ing the judgment of an individual's emotional state. It was once widely believed that expressive behavior is an unreliable indica- tor of emotional state and that ...
Journal of Personality and Social Psychology 1990, Vol. 59, No. 5, 1032-1039
Copyright 1990 by the American PsychologicalAssociation, Inc. 0022-3514/90/$00.75
Relative Contributions of Expressive Behavior and Contextual Information to the Judgment of the Emotional State of Another Makoto Nakamura, Ross Buck, and David A. Kenny University o f Connecticut This study used a technique for assessing the relative impact of facial-gestural expressions, as opposed to contextual information regarding the elicitor and situation, on the judgment of emotion. In Study 1, 28 undergraduates rated videotapes of spontaneous facial-gestural expressions and separately rated the emotionally loaded color slides that elicited those expressions. The source claritiesof the expressionsand slides were matched using correlation and distance measures, and 18 expressions and 9 slides were selected. In Study 2, 72 undergraduate receivers were shown systematic pairings of these expressions and slides and rated the emotional state of the expresser, who was supposedly watching that slide under public or private situational conditions. Expressions were found to be more important sources for all emotion judgments. For female receivers slides were relatively more important in the public than in the private situation.
One of the fundamental areas for debate in the study of emotional communication is the question o f the relative roles o f expressive behavior on the part o f the individual versus knowledge of the context in which the expression occurs in determining the judgment o f an individual's emotional state. It was once widely believed that expressive behavior is an unreliable indicator o f emotional state and that context (i.e, the eliciting stimulus and situation) contributes more than expression in emotion judgments (e.g, Bruner & Tagiuri, 1954; Hunt, 1941; Tagiuri, 1968). The detailed critique o f this belief by Ekman, Friesen, and Ellsworth (1972) did much to make the study o f emotional expression scientifically respectable. The critique by Ekman et al. (1972) was based on conceptual distinctions that forced a reevaluation o f previous experimental designs and past data. For example, these authors pointed out that the relative importance o f expressive behavior and contextual information (i.e, about the elicitor and situation) would be determined in great part by the clarity or ambiguity of the information from each of these sources. They defined source clarity as the "amount or type o f information available to observers when they are exposed to a single source" (p. 138). Ekman et al. noted the need systematically to measure and match source clarity in order to study meaningfully the relative importance o f expressive behavior and contextual information in the judgment of emotion.
Context In general, contexts are defined by senders, elicitors, situations, and receivers. Senders are stimulus persons whose fixed characteristics (e.g, age, gender, and physical features) provide information. Elicitors are stimuli that evoke emotion in senders. Situations are environments or fields that surround senders, elicitors, and sometimes receivers, and are characterized by the physical setting and social setting. Finally, receivers are people who are exposed to and receive information from senders, elicitors, or situations, or any combination thereof. One might think that receivers are usually observers and do not play important parts in characterizing contexts. Although this is generally true in experimental situations, and is in the present study, in everyday life the relationships between senders and receivers are interactive (Kenny & La Voie, 1984). In interactive settings, receivers are not mere observers but positive interactants who influence the contextual information) Why is a particular emotion expected to be associated with a certain pattern of contextual information? To answer this question, the effects o f display and decoding rules must be examined. Display rules were originally defined by Ekman and Friesen (1969; Ekman, 1973, 1977) to refer to the conventions, norms, and habits that develop regarding the management o f emotional responses. For example, Ekman, Friesen, and Malmstrom (1970; Ekman & Friesen, 1975) found that whereas American students expressed in an interview their negative emotion toward a tension-evoking movie, Japanese students tried to make positive expressions. This result is consistent with
This research was supported in part by National Institute of Mental Health Grant R01 MH-4075301 to Ross Buck and R01 MH-4029501 to David A. Kenny and by the University of Connecticut Research Foundation. Thanks go to Deborah A. Kashy, who provided critical comments on an earlier version of this article. Makoto Nakamura is now at the General Psychology Laboratory, Faculty of Human Sciences, Osaka University, Osaka, Japan. Correspondence concerning this article should be addressed to Ross Buck, Department of Communication Sciences, University of Connecticut, Storrs, Connecticut 06269-1085.
Contextual information is discussed in a general and interactive sense. Although it is important to take interactions between senders and receivers into account, as Kenny and La Vole (1984) suggested, in the following discussion it is assumed that the relationship between senders and receivers is static and noninteractive. Thus, in this study receivers do not affect contextual information, although they may be affected by that information. 1032
CONTRIBUTIONS OF BEHAVIORAND INFORMATION the fact that in Japan there exists a taboo (i.e, a display rule) against expressing negative emotions in public situations. Shimoda, Argyle, and Ricci-Bitti (1978) also reported that in comparison to English and Italian students, Japanese students tended not to express clearly emotions and attitudes. Decoding rules are the opposite side of the same coin: They are defined as cultural rules or expectations concerning the attention to and interpretation of emotional information (Buck, 1984). It is possible to interpret the ability to detect deception as a kind of ability to manage decoding rules. Blanck and Rosenthal (1982) and DePaulo and Rosenthal (1979) reviewed and studied the developmental changes in decoding discrepant messages from multiple sources. It was found that the ability to decode discrepant messages increases with age; that is, as they get older, children begin to attend to less controllable channels, which are more likely to "leak" or reveal the "true feelings," when they are confronted with inconsistent messages. Both display and decoding rules are learned through developmental processes and play important roles in interpersonal and social communication (Saarni, 1984). They come into play in the handling of the complex pattern of contextual information when making judgments of emotion. Expressive Behavior In contrast to context, expressive behavior includes information from different expressive channels of the sender (e.g, facial expressions, body movement, paralanguage, and verbal behavior). Emotional expressions are linked to judgments of emotion both directly (i.e, through spontaneous communication; Buck, 1984) and through decoding rules. Many studies have investigated the relative importance of different channels of emotional expression. The relative importance of channels is often defined in terms of the relative controllability or strength of inhibition of emotion expression in the channels. There seems to be a leakage hierarchy of nonverbal channels within a sender (cf. Blanck & Rosenthal, 1982; DePaulo & Rosenthal, 1979; Ekman & Friesen, 1974): Discrepancies between different channels are the most leaky (uncontrollable); micromomentary facial/bodily cues are next; and tone of voice and body movement follow. Facial expressions constitute the most controllable nonverbal channel. In regard to the need to control the expressive display, it is well-known that situations vary along a public-private (or formal-informal) dimension (cf. Knapp, 1984) and that the privacy of a situation may be critical to the expression or inhibition of true emotion. Indeed, one would expect that a common decoding rule is that expressive behavior is usually more inhibited in a public situation than in a private situation. In judging the relative contributions of expressive behavior and context (elicitors and situations) to emotion judgments, the relative importance of the information sources should be related to the clarity of the expression relative to the clarity of the elicitor, as well as the controllability of the emotional expression relative to the need to control such expression (e.g, in a public vs. private situation). Previous Research Previous studies in this area have produced conflicting results (Frijda, 1958; Goldberg, 1951; Goodenough & Tinker,
1033
1931; Knudsen & Muzekari, 1983; Munn, 1940; Spignesi & Shor, 1981; Vinacke, 1949; Wallbott, 1988; Watson, 1972). Only one study (Watson, 1972) found that facial cues contribute to emotion judgments more than contextual cues, and no study found context primacy. The controversy about the relative importance of face and context was renewed by Russell and Fehr (1987), who demonstrated that judgments of target facial expression can be influenced by the previous viewing of facial expression anchors, thus suggesting the importance of context. In reply, Ekman and O'Sullivan (1988) argued that Russell and Fehr's target faces may have been emotionally ambiguous or neutral (see also Russell & Fehr, 1988). Also, recently Wallbott (1988) found evidence that expression information was dominant when the expresser was female, whereas situation information was dominant when the expresser was male. Contextual information was not often systematically considered in these studies, and the relationship between elicitors and situations was often not stated. For example, Watson (1972) used the following sentences as contextual information: "He has just asked his girl to marry him and she says'yes' "and "At a party, he overhears an acquaintance making an insulting remark about his girl:' In the latter case, receivers can get information both of an elicitor (insulting remark about his girl) and a situation (at a party). However, the former sentence provides only information of an elicitor, that is, her answer"yes; which will presumably evoke happy feelings in the man. Situational information is not available in the sentence, and this may lead to inconsistent results. If the situation were private, judges may expect that the person will express emotion freely, whereas if the situation were public, facial expressions may be misleading. The Present Study One of the most critical problems with previous studies is in the definition and control of contextual information (e.g, elicitors, situations, and their relationship). The main purpose of the present study was to investigate the relative contributions of expressive behaviors and elicitors to emotion judgment when situational information is both specified and manipulated. As noted above, Ekman et al. (1972) emphasized the necessity of considering the clarity of information provided by both sources, face and context. If the clarifies of the sources are not known, it is impossible to compare the relative contributions of the information sources in a meaningful way. For example, if an informative face is combined with an ambiguous context and if the results show more contribution of face than context, it is not valid to conclude that facial expressions compose a more important source of information than contexts. The results may reflect only the difference of the amount of information available in those sources. The slide viewing technique (Buck, 1978) is useful in the study of this topic. This technique has been developed to study the spontaneous communication of emotion through facialgestural expressions. Sender subjects are shown a series of emotionally loaded slides and asked to describe their reactions to the slides verbally while, unknown to them, their facial-gestural reactions (including the face, shoulders, and upper torso) are videotaped by a hidden camera. The slide viewing technique makes it possible to specify not
1034
M. NAKAMURA, R. BUCK, AND D. KENNY
only the clarity of information provided by the expression and the elicitor (the slide), but also to define and manipulate the situation. In the present study, receivers were provided with explanations about the slide viewing situation that led them to view it as either a public or private (spontaneous) expressive situation. Instructing receivers that senders were watching slides by themselves and were recorded by a hidden camera should encourage the receiver to regard the situation as more informal and spontaneous, for it implies relatively freely expressed behavior on the part o f senders. On the other hand, instructing receivers that senders were watching slides with several experimenters in a laboratory and were being recorded by a visible camera should encourage receivers to regard the situation as more formal and public, and it should imply that the expressive behavior on the part o f the sender is more influenced by display rules, On the basis o f the preceding considerations, the following hypotheses were developed: First, we expected that facial expressions would generally have a larger effect on the emotion judgments than elicitors. This is based on the evidence o f the universality o f facial display o f emotion (Darwin, 1972; Ekman & Friesen, 1975), which would suggest that all else being equal, a sender's facial expression should be a more reliable indicator o f his or her particular emotional response than would the elicitor. For this reason, we expected that there would be a general tendency for facial expressions to be more important than contextual information in emotion judgments. Second, we expected that facial expressions would be relatively more important in the private situation than in the public situation, whereas elicitors would be relatively more important in the public situation than in the private situation. In the public situation, receivers should expect that the level of inhibition of expressive behavior is high relative to the private situation. On the other hand, the effect of elicitors may be rather stable across the situations. Two studies were conducted. In Study 1, the source clarities o f the facial-gestural expressions and slides were measured and equated. In Study 2 the above hypotheses were examined by using the pairings o f the expressions and slides that were selected in Study 1, and decoding rules were also analyzed. Study 1 The purpose o f Study I was to measure and equate the source clarities o f facial expressions and slides.
Method Subjects. A total of 28 undergraduates (14 men and 14 women) enrolled in an introductory communication sciences course at the University of Connecticut participated in Study 1. Stimuli. In all, 18 videotaped facial-gestural expressions (3 expressions for each of 6 senders) and nine slides were used as stimuli. The expressions and slides were selected in pilot studies from a pool of 232 videotaped facial expressions of undergraduates at the University of Connecticut taken by Wagner, Buck, and Winterbotham (1989) and the pool of the slides used in the slide viewing technique (Buck, 1978). Procedure. Subjects were told that the expressions were taken in a private situation; that is, the expressions were recorded by a hidden video camera while senders were watching slides alone.2 Subjects were
shown all 18 expressions and all nine slidesone-by-one by using a 20-in. TV monitor with a videotape recorder (for facial expressions) and a slide projector with a 30 × 25-cm back-lightedscreen (for slides). When viewing expressions, subjects were asked to rate how they thought the sender felt; when viewing slides, they rated how they thought people, in general, would feel about the slide. In both cases, the ratings were made along eight 7-point emotion scales. The eight scales were Pleasant-Unpleasant, Strong-Weak, Not at all Happy-Very Happy, Not at all Sad-Very Sad, Not at all Angry-Very Angry, Not at all Afraid-Very Afraid, Not at all Surprised-Very Surprised, and Not at all DisgustedVeryDisgusted. The mean ratings of the stimuli on these scales constitute the measure of the source clarities of the stimuli.
Results a n d Discussion Distance and correlation measures among stimuli were obtained to match the source clarities o f the stimuli. To measure the correlation between two stimuli, the eight scales were correlated for each subject. These correlations were then averaged across the 28 subjects. The distance measure was the square root o f the sum of squared differences between the means averaged across 28 subjects' eight scales. By examining the distance and correlation measures and also profiles o f ratings o f each slide and facial expression across the eight scales, 18 facial expressions (3 expressionsmpleasant, neutral, and unpleasant-each from 3 different male and 3 different female senders) and 9 slides (3 pleasant, 3 unpleasant, and 3 neutral) were selected. Those slides/expressions that provided closer distance and stronger correlation with one another and that met additional criteria were chosen. These additional criteria were as follows: Pleasant face/slide. The mean rating scores on the Pleasant and Happy scales were greater than the midpoint of the 7-point scales; the mean rating scores for negative emotions were less than the midpoint; and the differences o f the mean rating scores between the positive and negative emotions scales were more than one unit on the scales. Neutral face/slide. Mean rating scores (except for the Pleasant scale) were less than the midpoint and the ranges of the ratings were within two units on the scales. Unpleasant face/slide. Mean rating scores for Pleasant and Happy scales were less than the midpoint; mea.n ratings for negative emotions were greater than midpoints; and the differences o f the rating scores between the positive and negative emotion scales were more than one unit on the scales. Table 1 shows the average correlation (the top number in each cell) and distance measures (the bottom number) for each cell. The value in each cell is the mean across subjects o f the correlations and the distances between stimuli. Recall that the larger the correlation and the smaller the distance, the more similar the two stimuli are perceived. The results suggest that the neutral facial expressions were rated closer to the pleasant slides than to the neutral slides and that the neutral slides were rated closer to the pleasant facial expressions than to the neutral facial expressions. Therefore, both the neutral facial expressions 2 Earlier pilot studies varied the contextual information presented to subjects. We had public and private conditions as in Study 2. Because no differences were found between these two conditions in the pilot studies, only the private condition was used in this study. This supports the expectation that the effects of elicitors are stable across situations.
CONTRIBUTIONS OF BEHAVIOR AND INFORMATION Table 1
Means of Correlation and Distance Scores Measure 1. Neutral face 2. Pleasant face 3. Unpleasant face 4. Neutral slide 5. Pleasant slide 6. Unpleasant slide
1
2
.36 9 .30 20 -.16 21 .17 12 .34 21 - . 11 31
-.70 11 -.61 34 .76 15 .96 10 -.76 44
3
4
5
6
.99 7 -.84 47
m
-.86 11 -.29 22 -.71 37 .89 14
--.96 5 .66 18 -.47 32
m
B
.99 6
Note. The top and bottom numbers in each cell are correlation and distance scores, respectively.. Higher correlations and smaller distances indicate greater similarity.
and slides tended to be rated as pleasant. However, with the exception o f the neutral stimuli, the ratings o f the same types (pleasant and unpleasant) o f facial expressions and slides were highly consistent with each other, indicating that their source clarities were matched, and these facial expressions and slides served as the stimuli in Study 2. Despite our attempts to equate the source clarities o f the slides and facial expressions, there were s o m e differences both within the set o f slides or expressions o f the same type and across the two. To further equate the source clarities, we adjusted the ratings in Study 2 for these differences. For example, for ratings o f a pleasant facial expression, we c o m p u t e d the m e a n rating o f all pleasant facial expressions and slides. The adjustment consisted o f subtracting from the ratings in Study 2 for this pleasant facial expression the average rating for that expression from Study 1 minus the m e a n for all pleasant slides and facial expressions from Study 1. The same type o f adjustment was m a d e to the slide ratings. Study 2 The goal o f Study 2 was to e x a m i n e the hypotheses proposed in the introduction by using the stimuli whose source clarities were determined and matched in Study 1.
Method Design of the study. Independent variables were situation (situation information: public vs. private), order (the order of stimulus presentation: forward vs. backward), set (sets of pairings of facial expressions and a slide: three sets), and Rsex (receivers' sex) as between-subjects variables; therefore, there were 234 (2 × 2 × 3 × 2) between-subjects cells. The within-subjects variables were Ftype (type of facial expressions: pleasant, neutral, and unpleasant), Stype (types of slides: pleasant, neutral, and unpleasant), and Ssex (senders' sex). Dependent variables were the emotion scales used in both the pilot studies and Study 1. The within-subjects variables were combined by using three GrecoLatin squares: Ftype, Stype, and sender were counterbalanced. The six senders were divided randomly into three groups, each composed of one male and one female sender. Across all subjects, each sender's
1035
facial expression was paired with all three types of slides, and each sender was paired with all nine slides. Each subject viewed all nine cells of Ftype × Stype. Subjects. A total of 72 subjects (36 men and 36 women) enrolled in an introductory communication sciences course at the University of Connecticut pffrticipated in this study. Subjects were run in groups ranging from a minimum of 2 to a maximum of 12. Stimuli. The facial expressions and slides in which source elarities were examined and matched in Study 1 served as stimuli. There were 3 facial expressions (pleasant, neutral, and unpleasant) for each of six senders (3 men and 3 women; a total ofl 8 facial expressions) and three slides of three types (pleasant, neutral, and unpleasant; a total of 9 slides). Apparatus, A tape recorder was used to give instructions and to manipulate situational information. A 20-in. TV monitor, a videotape recorder, a slide projector, and a back-lighted screen (30 × 25 cm) were used to show facial expressions and slides. Booklets for the experimental session contained the instructions and the eight 7-point emotion scales for 2 practice and 18 experimental sequences. The eight 7-point scales were the same as those of Study 1. Six 7-point scales were used to examine the effects of the situational information (manipulation cheek): Public-Private, Formal-Informal, Natural-Artificial, Tense-Relaxed, Intentional-Spontaneous, and Nonexpressive-Expressive. These scales were selected to describe the general characteristics of situations in which facial expressions may occur. After the experimental session, a questionnaire was presented that contained a credibility cheek to examine whether subjects suspected that the pairing of facial expressions and a slide were not the actual ones. Finally, a Decoding Rule scale was given that included 36 sets of the following three 7-point scales: Do Not Express True Feelings At All-Express lhte Feelings Very
Much, Do Not Make Positive Expressions At All-Make PositiveExpressions VeryMuch, and Do Not Make Negative Expressions-Make Negative Expressions Very Much. These scales were rated to reply to the following description: "When a true feeling is (happy, sad, anger, fear, surprise, or disgus0 and a situation is (public, private, formal, informal, tense, or relaxed), people in general (followed by three scales): In this description, the emotions (e.g, happy) and situations (e.g, public) were systematically combined. Procedure. First, the following general instructions were provided by using both a written format and a tape recorder: This is a study of emotion. This study has two sessions: an experimental session and a questionnaire session. In the first session, you will see 18 pairs of a facial expression and a slide from a previous study. In the previous study, the facial expressions of 6 subjects (target persons) were recorded while they were watching various kinds of slides. The process of recording will be explained at the beginning of this experiment. There were nine slides, and each target person watched three of the nine. Therefore, some of the nine slides were watched by more than one target person. As you might expect, the reactions of the target persons are not exactly the same even to the same slides. In this experiment, we would like you to rate how the target persons felt about the slides along eight emotion scales provided. In the following session, you will get a questionnaire and answer several questions concerning emotion, which are not related to the first session. The manipulation oftbe situational information (public vs. private) followed. To do this, the setting and recording process of a previous study (i.e, slide viewing technique; Buck, 1978) was explained by using an explanation sheet, tape-recorded explanation, and three slides that illustrate the setting. In the public condition, subjects were informed that the facial expressions were recorded by a video camera visible to the sender while the sender was watching a slide in a laboratory with three experimenters (professors). In the private condition, subjects were informed that the facial expressions were recorded by a hidden
1036
M. NAKAMURA, R. BUCK, AND D. KENNY
video camera while a sender was watching slides alone in a room. Following the explanations, the first manipulation cheek was performed; that is, subjects were asked to rate the setting of the previous study along six situation scales (e.g, Public-Private). The main purpose of the manipulationcheck was to examine how subjects perceived the situation in which facial expressions were taken, that is, the effect of the situational information on the subjects' perception of the situations. The second purpose was to ensure that the subjects were explicitly aware of the public or private characteristics of the situation. In the experimental session, after two practice sequences and ensuring that subjects understood the procedure, 18 experimental sequences were presented. First, the videotaped facial expressions of a sender and a slide were simultaneously shown to subjects for about 5 s. After the pairing was turned off, subjects rated the emotional state of the sender along the eight emotion scales for about 30 s (the duration varied slightlyamong sequences and subjects).The next pairing was provided after all subjects had finished the rating, and this procedure was repeated until 18 experimental sequences were completed. Following the experiment, the second manipulationcheek (same as above) was performed. After the experimental session, a questionnaire session was conducted. At the beginningof the session, the credibility check was performed. Then subjects answered questions concerningdecoding rules. There was no time limitation in the questionnaire session. Analyses. Repeated measures analyses of variance (ANOVAs)were performed. Independent variables were Ftype, Stype, Ssex, situation, order, set, and Rsex. Dependent variables were the rating scores for the six situation scales (in the manipulation check), the eight emotion scales (in the experimental session), the credibility scale (in the credibility check), and the three expression scales (in the Decoding Rule scale). To measure the relative effects of the facial expressions versus the slides, a difference score was also used as a dependent variable. The difference score was calculated from emotion scale rating scores by the following equation: D = (PF - UF) - (PS - US), where D is a difference score, PF is a pleasant facial expression rating score, UF is an unpleasant facial expression rating score, PS is a pleasant slide rating score, and US is an unpleasantslide rating score. The PF, UF, PS, and US scores were obtained by averaging the rating scores for each category Source (facial expression or slide) x Pleasantness of Source (pleasant or unpleasant) for each receiver. A positive difference score for positive emotion scales and a negative difference score for negative emotion scales means that the effect of facial expressions is larger than that of slides.
Results Manipulation checks. Four-way (Situation x Order x Set x Rsex) repeated measures ANOVAs were performed with the mean scores of the first and second manipulation checks as dependent variables. A very large situation effect was obtained, F(1, 48) = 40.41, p < .001. That is, in the public condition the situation in which facial expressions were taken was rated to be more formal, public, artificial, tense, intentional, and nonexpressive than in the private condition. Therefore, the intended effect of the manipulation of situational information on the subjects' perception of the situation was obtained. No other effects were significant except for the interaction between order and set, F(2, 48) = 4.85, p < .01. This interaction was further examined by analyzing the data of the first and second manipulation checks separately. The Order x Set inter-
action was significant only in the first manipulation check. Because the first manipulation check was performed before the experimental session, the interaction must be attributed to chance differences between particular groups of subjects. Credibility check. The average rating score of credibility was 4.57 (7-point scale, m i n i m u m l to m a x i m u m 7, high score means high credibility). A four-way (Situation x Order × Set x Rsex) ANOVA was performed with the credibility rating score as a dependent variable. No between-subjects variable affected the credibility other than the significant interaction between set and Rsex, F(2, 48) = 5.18, p < .0 I. As set represents the three Greco-Latin squares, that is, the different pairings of a facial expression and a slide, the interaction means that the different pairings affect differently the credibility rating of male and female receivers. However, as will be shown later, this Set × Rsex interaction seems not to affect the qualitative aspect of the emotion scale ratings in the actual experiment. Accordingly, the entire set of data was analyzed in the following sections. Decoding Rule scale. Repeated measures ANOVAs were performed with three independent variables, emotion (six emotions), situation (public and private), and Rsex, for the three scale rating scores as dependent variables (express true feelings, express positive expressions, express negative expressions). All effects were highly significant (p < .00 l). The ratings of happy and surprise were significantly different from the negative emotions of sadness, anger, fear, and disgust. In the public situations, facial expressions were rated to be more inhibited than in the private situations, supporting the existence of a decoding rule to this effect. Emotion x Situation interactions occurred, so that in the private situations all emotions are judged to be expressed very much, whereas, in the public situations, negative emotions (especially sad) were judged to be less expressed than positive emotions. There was an interaction between Rsex and situation for the express true feeling rating scale. The effect of situation is larger for women than for men; thus, the decoding rule that one changes one's expression of true feelings in public versus private situations is apparently stronger for women. Contributions offacialexpressions andslides. Six-way (Situation x Rsex × Ftype X Stype x Set X Order) repeated measures ANOVAs were performed with the eight emotion scales as dependent variables. The main effects of Ftype and Stype were highly significant in all scales but one, the Stype main effect for the Angry scale. The main effects indicate that for the positive emotions scales, pleasant facial expressions and slides were rated higher than unpleasant ones. For negative emotion scales, including the Strong and Surprise scales, the relationship was the opposite. The Ftype X Stype interaction was significant for five scales--Pleasant, Strong, Happy, Sad, and Disgust. Table 2 presents the mean rating scores for the Pleasant and Sad scales on each type of facial expression and slide. On the Pleasant scale, facial expressions had a smaller effect when paired with unpleasant slides than with pleasant slides. The Happy scale showed a similar pattern. On the other hand, on the Sad scale (as well as disgust), the relationship was opposite: The effect of facial expressions was weakest for pleasant slides and strongest for unpleasant slides. The pattern for the Strong scale was quite different. When both slides and facial expressions were consistent and not neutral, the ratings were the highest.
1037
CONTRIBUTIONS OF BEHAVIOR AND INFORMATION Table 2
Table 4
Mean Rating Scores for the Pleasant and Sad Scales
Mean Rating Scores for Receiver Sex, Situation, and Slide Type on the Pleasant Scale
Facial expression Scale/slide Pleasant Pleasant Neutral Unpleasant Sad Pleasant Neutral Unpleasant
Pleasant
Neutral
Unpleasant
5.39 4.62 4.12
4.12 3.58 3.67
2.53 2.30 1.85
1.69 2.37 2.09
2.34 2.69 2.33
3.46 3.99 4.27
Type of slide Receiver sex/situation Male Public Private Female Public Private
Pleasant
Neutral
Unpleasant
3.90 3.86
3.59 3.30
3.50 3.18
4.31 3.98
3.51 3.61
2.77 3.40
Note. A rating score varies from i to 7.
Note. A rating score varies from I to 7.
In order to investigate directly the relative effects of facial expressions and slides, the difference scores were analyzed. As stated above, a positive difference score for positive emotion scales and a negative score for negative scales indicate that the effect of facial expressions is larger than the effect of slides. A series of t tests was performed, and the results are shown in Table 3. For all eight scales, mean difference scores were significantly different from zero. All of these effects indicate that facial expressions had larger effects than those of slides. For strong and surprise, negative information was seen as stronger and more surprising, which explains the negative signs in Table 3. Summarizing the results of this section, it appears that facial expressions generally tended to have a larger effect on emotion scale ratings than did slides. Effect of situational information. The six-way repeated measures ANOVA described above was examined to analyze the effect of situational information. A significant Stype x Situation interaction was found for the Angry and Disgust scales. On these two scales, slides had a larger effect in the public situation than in the private situation. Also significant was the Ftype x Situation interaction for the Strong scale. The effect of facial expressions was relatively larger in the private situation. It was also found that Stype and situation interacted with Rsex. This three-way interaction was significant for five
Table 3
Results oft Tests of Difference Scores Scale
M
t(48)
Pleasant Strong Happy Sad Angry Afraid Surprise Disgust
1.67 -0.80 1.48 - 1.45 -1.64 - 1.01 -2.01 -0.78
8.11"* -5.60** 8.10"* -7.35** -10.81" -5.29** - 12.26** -3.01"
Note. A positive difference score for positive emotion scales and a negative difference score for negative emotion scales means that the effect of facial expressions is larger than that of slides. *p