You are on page 1of 13

Journal of Personality and Social Psychology Copyright 1990 by the American Psychological Association, Inc.

1990, Vol. 59, No. 4, 730-742 0022-3514/9O/S00.75

Naturalistic Social Cognition: Empathic Accuracy in Mixed-Sex Dyads
William Ickes, Linda Stinson, Victor Bissonnette, and Stella Garcia
University of Texas at Arlington

The research paradigm that Ickes and his colleagues developed for the study of naturalistic social
cognition was used to explore the phenomenon of empathic accuracy in the initial, unstructured
interactions of 38 mixed-sex (male-female) dyads. The results indicated that an important aspect of
empathic accuracy—content accuracy—could be measured reliably (a = .94) with the procedure
used. The results further indicated that content accuracy was, to a large extent, an emergent prod-
uct of social interaction processes occurring at the level of the dyad. Although many of the findings
could be explained in informational terms, some significant motivational influences were ob-
served as well. For example, content accuracy was influenced by the partner's physical attractive-
ness and, more generally, by the perceiver's apparent interest in the partner (as indexed by various
thought/feeling measures). The individual difference variables of grade point average and self-
monitoring also predicted the subjects' levels of content accuracy; however, gender and self-report
measures of empathic skills and empathic accuracy did not.

If the study of subjective phenomena involving or occurring individual minds tested "one at a time," are related to "social"
within a single conscious mind is the domain of mainstream stimulus materials whose features and contents have been pre-
cognitive psychology, it follows logically that the study of inter- determined by the experimenter.
subjective phenomena involving or occurring between at least There are, however, some important and noteworthy excep-
two conscious minds is the proper domain of cognitive social tions to this rule. Group dynamics researchers have for some
psychology (Ickes, Tooke, Stinson, Baker, & Bissonnette, 1988). time studied intersubjective phenomena under headings such
The logic of this conclusion has been consensually validated by as brainstorming (Diehl & Stroebe, 1987; Street, 1974), group
both present and past reviewers of social cognition research. decision making (Janis & Mann, 1977; Miller, 1989; Stasser,
For example, Markus and Zajonc (1985) ended their Handbook Kerr, & Davis, 1989) groupthink (Janis, 1972; McCauley, 1989),
chapter on cognitive social psychology by stating that "the prop- group polarization (Moscovici & Zavalloni, 1969; Myers, 1982;
erties of social perception and social cognition that make them Myers & Lamm, 1976), group socialization (Moreland & Le-
distinct are reciprocity and intersubjectivity" (p. 213). They vine, 1982; 1989) and majority and minority influence (Latane
noted that "many earlier authors, such as Mead (1934), & Wolfe, 1981; Maass, West, & Cialdini, 1987; Moscovici &
Merleau-Ponty (1970), Asch, (1952), and Heider (1958)," have Mugny, 1983; Nemeth, 1986). More recently, social cognition
drawn essentially the same conclusion. researchers have begun to address intersubjective phenomena
In general, intersubjective phenomena can be characterized such as consensus and meta-accuracy in person perception
as those involving some form of interdependence between the (Funder, 1980,1987; Funder & Harris, 1986; Kenny & Albright,
contents or processes of at least two conscious minds (cf. 1987; Malloy & Albright, 1990), dyadic intersubjectivity (Ickes
Wegner, Giuliano, & Hertel, 1985). Given this definition, inter- et al, 1988; Krauss & Fussell, 1989), and transactive memory
subjective phenomena are clearly not the most frequently stud- (Wegner etal, 1985).
ied phenomena in cognitive social psychology (Ickes et al., We believe that intersubjective phenomena are of central im-
1988). In most studies of human social cognition, researchers portance to the study of social cognition. To facilitate research
have not inquired how the contents and processes of one mind on these phenomena, we recently extended the original dyadic
are interdependent with those of another. Instead, using as their interaction paradigm (Ickes, 1982,1983) to permit the study of
models the studies conducted in more traditional areas of psy- both the overt behavior and the covert thoughts and feelings of
chology, they have inquired how the contents and processes of dyad members during a period of unstructured interaction
(Ickes, Bissonnette, Garcia, & Stinson, 1990; Ickes, Robertson,
Tooke, & Teng, 1986; Ickes & Tooke, 1988). By adding thought/
We would like to express our appreciation to Dorothy Anderson, feeling assessment procedures of the type developed by re-
Kristine Arnold, Ray Bristow, Susan Ferry, Jenny Gatlin, Andrea De- searchers at Ohio State University (e.g, Brock, 1967; Cacioppo,
Bolt, Mujahid Khan, Laurie Luce, Carol Marangoni, Marianne
Glass, & Merluzzi, 1979; Cacioppo & Petty, 1981; Greenwald,
Mickan, Keith Milstead, and Jacquetta Talley for their assistance in
coding and analyzing the data from this study. We also wish to thank 1968; Petty & Cacioppo, 1979), we are now able to study a wide
David Kenny for his suggestions regarding data analysis and David range of phenomena that can be classed under the general head-
Funder, Tom Malloy, Carol Marangoni, Paul Paulus, and Brian Spitz- ing of naturalistic social cognition.
berg for their constructive comments on a draft of this article.
Correspondence concerning this article should be addressed to Wil-
Empathic Accuracy
liam Ickes, Department of Psychology, University of Texas, Arlington, In the present study, the dyadic interaction paradigm was
Texas 76019. further extended to permit the study of another important in-

and (e) their responses on per- In a social interaction context. These in- to understand. or facility (b) serve as predictors of content and valence accuracy. time-logged listing of those actual for example. which concerned their thoughts and feelings during the example. and view the videotape a second time. Danish and identical cubicles where they were each instructed to view a Kagan (1971). perceptual accuracy (Goldstein & Michaels. Greif and Hogan videotaped copy of the interaction. Rogers (1957). interpret (c) with accuracy (d) cluded various measures of the dyad members' (a) verbal and the private. Is empathic accuracy related to attributional inference? In ther modify the dyadic interaction paradigm so that dyad other words. for study. empathic ability (Davis. apprehend. Kagan. In clinical and counseling psychology. cific content of an interaction partner's thoughts and feelings . social acuity (Funder & Harris. When 1985).and videotaped. Crow and Hammond (1957). The subjects were then instructed to O'Keefe and Sypher (1981). Truax experimenter returned at the end of the observation period. Gladstein (1977). Cline (1964). we sought to address a range of questions that included the following. table variations across writers on what appears to be the same Consistent with the exploratory nature of the study. they were debriefed more completely expect. Chapin (1942). Valence accuracy refers to the degree to which some fundamental insights about the processes underlying em- one interactant's inferences about the emotional tone (positive. as well. self-moni- defined more simply as the degree to which one interactant is toring (Snyder. during this viewing of the tape was to infer the content of their These include accurate empathy (Rogers. it was necessary to fur. 1974). Goldstein and Michaels subjects were partially debriefed. logged listing of those inferred thought/feeling entries. 1973). the and Carkhuff(1967). thoughts and feelings (f) of some other per. conceptual definitions of the construct reveal the inevi. Kerr and Speroff interaction. ability. study. the subjects were seated in separate but (1969). is the accuracy of one's inferences about the spe- members could report not only their own thoughts and feelings. empathic accuracy During this interval in which the subjects were ostensibly wait- has been widely regarded as one of the most important nonspe. infer. because empathic study. content of an interaction partner's thoughts or feelings matches Balanced against the substantial costs involved in such an the actual content of the specific thoughts or feelings reported ambitious data collection effort was the possibility of achieving by the partner. their verbal and nonverbal cific factors common to all successful therapeutic outcomes— behaviors were unobtrusively audio. reported thoughts and feelings. and Honeycutt. a valid self-report measure of empathic accuracy has been data and that they would participate in the second part of the the elusive goal of several generations of researchers—see. pathic accuracy in naturally occurring social cognition. The common elements in these definitions erately assessed a wide range of variables that potentially could suggest that empathic accuracy is (a) a skill. present investigation. each sub- an element that is essential to effective communication—see. When the see. With neutral. ject made a written. In personality psychol. 1957). empathic accuracy has been viewed as remembered having had a specific thought or feeling. Pearce (1976). the interaction they experienced. Campbell. both subjects had completed this task and filled out a short gence (Sternberg. 1988). 1985. 1969). By stopping the videotape (1973). Mehrabian and Reed (1968). Dymond (1949). and Krathwohl (1971). pathic accuracy (Ickes. The members of each dyad were led into a waiting tion and has perhaps the longest history of interdisciplinary room and left there together in the experimenter's absence. Walker & Foley. They then were asked for (1985). can be summarized briefly as falls within the more general domain of interpersonal percep. and social intelli. (1954). negative) of an interaction partner's thoughts or feel. stopped for them at each of those points at which their interac- Several terms have been used in these literatures as synonyms tion partner reported a thought or feeling. during which the tape was Powers (1983). Can the major aspects of empathic accuracy—content ac- To study these aspects of empathic accuracy in the context of curacy and valence accuracy—be measured reliably? naturally occurring social interaction. 1986). As one would posttest questionnaire. Although there are a number of (GPA) scores and ratings of physical attractiveness were also different aspects of empathic accuracy that one might choose to obtained for each of the subjects in the study. covert. follows. Knapp. Current grade point average tant's thoughts and feelings. 1989). their signed consent that the tape could be released for use as ogy. The subject's task or near synonyms of empathic accuracy (Marangoni. nonverbal interaction behavior. we delib- common theme. 1981). Hogan If consent was given. and Marks and Tolsma (1986). This procedure. for example. And in with a remote start/pause control at those points at which they communication studies. Content accuracy refers to the intelligence of the perceivers and the attractiveness of their degree to which one interactant^ description of the inferred "target" partner. thought/feeling entries. 1983) and em- able to accurately infer the specific content of another interac. 2. Mehrabian and Epstein (1972). and Funder and Harris (1986). the two aspects we regard as most important are content accuracy could plausibly be determined in part by both the accuracy and valence accuracy. empathic accuracy can be sonality measures of shyness (Cheek & Buss. Barrett-Lennard (1962). Davis (1983). subjective (e) phenomenological reality. which is described in detail in the Empathic accuracy is an intersubjective phenomenon that Method section of this article. (d) their posttest perceptions of son^). time- (Carkhuff. Overview of the Present Study 1. EMPATHIC ACCURACY IN MIXED-SEX DYADS 731 tersubjective phenomenon that our research had not previously but their inferences about their partner's thoughts and feelings addressed: empathic accuracy. and then thanked and released. age strangers of the opposite sex that were the focus of the ings reported by the partner. respect to the initial unstructured interactions between college- ings match the actual valence of the specific thoughts and feel. (b) the content of their own mental experience. discrimination partner's thoughts and feelings and provide a written. ing for the experiment to begin.

and the subjects used their start/pause controls to indepen- the responses of both men and women using only a single dyad type. Once the subjects' written The subjects were 38 female and 38 male undergraduates enrolled in consent had been obtained. and the subjects were assured that the menter's control station. A video camera feeling (coded "I was thinking:" or "I was feeling:" on the coding sheet). instead of releasing it for use as data. negative. or 0 on the coding sheet). this lecture is no longer given during the first several weeks of the Collection ofthe videotape data. seated on the couch. and (d) whether the entry was This arrangement permitted the dyad members' interaction to be re- presumed to be positive. see Ickes et al. as well experimenter returned to the observation room. . The test cubicles were each equipped with a help button that allowed subjects to signal the experimenter (by means of a red light) that they needed 1 If the subjects in a given session displayed any sign that they were additional information or assistance of some other type. Along one wall of the control room was the experi- pletely candid was emphasized. Mixed. a research assistant in the control room activated Voie. This level of suspicion was so unusual Following the directions given on their respective sign-up sheets. 1985)? the video equipment to begin videotaping. The instructions the subjects received explicitly encouraged them to Adjacent to the observation room was a control room used to house report all of the thoughts and feelings they remembered having had as the video and audio equipment operated by the experimenter and a accurately and honestly as possible. as opposed that she had only one copy of the informed consent form. of the study. For a schematic dia- strangers. the Method content of the interaction would remain their own private concern. she left the observation room to get additional copies. With the instructor's cooperation.732 ICKES. to one of two 25 inch color TV monitors that were oriented to face through one-way mirrors into identical test cubicles.the subjects stopped the videotape at each point during the interaction that they recalled Setting and Equipment having had a specific thought or feeling. dently start and stop theirrespectivecopies of the videotape. AND GARCIA related to one's tendency to make more global inferences about ing satisfied that they were indeed strangers. means of a concealed video camera. instruction tapes to give them detailed instructions regarding the whose pairing into mixed-sex dyads was essentially random. To what extent does empathic accuracy depend on pro. form to complete the initial phrase). the instructions cautioned them to that were both connected (by means of a V-adapter) to the video cam- report only those thoughts and feelings they distinctly remembered era in the storage room across the hall.2 She further explained that if one or both of them exercised the right to have the tape erased immediately. STINSON. we found that a common element in three of the four cases waiting areas in the psychology building. enced during the 6-min interaction period. Kenny & La return in a few minutes. Promising to to those occurring at the dyad level of analysis (cf. Is empathic accuracy reliably predicted by individual dif. having had during the observation period. . In this part. they allowed us to study activated. Each test already acquainted with each other. rooms with minimal likelihood of the camera being detected. ment for them to interact or get to know each other.. thought/feeling data collection procedure. Is empathic accuracy influenced by motivational. In two subjects who were scheduled for each session reported to different doing so. Exactly 6 min later. the experimenter described the next part introductory psychology classes at the University of Texas at Arling. They recorded each thought or feeling on a thought/feeling coding form by entering (a) the time the The observation room in which the dyad members' interaction was thought or feeling occurred (available as a digitalreadouton the upper- recorded was furnished with a long couch and an accompanying coffee left part of the video image). After meeting the subjects and be. and the videotaping as by informational. and (b) that each scheduled pair of sub. (1986. Each VCR was also connected. the experimenter the partner's traits and dispositions? escorted them into the observation room and directed them to be 3. When the subjects had eral pretesting at the start of the semester. After probing for any evidence of suspicion. 68). focused on the area of the couch and coffee table was concealed in a (c) the specific content of the thought/feeling entry (written in sentence darkened storage room across a hallway from the observation room. the experimenter used video later used to schedule the nonpreselected male and female volunteers. jects would not meet and interact before their session began. These subjects had completed Davis's (1983) measure of empathic and make a written record of the thoughts and feelings they experi- abilities and Ickes's (1988) measure of empathic accuracy during a gen.) Subjects Collection of the thought/feeling data. This separation helped to was attendance at a lecture given by an introductory psychology in- ensure (a) that two friends or previously acquainted people would not structor who discussed the unobtrusive measurement of behavior by sign up for the same session. the experimenter con- ferences in gender or by self-reported differences in empathic ducted a partial debriefing and informed the subjects that their written skills and dispositions? consent was required for the tape to be used as data. Consistent with the instructions theyreceived. Separate sign-up sheets were been seated in their individual cubicles. (b) whether the entry was a thought or a table that concealed an FM wireless microphone. factors? was terminated. 2 Four dyads were dropped from the study after the probe for suspi- Procedure cion revealed that one or both of the dyad members had located the video camera while being taped. ducing them to each other or providing any instruction or encourage- p. or neutral in its overall affective tone corded through the open doorways of the storage and observation (coded +. They were not toreportany respectively. At this point. the subjects were asked to view the videotape ton. the experimenter continued the procedure without intro- gram of the laboratory and its arrangement. The two VCRS were then sex dyads were used for reasons of economy. The importance of being com- research assistant. semester. 5. the experimenter cited equipment cubicle also contained (a) a remote start/pause switch connected to the problems and gave each subject an experimental point credit without VCR and TV monitor system unique to that cubicle and (b) a supply of requiring any further participation in the study.1 The experimenter then pretended to discover cesses occurring at the individual level of analysis. the in our research that we were motivated to investigate it further. BISSONNETTE. the 4. If the subjects were thought/feeling coding forms (to be described). On the other hand. Here the experimenter could sit in front of a thoughts and feelings theyrecordedwould not be seen by their interac- single table that supported two identical videocassette recorders (VCRS) tion partners. (None of the subjects requested to have the tape erased.

other(s) entries. the sub." "what a weirdo!"). The interrater agreement viewing the videotape. or -). feeling (coded "he/she was thinking:" or "he/she was feeling:" on the event. cialized utility program in the COLLECT YOUR THOUGHTS software vided by the total number of personal pronouns used by the dyad package. as much as Empathic accuracy measures. These values were then di. This program pro- long enough to complete a posttest questionnaire. and released. & Patterson. ments to be made about the degree of similarity in the written content person pronouns). This questionnaire duces as output various summary indexes of thought/feeling content.97. the subjects were asked to remain in their respective cubicles ware program called COLLECT YOUR THOUGHTS. dyad members' own reported thoughts and feelings. first-. This program accessed the relevant raw data from a master member to create percentage scores that previously have been shown data file and computed the number of matches and mismatches be- to be free of any confound with individual differences in the amount of tween the rated valence of the actual entries and the corresponding talking (Ickes.. They also included the percentages of pleted the questionnaire and returned it to the experimenter. on one's current interaction partner (second. thoughts and feelings were recorded. expressive gestures. (For operational definitions of the measures. smiles and laughter). The judges first inference about the nature of the specific thought or feeling reported decided whether the entry was reported as a direct perspective (sub- by their partner at that point. partner. (b) whether the entry was presumed to be a thought or a concerned self.g. partner. On this second viewing. Among these was a measure of the percentage of partner attribution entries. point credit for their participation. The coded thought/feeling data for the subjects within each dyad Collection of the final self-report measures. and ex." "she is snobbish. or circumstance). negative entries. percentage of instances in which a dyad member's inference about the pressions of positive affect (i. These entries represented Behavioral Measures the dyad members' attributions about therelativelystable and endur- ing characteristics of their interaction partners (e. (b) the times. A few more fine-grained measures that were of special theoretical interest (e. Thefirstmeasure. direct perspective entries.87 for target. The valence accuracy measure was computed with the aid of a spe- ring in each dyad member's conversation. was an index of the ing turns). the experimenter collected these data and then initiated a new set of videotaped instructions. however. videotapes by independent judges who were kept unaware.e. the percentages of negative self entries and positive partner entries) were also computed. the subjects coded the inferred content of their partner's 1966). It then computed the measures provide a common metric for assessing the degree to which percentage of correct matches for each dyad member. In the final part of the were subsequently entered into a microcomputer as the input to a soft- study. They then decided on the target or object of the perception thoughts and feelings by entering (a) the time the thought or feeling expressed by the entry (i.99. 214-215. about their partner's reported thoughts and feelings. other person(s). feelings. be positive. and (d) whether the entry was presumed to values quite consistent with those previously reported by Ickes et al. directed gazes. subjects' perceptions of (a) the quality of their interaction and (b) the feelings. with a mean of negative) matched the actual valence label assigned to the entry by the . Then. Laing. These behaviors were computed from comparisons of the actual and inferred thought/ included the total frequency and duration of verbalizations (i. partner entries. were later coded by two independent judges who determined both the Each time the tape was stopped. When each subject had com. the subjects were required to make an perspective and the target of each thought/feeling entry. 1983) and empathic accuracy (Ickes.) degree to which a dyad member's written description of the inferred Personal pronoun usage was also coded from the videotapes by two content of his or her partner's entry matched the actual content of the independent judges who used tally marks to record the number of sentence that the partner had written to describe the entry.e. recorded in sentence form). valence accuracy. sponses to the previously described items on the posttest question- stopping them at each of those points in time at which the partner's naire. whether the thought or feeling primarily occurred. Two measures of empathic accuracy possible. given thought or feeling. and third-person singular and plural pronouns occur. self-monitoring (Snyder. or an environmental object. The interrater overall emotional tone of his or her partner's entry (positive. each subject would no Four types of self-report data were obtained that arerelevantto this longer use his or her remote start/pause control to start and stop the investigation: (a) the subjects' responses on personality measures of tape. and meta-perspective entries. cf. the experimenter (aided by a research assistant) would shyness (Cheek & Buss. was an index of the 1983. They were then thanked. content accuracy. speak. using the same type of coding ject's own perspective) or as a meta-perspective (subject's representa- form on which they had previously recorded their own thoughts and tion of his or her partner's perspective. 1981). neutral.92 to . jects were then debriefed more fully and were asked not to discuss the as determined by the judges' ratings of the primary target of each study with potential future subjects. mutual gazes. (c) the specific content of the thought or feeling (again categorical classifications were . 1974). The rates of interrater agreement for these coding sheet).75 to . self entries. When both subjects had completed their thought/feeling listings. 1974). EMPATHIC ACCURACY IN MIXED-SEX DYADS 733 thoughts and feelings that occurred to them for the first time while current interaction (third-person pronouns). a different proce- . included personality measures of shyness (Cheek & Buss. or on some other person(s) not involved in the of the actual and inferred thought/feeling entries. Self-Report Measures which informed the subjects that they would now view the videotape a second time. the content of the dyad members' conversations reflect a focus on self Because the content accuracy measure required subjective judg- (first-person pronouns). 0.90. rates for these measures ranged from .g. positive entries. or negative in its overall affective tone (again coded 1986. of any predictor variables being studied. The second measure. em- start and stop each subject's copy of the videotape at the appropriate pathic ability (Davis. "this guy is pretty The behavioral measures relevant to this study were coded from the creative. Collection of the empathic accuracy data. Reidhead.99 for perspective and . & Lee.e.. 1988). Instead. (c) their inferences The experimenter and the research assistant independently pre. and (d) their re- sented their respective copies of the tape to the two dyad members. or reliabilities of these measures ranged from . see Ickes & Turner. The subjects' instructions were in Thought/feeling measures. It also included items that assessed the subject's reported thought/feeling entries that were coded as thoughts. 1986). second-. pp. neutral. as +. these measures included the percentages of each self-monitoring (Snyder. The subjects' own thought/feeling data most respects analogous to those of the previous thought-listing task. feeling entries. and environment entries. These percentage score inferred entries for the members of each dyad. 1981) and of In the present study. degree to which they liked each other. Phillipson.

and Schneider. Obtaining baseline accuracy estimates for our measure of content cated that the six raters were using essentially the same judgment crite. sponses were significantly correlated (i£. (1986). both actual and inferred. a value Results and Discussion that justified using the mean of the ratings as the measure of physical attractiveness. so that each dyad member's actual thought/feeling entries measure was already computed with the aid of a specialized utility appeared on the left side of the page with the corresponding inferred program in the COLLECT YOUR THOUGHTS software package. the internal consistency of the six judges' con. A. a transcript of all the Obtaining estimates of the baseline accuracy levels for our measure thought/feeling entries reported in the study. Gage and Cronbach those reported by Ickes et al. Hastorf.94. matic teachers would probably credit a correct response in either case plest and most straightforward way to estimate the baseline accuracy and might deny credit only if they had reason to believe that the stu- levels that reflect the false accuracy component was to randomly pair dent had simply made a lucky guess. they could correctly infer the general theme or tenor of their partner's thought/feeling entries and. Funder (1980.g. separate Bissonnette modified the software so that the lab computer could gen- copies of the transcript were given to six independent judges. Kenny (personal communication. This logic suggested that our the posttest measures of liking and discomfort.4 tent accuracy ratings was . ity that the random pairing of the actual valence ratings and the in- fects of false accuracy have been removed.734 ICKES. they could correctly infer the justed individual-level correlations for which the dyad-level co- specific content of that specific entry (differential accuracy).. because it again required a ria. Prag- D. The high reliability of these ratings indi. baseline valence accuracy score. of valence accuracy was relatively easy. two tests of concurrent validity were conducted to replicate ception—see. for our present purposes it was necessary to estimate and credit for spelling the word perceive correctly (a) because he or she correct for only the third component in order to distinguish accuracy evidences the ability to spell that specific word (differential accuracy) that was based on some form of inferential ability from a false accuracy or (b) because he or she evidences the ability to apply the appropriate component that was due simply to chance. vided by his or her partner. the internal consistency of the content accuracy who viewed a clip of the first 20 s of each dyadic interaction with the judgments provided by the 11 raters was very high (a = . Their erate valence accuracy scores after randomly pairing each set of actual task was to compare the written content of each actual entry with that valence ratings with the corresponding set of partner inferences. Second. . uneducated guess about the content of the thoughts and feelings about self) was positively correlated at the entry and be correct on a few rare occasions because of chance alone dyad level with the posttest measure of discomfort during the (chance accuracy). the first two reflect some type of inferential ability. subject was small (e. and individual-level correlations.93. it is appro- subjects could accurately infer the content of a given thought/feeling priate to report only the dyad-level correlations and the ad- entry for any of three reasons. Victor entries appearing directly beside them on the right. by basing their educated guess on that Kenny & La Voie. Each dyad member's attractiveness was rated on a 10-point scale ranging from extremely unattractive (1) to extremely at- tractive (10). The logic underlying component analyses of accuracy in person per. (1986). First. and it justified the decision to use the mean of the ratings as our group of raters to make subjective judgments about the degree of simi- empirical measure of content accuracy. These estimated entries were reported). Of these three accuracy components. Kenny are reported in Table 1 in the form of dyad-level and adjusted (1986. The internal consistency of these ratings was . Because the dyad members' re- Ellsworth (1979)—was applied. however. 1985). Because the valence accuracy was created. The logic here is analogous ferred valence ratings would result in a larger percentage of the origi- to the logic of making a correction for guessing in the scores obtained nal pairings being preserved. AND GARCIA dure was used to derive this measure. whereas the third does not. To correct for this potential bias. Cronbach (1955. 4 each set of actual thought/feeling entries with the corresponding set of When the number of actual thoughts and feelings reported by a partner inferences and then determine the levels of content and va. pairings were then averaged to provide an estimate of each subject's In the present study. as typified by cases in which only three or four lence accuracy obtained with these random pairings. The results of these tests (1955).1958). make a correct inference about a given entry with a Consistent with the findings reported by Ickes et al. nonindependent) for to the present study of empathic accuracy. Although in some 3 studies an estimate of both the second and third components might be If this point is not clear. First. STINSON.1987). Then.91). Third. 1988. larity in the written content of the actual and inferred thought/feeling entries. con. BISSONNETTE. the estimated levels of baseline valence accu- baseline accuracy scores could then be subtracted from the original racy were likely to be slightly inflated because of the greater probabil- accuracy scores to create refined accuracy scores from which the ef. 1988) proposed that the sim. consider whether a student should be given required. content accuracy).. The time demands of this task were such that we could afford to have 11 raters (8 experienced raters and 3 novices) rate only one Physical Attractiveness Measure random pairing of the 76 sets of actual and inferred content entries. matching procedure. general stereotype.e. The valence accuracy scores across the 10 tent (1) to essentially the same content (2). probability greater than chance (stereotype accuracy). Kenny. variation has been statistically controlled (cf. with some necessary modifications. "i before e except after c" (stereotype accuracy). using a 3-point scale ranging from essentially subject made inferences about the set of valence accuracy ratings pro- different content (0) through somewhat similar. for example. accuracy was considerably more difficult. The physical attractiveness of each subject was assessed by 11 raters Fortunately. Harackiewicz and DePaulo (1982).1988). they the percentage of negative self entries (percentage of negative could simply make a lucky. Kenny and Albright (1987). Validity Tests of the Thought/Feeling Assessment Estimation of Baseline Accuracy Levels As checks on the validity of the thought/feeling assessment. but not the same. Ten of the corresponding inferred entry and rate the degree of similarity of these random pairings were run for each of the 76 cases in which a (i. the on an objective multiple-choice test in which subjects vary in the num. computer program allowed no original pairings to be preserved in the ber of items they complete. sound turned off.3 stereotypic rule.

The method of computation makes each Table 2 contains data that provide comparisons of (a) the component analogous to a percentage of a perfect accuracy score of original content and valence accuracy scores. the baseline. This finding indicates that the on the rank ordering of the subjects' content accuracy scores.0001) when tested against the content accuracy measure. and the re. Content accuracy 21. the analogous correlations for the valence accuracy were significantly greater than zero (ps < remaining variables revealed that many. Instead see Kenny and La Voie (1985).4 <.8 12. . As the data in the top portion of Table 2 the null hypothesis suggests that it was not a reliable compo- indicate. EMPATHIC ACCURACY IN MIXED-SEX DYADS 735 interaction. the percentage of positive partner en.44. correlated were ones for which the responses of the male and these valence accuracy data revealed that our original valence female dyad members were significantly correlated (i. nent of empathic accuracy but instead represented chance de- vised measures of content accuracy were all significantly viations from the valence accuracy baseline. the data were collapsed across gender and tested using the each measure (original score minus baseline). racy measure is not warranted. Suf- % negative self entries . On the other hand. In sharp contrast to the content accuracy data. 6 Although it would be logical and desirable at this point to present a Thought/feeling measure Dyad ADJIND Dyad ADJIND thorough discussion of the levels-of-analysis issue.. Table 2 tries (percentage of positive thoughts and feelings about Components of the Content and Valence Accuracy Measures partner) was positively correlated at the dyad level with the posttest rating of liking for partner. The F ratios test the significance of the mean for each component against the null hypothesis.0 <.15 -. Thus. but the mean for the revised (i. level of content accuracy achieved by the male dyad members Valence accuracy.16 (i. The corresponding ad. curacy.7 <. the estimated partial correlation between the percentage of Self-reported negative self entries and self-reported discomfort remained highly sig- discomfort Liking for partner nificant. the means for the original. for which a highly reliable level of null hypothesis. 4 3 * -. only the data for the combined sample are reported.91 for all subjects combined.0 <. 0 1 . greater than zero.13. Note. Because there individual as the unit of analysis.8 1. Because the scores of the male dyad members were not signifi- cantly correlated with those of their female partner for any of these levels for each of these measures. p < . of the . further discussion of the valence accu- accuracy scores contained a small baseline component in addi.0001 Baseline valence accuracy 36. though not all. and . For this reason.37). although the baseline component of content accuracy The revised content accuracy scores of the male dyad was statistically reliable for both the men and the women in the members were not significantly correlated with those of their sample.0O01 only at the individual level of analysis. Content accuracy.9 12. difference-score) mea. r. Further analysis revealed that the original and the revised content accuracy scores were correlated .07 . Accordingly. The fact that Kenny and La Voie's (1985) guidelines for determining whether this second component was not significant when tested against a given correlation exists and should be reported at the individ- ual level. (b) the baseline 100.005.e.5 In addition.005.49** . were no significant gender differences in the means or the pat- terns of intercorrelation of these measures. correcting for its influence had relatively little impact female partners (r = -.1 17.6 Table 1 5 Thought/Feeling Correlates of Self-Reported When the substantial correlation between the percentage of posi- Discomfort and Liking of Partner tive partner entries and self-reported discomfort was statistically con- trolled.91 for the women in Correlates of Content Accuracy the sample.6 646. respective dyad-level mean. . Dyad = dyad-level correlations (df= 36). Revised valence accuracy 3.2 94. the means for the original and baseline measures of partners. differential accuracy was found. or at both levels of analysis. and (c) the revised scores for measures.92 for the men. it was necessary to apply addition to a second.OO01 cating that the significant dyad-level relationships reported in Baseline content accuracy 5. the results reported next concern only the nents was highly reliable (all ps < .15 -.8 5.7 <.1885 Comparisons of Original and Revised Accuracy Scores Note.0001 Table 1 were emergent consequences of the dyad members' in. Empathic accuracy measure M SD F(l.6 121. As the data in the bottom portion of Table was essentially independent of that achieved by their female 2 indicate.0001 teractions that cannot be accounted for by processes occurring Valence accuracy 40.(35) = . because the subjects' scores on tion to a much larger component that was presumably a refined this measure provided no indication of differential valence ac- measure of differential content accuracy.8 <. much smaller component. **/><.51** . at the dyad level. space limitations preclude such discussion.1 242.75) P justed individual-level correlations were not significant.14 fice it to say here that dyad-level correlations are based on the mean % positive partner entries .e.1 419. predictor variables with which content accuracy proved to be sure was not.0001). ns). average) of the two dyad members' scores on each variable. deviations of the two dyad members' scores on each variable from its * p < . Revised content accuracy 15. ADJ IND = adjusted whereas the adjusted individual-level correlations are based on the individual-level correlations (df. indi. Each of these compo.e. These data indicate that our original content Given this finding.3 21.7 12.. were accuracy scores contained a very large baseline component in nonindependent).

29 indicates that the higher the percentage of questions were able to accurately infer each other's more transient asked by one dyad member relative to his or her partner.001.45**** feelings.35**** . their female partners for both the criterion variable (content It makes sense that these relationships with verbal behavior accuracy) and the predictor variable. 1989). Specifically. third-person pronouns were themselves negatively correlated at the dyad level. it was the relative By these guidelines. p < . have been placed in boxes and set in boldface type.e.15 measures reported.31) levels of analysis. criteria.11 ual dyad members who reported the largest percentages of % third-person pronouns -.001. lar pronouns (I. for cases in which the dyad themselves well to an informational interpretation. Kruglanski. *****/?<. revised content accuracy with the measure of how frequently sure was correlated with the measure of the percentage of the the dyad members looked at each other can also be interpreted . according to Kenny & La Voie's (1985) with the percentage of third-person pronouns (he.24** .31* that occurred in these initial.24 . those individ- %first-personsingular pronouns .12 . specific content of each other's thoughts and feelings.35** -. correlated with revised content accuracy. did the degree to which dyad represented the dyad member's scores as deviations from the members made attributions about each other's more stable and dyad's mean on each variable. reported under the heading of Thought/feeling mea- sures in Table 3. the significant correlates of content accuracy charac. the more that these strangers talked to each other about Kenny. my. ***p<. hand.e.03 inferences about the content of their partner's thoughts and Positive affect (frequency) .23** 38** . the more information that dyad members conveyed to able but were interdependent (i. sis the correlations should be reported and interpreted. the obtained positive correlation enduring characteristics correlate with the degree to which they of .14 . The remaining verbal behavior measure—the percentage of Partner attributions and content accuracy. the significant correlates would be found at the dyad level of analysis. they.04 partner attributions were also likely to make the most accurate Nonverbal behavior Directed gazes (frequency) .29 . BISSONNETTE. On the one members' responses were independent for the criterion vari. them. members elicit through questioning each other contribute to eral epistemic attempt to "understand" another person (cf.736 ICKES.34** -.01. We applied these third parties.47) and the adjusted individual Verbalizations (frequency) . Because the percentages offirst-personsingular and *p<.17 .05.. me. the revised content accuracy mea. Dyad = dyad-level correlations {df= 36).71. content accuracy was found to be positively Self-monitoring . AND GARCIA Table 3 dyad members' thought/feeling entries that concerned rela- Correlates of Revised Content Accuracy in tively enduring dispositions of their interaction partners. tion. The verbal behavior % partner attributions . personal communication. hand. Nonverbal behavior and content accuracy. Thisfindingsuggests that rela- making and empathic accuracy may not be required. Note also and should be interpreted only at the (unadjusted) individual that these conversational influences on empathic accuracy lend level of analysis. mixed-sex interactions had a pro- % partner entries . correlated) for the predictor each other through self-disclosing. ADJ IND = adjusted individ.24 nounced effect on the dyad members' content accuracy. but only at the ad- cerned the relationship between attributional inference and justed individual level of analysis.21 -. The correlation of To answer this question. On the other hand.005.24** . ****p<. even when the signifi- % questions asked .19 . for that was obvi- of content accuracy characterized relationships that occurred ously the level at which the conversation occurred.). A. as indi- % positive partner entries . these effects should not be viewed as independent of each other. r(36) = . In contrast. 1988). self-referenced conversa- variable.53***** cated in the first four rows of Table 3. for cases in which the responses of the use of first-person singular versus third-person pronouns that male dyad members were statistically independent of those of contributed to content accuracy. many of whom were known by only one of the guidelines to the data reported in Table 3 by using boldface dyad members. because tive differences in the amount of information that dyad both phenomena may be outcomes or products of a more gen. the less able they were to accurately infer the type and surrounding boxes to indicate at what level(s) of analy. indicated that the percentage of partner attri- Measures 1UNADJ Dyad ADJ IND bution entries was significantly correlated with revised content Verbal behavior accuracy at both the dyad (r = .33** correlated at the dyad level with a global measure of the total Partner's attractiveness . relative differences in content accuracy within the dyads. Thought/feeling Verbal behavior and content accuracy.16 (r = . etc.16 number of verbalizations (i. the thoughts and feelings? An affirmative answer to this question greater was that same dyad member's content accuracy as com- would suggest that separate theories to account for attribution pared with his or her partner's. The Initial Mixed-Sex Interactions results.03 . In other words. Rather. UNADJ = unadjusted individual-level correlations (df= 74).35**** . Of the verbal behavior Individual difference Grade point average .19 . STINSON. An initial and verbalizations that were coded as questions—was positively very important question we wished to ask of the data con.28** .29* cant dyad-level effect was statistically controlled. the better able they were to accurately infer the specific terized relationships that occurred and should be interpreted content of each other's thoughts and feelings. mine.47**** . On the other only at the dyad and adjusted individual levels ofanalysis (D. **/><.31* . Because the data at this level empathic inference. speaking turns) as well as with the more specific measure of the percentage offirst-personsingu- Note. The correlations that can be meaning- fully interpreted in this table. myself). content accuracy was negatively correlated at the dyad level ual-level correlations (df= 37).24** .35** -.07.41*** . she.

First. uncorrelated in this sample. rs = .09 . all of the correlations in this table are her partner. Worse yet. and percentage of partner attributions—appear to be man. 1986) and social per- tion at this level between smiling and content accuracy was not ceptiveness (e. though speculative.e. r(74) = .10 -. We found that partialing out the effect interaction partners (i. effects. p < . rps > . Revised Content Accuracy ture.43 and . reported in Self-report measures and content accuracy. pathic accuracy proved to be disappointingly poor predictors of tation of the verbal behavior data. an interest that also leads. These ship between smiling and content accuracy With these vari.05. & Patterson. ps < . to a higher level of con.21 -. the obtained positive correlation of . respectively (ps < . Our best guess is that smiling in mixed-sex vided direct support for this speculation: Content accuracy was dyads is an expression ofan underlying interest in the partner— found to be correlated with the Other-Directedness factor. & measure of gazing. How should we explain the adjusted individual-level correla. we tested and ruled out the possibility that this relation- ship was an indirect and socially mediated effect of the per- ceiver's smiling on his or her partner's motivation to be more fested the greatest interest in and positive sentiment for their open and self-disclosing. Consistent with Table 4. thoughts and feelings. lend themselves ables partialed out. the greater was that same dyad member's content reported at the individual level of analysis (df= 36 for the men and accuracy relative to his or her partner's. Analyses of adjusted individual-level relationship between smiling and the self-monitoring factor scores (Briggs & Cheek. 1984) substantially attenuated when the effects of gazing.01. but not with the Acting or Extraversion amount of looking. r(36) = . talking.15 -. Ickes. & Soldow. and so on enables the perceiver to further qualify the meaning of the other's words as well as to Revised content accuracy acquire additional information about the other's emotional Personality measures Men Women Combined state.13 -. racy for either the male or the female dyad members in this tions also suggest that the individual dyad members who mani.04 Fantasy identification (F) -. plausibly increase the motivation of individual dyad members ing. those reporting the largest percent- of the partner's talking did not eliminate the correlation be. ported by the pattern of measures that significantly correlate Third. to accurately read their partner's thoughts and feelings. of intelligence and motivation that is reflected in high GPA Third.e.45. 1980.06 affect measure)? Because the data at this level represent the Empathic accuracy (EA) . accuracy are reported in the last three rows of Table 3. is sup. However.14 tion of revised content accuracy with the measure of how fre. Buss. ns. The significant individual difference correlates of content ceiver's own rated liking for the partner mediated the relation. the positive Personal distress (PD) . 1986.16 -. We explored three possible mediational interpretations. frequency ofsmil.01. less. EMPATHIC ACCURACY IN MIXED-SEX DYADS 737 in informational terms. revised content accuracy. 1977. p < . by a route more subtle than the sheer r(74) = . the data revealed no obvious mediation for the achieve relatively high levels of content accuracy. p < . the physical attractiveness of their opposite-sex inter- with revised content accuracy at the adjusted individual level.03 dyad member's scores as deviations from the dyad's mean on each variable. factors.03. partner-relevant thoughts and feelings) were the tween the perceiver's smiling and his or her content accuracy most accurate in inferring the content of their partner's (rp = . Empathic concern (EC) . Rarick. GPA and self-monitoring scores were tent accuracy. Briggs.31. ifestations of an underlying interest in the partner.45 indicates Note. These (unadjusted) individual-level correla.11 quently the dyad members smiled at each other (i. Second. The relevant correlations. Because the scores of the male dyad members were not signifi- that the more often one dyad member smiled relative to his or cantly correlated with those of their female partners for any of these individual difference measures. all at the (unadjusted) individual level. Mill. the residual correlations were still to fairly straightforward interpretations. The datarevealedthat the measure of smiling was indeed Second. ns.32. Frequent gazing at one's interaction Table4 partner should facilitate content accuracy. the well-documented "other orientation" (e. 1986) pro- content accuracy. First. the combination significant. Geizer.09 and . this interpretation has a motivational/attentional flavor that The self-report personality measures of empathic skills and em- distinguishes it from the more strictly informational interpre.15 -.01). ages of positive.09 -. because the visual Nonsignificant Personality Correlates of monitoring of relatively rapid and subtle changes in body pos.g. Reidhead. study. gestures. being .00. most of the correlations reported in Table 4 . women separately and df= 74 for the combined sample).01). facial expression. or that are reflected in high self-monitoring scores may constitute use of verbal reinforcers were partialed out. we tested and ruled out the possibil. the correla. Briggs & Cheek. a second. we tested the possibility that the dyad members who scores may alsoreflectthe individual dyad members' more spe- smiled more frequently were more visually attentive or more cific ability and motivation to accurately "read" their interac- verbally responsive to their partners than those who smiled tion partners. Cheek. ity that either the partner's physical attractiveness or the per.12 .16 -.44. Note that Self-reported empathic dispositions and content accuracy.g. action partner is a highly salient external factor that also could All three measures—percentage ofquestions.43. independent source of motivation and ability to In summary. talking. This interpretation. indicated that none of the subscales of Davis's (1983) this speculation are the two remaining correlations reported in empathy measure nor the empathic accuracy measure devel- Table 3 between content accuracy and measures of thought/ oped by Ickes (1988) was a significant predictor ofcontent accu- feeling content. Perspective taking (PT) -. 1980. and so on. Ga- positively correlated at the adjusted individual level with the brenya & Arkin.

Hall. able to accurately infer the content of each other's more tran- however. The first was to estab. mixed-sex interactions. note that although gender is stereotypically regarded as an individual difference Correlates of the (Revised) Content Accuracy Measure variable that is related to empathic accuracy. (i.91) that our correc. attributed thoughts and feelings. This commonfindingsug. the women's re- vised content accuracy scores were not significantly greater The second goal of this study was also realized. Rogers. are relatively global and enduring (i. This novel. in large On the other hand. Tucker. dyad level. finding establishes a clear-cut link between the pro- would probably be most evident in long-term. the second was to what. not in initial interactions. a sient thoughts and feelings. Moreover. respec. Because Third. through directed gazes and by asking questions) and con- DiMatteo. they did replicate the common finding that women are Second. veyed to each other (i. in sign. Second. Such get's gender is perfectly confounded with the perceiver's gender a link may be important in developing a theoretical model that in this study. established rela. It is simply not warranted. the revised measure of content accuracy was so as a socially constructed meaning context that exists only at the highly correlated with the original measure (. subjects' it will have much utility in future studies of the initial. the generalizability of the null sex more these strangers talked to each other about third parties. the emergent product of social inter- lish the utility of our method for assessing empathic accuracy in action processes occurring at the dyad level of analysis. Third. question. the empathic accuracy of all parties present is.e. a pattern of individual-level correlations indicated the this measure failed to indicate a degree of differential accuracy importance to their content accuracy of the perceivers' degree that significantly exceeded baseline levels. f(37) < 1. the For all of these reasons. does this mean? Does it mean that two interacting use the method to explore some of the factors that determine dyad members should be more accurate in inferring each the level of empathic accuracy in this same interactional con. short- perceptions of their own empathic abilities do not predict their term interactions of strangers. & Archer. Apparently. dyad-level influence. present only as a passive (or even a hidden) observer of their interaction? Not at all. more information the dyad members elicited from each other Prince. 1979). The greater the dyad member's inter- . a product of this emergent. a pattern of dyad-level correlations revealed that the better decoders of facial expressions of emotion (Friedman. unstructured interactions of mixed-sex dyads. which stronger and more volatile emotions are expressed.e. Ms = 17. BISSONNETTE. These data patterns. it might still prove to actual empathic accuracy. First. STINSON. cognitive empathy task similar to ours (see also Archer & Akert. 1977).. content accuracy's relations with a range of other variables. the less able they were to accurately infer the specific content of each other's thoughts and feelings. which should not be generalized beyond significant gender differences are commonly found for fairly the initial. at least not in the context of an initial be reliable in longer term interactions or in relationships in interaction with a stranger of the opposite sex. the same correction for guessing had measure. But the context of initial.738 ICKES. First.94). cesses underlying attributional inference and the processes un- tionships. & DiMatteo. and Coffaro (1989) also found no dif. which proved to be highly reliable (a = members' thoughts and feelings emerges during the interaction . we are doubtful that of interest in their partner. derlying content accuracy as a form of empathic inference. difference obtained in the present study must remain open to many of whom were known by only one of the dyad members. Rosenthal. other stable dispositions) but also for those that are relatively ference between women and men in their performance on a specific and fleeting (i. because the tar.5. to draw so general a conclusion at this point.e. of each other's thoughts and feelings. tion for guessing appeared to have had only a negligible effect interdependent construction of the two dyad members and is on the correlations of the content accuracy scores with other available not only to them but to nonparticipating observers as variables. cf. the degree to which the dyad members made attribu- The most general interpretation of this nullfindingwould be tions about each other's more stable and enduring characteris- that the widely assumed relationship between gender and em.e. data patterns offered a coherent yet highly detailed view of tively. mixed-sex inter- There were two goals of this research. through self-referenced conversation).3 and 14. exactly. AND GARCIA were negative. Hall. These findings underscore what is clearly a major lesson of Summary and Implications the data in Table 3: Content accuracy in initial. attributed traits and although Riggio. However. Because this dyad-level meaning context is the joint. rather than positive. were obvious measures of self-reported empathy but are rarely found as follows: for behavioral measures of empathy. Riggio. the basis of their inferences about the content of the dyad sure of content accuracy. As a recent review by Eisenberg and Lennon (1983) suggests. It simply means that the information that any of the parties (interacting dyad members or passive Assessment of Empathic Accuracy observers) derive from the dyad members' behavior and use as The first goal was realized with respect to our original mea. actions is. Finally. On the other hand. ns. other's thoughts and feelings than should a third person who is text. 1979. it is possible that women may be more accurate is adequate to account not only for those social inferences that perceivers than men but that men may be harder to read. serious implications for the valence accuracy measure. to a large extent. in that the than those of their male partners. 1989). tics was positively correlated with the degree to which they were pathic accuracy is a cultural myth. Kruglanski. 1980. Gender and content accuracy. though not counterin- difference between women and men in empathic accuracy tuitive. This type of finding is by no means unique. well. gests that a reliable measure of valence accuracy (which ours was the better able they were to accurately infer the specific content not) might have revealed a gender difference favoring women.

1985) to communications theory between perspective-taking scores and accuracy was moder. thoughts and feelings. First. should be a topic of and feelings. in initial mixed-sex interactions. another pattern of individual-level correlations indi. Content accuracy was also positively correlated effects. Krippendorf. Barrett-Len- lack of interest in and attentiveness to one's partner during an nard. in general. On the other hand.g. Truax & Carkuff. our awareness of these limitations that led racy as the outcome of processes that are psychological (i. prove to be among the most theoretically provocative. 1972. Appar. This feelings). the higher was his or her level of content accuracy. How. in accepting differences variables in ways that sometimes did but sometimes our findings that gender and self-report measures of empathic did not support commonsense intuition. It cated that content accuracy was related to various individual would have somewhat more difficulty. the strength and gests that the counterintuitive null findings for gender and consistency of such emergent. too. In this study. Whereas all subjects were later encouraged by the em. We expect that these and interested in their partners during the interaction still ob. and somewhat counterintuitive. If so. the fact that we found significant effects for other individ. On the other hand. In the same vein. variety of theoretical perspectives rangingfromsymbolic inter- larly given Bernstein and Davis's (1982)findingthat the relation actionism (Stryker & Statham. 1972. evidence should be found. It is. with individual-level motivational factors such as the partner's pathy task instructions to accurately infer their partner's physical attractiveness and the perceiver's degree of interest in thoughts and feelings. nonsignificant) predictors ofcontent accuracy. In addition. 1967). they also under- score the limitations of more traditional research paradigms Implications for Theory and Research that fail to address many of the theoretical and methodological These findings have a number of important implications for issues that arise when social cognition is studied in the context the study of empathic accuracy and intersubjective phenomena of naturally occurring social interaction. at the dyad (or group) level of analysis. it is certainly not surprising that such ual difference predictors (e. Watzlawick. dispositional attributions about the of content accuracy is not just a distant cousin of the corre- partner.. ever. From the standpoint of a ity of these nullfindingsmust remain open to question. GPA and self-monitoring) sug. andreliableway. Ka- interaction cannot be easily compensated for by a later opportu. the subjects' gender and measures of their Third. The lack of correspondence between self-monitoring scores appeared to contribute independently to people's perceived and actual levels of ability to accurately infer the dyad members' ability and motivation to read each other the thoughts and feelings of others was quite striking in this accurately. specific thoughts and feelings and the extent to which one Thefindingswe obtained lead us to conclude that our measure makes more global. It. Penman. of course. and we regard this flawed estimation of one's own em- tractiveness appeared to increase the motivation of the individ. Buckley. cepting ourfindingsthat. the present findings also point to the content accuracy in the context of ongoing social interaction is . 1980. by individual difference variables in ways that do not always tions regarding the use of videotape review procedures for accord with conventional wisdom or cultural stereotype. Goldstein & Michaels. suggests that it is the subject's interest in the pathic accuracy are clearly important. to a large self-reported empathic skills proved to be disappointingly poor extent. GPA scores were associated with greater empathic accuracy. The general. complexity is evident in at least three respects. representation of the degree to which each dyad member can tional factors as determinants of empathic accuracy in naturally accurately infer the specific content of his or her partner's occurring social interaction. 1984. study. dyad-level influences in our data self-rated empathic skills should not readily be dismissed. (Fisher. with ently. EMPATHIC ACCURACY IN MIXED-SEX DYADS 739 est in his or her partner (as denned by thefrequencyof smiling complexity of empathic accuracy and the need to address this and the percentage of positive. they clearly portray empathic accu. these results Second. It was. particu. however. there is This pattern of correlations. For one thing. 1967) to social systems theory (Bateson. there is evidence that empathic accuracy is. like those for gazing and smiling at evidence that although informational determinants of em- the partner. 1977. there is evidence that empathic accuracy is affected may have important. the level of their partners' physical at. clearly underscore the intersubjective nature of empathic accu- racy and related phenomena. Con- training empathic skills in clinical and counseling psychology ventional wisdom would probably have little difficulty in ac- programs.e. 1985. Gladstein. ongoing social interaction. 1978.e. We believe that this method for assessing On the other hand. instead. extra-informational influences on empathic accuracy will tained higher accuracy scores than those who were not. one of our goals was to demon- link between the processes that determine both the accuracy of strate that it is possible to operationally define and measure one's inferences about the content of an interaction partner's empathic accuracy in a meaningful. our data indi. ated by the length of exposure to the target individual. the impairment in empathic accuracy resulting from a important clinical/applied implications as well (cf. the emergent product of interaction processes occurring (i. The specific ability skills and empathic accuracy were disappointingly poor predic- and motivational components that are reflected in GPA and in tors of content accuracy. most reassuring perhaps for the more rigorous sponding empathic understanding that naturally emerges in adherents of information processing approaches. gan. nity to view the interaction on videotape. By implication. pathic accuracy as one of the most fascinating and compelling ual dyad members to accurately read their partner's thoughts foci for future theory and research. they by no means tell the partner during the interaction that is crucial in mediating these whole story. 1962. those subjects who were most attentive to the partner during the interaction itself. & Jackson. partner-relevant thoughts and complexity in both theoretical and methodological terms. And. objective. thesefindingssuggest a social cognition. major clinical/applied interest. Beavin. us to conduct this most recent of our studies of naturalistic inferential) in nature. implica. 1967). high Fourth. a reasonably faithful cate that considerable weight should be assigned to informa.

and many other approaches as well. & Petty. perception and interpersonal behavior (pp. S. R. labor intensive. Y. & Michaels. & Lennon. dis. New York: Guilford Press. Ickes. Affective Communication Test. Journal of Personality. 3. suade as determinants of counterargument production. D. M. CA: Academic Press. 3. 41. problems in interpersonal perception. J. R. 221-284). (1962). Diehl. (1971).171- Other limitations concern the preliminary. & Krathwohl. kinesics. (1987). (1968). Englewood Cliffs. R.106-148. causal factors in therapeutic change. F. 193. A. and attitude change. M. 214-225. & Akert. Genest (Eds. and consequences.Funder. there is still and Social Psychology. NJ: Prentice. The develop. 52. E. 249-262. Sociology and systems theory.127-133. R. Journal of we must remember that it implicates communications. (1982). R. Brock. The Counseling Psychologist. 1982. 446-449. G. J. Maher (Ed. C. Prince. proxe. 407-412. Psychological Bulletin. we acknowledge that our own research has Carkhuff. Perspectives on human communication. Stanford. plied Social Psychology. (1980). Cognitive Therapy and Research. (1969). J. Hillsdale. M. Conceptual and methodological Experimental Social Psychology. 53. Gladstein. Deception and Cacioppo. Davis. J. anxiety. Rarick. In R.. Glass. Empathy and counseling outcome: An empiri- Cacioppo.. Gabrenya. tages. structure and correlates. (1981). Barrett-Lennard. W J. self. Briggs. G. S. Personality and and self-evaluations: A cognitive response analysis of heterosexual Social Psychology Bulletin. 309-342). Petrullo (Eds. R. Journal of Personality Eisenberg. (1983). Journal ofPerson. & Cheek. & Kagan. L. American Sociological Review. the method is so expensive. D. J. 443-449. Measurement of affective sensitivity: nomenon that is deeply rooted in cognitive social psychology. (1955). Tagiuri & L. Interpersonal perception. & Buss. Englewood Cliffs. K. (1986). 75-90. V B. & Buss. (1967). P. development and evaluation of personality scales. 473-493. H. Helping and human relations: A primer for lay some major limitations as well. Journal of Gage. Kagan. V (1979). & Davis. L. A. Toward a valid measure of interpersonal perception. R. Social psychology. H. Words and everything else: Verbal Journal of Consulting Psychology. (1952). (1986). 35. Journal of Counseling Psychology. NJ: Erlbaum. agreement and discrepancy in personality ratings. & Arkin. N. The role of factor analysis in the social judgment. C. K. Errors and mistakes: Evaluating the accuracy of Briggs. & Cronbach. J9. Cognitive assessment (pp. L. (1964). Evidence for a multidimensional approach. Shyness and sociability.. Although empathic accuracy is an inferential phe- Danish. for cognitive response assessment: The thought-listing technique. T. H. Steps to an ecology ofmind. Cheek. F. C. much we have to do. W (1967). & Stroebe. STINSON. Basic and Ap. nor does it adapt easily to the study of interaction in Diego. and nonverbal cues in social interpretation.1-19. Journal of the custom software we developed. 3. An analysis of the 528-550. accuracy. T. (1957). method is well suited to the study of short-term. T. J. 44.. Winston. of " others and "assumed similarity. T. & DiMatteo. In Goldstein. W. (1942). S. 333-351. New York: Holt. R. ality. M. R. Sex differences in empathy and and Social Psychology. 51-54. M. 384-390. 76.1983. Psychological Monographs. R. cognitive response to Campbell. C. M.. Books. BISSONNETTE. ress in experimental personality research (Vol. W. A scale for the measurement of empathic ability. The generality of accuracy and tional structure and content in determining levels of empathic response set in interpersonal perception. training. paralinguistics.740 ICKES. Greenwald. our measures of personal pronoun usage were Stanford University Press.). L. Cheek. Cronbach. (1955). Journal of Personality ing scratched the surface of this phenomenon. 330-339. Journal of Abnormal and Social Psychology. 6. We have obviously not yet taken full ad. M. J. NJ: Prentice-Hall. A.). T.. A. Prog- tions. systems theory.). Self-Monitoring Scale: Factor 38. and time Chapin. 1990). Brock. Merluzzi. (1977). it cannot be easily applied to study longer-term interac. (1981).. 101. Ickes et al. 54. 54. related capacities. (1958). R. On seeing ourselves as others see us: Self-other Bernstein. J. San tions. pp.. (1972). I. Riggio. F. M. ture of this research. Psychological Bulletin. (1985). Processes affecting scores on "understanding groups (cf." Psychological Bulletin. J. M. 7. 679-686. C.. N. M.Fisher. exploratory na. Communication discrepancy and intent to per. Funder. M.. Person relate content accuracy to a wide range of behavioral predictors. (1980). Understanding and assessing nonverbal expressiveness: The 1-33. Archer. On the several facets of personal- ality. (1978). Preliminary standardization of a social insight consuming that we would be reluctant to use it in the absence of scale. CA: In particular. Journal ofPersonality and Social Psychology. D. Despite its many advan. only a token acknowledgment of the importance of conversa. course analysis. N. G. 53. E. R. Glass. 94.. In B. dyadic interac- Cline. which subjects participate in more than one dyad or in larger Cronbach. Empathy: Development. 5. J. & Hammond. Cognitive learning. 70-79. L. A. C. D. G. & M. and professional helpers (Vols. Personality and Social Psychology Bulletin. 269-309. 18.. G. (1980). L. persuasion. H.13-22. (1980). 1-2). Productivity loss in brainstorming groups: Toward the solution of a riddle. ity assessment: The case of social acuity. Proposals leading to analytic treatment of so- vantage of the method's potential in allowing researchers to cial perception scores. R. T. E. Journal of Personality and Social Bateson. Moreover. (1949).. A. Psychological Bulletin. S. Crow. G. 54.. H. 1. & . Social psychological procedures cal and conceptual review. As a final note. Rinehart & cern the methodology we have used. & Merluzzi. Macmillan. Counseling Psychology. Buckley. New \brk: Hall. 258-266. G. Funder. 18. & Soldow. M. 13. (1977). C. although the Personality and Social Psychology. S. 353-379). Measuring individual differences in empathy: mics. Self-Monitoring Scale. 497-509. T. New York: Ballantine Psychology. (1977). Greenwald. S. D. Journal ofPerson- consciousness. (1983). R. B. Perspective-taking. S. & Harris. 113-126. 48. Self-statements judgment accuracy: A study in person perception. Some of these limitations con. W (1987). 6. R.100-131. In A. linguistics. Journal of Personality and Social Psychology. and accuracy in person perception. (1971). References Dymond. Dimensions of therapist response as Friedman. R. D. Hav. Geizer. Asch. AND GARCIA a viable one that should be applicable to work in other social ment and validation of a scale to measure affective sensitivity (em- science disciplines as well as in other areas of psychology pathy). N.

Paulus(Ed. Knapp. influence.. L. tion accuracy. & Speroff. & Stinson. B. Arlington. small groups. & Wolfe. 247-256. In S. Development of an empathy scale. A. E.). Unpublished doctoral dissertation. Gunn. 288-309). A. Ickes. Journal ofPersonality and Social Psychology. 23-32. 52. & Albright. psychology. Interpersonal perception in a so- and social behavior. 388. & Epstein. 40. 419- roles. Machiavellianism Mead. K. Phillipson. New perception: A component analysis according to Cronbach. 21-36. pp. In W (2nd ed.). Minority influence. Hendrick (Ed. Psychological Bulletin. Advances in experimental social psychology (Wo\. (1989).. 372- Naturalistic social cognition: Intersubjectivity in same-sex dyads.157-174. (1969). Psychological Bulletin. New York: Random House. L.). Kagan. J. Mind. 41 -64). B. University of Texas at Arlington. ity and Social Psychology Bulletin. & Lee.. A basic paradigm for the study of personality. In W. W. (1985). Group socialization in small Janis. L. Clark (Eds. (1954). L.Kerr. F. (1972). 8. Journal of Personality and Social Psychology. Gender. & Levine. B.Marks. M. A. 16-44). B. The psychology ofinterpersonal relations. The observational method: Studying the Merleau-Ponty. Personal. Basic group processes (pp. L. Chichester. Wiley. Maass. West. Bissonnette.). (1988). .. & Zajonc. B. W(1989). (1982). (1990). G. R. Knowles (Eds. (1990). The group polarization phenome- 402. Hendrick & M. (1985). dyads. The cognitive perspective in social Ickes.).. M. M. Psychotherapy. Empathy research: Some method- lished scale and associated psychometric data. J. CA: Sage.). roles. W. ing skills and empathic expression. (1986). & Patterson. 305-341).. Mill. University ofTexas at ological considerations. opposite-sex sibling: Birth-order influences in mixed-sex 327-355). decision making (pp. Ber- Janis.. Baker. Accuracy in interpersonal percep. R. (1988). (1986). Paulus search (pp.). M. Boston: Allyn & Bacon. cial context. D. New directions for methodology of social objectively assessed empathic accuracy in a clinically relevant context. New York: Humanities Press. Naturalistic Journal ofPersonality. C. Journal ofCounseling Psychology. Journal of Con. Brandstatter. pp. A. (Colin Smith. & Hogan. England: Wiley. A basic paradigm for the study of unstructured dyadic Marangoni. 12. M. Stinson. Aronson (Eds. 8. & Mugny. Psychological foundations of attitudes (pp. J. B. (1958). S. Journal ofCommunication. (1976). W. B. Markus. Journal ofPersonality. S. 57-77). Skill in nonverbal communication (pp. Sage. R. pp. 58. 11. cago Press. Tooke. G. NJ: Erlbaum. Principles and methods of social re.). & Bissonnette. (1983). Davis. M... & Albright. East Lansing: Michi. T.. 58-74. Lindzey & E.. Ickes. The handbook of personal relationships: attitudes. Implementing architectural de- empathy. In C. Tooke. research. M. J. CA: Academic Press. older. Mehrabian. W. (1968). L. NJ: Erlbaum. & Fussell. Garcia.125-135. 57. & Mann. Rosenthal (Ed. D. (1983). The Journal ofGeneral Psychology. (1989). Journal ofPersonality and Social Psychology. Psychological Bulletin. Accuracy of person Kruglanski. minorities. & G. Newbury Park. In P. non. D. and interventions (pp. W. pp. W A. Some determinants of communica- of Personality and Social Psychology.\5.. and commitment. 66-82. A. 55-79). (Ed. C. (1966). S. High and low self-monitoring individuals: Their decod- Ickes. H. & Cialdini. Ickes. (1983). social cognition: Methodology. & Tolsma. On knowing conversion. 51. & Powers. Crano & M. (1972). 525-543. 125-161). & Tooke.1962) 97). cation. Stocker-Kreichgauer (Eds. Kenny. Paper prepared for a Hall. L. & Lamm. and behavioral science (pp. Influencing human interaction. Minority influence and Honeycutt. In R. C.. The social impact of majorities and Hogan. 307-316.Myers. (1969). Interpersonal percep- Heider. & Reed. 34. The facilitative effect of feedback training on interaction. Newcomers and oldtimers in gan State University Press. 137-192). Journal of Personality and Social Psychology. (1984). and nonverbal communication festschrift to honor Ragnar Rommetveit. D. New York: Harper & Row. Hillsdale. G. L. New York: Springer-Verlag. Latane. B. (1972). C. R. L.). Duck (Ed. 137-230). (1970). N. E. (Original work published personal relationships: Theory. (1982).. E. 20. and interventions (pp. R. W. Psychology. Journal Mehrabian. San Diego. Implement. 93. empathy test. & La Voie. In P. An 18-item measure of empathic accuracy. The group as a polarizer of ships. N. 12. L. 4. D. R. Psychological Review. Duck (Ed.Moscovici. A. D. CA: Communication. H. 23. and social behavior (pp. B. Group processes: Review ofperson- others and predicting what they say. Group Kenny. San Diego. 102. (1987). Wiley. The analysis of data from two-person relation.). 48. (1984).). signs: On the carpentry of intersubjectivity. Psychology of group influence Kenny. & Zavalloni. San Francisco: Jossey-Bass. (1986). M. Differential contributions of majority and minority effects. gender roles. (1988). 339-348. 250-260. 390. The handbook of ing and using the dyadic interaction paradigm. In P. Krippendorf..). (1982). Ostrom (Eds. Paulus (Ed. 210-222. J.. In G. selfand society. 428. I." Social Cogni. McCauley. sis ofconflict. Harackiewicz. 45.). M. The theory and measurement of Krauss. E. San Diego. S. The Western Journal of Speech ality and social psychology (Vol. EMPATHIC ACCURACY IN MIXED-SEX DYADS 741 T. & Turner. H. E. V L. Journal of Personality and Social Newbury Park. Psychological Review. The handbook of Trans. W. J.. Validation and evaluation of the 170). research. Miller. In C. groups: Temporal changes in individual-group relations. (1987).' Compliance and internalization. (1989). 70. New York: Free Press. & Levine. skills.). M. 365-384. Chichester. (1989). On the social advantages of having an rules. Methods for measuring dyads and groups. In H. W. D. Decision making: A psychological analy. 147. MA: Oelgeschlager. Unpub. An epistemological foundation of communi- 32-67). H. Ickes. Kenny. sulting and Clinical Psychology. W. (1981). 47.). R. (1982). H. Journal ofNonverbal Behavior. A. H. Laing. assessment. A measure of emotional empathy. Ickes. J. L. A. (1934). 602-627. In H. 33. 50. Hillsdale. Psychology of group influence (2nd ed. B. J. 4-20. Greif. (1986). Robertson. V (1988). Boston: Houghton-Mifflin. 88. 269-276. The nature of social influence in groupthink: Research methods in personality and social psychology (pp. R. CA: Academic Press. (1986). W. L. R. Separating individual and group Nemeth. 280-284. V. W. New York: Springer-Verlag. & Teng. S. CA: Academic Press. 143-186). Chicago: University of Chi- and self-monitoring: As different as "me" and "you. (1983). J. S.. S. W. (1973). Reis (Ed. & DePaulo.. Brewer (Eds. (1979). 79. M. A. R. (1989). York: Plenum Press. tion.kowitz(Ed. pp. Polarizing effects of social interaction. Lay epistemics and human knowledge. G. Victims ofgroupthink. G. W G. Personality. tion: A social relations analysis. Moreland. Malloy. S. R. Moscovici. S.Moreland. Reidhead. I. (1977). interaction of minds and bodies. 83. England: Myers. choice. Ickes. and validation.. 5-21). Phenomenology of perception. Review of personality and social psychology: Vol. 438-453. social psychology (3rd ed. & Hain. The social psychological effects of group decision Ickes. Ickes & E. 58-84.). In S. New York: tion: A theory and method ofresearch. M. Theory. G. Cambridge.. In L. A.

Schneider. Cognitive complexity measures consensus models in decision-making groups. Psychological Reports. A. (1979). Beavin. Handbook of social psychology Petty. D. P. D.communication: A study ofinteraction patterns. New York: Random House.). R. & Cacioppo. Communication processes and relationships. A. & Ellsworth. Miller gence. R. Journal ofApplied Psychology. M. acting groups. (1974). C. Kerr. incompatible relationships (pp. 59. C. R. measurement. B. NJ: man Communication Research. J. Person per- ception (2nd ed. In W Ickes (Ed. Truax. (1974). R. R. P. W B. R. 33. & Jackson. J. & Hertel. Snyder. 72-92. S. 30. and the relationship of cognitive complexity to communication.Watzlawick. Beyond IQ: A triarchic theory of human intelli- rules-based theory of interpersonal communication. Journal of Received June 15. In P. 433-436. MA: Addison-Wesley. New \brk: Norton. E. (3rd ed. 253-276). AND GARCIA O'Keefe. pathologies.& Sy pher. (1976). R. Brainstorming by individuals. P. adoxes. J. Self-monitoring of expressive behavior. E.95. Paulus (Ed. In G. Bal. Symbolic interaction and role theory. New York: Cambridge University Press.pendence in close relationships. Revision received May 1. 5. Riggio. Walker. pp. Sensitivity to nonverbal communication: The PONS Test. Erlbaum. 10. coacting and inter- erly Hills. G. M. CA: Academic Press. New York: Springer-Verlag. & Davis. Compatible and timore: Johns Hopkins University Press. E.173-176. T. 93-99. Penman. STINSON. (1985). & Foley. Hillsdale.1990 Stasser. Hu. suasive intent and involvement on cognitive responses and persua. (1985). 311-378). Aronson (Eds.). 839-864. Hall. DiMatteo. (1967). Chicago: Aldine. C. & CarkufT. T. H. Explorations in interpersonal communication (pp.... Pragmatics of human peutic personality change. (1973). San Stryker. Rogers. and par- 103. Influence processes and Accepted May 16. 17-36). Giuliano. Social intelligence: Its history and pathy. (Ed.1990 • . Cognitive interde- (1979). J. The coordinated management of meaning: A Steinberg. R. D (1989). M. & Archer. 526-537. J. DtJ. M. (1985). L. H. Tucker. T. A.). Effects of forewarning and per. Pearce.1989 Personality and Social Psychology. (1979). Psychology ofgroup influence (2nd ed. N. Lindzey & E. The necessary and sufficient conditions of thera. Toward effective counseling and sion.). J. Social skills and em. & Coffaro. Rosenthal. psychotherapy: Training and practice. 8.). Journal ofCounseling Psychology. Bev-Street. L. H. Personality and Individual Differences. Rogers. Personality and Social Psychology Bulletin. D. & Statham. J. 21. CA: Sage. In G. (1980). E. D. Hastorf. Wegner.742 ICKES. W R. R. (1989). (1981). pp. (1957). J. R. (1967). Diego. 279-326). BISSONNETTE. P. R. Reading. B.