in Table

6, added substance to the conclusions suggested by the graphs

in Figure 2. The following findings were obtained:

1. The positive correlations between the %POS index and in-

teractional involvement were significant at the high, but not at

the low, levels of both moderator variables.

2. The difference in the size of these correlations at the high

versus low levels was significant (using the independent corre-

lations estimate) for the internal correspondence variable, rs =

.68 versus .23, z = 2.23, p < .04, but not for the private self-

consciousness variable, rs = .64 versus .39, z = 1.24, ns.

3. The negative correlation between the %NEG index and

interactional involvement was significant at the high level of

internal correspondence, r = —.47, and differed with marginal

significance from the near-zero correlation obtained at the low

level, r = -.01, z = -1.87,;) < .075.

4. By contrast, the correlation between the %NEG index and

interactional involvement was not significant at the high level

of private self-consciousness, r = —.20, but was significant at

the low level, r = -.32, although the difference between the two

was not reliable, z< \,ns.

5. The %NEU index was unrelated to interactional involve-

ment at the low level of private self-consciousness, r = .01, but

was negatively correlated with it at the high level, r = —.40,

although the difference again fell short of being significant, z =

-1.58, p<.12.

In attempting to interpret these data, we relied heavily on

Brickman's (1978) and Carver and Scheier's (1981) theoretical

conceptualizations. In both conceptualizations, a prototypic

instance of correspondence between affectively relevant disposi-

tions and behavior is one in which the disposition is antecedent

to some behavior that subsequently expresses or derives from

it.4 If we view the Table 6 data with this causal ordering in mind,

it appears that the internal correspondence variable moderates

the relation between affectively relevant dispositions and behav-

ior in essentially the way that Brickman (1978) proposed: The

greater the internal correspondence, the greater the impact of

the predominant affective state on behavior. Although Brick-

man did not specify the process(es) by which the translation of

affective state into behavior might occur, the theoretical flavor

of his analysis suggests (to us, at least) that this translation oc-

curs in an essentially mechanistic way—via automatic, as op-

posed to controlled, information processing (cf. Schneider &

Shiffrin, 1977; Shiffrin& Schneider, 1977).

Carver and Scheier (1981), on the other hand, have been quite

explicit in characterizing the moderating function of the private

self-consciousness variable in terms of controlled information

processing. Individuals who are high, as opposed to low, in pri-

vate self-consciousness are assumed not only to be more aware

3 Clearly, because the %POS, %NEU, and %NEG indexes must sumto 100%, they are not independent of each other. For this reason, it is

not appropriate to interpret any of the rs, r\ or ps reported for any oneof these indexes in Figures 2 or 3 or in Tables 5-6 as independent ofthose for the remaining two indexes. Although we have elected to report,and occasionally to discuss, these values individually in order to providea complete and detailed description of the data, the reader should keep

in mind that it is the interdependent patterns of the %POS, %NEU, and%NEG data that are significant, rather than the data for a given indextaken separately. In general, the consistency and multiple corroboration(Lykken, 1968) of these and the other reported findings tend to mini-mize the possibility of Type I error, but additional safeguards (e.g., nor-malizing transformations of percentage-score measures) should proba-

bly be added in future research.4 As before, an interpretation of these data that reverses the implied

causal ordering is also plausible (i.e., that as a consequence of increasinginteractional involvement, high private self-conscious subjects—being

NATURALISTIC SOCIAL COGNITION 77

Table 6

Independent Moderating Effects of Internal Correspondence and Private Self-Consciousness on the Correlations

Between Selected Thought-Feeling Measures and Interactional Involvement

Thought-feeling measure

Internal correspondence Private self-consciousness

Low High Low

"p< .05. **p< .01. *"p< .005.

High

% positive entries (%POS)% neutral entries (%NEU)% negative entries (%NEG)% meta entries (%META)

.23-.20-.01

.15

.05

.04

.00

.02

.68"*-.14-.47**-.04

.47

.02

.22

.00

.39**

.01-.32**-.27

.15

.00

.10

.08

.64***-.40*-.20

.37*

.41

.16

.04

.14

of their cognitive and affective dispositions but to exert more

conscious control over how, or even whether, these dispositions

are expressed in behavior. If, for example, the behavioral ex-

pression of such a disposition is judged to be incorrect or inap-

propriate with respect to some situational or personal "standard

of behavior" (Carver &Scheier, 1981, pp. 119-141), behavioral

s»/>pression rather than expression of the disposition is likely

to occur. This view of private self-consciousness as exerting its

influence as a personal moderating variable by means of con-

trolled rather than automatic information processing is con-

vincingly supported by the results of several experimental stud-

ies reported by Carver and Scheier.

These theoretical considerations led us to ask, What form

would controlled information processing take in the context of

an initial interaction with a stranger? The answer that suggested

itself was, Taking the interaction partner's perspective into ac-

count through the use of meta-perspective thoughts and feel-

ings. Accordingly, we computed for each subject grouping in

Table 6 the correlations between the percentage of meta-per-

spective entries (%META) the subjects reported and their level

of interactional involvement.

When we examined the independent moderating variable

effects reported in the last row of each panel of Table 6, we were

excited to find evidence for what was indeed a differential use

of controlled information processing via meta-perspective tak-

ing by subjects high versus low in private self-consciousness. For

subjects low in private self-consciousness, the %META index

tended to be negatively correlated with interactional involve-

ment, r = -.27. In contrast, for subjects high in private self-

consciousness, the comparable correlation was significantly

positive, r = .37, and the difference between these correlations

predisposed to engage in controlled information processing—are par-ticularly likely to think and feel about things from their partners' per-spective. In fact, although Brickman (1978) and Carver and Scheier(1981) both gave greater emphasis to the causal sequence in whichthoughts and feelings precede and give rise to behavior, they readilyacknowledged—along with theorists such as Klinger (1978)—that the

reverse causal sequence, in which behavior precedes and gives rise tothoughts and feelings, also commonly occurs. The notion of reciprocalcausality encompasses both of these possibilities and is compellinglyarticulated by Klinger(1977, 1978) in his discussion of the distinction

between operant and respondent thought-feeling content.

was also significant, z = 2.44, p < .02. No such effects were

evident in the data for the internal correspondence variable

(both rs were nonsignificant).

Viewed collectively, the private self-consciousness data offer

compelling support for Carver and Scheier's (1981) assumption

that subjects high in private self-consciousness are particularly

likely to control the behavioral expression/suppression of their

predominant affective states. When compared to low private

self-conscious subjects, high private self-conscious subjects dis-

played both a greater behavioral expression of positive thoughts

and feelings and a greater behavioral suppression of negative

thoughts and feelings. In doing so, they made it possible for us

to observe a data pattern in which (a) the percentage of positive

thoughts and feelings was highly correlated with behavior, (b)

the percentage of negative thoughts and feelings was essentially

uncorrelated with behavior, and (c) the percentage of neutral

thoughts and feelings was statistically constrained to be nega-

tively correlated with behavior (by virtue of the fact that, of

the remaining two percentages, one revealed a strong positive

correlation whereas the other revealed essentially no correla-

tion).

The data for the %META index not only reinforce this con-

trolled expression/suppression interpretation but also help to

specify the process involved. They do so by suggesting that sub-

jects high in private self-consciousness (a) use the potential for

control inherent in meta-perspective taking in order to facilitate

their involvement in an initial interaction with a stranger, and

(b) succeed in their goal according to the extent of this use. On

the other hand, subjects low in private self-consciousness, who

presumably are not predisposed to engage in controlled infor-

mation processing, appear to find that adopting the partner's

perspective can actually impede, rather than facilitate, their in-

teractional involvement (see Footnote 4).

Tests for "interactive" moderating variable effects. In addi-

tion to replicating the Figure 2 analyses with the unitary inter-

actional involvement measure, we decided to take our explora-

tions a step further by testing for possible "interactive" moder-

ating variable effects. Because internal correspondence and

private self-consciousness were empirically independent in this

sample, it was possible to sort the subjects into the cells of a 2 X

2 factorial, representing all four combinations of the high and

low levels of these two moderating variables. Correlations were

then computed within each cell between the %POS, %NEU,

78 ICKES, ROBERTSON, TOOKE, AND TENG

Internal Correspondence

Low High

Low

High

(n = 16)

% positive

% neutral

% negative

% meta

(n = 14)

% positive

% neutral

% negative

% meta

r

.04

-.12

.07

.29

r

.55'

-.13

-.27

.16

r2

.00

.01

.00

.08

r2

.30

.02

.07

.03

(n = 16)

% positive

% neutral

% negative

% meta

(n = 14)

% positive

% neutral

% negative

% meta

r

.65'"

.14

-.58"

-.45'

r

.70'"

-.63"

-.13

.56'

r2

.42

.02

.33

.20

r2

.48

.39

.02

.31

Figure 3. Interactive moderating effects of internal correspondence andprivate self-consciousness on the correlations between selectedthought-feeling indexes and interactional involvement. (*p < .05.**p<.0l.***/><.005.)

%NEG, and %META thought-feeling indexes and the singlesummary measure of interactional involvement, with theeffects of gender partialed out as before.

The results of this analysis, represented in Figure 3, indicatedthat private self-consciousness (PSC) and internal correspon-dence (1C) interacted in moderating the relations between the%NEG and %META indexes and interactional involvement.The difference between independent correlations that provideda measure of the "interaction effect" was marginally significantfor the %NEG data, z = -1.97, p< .06, and was clearly signifi-cant for the %META data, z = 2.53, j? < .02. The major compo-nent of the interaction effect in the %NEG data was that thecorrelation between the %NEG index and interactional involve-ment was significantly negative only for the low-PSC-high-ICsubjects, r = -.58. The basis of the interaction effect in the%META data was that the correlations were significantlydifferent in the right two cells (i.e., the high 1C cells) of the 2 X2 design in Figure 3, rs - -.45 and .56, z - -2.72, p < .01, butdid not differ in the left two cells (i.e., the low 1C cells), rs = .29and.!6,z< \,ns.

Given the personality characteristics of the subjects in eachof the four groups, these interaction effects make sense. Of thefour groups of subjects, only those in the low-PSC-high-ICgroup (Figure 3, upper-right cell) would be expected to displaya significant degree of correspondence between negative affectand behavior because they are the only subjects who (a) are pre-disposed to express both negative and positive affect in theirbehavior (high 1C), and (b) are not predisposed to reflectivelymonitor and suppress negative affect whenever its expression isjudged to be inappropriate or maladaptive (low PSC).

By the same token, the interaction effect in the %META dataseems to follow from the differences among the four types ofsubjects. Only the low-PSC-high-IC subjects, who appear tospontaneously "act out" their prevailing positive or negativefeelings (high 1C) without reflecting first on the social conse-quences (low PSC) would be expected to display a negative cor-

relation between meta-perspective use and interactional in-volvement. The reason is that, for these subjects in particular,adopting the partner's perspective and thereby coming to evalu-ate the appropriateness of expressing their current affectivestate is an activity likely to disrupt their usual tendency to spon-taneously act out their feelings as soon as they occur. The moreoften their typical mode of relating to others is disrupted in thisway, the more difficult it may be for them to maintain an opti-mal level of interactional involvement.

In contrast, using the partner's perspective in order to moni-tor their own affective expression should tend to facilitate,rather than impair, the interactional involvement of high-PSC-high-IC subjects (Figure 3, lower-right cell). These subjects (a)are predisposed to monitor and control which affective statesare expressed in their behavior (high PSC) and (b) have goodreason to be concerned that their "inappropriate" negativeaffect might indeed be expressed in the absence of such control(high 1C). Accordingly, they should be particularly likely toadopt their partner's perspective in order to facilitate interac-tional involvement by carefully controlling what they say anddo. In fact, the average number of meta-perspective entries re-ported by the high-PSC-high-IC subjects, M = 0.86, was sig-nificantly greater than that reported by the low-PSC-high-ICsubjects, M = 0.31, ̂ (28) = 7.96, p < .01, although the averagepercentage of meta-perspective entries reported by the twogroups did not differ significantly, Ms ~ 7.4 and 6.1, F(2%) =1.31, JW.

On the one hand, these data offer some fascinating insightsinto the personality dynamics that determine how negativeaffect and meta-perspective taking are related to the initial in-teraction behavior of different subject types. On the other hand,they complicate the predictive and explanatory uses of personalmoderating variables by revealing that these variables (a) do notall function in the same way, and (b) may exert interactive, aswell as independent, influences on the correspondence betweendispositions and behavior.

Behavioral and thought-feeling correlates of gender. Al-though the subjects' internal correspondence scores were sig-nificantly related both to their gender and to their report of thepredominance of feelings, gender per se was not significantlycorrelated with either the number or percentage of feelings re-ported (or with the number and percentage of thoughts re-ported); rs ranged from -.04 to .07. These findings do not sup-port the stereotype that men are thought-oriented whereaswomen are feeling-oriented. Instead, they suggest that the pre-dominance of feelings over thoughts in an individual's mentallife may have more to do with the individual's psychologicalorientation (e.g., degree of internal-external correspondence,sex role orientation) than with his or her biological gender.

Additional analyses revealed that the subjects' gender VMS sig-nificantly correlated with several other thought-feeling indexes,however, and these correlations are reported in Table 7 (upperportion). The findings they represent may be summarized asfollows:

1. Relative to men, women reported a greater percentage ofmeta-perspective entries (i.e., thoughts and feelings from theirpartners' perspective) and a correspondingly smaller percentageof direct perspective entries.

2. In addition to adopting their partners' perspective in agreater percentage of entries, women provided further evidence

NATURALISTIC SOCIAL COGNITION 79

Table 7

Thought-Feeling and Behavioral Correlates

of the Subjects' Gender

Measure Gender

Thought-feeling (%)Direct perspective entries (D)Meta-perspective entries (M)Positive entriesPositive partner entriesPartner entriesPartner entries (D)Negative entriesNegative self entriesNegative self entries (D)Negative self feelings (D)

BehavioralVerbalizations (frequency)Directed gazes (duration)Mutual gazes (frequency)"Mutual gazes (duration)8

Positive affect (duration)Verbal reinforcers (frequency)Body posture (closed vs. open)Body orientation (toward partner)Interpersonal distance"

-.25*.25*.33***.24*.27**.25*

-.22*-.25*-.28**-.31***

.25*

.36***

.24*

.34***

.25*

.28**-.22*

.23*-.23*

Note. D = direct perspective; M = meta perspective. Positive corre-lations indicate higher scores by women; df= 58 in all cases.• Denotes a dyadic rather than an individual behavior.*p < .05. **p < .025. ***p < .01.

of their involvement with their partners by reporting a greater

percentage of direct perspective thoughts and feelings about

their partners than did men.

3. The affective or emotional tone of the thoughts and feel-

ings reported by women during the 5-min observation period

was both more positive and less negative than that reported

by men.

4. In particular, women reported a greater percentage of pos-

itive thoughts and feelings about their interaction partner, and

a smaller percentage of negative thoughts and feelings about

themselves, than did men.

The subjects' gender was also related in this study to several

behavioral measures of interactional involvement (Table 7,

lower portion). These differences in the overt interaction behav-

ior of men and women generally replicate those found in previ-

ous research. In earlier studies employing the unstructured dy-

adic interaction paradigm, men have been observed to sit fur-

ther apart from their partners, orient their bodies less toward

them, and talk, look, and smile less, as well (Ickes & Barnes,

1977,1978; Ickes etal., 1979; Ickes & Turner, 1983).

Taken together, the patterns of correlation represented in Ta-

ble 7 suggest that the differences in the covert interaction expe-

riences of men and women may parallel the differences in their

overt interaction behavior. Both cognitively and behaviorally,

the men in this study appear to have been relatively uncomfort-

able, self-absorbed, and detached from their interaction part-

ners. By contrast, in their thoughts and feelings as well as in

their behavior, the women appear to have been more at ease and

more involved with their interaction partners.5

A gender-based moderating variable effect. The striking gen-

der differences just reported are complemented by a final pat-

tern of correlations that were evident in the data. This pattern,

represented in Table 8, indicated that the use of meta-perspec-

tive taking was positively correlated with interactional involve-

ment for men but not for women. Although 10 of the 11 corre-

lations for men in Table 8 are statistically reliable, the estimates

of their magnitude may be somewhat inflated owing to the fact

that only 4 out of 29 men (vs. 12 out of 31 women) reported

any meta-perspective use, x2(l) = 4.76, p <. 03. Nevertheless,

these data suggest an interesting parallel in the way that gender

and private self-consciousness function as personal moderating

variables.

Men, like subjects high in private self-consciousness, ap-

peared to use their partner's perspective in a controlled, instru-

mental way, to facilitate their interactional involvement.

Women, like subjects low in private self-consciousness, ap-

peared to use their partner's perspective in a less controlled,

more automatic way that was essentially unrelated to their level

of interactional involvement. Although gender and private self-

consciousness are empirically unrelated, they display at least

some functional similarity in the way they moderate the rela-

tion between the subjects' meta-perspective taking and the level

of their involvement in these initial, unstructured interactions.

Taken in sum, the findings reported in Tables 7 and 8 suggest

that women have an orientation toward social interaction that

is more involved, more spontaneous, more open, more positive,

and more partner-oriented than that of men. Men seldom adopt

their partner's perspective at all, and then do so for apparently

instrumental reasons (e.g., to cope with their own negative feel-

ings of awkwardness and anxiety, to achieve better expressive

control, to facilitate interactional involvement). In contrast,

women not only adopt their partner's perspective more fre-

quently, but do so in an apparently spontaneous, empathetic

way that may have more to do with the communal satisfaction

of sharing thoughts and feelings than with the instrumental goal

of making the interaction work. This speculation is consistent

with the view of other writers who have proposed that men are

socialized to exert more control over their affective states than

women do, by suppressing and/or ignoring any feelings that

might interfere with the goals they are trying to attain (e.g., Ba-

kan, 1966; Henley, 1977; Pleck& Sawyer, 1974;Sattell, 1976).

Conclusions and Implications

Viewed collectively, the data we have reported provide con-

siderable evidence for the reliability and validity of the thought-

5 It might be argued that the thoughts and feelings of men and womentend to parallel their interaction behavior because men and women usedifferent criteria in defining what constitutes a thought, a feeling, andso forth. This explanation, however, is not supported by the data. First,

most of the thought-feeling indexes that correlated significantly withgender in Table 7 represent dimensions (valence, target, perspective)judged with high interrater reliability by independent raters (1 man,1 woman) and not by the subjects themselves. Second, on the single

dimension that the subjects rf;rfrate themselves (thoughts vs. feelings),there was no hint of a gender difference. These data are therefore consis-tent with those reported by Reis, Senchak, and Solomon (1985) in find-ing no evidence that men and women use different criteria to evaluateeither the objective or the subjective aspects of their interaction experi-

ence.

80 ICKES, ROBERTSON, TOOKE, AND TENG

TablesGender Differences in the Correlations Between the Percentage ofMeta-Perspective

Thought-Feeling Entries and Behavioral Measures

Behavioral measure Men Women

Verbalizations (frequency)Verbalizations (duration)Directed gazes (frequency)Directed gazes (duration)Mutual gazes (frequency)Mutual gazes (duration)Gestures (frequency)Gestures (duration)Positive affect (frequency)Positive affect (duration)Verbal reinforcers (frequency)

.20

.36*

.35*

.31*

.45**

.59***

.77****

.76****

.46*»

.32*

.34*

-.09-.17-.29-.11-.18

.08-.13-.12-.07

.06-.27

1.072.022.441.592.472.184.184.112.081.012.33

as.05.02ns

.02

.04

.001

.001

.05ns

.03

Note. df= 27 (men) and 29 (women).*p < .05. ** p < .01. ***p < .002. **** p < .001.

feeling assessment used in the expanded dyadic interaction

paradigm. Our procedure for coding the content of subjects'

reported thoughts and feelings has been shown to yield high

interrater reliabilities, and various correlational findings have

supported the concurrent, convergent, and discriminant valid-

ity of several representative thought-feeling measures. In addi-

tion, other findings too numerous to reiterate have documented

the paradigm's utility in revealing the ways that specific

thought-feeling indexes are related to interaction behavior.

At least three broad conclusions are suggested by the results

of this research. First, the thoughts and feelings that occur to

two individuals during their initial interaction are in part deter-

mined by their own respective personality traits and disposi-

tions. Second, the general affective tone of the individuals'

thoughts and feelings is related to their degree of behavioral in-

volvement in the interaction, although the strength and nature

of this linkage is further moderated by personal moderating

variables such as internal correspondence and private self-con-

sciousness. Third, the relation between individuals' thought-

feeling content and the behavior they display is frequently one

of causal reciprocity. Just as thoughts and feelings can serve as

the stimuli evoking certain behaviors, so also can behaviors

serve as the stimuli evoking certain thoughts and feelings.

The expanded dyadic interaction paradigm is clearly success-

ful in enabling the intensive study of both the covert and overt

aspects of naturally occurring social interaction. As a conse-

quence of its success, however, the paradigm requires us to con-

front the fact that thoughts and feelings are related to behavior

in ways that are often complex and causally unruly. In contrast

to methodologies that artificially constrain the apparent com-

plexity of such relations in order to ensure a degree of theoreti-

cal simplicity, the expanded dyadic interaction paradigm is de-

signed for the more realistic task of making these complex re-

lations more accessible to scientific observation and analysis.

Toward this end, we would like to broadly define two major

paths for future research using the expanded dyadic interaction

paradigm. These two paths are represented by structural versus

external variable studies of social interaction (Duncan, 1969;

Duncan & Fiske, 1977).

Regarding the first of these paths, the expanded paradigm

seems to be ideally suited to structural studies of social interac-

tion. It permits the researcher to study not only the overt behav-

ioral structure of the interaction but its covert cognitive struc-

ture as well. A particular advantage of the paradigm in such

structural studies is the ability to specify the time-location of

the subjects' thoughts and feelings as their behavioral interac-

tion unfolds. In future studies, it may be possible to examine

the linkages between specific thoughts and feelings and specific

antecedent and consequent behaviors in extremely fine detail.

Regarding the second of these paths, the paradigm is also well

suited to external variable studies of social interaction (Duncan

& Fiske, 1977). The findings reviewed by Cacioppo and Petty

(1981) and those reported in the present article offer convincing

evidence that the thought-listing technique is sensitive to indi-

vidual difference variables. Employing this technique in the

context of the unstructured dyadic interaction paradigm makes

it possible to assess the influences of personality and individual

difference variables on social cognition and social behavior at

both the individual and dyad levels of analysis. Moreover, it

should be possible to study the influence of a wide range of other

external variables (e.g., the role relations of the dyad members,

preinteraction expectancies or stereotypes, environmental or

other situational variables) in a similar manner.

In summary, the expanded dyadic interaction paradigm pro-

vides an exceptionally flexible and powerful methodological ap-

proach to the study of social cognition and social behavior. It

represents a serious response to the challenge to study these

phenomena in the most naturalistic way possible without sacri-

ficing the objectivity of observation or the precision of measure-

ment that are the hallmarks of scientific inquiry.

References

Bakan, D. (1966). The duality of human existence. Chicago: Rand-

McNally.Bargh, J. A. (1982). Attention and automaticity in the processing of self-

relevant information. Journal of Personality and Social Psychology,

43,425-436.Bern, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Ad-

vances in experimental social psychology (Vol. 6, pp. 1-62). New

York: Academic Press.Brickman, P. (1978). Is it real? In J. H. Harvey, W. Ickes, & R. F. Kidd

NATURALISTIC SOCIAL COGNITION 81

(Eds.), New directions in attribution research (Vol. 2, pp. 5-34). Hills-

dale, NJ: Erlbaum.

Brock, T. C. (1967). Communication discrepancy and intent to per-

suade as determinants of counterargument production. Journal of

Experimental Social Psychology, 3,269-309.

Buss, A. H. (1980). Self-consciousness and social anxiety. San Fran-cisco: Freeman.

Byrne, D. (1971). The attraction paradigm. New York: Academic Press.

Cacioppo, 3. X, Glass, C. R., & Merluzzi, T. V. (1979). Self-statements

and self-evaluations: A cognitive response analysis of heterosexual

anxiety. Cognitive Therapy and Research, 3,249-262.

Cacioppo, J. T., & Petty, R. E. (1979). Attitudes and cognitive response:

An electrophysiological approach. Journal of Personality and Social

Psychology, 37, 2181-2199.

Cacioppo, J. X, & Petty, R. E. (1981). Social psychological procedures

for cognitive response assessment: The thought-listing technique. In

T. Merluzzi, C. Glass, & M. Genest (Eds.), Cognitive assessment (pp.

309-342). New York: Guilford.

Carter, R. F., Ruggels, W. L., Jackson, K. M., & Hefner, M. (1973).

Application of signaled stopping technique to communication re-

search. In P. Clarke (Ed.), New models for mass communication re-

search (pp. 15-43). Beverly Hills, CA: Sage.

Carver, C. S., & Scheier, M. F. (1981). Attention and self-regulation: A

control-theory approach to human behavior. New York: Springer-

Craighead, W. E., Kimball, W. H., &Rehak, P. J. (1979). Mood changes,

physiological responses and self-statements during social rejection

imagery. Cognitive Therapy and Research, 3,249-262.

Davison, G. C, Robins, C, & Johnson, M. K. (1983). Articulated

thoughts during simulated situations: A paradigm for studying cogni-

tion in emotion and behavior. Cognitive Therapy and Research, 7,

17-40.

Duncan, S. D., Jr. (1969). Nonverbal communication. Psychological

Bulletin, 72, 118-137.

Duncan, S. D., Jr., & Fiske, D. W. (1977). Face-to-face interaction: Re-

search, methods, and theory. Hillsdale, NJ: Erlbaum.

Fenigstein, A., Scheier, M. E, & Buss, A. H. (1975). Public and private

self-consciousness: Assessment and theory. Journal of Clinical and

Consulting Psychology, 43, 522-527.

Freud, S. (1939). The interpretation of dreams. In A. A. Brill (Ed. andTrans.), The basic writings ofSigmund Freud. New York: Random

House. (Original work published 1900)

Geller, V., & Shaver, P. (1976). Cognitive consequences of self-aware-

ness. Journal of Experimental Social Psychology, 12, 99-108.

Godden, D. R., & Baddeley, A. D. (1975). Context-dependent memory

in two natural environments: On land and under water. British Jour-

nal of Psychology, 66, 325-331.

Goffman, E. (1969). Strategic interaction. Philadelphia: University ofPennsylvania Press.

Greenwald, A. G. (1968). Cognitive learning, cognitive response to per-

suasion, and attitude change. In A. G. Greenwald, X C. Brock, &T. M. Ostrom (Eds.), Psychological foundations of attitudes (pp. 147-

170). New York: Academic Press.

Greenwald, A. G. (1981). Self and memory. InG. H. Bower (Ed.), The

psychology of learning and motivation. New York: Academic Press.

Griffin, D. R. (1984). Animal thinking. American Scientist, 72, 456-464.

Harvey, J. H., Yarkin, K. L., Lightner, J. M., & Town, J. P. (1980).

Unsolicited interpretation and recall of interpersonal events. Journal

of Personality and Social Psychology, 38, 551-568.

Hastie, R. (1981). Schematic principles in human memory. In E. XHiggins, C. P. Herman, & M. P. Zanna (Eds.), Social cognition: The

Ontario Symposium (Vol. 1, pp. 39-88). Hillsdale, NJ: Erlbaum.

Henley, N. M. (1977). Body politics: Power, sex, and nonverbal commu-

nication. Englewood Cliffs, NJ: Prentice-Hall.Hovland, C. L, Lumsdaine, A. A., & Sheffield, F. D. (1949). Experi-

ments on mass communication. Princeton, NJ: Princeton University

Press.

Hull, J. G., & Levy, A. S. (1979). The organizational functions of the

self: An alternative to the Duval and Wicklund model of self-aware-

ness. Journal of Personality and Social Psychology, 37,756-768.

Hurlburt, R. T. (1979). Random sampling of cognitions and behavior.

Journal of Research in Personality, 13, 103-111.

Hurlburt, R. T. (1980). Validation and correlation of thought sampling

with retrospective measures. Cognitive Therapy and Research, 4,

235-238.

Ickes, W. (1982). A basic paradigm for the study of personality, roles,

and social behavior. In W. Ickes & E. S. Knowles (Eds.), Personality,

roles, and social behavior (pp. 305-341). New York: Springer-Verlag.

Ickes, W. (1983). A basic paradigm for the study of unstructured dyadic

interaction. In H. Reis (Ed.), New directions for methodology of social

and behavioral science (pp. 5-21). San Francisco: Jossey-Bass.

Ickes, W. (1984). Compositions in black and white: Determinants of

interaction in interracial dyads. Journal of Personality and Social

Psychology, 47, 330-341.

Ickes, W., & Barnes, R. D. (1977). The role of sex and self-monitoring in

unstructured dyadic interactions. Journal of Personality and Social

Psychology. 35, 315-330.

Ickes, W., & Barnes, R. D. (1978). Boys and girls together—and alien-

ated: On enacting stereotyped sex roles in mixed-sex dyads. Journal

of Personality and Social Psychology, 36, 669-683.

Ickes, W., Patterson, M. L., Rajecki, D. W., & Tanford, S. (1982). Behav-

ioral and cognitive consequences of reciprocal versus compensatory

responses to preinteraction expectancies. Social Cognition, 1, 160-

190.Ickes, W., Schermer, B., & Steeno, J. (1979). Sex and sex-role influences

in same-sex dyads. Social Psychology Quarterly, 42, 373-385.

Ickes, W., & Turner, M. (1983). On the social advantages of having an

older, opposite-sex sibling: Birth-order influences in mixed-sex dyads.

Journal of Personality and Social Psychology, 45,210-222.

Kagan, N. I. (1977). Interpersonal process recall: A method of influenc-

ing human interaction [Training manual]. East Lansing: Michigan

State University.

Kenny, D. A., & La Voie, L. (1985). Separating individual and group

effects. Journal of Personality and Social Psychology, 48, 339-348.

Klinger, E. (1977). Meaning and void: Inner experience and the incen-

tives in people's lives. Minneapolis: University of Minnesota Press.

Klinger, E. (1978). Modes of normal conscious flow. In K. S. Pope &J. L. Singer (Eds.), The stream of consciousness: Scientific investiga-

tions into the flow of human experience (pp. 225-258). New York:

Plenum.

Kuiper, N. A. (1981). Convergent evidence for the self as a prototype.

Personality and Social Psychology Bulletin, 7, 438-443.Laing, R. D., Phillipson, H., & Lee, A. R. (1966). Interpersonal percep-

tion: A theory and method of research. New York: Harper & Row.

Lingle, J. H., & Ostrom, T. M. (1979). Retrieval selectivity in memory-

based impression judgments. Journal of Personality and Social Psy-

chology, 37, 180-194.

Lykken, D. T. (1968). Statistical significance in psychological research.

Psychological Bulletin, 70, 151-159.

Markus, H. (1977). Self-schemata and processing information about

the self. Journal of Personality and Social Psychology, 35, 63-78.Meichenbaum, D. (1977). Cognitive-behavior modification: An integra-

tive approach. New York: Plenum.Neisser, U. (1980). On "Social Knowing." Personality and Social Psy-

chology Bulletin, 6, 601-605.

Patterson, M. L. (1976). An arousal model of interpersonal intimacy.

Psychological Review 83, 235-245.Peterman, J. N., III. (1940). Program research. The Program

Analyzer—a new technique in studying liked and disliked items in

radio programs. Journal of Applied Psychology, 23, 725-741.Petty, R. E., & Cacioppo, J. X (1977). Forewarning, cognitive respond-

82 ICK.ES, ROBERTSON, TOOKE, AND TENG

ing, and resistance to persuasion. Journal of Personality and Social

Psychology, 35, 645-655.

Petty, R. E., & Cacioppo, J. T. (1979). Effects of forewarning and persua-

sive intent and involvement on cognitive responses and persuasion.Personality and Social Psychology Bulletin, 5, 173-176.

Pleck, J. H., & Sawyer, J. (Eds.). (1974). Men and masculinity. Engle-

wood Cliffs, NJ: Prentice-Hall.

Rajecki, D. W., Ickes, W., & Tanford, S. (1981). Locus of control andreactions to a stranger. Personality and Social Psychology Bulletin, 7,

282-289.

Reis, H. T., Senchak, M., & Solomon, B. (1985). Sex differences in the

intimacy of social interaction: Further examination of potential ex-planations. Journal of Personality and Social Psychology, 48, 1204-1217.

Rogers, T. B., Kuiper, N. A., & KJrker, W. S. (1977). Self-reference and

the encoding of personal information. Journal of Personality and So-

cial Psychology, 35, 677-688.

Sattel, J. W. (1976). The inexpressive male: Tragedy or sexual politics?Social Problems, 23,469-477.

Schachter, S. (1964). The interaction of cognitive and physiological de-

terminants of emotional state. In L. Berkowitz (Ed.), Advances in ex-

perimental social psychology (Vol. 1, pp. 49-80). New York: Aca-

demic Press.

Schachter, S., & Singer, J. E. (1962). Cognitive, social, and physiologicaldeterminants of emotional state. Psychological Review, 69. 379-399.

Scheier, M. E, & Carver, C. S. (1983). Two sides of the self: One for you

and one for me. In J. Suls & A. G. Greenwald (Eds.), Psychological

perspectives on the self (Vol. 2). Hillsdale, NJ: Erlbaum.

Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic hu-

man information processing: I. Detection, search, and attention. Psy-

chological Review, 84, 1-66.Schwartz, G. E. (1975). Biofeedback, self-regulation, and the patterning

of physiological processes, American Scientist, 63. 314-324.

Schwartz, G. E., Fair, P. L., Salt, P., Mandel, M. R., & Klerman, G. L.(1976). Facial expression and imagery in depression: An electromyo-

graphic study. Psychosomatic Medicine, 38, 337-347.Shim-in, R. M., & Schneider, W. (1977). Controlled and automatic hu-

man information processing: [I. Perceptual learning, automatic at-tending, and general theory. Psychological Review. 84, 127-190.

Smith, S. M., Glenberg, A., & Bjork, R. A. (1978). Environmental con-text and human memory. Memory and Cognition, 6, 342-353.

Snyder, M., & Ickes, W. (1985). Personality and social behavior. In G.

Lindzey & E. Aronson (Eds.), The handbook of social psychology:

Third edition. (Vol. 2, pp. 883-947). New York: Random House.Tulving, E. (1983). Elements of episodic memory. New \ork: Oxford

University Press.Turner, R. G., Scheier, M. F, Carver, C. S., & Ickes, W. (1978). Corre-

lates of self-consciousness. Journal of Personality Assessment, 42,

285-289.

Wegner, D. M., & Giuliano, T. (1982). The forms of social awareness.In W. Ickes & E. S. Knowles (Eds.), Personality, roles, and social be-

havior(pp. 165-198). New York: Springer-Verlag.

Wright, P. L. (1974). On the direct monitoring of cognitive response toadvertising. In G. D. Hughes & M. L. Ray (Eds.), Buyer/consumer

information processing. Chapel Hill: University of North Carolina

Press.

Received September 13,1985

Revision received February 14, 1986 •