THFQIC Date 0-7639 .9“. ............_.1. LIB R A R Y Michigan State University This is to certify that the thesis entitled RELATIONSHIP BETWEEN COMMUNICATORS' ABILITY TO PERCEIVE AND INTERPRET FACIAL AFFECT CUES AND PERCEPTIONS OF THEIR BEHAVIOR AS EMPATHIC presented by Lynn Elizabeth Fraedrich Aho has been accepted towards fulfillment of the requirements for Master of Arts degreein Corrmum'cation Mnjor professor June 23, 1980 ,L OVERDUE FINES: 25¢ per «y per item RETURNING LIBRARY MATERIALS: P‘lace in book return to rem charge from circulation reco RELATIONSHIP BETWEEN COMMUNICATORS' ABILITY TO PERCEIVE AND INTERPRET FACIAL AFFECT CUES AND PERCEPTIONS OF THEIR BEHAVIOR AS EMPATHIC By Lynn Eiizabeth Fraedrich Aho A THESIS Submitted to Michigan State University in partia] fuifiilment of the requirements for the degree of MASTER OF ARTS Department of Communication 1980 ABSTRACT RELATIONSHIP BETWEEN COMMUNICATORS' ABILITY TO PERCEIVE AND INTERPRET FACIAL AFFECT CUES AND PERCEPTIONS OF THEIR BEHAVIOR AS EMPATHIC By Lynn Elizabeth Fraedrich Aho Counselors and clinicians have long suggested that observation of clients'nonverbal emotional cues is extremely helpful in understand- ing and reacting to them. In this thesis, the relationship between ability to decode facial expressions of emotion, and to perceive dif- ferences in visual cues and being perceived as rewarding to interact with is examined. One-hundred-fifty-eight undergraduates were tested for decoding ability engaged in a role-play counseling exercise, and were rated on their expression of level of regard, empathic understanding, uncondi- tionality of regard, and general rewardingness by their partners. No significant relationships supporting the hypothesis of a po$itive re- lationship between decoding ability and interaction outcomes were found. Significant sex differences were found for level of regard and general rewardingness. Women were perceived as showing higher regard for partners and as being more rewarding interactants than men by both men and women. ACKNOWLEDGMENTS The help and encouragement of my guidance committee, Dr. Gerald Miller, Dr. Judee Burgoon, and Dr. Norman Fontes must be acknowledged here. I sincerely thank them, especially my chairman, Dr. Miller, whose advice and assistance were invaluable. My gratitude is also expressed to Dr. Michael Burgoon. I owe a big "thank-you" to my colleagues, especially Jim Dillard and Scott Garrison. Thanks are also expressed to research assistants Mike Manzoni, LuAnn Van Theime, Lorna Akins, and especially Jenny Moore. I am grateful to and would like to thank my parents, siblings, and other members of my family for their interest and support. Special thanks are expressed to Julie for typing this manuscript. Finally, my most heartfelt appreciation and thanks are expressed to my husband. ii TABLE OF CONTENTS Page LIST OF TABLES ......................... iv LIST OF FIGURES ........................ v Chapter I. RATIONALE ....................... I II. METHOD ........................ 9 Participants .................... 9 Operationalization of the Variables ......... 9 Procedures ..................... l5 III. RESULTS ........................ 20 IV. DISCUSSION ...................... 28 LIST OF REFERENCES ....................... 38 LIST OF TABLES Table Page 1. Additional Items Used to Measure Rewardingness of Interaction with Partner ................. l0 2. Response Choices Used in the Cue Interpretation Test . . . l5 3. Univariate Multiple Regression of Empathic Understanding Scale Scores Predicted by Assignment to Dyad and Cue Discrimination ...................... 21 4. Univariate Multiple Regression of Unconditionality of Regard Scale Scores Predicted by Assignment to Dyad and Cue Discrimination .................... 22 5. Univariate Multiple Regression of Level of Regard Scale Scores Predicted by Assignment to Dyad and Cue Discrimination ...................... 23 6. Univariate Multiple Regression of Additional Items Scores Predicted by Assignment to Dyad and Cue Discrimination . . 24 7. Results of Cross-sex Comparisons ............. 25 8. Two-way Analysis of Variance in Level of Regard Scale Scores by Participant Sex and Partner Sex ........ 26 9. Two-way Analysis of Variance in Additional Item Scores by Participant Sex and Partner Sex ............. 27 iv LIST OF FIGURES Figure Page 1. Example of Slides Used in the Cue Perception Test . . . l2 CHAPTER I RATIONALE Most extant work on the concept of empathy has been done by psy- chologists, who have characteristically taken one of three approaches in defining empathy. Some have described empathy as a mystic quality (Katz, l963; Stein, 1964), or have left it undefined (Sullivan, l953). Others, whose work is thoroughly summarized by Stotland (l969), have dealt with empathy as a psychOphysical response. A final group has considered em- pathy to be a social perception skill, dealing with individuals' ability to predict others' responses to selected stimuli or to infer the emo- tional states of others based on either behavioral or situational cues (Dymond, l949; Truax & Carkhuff, T967; Hogan, 1969; Natale, 1972). Dymond's (1949) paper engendered a line of research in which empathy was operationalized as an individual's ability to predict the responses of a friend or acquaintance to a trait questionnaire. Truax and Carkhuff (T967) dealt with the ability of therapists to infer clients' emotional states. Researchers examining the development of empathy in children (Cohen, l973; Partyka, T974; Brandt, l976) have defined it as the abil- ity to take the role of another. This has been operationalized either as the ability to describe a situation from the viewpoint of another or the ability to identify emotions another would be likely to feel in a given situation. The common keystone of all these approaches is that they deal with an individual's ability to understand another's cognitive or emotional state based on information provided by observable cues. l Because they deal with the ability to receive and process cues, these definitions have the most significance for communication. The social perception view of empathy also most closely resembles a communication definition of empathy presented by Miller and Steinberg (1975). In their definition, Miller and Steinberg propose that empathy consists of two steps: l. The prospective empathizer must be able to discrimi- nate accurately the ways that the individual's moti- vational and attitudinal posture differs from others. 2. Accurate discriminations must be followed by behaviors that are viewed as desirable, or rewarding by persons who are the objects of prediction. (p. 175) Unlike several of the authors who viewed empathy as a social perception skill, most notably Hogan (l969) and Dymond (l949, 1950), Miller and Steinberg have not defined empathy as the result of one or more under- lying personality traits. Rather, their definition specifies a series of events culminating in the perception by the person who is the object of empathy, or empathizee, that the empathizer's behaviors are desir- able or rewarding. In contrast, the social perception theorists would view this perception as the result of empathy. Empathy itself, in their view, is the empathizer's ability to understand the empathizee. Unlike Miller and Steinberg, they do not state that a person must base future communication on that understanding in order to be considered empathic. Shifting the definitional focus on empathy to the desired end state, as Miller and Steinberg have done, provides several advantages for the communication theorist. First, it emphasizes the importance of the perceptions of the empathizee and of communication between people. Unfortunately, none of the social-perception theorists re- viewed determined whether empathic skill, as they defined it, resulted in the empathizee feeling more understood or rewarded. Social percep- tion skill has limited utility if it does not increase people's abili- ties to reach their social goals. Second, viewing empathy as a communication process emphasizes the need to study each step in the process rather than limiting study to one isolated aspect. Thus, this view prompts empirical investigation of re- lationships of perceptual skills and personality traits to empathy. While taking a social perception-view of empathy does not prevent a re- searcher from doing this, social-perception researchers have typically either focused on measuring a person's perceptiveness in a particular relationship (Dymond, T949; Truax & Carkhuff, l967) or on the personal- ity traits of the empathic person (Hogan, T969). They have not dealt with either raw perceptual ability, such as visual or auditory acuity, or with skill in processing linguistic or nonverbal cues. Research linking personality traits with perceptual skills and both of these with empathic effectiveness would be a step toward an empirically, rather than philosophically based theory of empathy. Third, Miller and Steinberg's description of_the process of em- phasizing has implications for developing empathic ability. If, as they propose, the perceptual and inferential abilities needed to dis- criminate accurately among individuals' attitudinal and motivational postures allow a communicator to behave more consistently in ways that are rewarding to others, then developing these skills should allow a person to be more empathic. This skill development would have great utility in the training of counselors, therapists, and medical person- nel. Discriminating among individuals' attitudinal and motivational postures involves an inferential process of using available information to draw conclusions about underlying motives and attitudes. Observations of the empathizee's behavior, particularly his or her communication, are the chief source of information available to prospective empathizers. It is anticipated that the greater the individual's accuracy in perceiv- ing and interpreting verbal and nonverbal cues, the greater the amount and accuracy of the information she or he derives from observation of communication behavior. Two conceptually distinct abilities constitute skill in processing communication: the ability to perceive subtle vari- ations in verbal and nonverbal cues, and the ability to interpret, or accurately assign meaning to, these cues as indicators of the other's emotional or cognitive state. These two abilities tagether will be called cue discrimination. Accurate cue perception is a necessary con- dition for interpreting cues, since cues which are not perceived or are misperceived will not produce accurate interpretations, except by chance. Two other factors are also needed to produce accurate discriminations: motivation and inferential accuracy. Like any other inferential problem, discriminating people's attitudinal and motivational postures is sub- ject to error even when based on accurate information. For example, a person might accurately perceive and interpret a cluster of verbal and nonverbal cues as signalling anger and frustration, yet mistakenly infer the source of the frustration and anger, leading to an erroneous conclusion about the other's attitudes. Also, a person might have the ability to make highly accurate discriminations, yet lack the motivation to do 50. People do not always operate at the peak of their abilities nor do they invest maximal emotional and cognitive energy in all re- lationships--or even in all interactions with a given relational partner. The preceding reasoning culminates in the following theoretical proposition: 1. The greater the cue discrimination ability, the greater the accurate discrimination, given equal levels of inferential error and motivation. If the discussion is limited to situations in which communicators desire to reward each other, a second theoretical preposition may be formulated from Miller and Steinberg's (l975) definition of empathy. 2. Accurate discriminations lead to the empathizee per- ceiving that the prospective empathizer's behaviors are desirable and rewarding, that is, greater discrimi- nation produces greater empathy. Limiting the sphere of discussion in this way is suggested by Miller and Steinberg (l975) on the grounds that persons motivated to punish others are not considered empathic. Naturally, this proposition will not be true when persons desire to punish each other or are apathetic. Taken together, these two propositions lead to the first hypoth- esis of this study: H]: All other things being equal, the greater the cue discrimination ability, the greater the empathy. It must be emphasized that empathy, as defined by Miller and Steinberg (l975) is a function of the perceptions of the empathizee. Empathy is a relational phenomenon, not a characteristic of individuals. Stated differently, empathy only exists when people are communicating and can only be measured by ascertaining how the interaction is per- ceived by the communicators. Cue discrimination ability, on the other hand, is an individual characteristic. Its component skills exist and can be measured using objective tests when no interaction is taking place. In this conceptualization, empathy is a relational variable while cue discrimination is a psychological variable. Although the two constructs are hypothesized to be positively related, they are quite different in nature. ..Since Miller and Steinberg (1975) indicate that discrimination ability may be related to aspects of cognitive style (p. 183), it is deemed of theoretical interest to test the relationships between dis- crimination ability, being perceived as empathic, and certain person- ality traits which have been suggested as factors that contribute to a person interacting empathically. Hogan (l969) developed a measure of empathic disposition which includes items assessing the person's toler- ance, social awareness and affiliative tendencies, ethnocentricity, and authoritarianism. If these dispositional traits are related to empathy, high positive associations between the empathic disposition scores and scores for empathy and cue discrimination would be expected. This ex- pectation leads to the second and third hypotheses of this study: H2: Empathy is positively related to empathic disposition. H3: Ability to discriminate cues accurately is positively related to empathic disposition. Finally, whenever cross-sex differences in empathy have been re- ported (Dymond, 1949; Dymond, 1950; Hogan, l969; Mehrabian & Epstein, 1972), women have displayed greater empathy than men. Is this primarily an effect of social stereotyping, a result of more accurate discrimina- tion ability, a general motivational tendency to try to please others, or some combination of these factors? The literature on sex differences (Maccoby & Jacklin, 1974) is inconclusive concerning each of these pos- sibilities. There are no consistent findings of sex differences in visual perception. However, some studies of adolescents and adults re- port that males are more proficient at visual-spatial tasks, while others report no differences. Self-concept research suggests that fe- males tend to view themselves as more socially oriented than do males. Studies of children's social behavior, on the other hand, yield no con- sistent pattern of differences in the affiliativeness or social orienta- tion of girls and boys. No differences in accuracy of interpretation of emotion in faces for children ages three to adult have been reported. Hall (1978) reports that the overwhelming majority of studies which re- port examining sex differences found that females were more accurate .at interpreting emotional expressions, but notes that a great many studies do not comment on sex differences at all. Because knowledge of this area is rather limited, a series of cross-sex comparisons is warranted. First, a comparison of the perceived empathy of men and women would provide a test of the idea that women are more empathic. This leads to a fourth hypothesis: H4: Women are perceived as more empathic than men by both men and women. In addition, cross—sex comparisons of cue perception and inter- pretation and of empathic disposition could shed some light on the source of the differences, if any exist. This yields three additional hypotheses: H5: Women display greater ability to perceive cues than men. H6: Women display greater ability to accurately interpret cues than men. H7: Women show more empathic disposition than men. These comparisons have theoretical import beyond simply comparing the performance of men and women because of the extent to which sex differ- ences could mask the hypothesized relationships. For instance, if women are viewed as uniformly empathic due to social stereotyping, regardless of the quality of the interactions in which they participate, the hypothe- sized relationship between empathy and accurate discrimination ability will be obscured. Finally, comparisons among dyads of differing sex composition may be made by summing partners' empathy scores. Basing predictions solely on the sex differences already predicted, the hypothesis for this com- parison is: H8: the most perceived empathy is found in a female-female When scores for members of each dyad are combined, dyad, the next most in a mixed-sex dyad, and the least in a male-male dyad. If the similarity of the communicators is a salient factor in the amount of empathy generated, significant differences will be found, but mixed- sex dyads will show the least empathy, male-male dyads an intermediate amount, and female-female dyads the most. CHAPTER II METHOD Participants The participants were 158 students, 59 men and 99 women, enrolled in either an introductory communication course or an introductory psy- chology course at Michigan State University. The students, who were given extra credit for volunteering to participate, were assigned ran- domly to same-sex or mixed-sex dyads. There were 13 male-male dyads, 29 female-female dyads, and 30 mixed dyads. No dyad consisted of per- sons who were previously acquainted. Operationalization of the Variables Empathy, as conceptually defined by Miller and Steinberg (1975) was operationalized as the ratings of each participant's behavior made by his/her partner on scales from the Relationship Inventory (Barrett- Lennard, 1962) and additional items of the same format constructed to refer specifically to the research situation. The Relationship Inven- tory was selected because its Empathic Understanding Scale provides a brief and clear measure of empathy pgr_§g, Items form the Level of Regard and Unconditionality of Regard scales were used as additional indicators of rewarding behavior. This scale was based on concepts of counseling developed by Carl Rogers. Unconditionality is considered rewarding in that context on the basis that consistent behavior, that is, that expressed emotions toward another are not changed by or 10 dependent on transitory events, is a necessary condition for positive change in psychotherapy (Barrett-Lennard, 1962). Several items used to obtain ratings of participants' enjoyment of the discussion task and their willingness to participate in another experiment involving the same type of task with the same or another partner were framed in the same format as the Relationship Inventory items and included in the questionnaire. Each Relationship Inventory item consists of a statement about the partner's behavior and six response categories ranging from "I strongly feel that it is true,“ to "I strongly feel that it is not true,“ with no neutral point. There were 16 items in the Empathic Understanding scale, 17 in the Level of Regard scale, 17 in the Uncondi- tionality of Regard scale, and 6 additional items. The additional items are listed in Table 1. Table 1 Additional Items Used to Measure Rewardingness of Interaction with Partner 51. I would enjoy participating in another experiment involving a discussion period similar to the one in this statement. 52. I would be willing to participate in another experiment like this one if I could receive extra credit for it. 53. I felt very uncomfortable during the discussion period. 54. If I participated in another experiment like this one, I would not want to have the same partner. 55. If I have a chance, I will try to get to know my partner in this experiment as a friend. 56. I would have enjoyed the discussion task more if I had had a different partner. 11 Reliabilities for the Relationship Inventory are reported by Barrett-Lennard (1962), who also describes content-validity tests of the inventory. Comparisons between the Relationship Inventory and other empathy measures are reported by Caracena and Vicory (1969). Barrett-Lennard used the Spearman Brown formula to calculate split-half reliabilities for each scale with the following results: Level of Re- gard, .93; Empathic Understanding, .86; Unconditionality of Regard, .82. Test-retest correlations in another sample were, Level of Regard, .84; Empathic Understanding, .89; and Unconditionality of Regard, .90. Re- liability analysis of data collected in this study yielded the follow- ing Cronbach's Alphas: Level of Regard, .90; Empathic Understanding, .80; Unconditionality of Regard, .77. The additional items had a reli- ability of .57. Cue perception, the sensory capacity to perceive differences in visual or audible cues, was operationalized as performance in response to a series of 20 visual discrimination items presented on slides (Fig- ure 1). Each slide consisted of a stimulus figure and two matching figures. Participants were to select which, if either, of the two match- ing figures was different from the stimulus figure after the slide was presented for one half second. Response categories were: a) Both_fig- ures are the same as the stimulus figure. b) The figure on the left_ differs from the stimulus figure. c) The figure on the right differs from the stimulus figure. d) Both_figures diffgr_from the stimulus fig- ure. The choices were displayed on the screen for about 20 seconds after each stimulus slide was presented. Reliability analysis yielded a Cronbach's Alpha of .47 for the cue perception test. 11 Reliabilities for the Relationship Inventory are reported by Barrett-Lennard (1962), who also describes content-validity tests of the inventory. Comparisons between the Relationship Inventory and other empathy measures are reported by Caracena and Vicory (l969). Barrett-Lennard used the Spearman Brown formula to calculate split-half reliabilities for each scale with the following results: Level of Re- gard, .93; Empathic Understanding, .86; Unconditionality of Regard, .82. Test-retest correlations in another sample were, Level of Regard, .84; Empathic Understanding, .89; and Unconditionality of Regard, .90. Re- liability analysis of data collected in this study yielded the follow- ing Cronbach's Alphas: Level of Regard, .90; Empathic Understanding, .80; Unconditionality of Regard, .77. The additional items had a reli- ability of .57. Cue perception, the sensory capacity to perceive differences in visual or audible cues, was operationalized as performance in response to a series of 20 visual discrimination items presented on slides (Fig- ure 1). Each slide consisted of a stimulus figure and two matching figures. Participants were to select which, if either, of the two match- ing figures was different from the stimulus figure after the slide was presented for one half second. Response categories were: a) 89th fig- ures are the same as the stimulus figure. b) The figure on the left differs from the stimulus figure. c) The figure on the righthiffers from the stimulus figure. d) Both_figures giffgr_from the stimulus fig- ure. The choices were displayed on the screen for about 20 seconds after each stimulus slide was presented. Reliability analysis yielded a Cronbach's Alpha of .47 for the cue perception test. 12 Figure l--Examp1e of Slides Used in the Cue Perception Test. 13 Cue interpretation, conceptually defined as the ability to accu- rately assign meaning to emotional cues, was operationalized as the ability to identify the emotion displayed on each of 30 slides showing facial expressions of "pure" emotion (Ekman & Friesen, 1976). Each slide consists of a black-and-white closeup of the face of a poser dis- playing one of six emotional expressions, happiness, sadness, fear, anger, surprise, and disgust. Based on their extensive research on the nature of facial expressions of emotion, Ekman and Friesen (1975) con- sidered these six expressions basic or pure emotional expressions which are blended to produce more subtle or complex expressions of emotion. Most of the time, they found, interactants mask their emotions with smiling or neutral expressions. Pure expressions of emotion occur only momentarily, rarely enduring for longer than .8 second, and frequently disappearing much more quickly. The posers appearing in the slides were trained to contract or relax specific facial muscles to duplicate the expressions isolated in earlier research. Ekman and Friesen (1976) report that all the photographs from which those used in this study were selected had reliabilities of more than .70, the majority having reliabilities of .90 or greater. A Cronbach's Alpha of .67 was obtained for this test. Five slides depicting each of the six emotions were selected. Each emotional expression was portrayed by both male and female posers. Half the slides in the set showed female posers and half male. Slides were ordered so that neither the same emotion nor the same poser were presented twice in a row, and the order of presentation was the same for all subjects. For each slide, five response choices were displayed 14 for about 30 seconds after the stimulus slide had been displayed for one half second. The very short exposure time for the stimulus slides in this test and the cue perception test is based on Ekman and Friesen's findings on the duration of emotional expressions. Because emotional expressions appear very fleetingly, one must be able to perceive them very quickly while interacting, and a brief test display approximated this situation better than a more lengthy one. Four different response sets, which are displayed in Table 2, were used. Because expressions of fear are difficult to distinguish from those of surprise, and expressions of anger are difficult to dis- tinguish from those of disgust, response choices were assigned so that half of the slides of these emotions were easy discriminations by vir- tue of not presenting the incorrect member of the pair as a choice. Otherwise, response sets were distributed randomly. To facilitate recognition that different sets of response choices were being pre- sented, each set was photographed on a different color background. Empathic disposition was conceptually defined as those person— ality traits associated with empathy in the social-psychological literature, including tolerance, lack of authoritarian tendencies, self-acceptance, and extraversion. This construct was assessed using Hogan's (l969) Empathy Inventory, a personality inventory consisting of 64 self-descriptive statements to which participants are to respond either "true" or "false." Sample statements are: "I always try to consider the other fellow's feelings before I do something." "I am not easily angered,“ and “I would like to belong to a singing club." The entire instrument is available from the Psychology Department of 15 Table 2 Response Choices Used in the Cue Interpretation Test Anznggeiet Backgrog??d:010r Response choices 1 white a. disgust b. happiness c. sadness d. surprise e. fear 2 orange a. fear b. sadness c. anger d. happiness e. surprise 3 yellow a. anger b. fear c. happiness d. disgust e. sadness 4 light blue a. surprise b. disgust c. anger d. happiness e. sadness The Johns Hapkins University. Hogan (1969) reports that Spearman- Brown reliabilities for composite ratings using his scale ranged from .68 to .86 with a mean reliability of .80. A test-retest correlation over a two month interval yielded a reliability of .84. A Cronbach's Alpha of .72 was obtained for the Empathy Inventory in this study. Procedures Upon arriving at the lab, participants signed a consent form and were given an identification number card. The use of this number 16 to identify all experimental materials instead of a name or student ID number helped ensure the participants' anonymity. The numbers also served to assign participants to dyads and to match up the scores for members of each dyad. Participants were told that the topic of the first study was "counseling by non-professionals" and that they would be asked to discuss with their partners a "moderately stressful problem, of the sort you might discuss with friends." Each participant acted as counselor for one 10-minute discussion period and counselee for the other. "Having a disagreement with one's roommate over housekeeping chores" was given as an example of a moderately stressful problem. Em- phasis was placed on discussing a real problem, and any participant who found this too threatening was given a chance to leave at this point. After the two "counseling" sessions, each participant rated her or his partner on scales from the Relationship Inventory (Barrett- Lennard, 1962). Participants were told that this completed the first experiment. The cue perception and cue interpretation tests were introduced as part of a second experiment, "Interpretation of Nonverbal Cues." The cue perception test was administered first. Participants were shown a sample stimulus slide and instructed that for the first test, they would have to choose which figure, if any, was different from the stimulus figure on the far left, and the response choices were re- viewed for them. They were told that each set of figures would be displayed on the screen for about half a second and followed by a longer display of the answer choices during which they were to mark their answers. A sample item was shown in cadence to familiarize par- ticipants with the procedure. 17 When the cue perception test was completed, instructions were given for the cue interpretation test. Participants were told that this test was similar to the first one, except that instead of picking which item was different, they would see a single photograph of a face and select the emotion they thought was being expressed. They were warned that unlike the first test, several different answer sets were used in this exercise, and that they would be cued in to this fact by the use of a different background color for each different answer set. A practice item was shown in cadence to familiarize participants with the procedure. Then they completed the 30-item test. Next, participants completed Hogan's (l969) Empathy Inventory, which was introduced as an assessment of "the way you view yourself and the world around you." Before beginning the inventory,participants were reminded that their responses were both anonymous and confidential and asked to answer as honestly as possible. Finally participants were debriefed, thanked for their participation, and dismissed. Although it was expected that the procedures would allow for few, if any, order effects, one potential problem was noted. At the begin- ning of the cue perception test many of the participants, who were seated near their role-play partners, continued to talk to each other and had difficulty attending to the slides shown for the test. If, as seems likely, this was most likely to happen in those dyads with the most rewarding outcomes, then members of these dyads would show arti- ficially diminished scores for cue perception, attenuating the pro- posed relationship between cue perception and empathy. Since accurate discrimination ability is composed of cue per- ception ability and the ability to interpret cues accurately, scores 18 for cue discrimination were created by summing scores from the cue per- ception and interpretation tests. Hypothesis 1 was tested with a multi- variate multiple regression analysis using accurate discrimination and assignment to dyad as predictors of empathy, the latter being measured by the partners' ratings on the three scales of the Relationship Inven- tory and the additional items. These four measures were correlated, but the Relationship Inventory (Barrett-Lennard, 1962) is not summative across scales. Therefore, scores for the three Relationship Inventory scales and the additional items were considered a vector of related dependent variables in the analysis. On the basis of work by Pillai and Jyachandran (1967) indicating that Hotelling's trace statistic has greater power than Pillai's criterion, Wilk's Lambda, or Roy's Largest Root for large samples with highly divergent roots, Hotelling's trace statistic was chosen as the test statistic. Univariate regression analy- ses were conducted to obtain regression coefficients of accurate dis- crimination and assignment to dyad, which was included to control for the fact that partners' ratings of each other are not independent. The relationships between empathic disposition and perceived em- pathy and cue perception ability specified in Hypotheses 2 and 3 were tested by computing two partial correlations: a correlation between empathy and empathic disposition controlling for the effects of cue discrimination ability, and a correlation between discrimination ability and empathic disposition controlling for the effects of perceived em- pathy. For the purposes of these tests, perceived empathy was measured by the score on the Empathic Understanding scale of the Relationship Inventory. 19 Hypotheses 4, 5, 6, and 7, which specify sex differences for indi- viduals, were tested using one-tailed t-tests between means for empathy, cue perception, cue interpretation, and empathic disposition. The differences in empathy generated in mixed and same sex dyads posited by Hypothesis 8 were tested for overall significance using an analysis of variance. All tests were conducted at the .05 level of sig- nificance. CHAPTER III RESULTS The multivariate multiple regression analysis conducted to test Hypothesis 1 showed trends, but no significance. The value of the test statistic, Hotelling's trace statistic, was .125; its approximate F was 1.96 at 8/250 degrees of freedom. Its probability was .052. Univariate regression analyses were performed for each of the scales from the Relational Inventory (Barrett-Lennard, 1962) to further eXplore the nature of the trends indicated by the multivariate analysis. The analyses performed on the empathy scores and the additional items yielded no significant results or trends. The analysis of the Level of Regard scores was not significant overall, but did yield a significant regression coefficient for discrimination of -.22. The R2 for discrimi- nation in this analysis was .03. The analysis of the unconditionality of regard scores, which was also nonsignificant overall, yielded a re- gression coefficient of .19 at a significance level of p <.55. The R2 for discrimination in that analysis was .03. Thus,no support can be claimed for the hypothesized positive relationship between discrimi- nation and empathy. The partial correlation between empathy and empathic disposition controlling for discrimination ability was -.05 and was not signifi- cant (p :n05; df = 148). The partial correlation between empathic disposition and discrimination ability controlling for empathy was -.O4, 20 21 Tab1e 3 Univariate Multiple Regression of Empathic Understanding Scale Scores Predicted by Assignment to Dyad and Cue Discrimination Y B 11X] + B 12X2 Multiple R = .044 d.f. = 2/129 p:>.05 R2 = .002 F = .122 Variable beta R2 change F p Assignment to dyad (X1) -.016 .0017 .198 p >.05 Cue discrimination (X2) .019 .0025 .032 p >.05 22 Table 4 Univariate Multiple Regression of Unconditionality of Regard Scale Scores Predicted by Assignment to Dyad and Cue Discrimination A Y2 = B 21x1 T B 22X2 Multiple R = .171 d.f. = 2/129 p >.05 R2 = .029 F = 1.93 Beta§_ Assignment to dyad beta = .004 F = .012 R2 change = .00003 p > .05 Cue discrimination beta = .192 F =3.86 R2 change = .029 .05 >p < .06 23 Table 5 Univariate Multiple Regression of Level of Regard Scale Scores Predicted by Assignment to Dyad and Cue Discrimination A Y3 ‘ B31x1 + B 32X2 Multiple R = .179 d.f. = 2/129 p:>.05 R2 = .032 F = 2.13 Beta; Assignment to dyad beta = .006 F = .023 R2 change = .0008 p > .05 Cue discrimination beta =-.224 F =4.15 R2 change = .032 p < .05 24 Table 6 Univariate Multiple Regression of Additional Items Scores Predicted by Assignment to Dyad and Cue Discrimination ? 4 = B41x1 + B42X2 Multiple R = .119 d.f. = 2/129 p >.05 R2 = .014 F = .932 Betas Assignment to dyad beta = .012 F = .524 R2 change = .003 p > .05 Cue discrimination beta = .053 F =1.47 R2 change = .011 p > .05 25 also nonsignificant (p >.05; df = 148). Thus, the data fail to support either Hypothesis 2 or Hypothesis 3. The sex differences Hypotheses, 4-7, were tested using t-tests, the results of which are presented in Table 7. Table 7 Cross-sex Comparisons Group means Variable females males d.f. t Empathic Understanding 8.6 6.4 149 1.23 Unconditionality of Regard 0.9 0.03 145 0.44 Level of Regard 27.6 23.3 155 2.04* Additional Items 6.4 4.2 147 2.57* Cue perception 38.8 38.1 156 0.54 Cue interpretation 67.1 66.4 156 0.63 Empathic diSposition 52.5 50.7 156 1.93* p <.05 As can be seen from these results, the data fail to support Hypothesis 5, that women would demonstrate greater ability than men to perceive cues, and Hypothesis 6, that women would show greater ability than men to interpret cues. The data do support Hypothesis 7, that women show higher levels of personality traits related to empathy. Hypothesis 4 is also partially supported; i.e., that women show more empathy than men. This support 26 is indicated by significant differences for two of the four scales used as indices of empathy. Since the Level of Regard and experimental partner evaluation scores (additional items) were ratings made by the participants' part- ners and since there were far more female-female dyads than male- female dyads, these results can conceivably be interpreted as a sex difference in response tendencies, i.e. females may tend to rate their partners more positively than do males. To eliminate this possibility, two-way analyses of variance using participant sex and partner sex as variates were computed for each measure. Neither measure showed effects for partner sex or for interaction. Therefore, these results do not appear to reflect sex differences in the response tendencies of the raters. Table 8 Two-way Analysis of Variance in Level of Regard Scale Scores by Participant Sex and Parnter Sex Source SS d.f. MS F Participant sex 541.62 1 541.62 3.17a Partner sex 27.61 1 27.61 .16a Interaction 34.08 1 34.08 .202 Error 21868.22 128 170.85 Total 22448.24 131 171.36 27 Table 9 Two-way Analysis of Variance in Additional Item Scores by Participant Sex and Partner Sex Source SS d.f. MS F Participant sex 118.93 1 118.93 4.55 Partner sex 6.96 1 6.96 .28a Interaction 31.18 1 31.18 1.19a Error 3343.09 128 26.12 Total 3494.55 131 26.68 E'p > .05 Hypothesis 8, which specified sex-composition differences in the total amount of empathy generated in the dyad was tested using an anal- ysis of variance. No significant effect of sex composition was found on the amount of empathy generated in the dyad (F