INDIVIDUAL DIFFERENCES IN THE NONVERBAL COMMUNICATION OF AFFECT: THE DIAGNOSTIC ANALYSIS OF NONVERBAL ACCURACY SCALE Stephen Nowicki Jr and Marshall P. Duke ABSTRACT: The Diagnostic Analysis of Nonverbal Accuracy (DANVA) was de- signed to measure individual differences in the accurate sending and receiving of nonverbal social information. The DANVA consists of four receptive and three expressive subtests that measure nonverbal processing accuracy in children from 6 to 10 years of age. Four propositions were offered to guide the gathering of construct validity data for the DANVA. In support of the propositions, researchersfound that DANVA accuracy scoresincreased with age, were internally consistent and reliable over time, and showed significant relationships with indices of personal and social adjustment and academic achievement but were not related to IQ. Evidence for construct validity was stronger for receptive, as compared to expressive, subtests. Future research should include additional populations of subjects and study of the impact of intensity of emotion being sent or received. The study of nonverbal behavior has a long history, but the beginning of modern scientific interest is often traced to Darwin (1872/1965), who focused on the possible modifying impact of body and facial expressive movements on verbal language. The past 30 years have seen a geometric increase in studies of nonverbal behavior, with more than two-thirds of the total references regarding nonverbal behavior having been published after 1970 (Ellyson & Dovidio, 1985). Only a limited number of these studies have been concerned with the development of standardized instruments to measure individual differences in the ability to correctly recognize and send information about affect through nonverbal means. Fewer still have focused on the measurement of nonverbal social communication in children. Address.correspondence to Stephen Nowicki Jr., Department of Psychology, Emory University, Atlanta, GA 30322. Journal of Nonverbal Behavior 18(I), Spring 1994 9 1994 Human Sciences Press, Inc. 9 10 JOURNAL OF NONVERBALBEHAVIOR The focus of the present paper is a research instrument designed to assess individual differences in children's ability to accurately send and receive emotional information through nonverbal means: the Diagnostic Analysis of Nonverbal Accuracy (DANVA). Individual differences in processing nonverbal information may be related to indicators of personal and social adjustment because skill in nonverbal communication is assumed to be a necessary ability for effective social interaction. Perhaps the most ambitious prior attempt to construct a comprehensive assessment instrument of nonverbal receptive abilities and the most frequently used is the Profile of Nonverbal Sensitivity (PONS) developed by Rosenthal and his colleagues (Rosenthal, J. Hall, DiMatteo, Rogers, & Archer, 1979). The original full PONS consists of a film with 220 two-second audio and/or visual segments that represent an emotional response. In each segment, a young adult woman portrays one of 20 different situations that are categorized into four different groupings created by crossing two affectire dimensions: positivity-negativity and dominance-submission. These categories are represented in three different video channels (full figure, neck to knee, and face) and two audio channels (content filtered and random spliced). The visual and auditory presentations can be given alone or in combination with one another to form 11 channel categories. Each 2-second segment is followed by a 5-second interval during which time the subject selects from a standard answer sheet one of two items that better describes the scene. For example, in one scene (#22) from the children's form the choice for the subject is (A) Judy is mad at her friend for being late, or (B) Judy is saying, "Thank you." There have been a variety of forms derived from the original full PONS including separate Face, Body, Tone, and Discrepancies measures (as cited in Blanck, Rosenthal, Snodgrass, DePaulo, & Zuckerman, 1981), the Test of Micromomentary Movement Effects (McLeod & Rosenthal, 1983), and, most pertinent to the present paper, the children's PONS (Rosenthal, J. Hall, Archer, DiMatteo, & Rogers, 1977). While the visual and auditory stimuli are the same, the major difference between the Children's PONS and the original full PONS is construction of response items that are more applicable and readable for younger age subjects. The PONS was originally constructed for use with older children and adults and it has most frequently been used with older subject populations (Rosenthal et al., 1979). Research using the full PONS or its derivations with younger children has been less frequent (e. g., C. Hall & Richmond, 1985; Jackson, Enright, & Murdock, 1987; Stone & la Greca, 1984). One difficulty in using the PONS with children is that it requires subjects not 11 STEPHEN NOWICKI JR., MARSHALL P. DUKE only to decode the valence of the affect being presented, but then to match that judgment, via a forced choice framework, to the correct social situation. Thus the PONS potentially confounds sensitivity to nonverbal social cues and knowledge of adult social situations (O'Sullivan, 1982). Subjects could conceivably be able to accurately identify nonverbal emotional cues and still score low on the PONS. Another difficulty using the PONS with children is that the audio portion uses content-filtering or random-splicing procedures which Axelrod (1982) suggests "resu}t in sound patterns not normally found in life situations" (p. 611). Finally, while a taxonomy of positive, negative, dominant, and submissive quadrants may have theoretical benefits, it does not help researchers who may be more interested in determining subjects' ability to decode particular basic emotions such as happiness, sadness, anger, and fear. Other attempts to measure individual differences in the ability of children to process nonverbal affect information also seem to fall short in one way or another. The Social Skills Inventory (Riggio, 1986), like the PONS, is designed primarily for use with adults. Further, it is not actually a measure of nonverbal processing ability, but a self-report instrument of "perceived" nonverbal performance. While the Social Interpretations Test (SIT, Akert & Archer, 1977) uses videotaped interpersonal situations as test stimuli, it is also appropriate primarily for adults, focuses only on receptive skills, and does not allow for the measurement of the separate basic emotions (Maheady & Sainato, 1986). Videotape sequences are used to measure expressive communication skills in the Communication and Reception of Affect Test (CARAT, Buck, 1980) and there are forms for children (CARAT-C) as well as adults. But the CARAT does not yield scores for separate channels of nonverbal communication or provide age norms. Other promising tests have appeared, but seem also to have fallen short of obtaining the necessary construct validity data to support their continued use (e.g., The Test of Emotional Styles (TES), Allen & Hamsher, 1974; Understanding Our Feelings, Elmore, 1985; The Feldstein Affect Judgment Test, Wolitzky, 1973). Based on this brief review of existing measures, there appears to be a need for an instrument that measures, especially in children, individual differences in the ability to receive and to send affective information through nonverbal means. The present paper describes one possible instrument, the DANVA. The DANVA measures individual differences in four types of nonverbal receiving ability (facial expressions, postures, gestures, and tones of voice) and three types of nonverbal sending accuracy (facial 12 JOURNAL OF NONVERBAL BEHAVIOR expressions, gestures, and tones of voice). The aim in developing the DANVA was to help in identifying children who may be having difficulties in processing nonverbal information about affect. Nowicki and Duke (1992a) have termed such a nonverbal social processing deficit dyssemia, meaning a difficulty (dys) in using nonverbal signs or signals (semes). DANVA subtests sample emotions, such as happiness, sadness, anger, and fear, that appear to be necessary for children's everyday interpersonal functioning. Rather than being chosen on the basis of any particular theory of emotional development or on "anatomically objective" criteria such as position of facial features or specific facial muscle tensions (e.g., Ekman & Friesen, 1975; Izard, 1971), DANVA test items were selected primarily on empirical-normative grounds. That is, if a designated number of members of a norm group judged a nonverbal behavior as communicating a particular type of affect it was included in preliminary forms of the test (Cook, cited in Rosenthal et al., 1979). The basic assumption was that the more frequently children receive or send nonverbal emotional information differently from most of their peers, the more likely it is that they will be at an interpersonal disadvantage. While the implications of this assumption are most obviously seen in the children who read or send nonverbal information more poorly than their peers, it is also true that subjects who are much better than their peers at processing such information may also suffer rejection (e.g., they "read too much" into a situation). This latter point has already been made by Rosenthal and his colleagues, who argued that social relationships were troubled when people were especially skillful at decoding nonverbal messages that they were not intended to receive (Rosenthal & DePaulo, 1979; Snodgrass & Rosenthal, 1985). More will be said about the criteria established to select items for the DANVA in the measures section. The objectives of the present paper are (a) to describe the development of the receptive and expressive subtests that make up the DANVA and to provide their psychometric characteristics; and (b) to present and to evaluate construct validity evidence available for the DANVA subtests. The following propositions guided the gathering of construct validity information. Proposition I: DANVA accuracy scores will increase with age. Because children's perceptual and informational processing skills improve with age it was expected they would become more accurate senders and receivers of nonverbal information the older they become (Zuckerman, Blanck, DePaulo, & Rosenthal, 1980). Proposition 2: DANVA subtests will show acceptable estimates of in- 13 STEPHEN NOWICKI JR., MARSHALL P. DUKE ternal consistency (coefficient alpha of at least .70) and test-retest reliability (at least r = .70 for four weeks). Because items are not arranged according to difficulty, it was expected that estimates of internal consistency would be underestimates of the true internal consistency (Rotter, 1990). It was expected that estimates of reliability will increase with the age of children tested. Proposition 3: Individual differences in DANVA accuracy scores will be related to indices of personal and social adjustment. Because the sending and receiving of ~ffective information through nonverbal means is assumed to be necessary for effective interpersonal interactions, children who are less accurate than their peers would be expected to experience personal and interpersonal difficulties. This proposition was evaluated in three ways. Individual differences in DANVA accuracy scores were predicted to be related: (3a) to self-report inventory scores that indicate personal functioning such as locus of control of reinforcement and self-esteem; (3b) to indicators of interpersonal functioning such as peer-rated popularity and unpopularity and teacher ratings of social functioning; and (3c) to indicators of interpersonal maladjustment such as being classified as behavior disordered or emotionally disturbed. Proposition 4: Individual differences in DANVA accuracy scores will be associated with academic achievement as measured by standardized test scores but not with IQ. It was assumed that difficulties in processing nonverbal information would create anxiety that would interfere with the processing of information necessary for achievement in an academic setting. In support of this proposition, Halberstadt and J. Hall (1980) reviewed 22 studies and found a small but positive correlation between cognitive ability (as measured by standard tests and school performance) and the ability to identify nonverbal expressions of emotion. However, only four of these studies were with subjects younger than ten years of age and in none of them did the authors attempt to separate the relation of nonverbal processing ability with IQ from that with achievement. Method Subjects The subjects for the evaluation of the four propositions came from the studies described in Table 1. Most of the subjects were between the ages of 6 and 10, although in a few studies older subjects were included. 14 JOURNAL OF NONVERBALBEHAVIOR TABLE 1 Construct Validation Studies for the Diagnostic Analysis of Nonverbal Accuracy (DANVA) Author Subjects Danva subtests Brumley (I 993) 181 3rd and 6th grade students All receptive subtests Cooley & Treimer (1991) 16 emotionally disturbed subjects between 11 and 15 years of age All receptive subtests Feig (1993) 152 third grade students All receptive subtests All expressive subtests given to 38 children Hallin (1991) 30 emotionally disturbed subjects between 8 and 14 years of age and 30 nondisturbed subjects matched for age, race, and IQ All receptive subtests Hill (1992) 62 third grade students Receptive faces Nowicki (1990) Nowicki (1992) 34 third grade students 42 third grade students All receptive subtests All receptive subtests Nowicki & Carton 221 first through fifth grade students (1993) Nowicki & DiGirolamo (1989) 20 emotionally disturbed subjects between 9-I I years of age and 20 nondisturbed subjects matched for age, sex, and race FACES 2:48 adult and 48 child facial expressions Short forms of receptive faces and listening subtests 15 STEPHEN NOWICKI JR., MARSHALL P. DUKE TABLE 1 Author (continued) Subjects Danva subtests Nowicki & Duke (1989) 1001 first through fifth grade students All receptive subtests All expressive subtests (n = 189) Nowicki & Duke (1992b) 455 first through fifth grade students Receptive subtests faces and listening Rowe (1992) 20 girls and 36 boys between ages 8 to 11 All receptive subtests Zanier (1991) 30 six-year-old boys 30 ten-year-old boys All receptive subtests Zenker (1992) 39 subjects between 8 and 16 years of age who were in residential treatment All receptive subtests Measures Nonverbal social processing abilities were measured by the Diagnostic Analysis of Nonverbal Accuracy scales (DANVA). The DANVA consists of four subtests that measure individual differences in receptive abilities in facial expressions, postures, gestures, and tones of voice and three that test for individual differences in expressive accuracy in facial expressions, gestures, and tones of voice. Although field work continues on it, the expressive postures subtest did not receive sufficient psychometric support to include it in the DANVA at the present time. While there are some differences in how specific items were selected for specific subtests, some common procedures were used across subtests. First, each subtest was constructed independently. There was no attempt to select items from any one subtest so that they would or would not intercorrelate with items from another subtest. This was done because there has been little theoretical or empirical agreement on the underlying relationship of nonverbal social processing skills to one another. Further, while it seems sensible that some nonverbal processing skills ought to cluster together, especially within the expressive or receptive areas, empirical re- 16 JOURNAL OF NONVERBALBEHAVIOR suits suggest they may be independent (e.g., Feldman & Thayer, 1980). The same seems to be true of the relation between expressive and receptive abilities. For example, DePaulo and Rosenthal (1979) reported that a review of 19 studies of the relation between sending and receiving accuracy found a median r of .16. Second, items for each of the subtests were selected primarily on the basis of percentage of judges agreeing on the identification of a particular emotion. Of the five general ways to establish a criterion of accuracy for nonverbal social processing described by Cook (cited in Rosenthal et al., 1979, p. 19), this was the one that most closely reflected the ecological situation subjects faced in their daily interactions. Third, judges were obtained from across a wide age range to assure that the items measured consistently across age. If an item was identified by a certain percentage of subjects at one age, we wanted to check to make sure that the same or greater percentage of older subjects would also identify that item similarly. In choosing items for particular age groups, the major focus was on that age's reference group and the younger and older age reference groups were used as reliability checks. While the DANVA was primarily constructed for use with the 6- to 10-year-old age group, it probably could be used with older subjects to identify those experiencing difficulties in processing nonverbal information by comparing them to the 10-year-olds' performance norms. However, because of its low ceiling, the DANVA would not be a good test to use in discriminating among older subjects who were average or better in processing nonverbal social information. Fourth, a relatively high percentage of interjudge agreement was used for item selection because our major interest was to identify those children who were having difficulty with processing the basic elements of nonver~ bal social information. If greater difficulty is desired, items can be shown with a briefer exposure time as has been done with the PONS (McLeod & Rosenthal, 1983). Fifth, four basic core emotionswere chosen for identification and for sending: happiness, sadness, anger, and fear. They were chosen because they are the most frequently encountered in everyday interactions and usually are learned by the time children are 10 years of age. Without the ability to identify and send these emotions nonverbally, children would be at a disadvantage interpersonally. Of course, much more is required of children's nonverbal communication skills than accurately identifying and sending the'four core emotions nonverbally. But without this knowledge, recognition of the social situation and accurately processing more complex emotions like guilt and shame would be difficult if not impossible. 17 STEPHEN NOWlCKI JR., MARSHALL P. DUKE Sixth, the DANVA subtest items primarily rely on posed stimuli. However, encoding of posed and spontaneous stimuli are significantly related (median r over six studies = .43, Halberstadt, 1986) and the subjects used to generate the nonverbal responses were aided by stories and descriptions of events consistent with being happy, sad, angry, or fearful. In many instances, subjects were photographed as they moved from one "pose" to another or were engaged in conversation with the experimenter, and some of these relatively more unposed stimuli passed through the judges' item selection procedure into the final form of the subtest. Responses for facial expressions, postures, and gestures were not only posed but static. However, since most spontaneous nonverbal cues are relatively short-lived, their meaning must be grasped in a moment. With these common requirements in mind, the seven DANVA subtests are described next. The four receptive subtests are presented first, followed by the three expressive ones. Receptive Facial Expressions Subtest. This scale consists of 20 slides of children's and 20 slides of adults' facial expressions. Children's facial expressions were sampled from the Children's Affect Test (Kay, 1984; McCauley, Kay, Ito, & Treder, 1987) and adults' facial expressions were taken from the Brief Affect Recognition Task (BART, Ekman & Friesen, 1974). Facial expressions were viewed by two samples of raters: college students (n = 100) and second grade students (n = 100). Raters viewed each slide and indicated if they thought the expressions were happy, sad, angry, fearful, or other. They also rated the certainty of their judgment on a 5-point scale with higher scores related to greater certainty. Slides with ratings of 4 or more were selected for further use in the DANVA. The BART slides that constituted the final form of the adult facial expressions were in order of presentation: 21, 84, 23, 17, 44, 28, 51, 22, 68, 18, 47, 10, 75, 29, 24, 92, 50, 25, 2, 42. The facial expression subtest consists of four slides (two males and two females) for each of four emotions (happy, sad, angry, and fearful) as portrayed by children and adults. Four children's and four adults' slides were neutral. Scores could range from 0 to 16 for both children's and adults' facial expressions scales and from 0 to 32 for the total facial expressions subtest. Neutral expressions were not scored. Facial expressions were shown for one second with children having as much time as they needed for selecting one of four emotions: happy, sad, angry, or'fearful. Subjects were cued by the administrator saying "ready" before showing the next slide. Subjects usually took five seconds or less to respond. Because it is important for sight lines to be free when looking at 18 JOURNAL OF NONVERBAL BEHAVIOR facial expressions, it is suggested that the facial expression subtest be given individually or in small groups where the angle of viewing is not severe. Receptive Postures Subtest. This scale consists of 12 slides of a woman dressed in a solid dark colored sweatsuit who kept her face hidden from view. She showed various postures that reflected happy, sad, angry, or fearful emotions. For example, a posture with both arms raised high and the knees flexed was used to indicate "happy." The decision to use the slides of this particular woman model was made because she appeared to be the more reliably "read" by judges than the other models. The slides resulting from these sessions were given to four groups of 20 subjects (college, tenth grade, fifth grade, and preschool students). Only slides that obtained at least 80% agreement from all four groups were included for further consideration, with this exception: For preschoolers only; agreement was 75% for two of the postures reflecting fear, while the rest of the postures attained at least 80% agreement. (Preschool raters were worked with individually and their responses were recorded by an examiner rather than being self-reported on a rating form.) Subjects were shown the posture for one second and given as much time as they needed to check one of four possible emotions listed next to the number of the slide. Responses were usually made in five seconds or less. The administrator cued the subject for the next slide by saying "ready." Scores can range from 0 to 12. Receptive Gestures 5ubtest. This scale consists of 12 slides of two models, one a male child eight years of age (appears on eight slides) and one a young male adult (appears on four slides), portraying gestures that communicate happy, sad, angry, and fearful emotions. Male and female models were photographed portraying selected gestures. Slides of these gestures were viewed by the same subjects who rated the posture stimuli. They were asked to identify what emotion they believed was being communicated by the gesture they viewed. For example, a slide showing two closed fists with one pulled back was used to signify anger. The slides selected for the present receptive gesture scale received at least 80% agreement from all four groups (ranging from 80% in the preschool group to 100% in the college student group). The gesture slides are presented for one second. The subjects have as long as they need to select one of four emotions, happy, sad, angry, or fearful. The~ usually responded in less than five seconds. The administrator cued the next slide by saying "ready." Scores can range from 0 to 12. 19 STEPHEN NOWICKI JR., MARSHALL P. DUKE Receptive Paralanguage Subtest. This scale consists of 16 trials during which subjects listen to a person say the following sentence, "1 am going out of the room now and I will be back later," to reflect happy, sad, angry, or fearful feelings. Although some researchers have used sound filters (e.g, Rosenthal et al., 1979) to remove the potential biasing impact of words, others have used the more naturalistic and ecologically valid procedure of having subjects recite a neutral sentence in a way to give it emotional meaning (Davitz & Davitz, 1959; Maitland, 1977). Two boys and two girls between the ages of 10 and 12 were trained by the experimenters to say the sentence to communicate happy, sad, angry, and fearful emotions. Their attempts were then recorded and played for the four different age groups used in the development of the receptive posture and gesture scales. One girl's recordings were identified at least 80% of the time. Because the judges could not identify the sex of the sender, her vocalizations were chosen for the final form of this subtest. There are 16 trials with each emotion being expressed four times. Subjects are cued by the administrator saying "ready" before the next sentence is played. Scores can range from 0 to 16. Expressive Facial Expressions Subtest. This test consists of eight trials. Subjects were asked to sit in a comfortable chair 10 feet from a video camera. They were told that various situations would be described to them and they were to make a facial expression that was consistent with the situation they heard. There were two situational descriptions for each of the four emotions of happy, sad, angry, and fearful. An example of a situational description is, "You have received a gift for your birthday that you have always wanted. You feel happy." The videotaped facial expressions were then rated for their accuracy in portraying the requested emotion. Although any rating scale can be used to rate the emotions shown in the videotaped facial expressions, the one used in the DANVA was based on the features associated with facial expression emotion described by Ekman and Friesen (1975). Generally, subjects were judged as either accurate or not for the requested emotion. Two undergraduate men and women applied this rating scheme on practice items until it produced 90% agreement between any two of the four raters for each of the four emotions. Scores can range from zero to eight. Expressive Gestures Subtest. This test consists of 12 trials. Subjects were asked to sit in a comfortable chair 10 feet from a video camera. They were told that the examiner was going to ask them to describe something 20 JOURNAL OF NONVERBALBEHAVIOR by using only his or her arms and hands. For example, happiness could be shown by a "thumbs up" or "OK" signs. Preliminary testing with pilot subjects suggested that it would be difficult to repeatedly ask for different gestures to reflect happy, sad, angry, and fearful emotions. Therefore, frequently used information for facilitating interactions was chosen such as "inviting someone to play," "asking someone to stop," "liking someone," or "wanting to be left alone." The gestures used by the subjects were rated by four raters who were trained to code accuracy of gesture communication to the point of 90% agreement between any two of them. Scores could range from 0 to 12. Expressive Paralanguage Subtest. This test consists of eight trials. Subjects were asked to sit in a comfortable chair close to an audio recorder's external microphone (a highly sensitive "Sound Grabber"). They were handed a cardboard pad of paper upon which the following sentence was printed in 3/4-inch block letters: "1 am going to get my bike now and go for a ride." Subjects were given as much time as they needed to practice reading the sentence aloud. When subjects were ready, experimenters read a description of a situation that was associated with a particular affect. For example, "You have found out that you have won a nice prize in a contest. You feel happy. Say the sentence in such a way that it sounds like you are happy." There were two trials for each of the emotions of happiness, sadness, anger, and fear. Each of the eight trials was rated for subjects' accuracy in portraying the desired affect. For example, to receive a rating of "1" for "happy," the subjects had to produce a high, rapid voice that fluctuated and communicated positive affect. Four raters were trained to the point where any one set of ratings were in at least 90% agreement with the ratings of any one of the other three judges. Scores could range from zero to eight. Results and Discussion Fourteen studies were used to establish preliminary evidence of construct validity for the DANVA. They are briefly described in Table 1 and will be referred to with regards to propositions one through four. Propositiom 1 Nowicki and Duke (1989) administered the DANVA to 1001 children from the first through fifth grades. Subjects' ethnicity was comparable to 21 STEPHEN NOWICKI JR., MARSHALL P. DUKE community rates. They were from public and Roman Catholic schools that served lower-middle and middle-middle class families as indicated on school records by the highest occupational status presented by either parent. Eighty-six percent of the children's parents gave permission for their children to participate in the study. Based on data from class records, no social class or IQ differences were found to exist between the white and nonwhite subjects. Subjects were tested in small groups within each grade. Table 2 presents data pertinent to the assertion of Proposition I that DANVA accuracy sc~es should increase with age. To evaluate this proposition, the 1001 subjects from the first through fifth grades completed the DANVA receptive tests and 189 of them also completed the DANVA expressive tests. Since factorial analyses of variance revealed that there were no sex differences, DANVA accuracy scores were collapsed across sex of subject. Analyses were computed separately for the receptive and the expressive subtests. Linear trend analyses for each of the receptive subtests indicated that mean accuracy increased with age (p < .05; F ranged from 4.80 for gestures to 10.66 for facial expressions; r ranged from .26 for gestures to .44 for facial expressions). Linear trend analyses computed for each of TABLE 2 Accuracy Score Means and Standard Deviations for Diagnostic Analysis of Nonverbal Accuracy (DANVA) Subtestsin Grades One through Five Grade Receptivesubtests~ Faces Gestures Postures Paralanguage One (n --- 199) 27.3 Two (n = 206) Three (n = 176) Four (n = 208) Five (n = 219) Possible range (2.6) 27.9 (2.6) 28.6 (2.5) 29.2 (2.2) 29.2 (1.9) 0-32 7.3 (2.4) 7.3 (2.5) .7.9 (2.1) 8.4 (1.7) 8.5 (1.7) 6.6 (2.3) 7.2 (2.2) 7.5 (2.3) 8.0 (1.8) 7.9 (1.9) 11.8 (3.2) 11.4 (3.4) 13.9 (1.9) 14.9 (2.7) 14.8 (2.2) 0-12 0-12 0-16 Expressive subtestsb One (n=35) Two (n =40) Three (n=41) Faces Gestures Paralanguage 4.2 (1.8) 7.9 (3.1) 5.3 (2.0) 5.5 (2.0) 8.1 (2.8) 5.3 (2.0) 5.7 (2.0) 6.5 (1.7) 7.2 (1.0) 8.8 (2.5) 10.4 (2.4) 11.2 (2.1) 5.8 (1.6) 6.0 (1.4) 7.1 (.9) Note. Standard deviations are shown in parentheses. an = 1001 Un= 184 Four (n=31) Five (n--- 37) Range 0-8 0-12 0-8 22 JOURNAL OF NONVERBAL BEHAVIOR the expressive subtests also indicated that the mean accuracy increased with grade level (p <: .05; F ranged from 4.33 for gestures to 7.30 for facial expressions; r ranged from .17 for gestures to .45 for facial expressions). However, inspection of the means for both the receptive and expressive subtests showed that accuracy did not increase at the same rate from first to fifth grade for any of them. Perhaps the relatively small number of subjects per grade level played some part in the somewhat inconsistent rise in accuracy with age. Therefore, while these data can be seen as supporting proposition I, they also can be taken to suggest that grade norms ought to be used with some caution until additional subjects are tested.' Proposition 2 Table 3 presents the overall internal consistency and test-retest reliability estimates for the DANVA subtests. Cronbach's coefficient alphas for the receptive tests (n = 1001) ranged from .77 to .88, and for the expressive tests (n = 189) from .68 to .82. Coefficient alphas computed for each grade ranged from lows of .67 for expressive paralanguage in the first grade to .75 or higher at all the other grade levels. Test-retest reliabilities were computed after four weeks for 20 students at each grade level. The overall test-retest reliabilities ranged from a low of r = .70 for expressive paralanguage to r = .86 for receptive gestures. The DANVA subtests ap- TABLE 3 Internal Consistency and Test-Retest Reliabilities for the Diagnostic Analysis of Nonyerbal Accuracy (DANVA) Subtests DANVA subtests Receptivea Faces Gestures Postures Paralanguage ExpressiveLb Faces Gestures Paralanguage ~n=1001 bn = 189 Coefficient alpha Test-retestr (4 weeks) .88 .79 .82 .77 ,84 .86 ,77 .74 .82 .78 .68 .80 .72 .70 23 STEPHEN NOWICKI JR., MARSHALL P. DUKE peared to possess preliminary evidence of internal consistency and testretest reliability required by proposition 2. In support of proposition 2, Zanier (1991) also found alpha coefficients that were comparable to those reported in the original Nowicki and Duke (1989) DANVA sample. For six-year-old boys the alphas for the DANVA receptive subtests ranged from .74 to .82 and for I 0-year-old boys they ranged from .77 to .88. Now that Propositions I and 2 involving the psychometric integrity of the DANVA subtests ~are supported, the following data focus on the relationship between DANVA subtest scores and a variety of relevant construct validity criteria. The criteria relate to important aspects of the personal and social lives of children and reflect how individual differences in nonverbal social processing may impact on children's adjustment. Proposition 3a There was support for the proposed relationship between accuracy scores on the DANVA receptive and expressive subtests in Nowicki and Duke's original sample and locus of control scores as measured by the Children's Nowicki-Strickland Internal-External scale (CNSIE, Nowicki & Strickland, 1973). The correlations between locus of control (lower scores indicating internality and perceived contingency between actions and outcomes and higher ones, externality) and DANVA receptive subtests scores were consistently significant, but generally low, ranging from r(989) = -.14, p ( .05, for receptive posture to r(989) = -.28, p ( .05, for receptive facial expressions. Of the three DANVA expressive subtests, only the correlation between locus of control and expressive paralanguage was significant, r(186) = -.32, p < .05. Additional support for proposition 3a was obtained by Nowicki and Duke (1992b) who used subjects primarily from the third through fifth grades and found significant correlations between locus of control and receptive facial expressions, r(453) = -.34, p ~ .05, and receptive paralanguage, r(453) = -.32, p < .05. Finally, Hill (1992) in a sample of third-grade students found a significant correlation for CNSIE and Harter (1985) Self-Perception Profile for Children scores (lower scores indicating more internal attributions for outcomes and higher scores more external ones) and DANVA receptive facial expressive performance, r(60) = -.34, p < .05, and r(60) = -.44, p < .05, respectively. Besides being related to internal control expectancies, greater DANVA accuracy also was related to higher self-esteem. In one study (Nowicki, 1992), Coopersmith (1981) self-esteem scores were found to be signifi- 24 JOURNAL OF NONVERBALBEHAVIOR cantly correlated with higher accuracy on DANVA receptive subtests in a sample of third-grade children: for facial expression, r(40) = .55, p < .05; for posture, r(40) = .41, p < .05; for gestures, r(40) = .38, p < .05; and for paralanguage, r(40) = .62, p < .05. In another study (Cooley & Treimer, 1991), a sample of children between the ages of 11 and 15 who attended a center-based school program for the severely emotionally disturbed showed significant correlations between Coopersmith self-esteem scores and receptive facial expression scores, r(14) = .52, p < .05; receptive paralanguage, r(14) = .61, p < .05; and nonsignificant trends for receptive postures, r(14) = .39, p > .05, and receptive gestures, r(14) = .42, p > .05. In a third and final study regarding self-esteem (Feig, 1993), there was a significant correlation between Piers-Harris self-concept (Piers, 1984) and receptive DANVA facial expression scores, r(150) = .28, p < .05, in third-grade children. These data provide preliminary support for Proposition 3a, especially for the DANVA receptive subtests. DANVA expressive subtests were less often given, but when they were administered they also were less often found to be related to personality variables. In contrast, individual differences in DANVA receptive accuracy scores were related in predicted ways to locus of control of reinforcement and self-esteem. Proposition 3b To evaluate the proposition that DANVA accuracy scores would be related to sociometric status, children in the original Nowicki and Duke (1989) sample (n = 1001) were given sheets of paper with the names of all the children in their class. They were asked to put a circle around the names of three children they liked the best and a line through the names of three children they liked the least. Because of potential problems in reading, children in grades one and two were tested individually so that those with reading problems could obtain help with names if needed. Like and dislike scores were transformed into z-scores to control for differences in class size. The like and dislike scores were independent of one another (median r = .09) and constituted different aspects of children's social status. While some researchers have separated children into popular, average, neglected, rejected, and controversial subgroups, we were more interested in examining the relation of nonverbal processing ability to social status throughout the range of scores and therefore computed correlations between like and dislike z-scores for each of the receptive and expressive subtests. While the correlations between DANVA receptive subtest scores and 25 STEPHEN NOWICKI JR., MARSHALL P. DUKE sociometric like and dislike z-scores tended to be low, 11 out of a possible 16 were significant and in the predicted direction with less accuracy related to more negative peer ratings and greater accuracy related to higher positive peer ratings. In contrast, only two of a possible 12 correlations were in the predicted direction for DANVA expressive subtest scores and sociometric indicators. More specifically, for boys (n = 485), significant correlations for like ratings at the p < .05 level were found with receptive facial expression, r = .21; receptive gestures, r = .18; receptive paralanguage, r = .24; al~d expressive gestures, r = .27. For girls (n = 526), significant correlations were found for receptive facial expressions, r = .26; receptive postures, r = .12; receptive paralanguage, r = .31; and expressive paralanguage, r = .27. In terms of dislike scores, for boys greater accuracy was correlated negatively with scores for receptive facial expressions, r = - . 2 0 ; receptive postures, r = - . 1 9 ; and receptive paralanguage, r = - . 2 2 ; and for girls receptive facial expressions, r = - . 1 2 , and receptive postures, r = - . 1 1 . Additional support for Proposition 3b was found in four studies. Nowicki and Duke {1992b) found that third- through fifth-grade children (n = 455} who received fewer like votes had lower accuracy scores on the receptive facial expressions subtest, r = .27, p < .05, and on the receptive paralanguage subtest, r = .30, p < .05. Further, more dislike votes were related to lower accuracy scores on the receptive facial expressions subtest, r = - . 3 0 , p < .05, and on the receptive paralanguage subtest, r = - . 2 7 , p < .05. Rowe (1992} administered the DANVA receptive subtests and a peer role-play nomination procedure to assign sociometric status to children between the ages of 8 and 11 who were attending a summer camp. She found the following significant (p < .05) correlations for girls: Receptive adult facial expressions with votes received for "negative feeling," r(18) = - . 4 2 , and for "sad feelings," r(18} = - . 6 2 ; receptive child facial expressions and votes received for "negative feelings," r(20} = - . 4 8 , and for "being left out," r(18) = - . 4 8 . There were three significant correlations for boys: receptive total facial expressions and votes received for "being bossy," r(34) = - . 3 9 , "being picked for games," r(34) = .29, and "negative feelings," r(34} = - . 3 7 . Feig (1993) found significant correlations between like votes received in a third-grade class and DANVA receptive subtest scores on total facial expressions and paralanguage, r(150} = .38, p < .05, and r(150} = .43, p < .05, respectively. In addition, she found a significant correlation between DANVA expressive paralanguage subtest scores and like votes, r(36} = .48, p < .05. 26 JOURNAL OF NONVERBAL BEHAVIOR In another study of the relationship between DANVA scores and peer popularity ratings, Brumley (1993) tested 181 third and sixth graders. Using a total composite score derived from a combination of DANVA subtest scores, she found a correlation of .77 between popularity ratings and nonverbal language efficacy. Brumley also developed a teacher-usable, 75-item checklist form of the DANVA, the DANVA-QSP (Quick Screening Profile), and found that it, too, correlated strongly with peer popularity ratings among her subjects. Besides peer-rated interpersonal status, teacher ratings of the social behavior of children also were found to be related to individual differences in DANVA subtest accuracy scores. To rate social behavior, teachers completed the Social Judgment Scale (SJS, Nowicki, 1987). The SJS consists of seven characteristics (e.g., Child doesn't understand rules for games) that were derived from descriptions offered by Johnson and Myklebust (1967) of children with problems in processing nonverbal social cues. Each characteristic was rated on a scale that ranged from 1 (not very characteristic of the child) to 5 (very characteristic of the child). A sample of fourth-grade children was given the DANVA receptive subtests and was rated by their teacher on the SJS. Significant correlations were found between SIS scores and the following receptive subtests: total facial expressions, r(41) = -.48, p < .05, paralanguage, r(41) = -.39, p < .05, and postures, r(41) = -.42, p < .05. These results constitute preliminary support for Proposition 3b. Individual differences in DANVA accuracy scores were related to indices of interpersonal and teacher ratings of peer functioning. However, most of the reported research involved receptive rather than expressive nonverbal processing abilities. Proposition 3c Support for Proposition 3c, that individual differences in DANVA subtest accuracy scores are related to indicators of interpersonal adjustment, came from several studies. Hallin (1991) tested children between the ages of 8 and 13 (n = 14) and adolescents between the ages of 14 and 16 (n = 16) who were in residential treatment for psychological problems. When children and adolescents who were emotionally disturbed were combined into one group it was found that they had significantly (p < .05) lower scores than _did nonemotionally disturbed subjects matched with them for age, sex, race, and IQ on the receptive "adult" facial expressions subtest (M = 13.1 and M = 15.0, respectively) and receptive paralanguage subtest (M = 12.9 and M = 14.5, respectively). Further, adolescents who 27 STEPHEN NOWICKI JR., MARSHALL P. DUKE were emotionally disturbed had significantly (p < .05) lower accuracy scores than their matched controls who were not emotionally disturbed on the receptive child facial expressions subtest (M = 10.7 and M = 15.5, respectively), the receptive postures subtest (M = 8.1 and M = 11.7, respectively), and receptive gestures subtest (M = 8.3 and M = 11.3, respectively). These results are consistent with the prediction that children and adolescents who have psychological problems are less accurate in decoding nonverbal social information than their nondisturbed peers. Consistent with He above findings of Hallin (1991), Cooley and Treimer (1991) reported that subjects between the ages of 11 and 15 who were in a class for severely emotionally disturbed students were significantly less accurate than nondisturbed peers on the receptive facial expression and receptive paralanguage subtests. These investigators did not report subtest means, however. In another study, Zenker (1992) found results consistent with the hypothesis that lower accuracy of nonverbal social processing was related to greater number of indicators of maladjustment. The subjects (n = 39) ranged in age from 8 to 16 years of age and were in residential psychiatric treatment. Boys and girls did not differ in IQ or age. Based on case file information, subjects were rated for indicators of conduct disorder or emotional disturbance according to the Georgia Psychoeducational Network Rating Scale (Hinley et al., 1990). Zenker found that a higher number of indicators of conduct disorder in girls was significantly correlated with lower accuracy scores on the receptive child facial expressions subtest, r(12) = -.74, p < .01, the combined receptive adult and child facial expressions subtest, r(12) = -.53, p < .05, and the receptive postures subtest, r(12) = -.63, p < .01. In this population, indicators of emotional disturbance were not related to individual differences in DANVA receptive performance. While Zenker (1992) found individual differences in DANVA accuracy scores related to the presence of indicators of conduct disorder, Nowicki and DiGirolamo (1989) tested 20 children between the ages of 9 and 11 years of age who were in classes for the emotionally disturbed and 20 nonemotionally disturbed subjects who were matched for age, sex, and race on the receptive paralanguage test. They found that emotionally disturbed children were significantly less accurate in identifying affect communicated paralinguistically, t(38) = 2.27, p < .05. Finally, Zenker (1992) asked teachers to rate children (aged 8 to 16 years of age) who were in a residential treatment center on the Social Judgment Scale. For girls, but not boys, lower SJS ratings were related to less accurate total DANVA receptive scores, r(12) = .66, p < .05. 28 JOURNAL OF NONVERBALBEHAVIOR The results of these studies suggest an association between DANVA receptive performance scores and indicators of psychological maladjustment. Receptive facial expressions and paralanguage were most often found to be involved in significant predicted relations. In contrast to the more extensive research reported for the DANVA receptive subtests, there has been relatively little investigation of the relationship between individual differences in expressive DANVA subtest accuracy and personal and interpersonal adjustment. Proposition 4 To examine the proposition that individual differences in DANVA accuracy scores were related to academic achievement as reflected by standardized test scores, data were obtained from the Nowicki and Duke (1989) sample of children from grades one and five. Table 4 presents the correlations between various DANVA receptive (n = 968) and expressive (n = 176) subtest scores and the Comprehensive Tests of Basic Skills scores (CTBS, McGraw-Hill, 1983). The CTBS is a set of standardized TABLE 4 Correlations Between Nonverbal Accuracy and Academic Achievement Word Reading Math Math Vocabu- recog- compre- Math count- compreDANVAsubtests lary nition hension concepts ing hension Spelling Receptivea Faces Gestures Postures Paralanguage .31 .30 .30 .45 .29 .27 .27 .36 .31 .32 .32 .45 .28 .29 .29 .40 .30 .26 .26 .41 .31 .30 .20 .41 .28 .22 .22 .36 Expressiveb Faces Gestures Paralanguage .15 .11 .26 .I I .07 .17 .23 .08. .16 .I I .16 .12 .23 .06 .25 .09 .14 .23 -.05 .10 .17 Note. Academic achievementsubtestsare from the ComprehensiveTestsof BasicSkills. Significant(p <.05) receptivecorrelationsare. | 2 and above. Significant(p < .05) expressive correlations are .20 and above. an= 968 bn=176 29 STEPHEN NOWICKI JR., MARSHALL P. DUKE achievement tests that have accumulated considerable evidence of construct validity for elementary grade school children. All 48 possible correlations between scores from the four DANVA receptive subtests and 12 achievement subtests were significant and in the predicted direction. The findings between DANVA expressive subtest scores and the CTBS were less extensive. Although there were a greater number of significant relationships than would have been expected by chance (9 out of 36), the proportion was far less than that found for the DANVA receptive subtest scores. While most of the e~pressive subtest correlations were in the predicted direction, only those involving the expressive paralanguage and expressive facial expression subtests reached significance. Interestingly, unlike the significant associations found in the past between individual differences in other measures of receptive nonverbal processing abilities like the PONS (Halberstadt & J. Hall, 1980), there were no significant correlations between DANVA receptive or expressive scores and IQ (as measured by the Henmon-Nelson IQ test, 1983). Additional support for the significant association between individual differences in DANVA receptive scores and academic achievement was found by Nowicki and Duke (1992b) in a population of elementary school age youngsters (n = 455). Using the Comprehensive Tests of Basic Skills, they found higher receptive facial expression and receptive paralanguage scores significantly related to higher total achievement test scores, r(453) = .32, p < .01 and r(453) = .47, p < .01, respectively. In sum, individual differences in DANVA receptive subtest scores were related to academic achievement. The supportive evidence again was more consistent for the relationship between individual differences in receptive than in expressive DANVA subtest scores and academic achievement. While DANVA receptive and expressive abilities were clearly related to academic achievement, they were not related to IQ as assessed by a standardized IQ measure. Other Evidence of Construct Validity for the DANVA Subtests Gathering convergent validity evidence for the DANVA is a complex task, in part because researchers frequently create different individual difference measures of nonverbal social processing ability from one study to the next. One exception to this tendency is the PONS, which has been used in a number of studies. However, as mentioned earlier, the PONS has been primarily used with adults and has not accumulated impressive construct validity data supporting its use in children. Still, since the PONS is probably the most popular present day measure of receptive nonverbal 30 JOURNAL OF NONVERBALBEHAVIOR processing ability, its relationship to the DANVA needed to be explored. To this end, Nowicki (1990) administered the PONS and the DANVA scales to a sample of 34 third-grade children. Convergent validity was provided by two significant correlations between total accuracy scores on the PONS and scores on the DANVA receptive facial expressions, r(32) = .44, p < .05, and receptive paralanguage subtests, r(32) = .53, p < .05. The other two receptive subtests, receptive gestures and postures, were not significantly related to total PONS scores. Another complex aspect of construct validity is evaluating the nature of the intercorrelations among nonverbal processing abilities. Since there is little theoretical agreement as to how nonverbal social processing abilities are related to one another, no prediction was made for DANVA subtest scores. To assess the intercorrelations among the DANVA subtests, the accuracy scores from the receptive (n = 1001) and expressive (n = 184) subtests of Nowicki and Duke's (1989) original sample were intercorrelated with one another. It was found that there were significant, but low, intercorrelations among receptive subtest scores (ranging from r = .I 7 between postures and gestures to r = .33 between total facial expression and paralanguage), and among expressive subtest scores (ranging from r = .27 between gestures and facial expressions to r = .37 between facial expressions and paralanguage). Further, the intercorrelations across receptive and expressive subtests were either low or negative. For example, receptive and expressive facial expression subtest scores correlated - . 0 6 ; paralanguage subtest scores correlated .04; and gesture subtest scores correlated . 15. The low intercorrelations found among subtests suggest that, at least as measured by the DANVA, there was no general nonverbal social processing ability and that performance on one subtest does not necessarily predict performance on another within a receptive or expressive cluster or in the same nonverbal channel across modalities. Conclusion A number of studies have provided preliminary evidence of construct validity for the DANVA. DANVA subtest accuracy scores improved with age, were internally consistent and reliable over time, and showed significant relationships with indices of personal and social adjustment and academic achievement. Interestingly enough, IQ scores were not found to be related to individual differences in DANVA receptive or expressive skills. Such discriminative validity is helpful in responding to the possible criticism that 31 STEPHEN NOWICKI JR., MARSHALL P. DUKE the DANVA is measuring impoverished functioning in general and not a set of more circumscribed nonverbal social processing abilities. The supporting evidence of construct validity was more plentiful for DANVA receptive, than for expressive, subtests. One probable reason for more studies using the DANVA receptive subtests is that they are easier to administer and score compared to the DANVA expressive subtests. DANVA expressive subtests require videotape equipment, individual administration, and extensive coding to arrive at scores. Because of this, sample sizes were small, and thus, larger effect sizes were needed to reach significance in expressive processing skills. In spite of the difficulties inherent in assessing expressive nonverbal accuracy, the pursuit of reliable and valid instruments for measuring individual differences in the ability to express emotions nonverbally still remains a valuable goal. Errors in sending emotional meanings can create as many personal and interpersonal problems as do mistakes in reading emotions. In this regard, the dimension of intensity of the nonverbal cue needs to be controlled for and evaluated in future tests of both nonverbal sending and receiving abilities. In most tests of nonverbal social processing ability, the intensity of stimuli or expressions is not systematically varied or controlled for. Children may have problems reading or sending emotions at lower levels of intensity. Such accuracy problems may become more significant as children become older because nonverbal cues become more subtle and important with age. To this end, Nowicki and Carton (in press) have reported work on a receptive test, FACES 2, that varies intensity of affect in facial expressions presented by adults and children. The facial expression stimuli reflect high, middle, and low intensity levels of the emotions of happiness, sadness, anger, and fear and allow for the possible identification of the degree of intensity at which subjects begin to become inaccurate. Continued research to develop other assessment instruments that vary intensity of affect in other nonverbal channels of sending or receiving of emotional information is encouraged. In closing, it needs to be emphasized that the DANVA was designed to be a screening test. DANVA subtest scores identify strengths and weaknesses in abilities to process nonverbal information about basic core emotions. However, DANVA scores do not reveal why' children obtained the scores they did. Children who obtain low DANVA scores (i.e., those with some form of dyssemia) should be observed further to uncover the possible source of their nonverbal social processing difficulties. Among these sources may be any number of learning or biological factors that can produce difficulties in discriminating, using, expressing, or applying nonver- 32 JOURNAL OF NONVERBAL BEHAVIOR bal social information. Understanding the source of dyssemias can help determine how and where to focus remediation efforts (Minskoff, 1980a, 1980b). Note 1. Several studies also suggest the stability of the norms established by Nowicki and Duke (1989). Zanier (1991) had 30 six-year-old and 30 ten-year-old boys complete the DANVA receptive tests and found means and standard deviations similar to those presented in the original sample. The only receptive test not comparable was the one measuring paralanguage but that was largely because it included the additional sixteen sentencesspoken by a male. Consistent with proposition 1, Zanier found that ten-year-old boys were significantly more accurate than six-year-old boys on all the receptive subtests. Feig (1993) administered the DANVA receptive subteststo a sample of 152 subjects (M = 9.5 years, SD = .62) and found that their scores did not differ significantly from those of the third graders in the original sample of subjects except for the DANVA receptive gestures subtest. On this subtest, Feig's sample scored significantly higher than the original DANVA sample (M -- 9.1 and M = 8.0, respectively). Feig also administered the DANVA expressive subtests to a smaller sample of 38 subjects and found no significant differences between the scores of her subjects and those of third graders from the original sample. Lastly, Hill (1992) had 62 third graders take the DANVA receptive faces subtest. She found the scores did not differ significantly from those of third graders in the Nowicki and Duke (1989) sample. References Akert, R. M., & Archer, D. (1977). Words and everything else: Verbal and nonverbal cues in social interpretation. Journal of Personality and Social Psychology, 35, 443-449. Allen, J. R., & Hamsher, J. H. (1974). The development and validation of a test of emotional styles. Journal of Consulting and Clinical Psychology, 42, 663-668. Axelrod, L (1982). Social perception in learning disabled adolescents. Journal of Learning Disabilities, 1S, 610-613. Blanck, P. D., Rosenthal, R., Snodgrass, S. E., DePaulo, B. M., & Zuckerman, M. (1981). Sex differences in eavesdropping on nonverbal cues: Developmental changes. Journal of Personality and Social Psychology, 4 I, 391-396. Brody, L. R. (1985). Gender differences in emotional development: A review of theories and research. Journal of Personality, 53, 102-149. Bruck, M. (1986). Social and emotional adjustments of learning disabled children: A review of the issues. In S. Ceci (Ed.), Handbook of cognitive, social and neuropsychologial aspects of learning disabilities (Vol. 1, pp. 361-379). Hillsdale, NJ: Erlbaum. Brumley, B. (1993). Application of a teachers rating scale of nonverbal behavior. Unpublished doctoral dissertation, Department of Psychology, Emory University, Atlanta. Bryan, T. H. (1977). Learning disabled children's comprehension of nonverbal communication. Journal of Learning Disabilities, 10(8), 36-42. Buck, R. (1980)'. Nonverbal behavior and the theory of emotion: The facial feedback hypothesis. Journal of Personality and Social Psychology, 38, 811-824. Burns, K. & Beier, E. (1973). Significance of vocal and visual channels in the decoding of emotional meaning. Journal of Communication, 23, 118-130. 33 STEPHEN NOWICKI JR., MARSHALL P. DUKE Butler, S. (1992). Sex differences in nonverbal communication. Unpublished manuscript, Department of Psychology, Emo~ University, Atlanta. Comprehensive tests of basic skills, Form S (CTgS-5). (1983). New York: McGraw-Hill. Cooley, E., & Treimer, D. (1991). Nonverbal processing abilities in emotionally disturbed students. Paper presented at the meetings of the Southeastern Psychological Association, New Orleans. Coopersmith, S. (1981). Antecedents of self-esteem. San Francisco: Freeman. Creasy, G. L., & Jarvis, P. A. (1987). Sensitivity to nonverbal communication among male learning disabled adolescents. Perceptual and Motor Skills, 64, 873-874. Darwin, C. (1965). The expression of the emotions in man and animals. Chicago: University of Chicago Press. (Original work published in 1872.) Davitz, J. R., & Davitz, Lr (1959). The communication of feelings by content-free speech. Journal of Communication, 9, 6-13. Deutsch, R. (1975). Effects of sex of subject and story character on pre-schoolers' perceptions of affective responses and interpersonal behavior in story sequences. Developmental Psychology, I1,112-113. Duke, M. P., & Nowicki, S., Jr. (1972). A new measure and social learning model for interpersonal distance. Journal of Experimental Research in Personality, 6, 1-16. Ekman, P., & Friesen, W. V. (1974). Nonverbal behavior and psychopathology. In R. J. Friedman & M. Katz (Eds.), The psychology of depression: Contemporary theory and research (pp. 3-31). Washington, DC: Winston & Sons. Ekrnan, P., & Friesen, W. V. (1975). Unmasking the face. Englewood Cliffs, NJ: Prentice-Hall. Ellyson, S. L., & Dovidio, J. F. (1985). Power, dominance, and nonverbal behavior: Basic concepts and issues. In S. L. Ellyson & J. F. Dovidio (Eds.), Power, dominance, and nonverbal behavior (pp. 1-27). New York: Springer-Verlag. Elmore, B. C. (1985). Emotionally handicapped comprehension of nonverbal communication. Journal of Holistic Medicine, 7, 194-201. Feig, S. (1993). Nonverbal social communication and popularity. Unpublished doctoral dissertation, Department of Psychology, Emory University, Atlanta. Feldman, M., & Thayer, S. (1980). A comparison of three measures of nonverbal decoding ability. Journal of Social Psychology, 112, 91-97. Felleman, E. S., Barden, R. C., Carlson, C. T., Rosenberg, L., & Masters J. C. (1983). Children's and adults' recognition of spontaneous and posed emotional expressions in young children. Developmental Psychology, 19, 405-413. Fingeret, A. L., Monti, P. M., & Paxson, M. A. (1985). Social perception, social performance, and self-perception: A study with psychiatric and non-psychiatric groups. Behavior Modification, 9, 345-356. Frank, H., & Hoffman, N. (1986). Borderline empathy: An empirical investigation. Comprehensive Psychiatry, 27, 387-395. Gottesman, J., Gonso, J., & Rasmussen, B. (1975). Social interaction, social competence and friendship in children. Child Development, 48, 513-517. Halberstadt, A. G. (1986). Family socialization of emotional expression and nonverbal communication styles and skills. Iournal of Personality and Social Psychology, 31,827-836. Halberstadt, A. G., & Hall, J. A. (1980). Who's getting the message? Children's nonverbal skill and their evaluation by teachers. Developmental Psychology, 16, 564-573. Hall, J. A. (1978). Gender effects in decoding nonverbal cues. Psychological Bulletin, 85, 845-857. Hall, C. W., & Richmond, B. O. (1985). Nonverbal communication, self esteem and interpersonal relations of LD and non-LD students. The Exceptional Child, 32, 87-91. Hallin, K. (1991 ). Nonverbal social processing deficits and external locus of control in children in. residential treatment. Unpublished honors thesis, Department of Psychology, Emo~, University, Atlanta. Hatter, S. (1985). Manual for the Self-Perception Profile for Children. University of Denver, Denver, CO. 34 JOURNAL OF NONVERBAL BEHAVIOR Henmon-Nelson Test of Mental Abilities (1973). Boston: Houghton Mifflin. Hill, J. (1992). Deficits in nonverbal social processing: The effects of an intervention with elementary school children. Unpublished doctoral dissertation, Department of Psychology, Emory University, Atlanta. Hinely, P., et al. (1990). Georgia Psychoeduational Network. Georgia Psychoeducational Network Publication Monograph # 1. Izarcl, C. E. (1971). The face of emotion. New York: Appleton-Century-Crofts. Jackson, S. C., Enright, R. D., & Murdock, ]. Y. (1987). Social perception problems in learning disabled youth: Developmental lag versus perceptual deficit. Journal of Learning Disabilities, 20, 361-364. Johnson, D. J., & Myklebust, H. R. (1967). Learning disabilities: Educational principles and practices. New York: Gruen & Stratton. Kagan, N. (1977). Affective Sensitivity Test: Validity and reliability. Paper presented at the meeting of the American Psychological Association, San Francisco. Kay, T. (1984). Individual differences in children's abilities to discriminate positive and negative affect from facial cues. Unpublished doctoral dissertation, Emory University, Atlanta. Maheady, L., & Sainato, D. M. (1986). Leaming~isabled students' perception of social events. In S. Ceci (Ed.), Handbook of cognitive, social and neuropsychological aspects of learning disabilities (Vol. 1, pp. 1-38). Hillsdale, N]: Erlbaum. McCauley, E., Kay, T., Ito, ]., & Trecler, R. (1987). The Turner Syndrome: Cognitive deficits, affective discrimination, and behavior problems. Child Development, 58, 464-473. McLeod, P. L., & Rosenthal, R. (1983). Micromomentary movement and the decoding of face and body cues. Ioumal of Nonverbal Behavior, 8, 83-90. Maitland, G. F. (1977). The perception of facial and vocal expressions of emotions by learning disabled, emotionally disturbed, and normal children. Dissertation Abstracts International, 38, 5396-A. Minskoff, E. H. (1980a). Teaching approach for developing nonverbal communication skills in students with social perception deficits: Part I. The basic approach and body language cues. Journal of Learning Disabilities, 13, 118-124. Minskoff, E. H. (1980b). Teaching approach for developing nonverbal communication skills in students with social perception deficits: Part II. Proxemic, vocalic, and artifactual cues. Journal of Learning Disabilities, 13, 203-208. Nowicki, S. Jr. (1987). Use of the Social ludgment Scale with children. Unpublished manuscript, Department of Psychology, Emory University, Atlanta. Nowicki, S. Jr. (1990). The relation of the DANVA to the PONS. Unpublished manuscript, Department of Psychology, Emory University, Atlanta. Nowicki, S. Jr. (1992). Self-esteem and DANVA subtest scores. Unpublished manuscript, Department of Psychology, Emory University, Atlanta. Nowicki, S. Jr., & DiGirolamo, A. (1989). The association of external locus of control, nonverbal processing difficulties and emotional disturbance. Behavioral Disorders, 15, 28-34. Nowicki, S. Jr., & Carton, J. (in press). The measurement of emotional intensity from facial expressions: The DANVA FACES 2. Journal of Social Psychology. Nowicki, S. Jr., & Duke, M. P. (1989). A measure of nonverbal social processing ability in children between the ages of 6 and 10. Paper presented as part of a symposium at the American Psychological Society, Alexandria, VA. Nowicki, S. J., & Duke, M. P. (1992a). Helping the child who doesn't fit in. Atlanta: Peachtree Press. Nowicki, S. Jr,, & Duke, M. P. (19921o). The association of children's nonverbal decoding abilities with their popularity, locus of control, and academic achievement. Journal of Genetic Psychology, 153, 385-394. Nowicki, S., & Strickland, B. (1973). A locus of control scale for children. Ioumal of Consulting and Clinical Psychology, 40, 148-IS4. O'Sullivan, M. (1982). Measuring the ability to recognize facial expressions of emotion. In P. 35 STEPHEN NOWICKI JR., MARSHALL P. DUKE Ekman & K. R. Scherer (Eds.), Emotion in the human face (pp. 281-317). Cambridge: Cambridge University Press. Penn, D. L. (1991). Cognitive rehabilitation of social deficits in schizophrenia: A direction of promise or following a primrose path. Psychosocial Rehabilitation Journal, 15, 27-41. Piers, E. V. (1984). Piers-Harris Children's Self-Concept Scale. Nashville, TN: Counselor Recordings and Tests. Quay, H. C. (1977). Measuring dimensions of deviant behavior: The Behavior Problem Checklist. Journal of Abnormal Child Psychology/, 5, 277-288. Reiff, H. B., & Gerber, P. J. (1990). Cognitive correlates of social perception in students with learning disabilities. Journal of Learning Disabilities, 23, 260-262. Riggio, R. E. (1986). Assessmentof social skill. Journal of Personality and Social Psychology, S1,649-660. Rosenthal, R., & DePaulo, B. M. (1979). Sex differences in eavesdropping on nonverbal cues. Journal of Personality and Social Psychology, 37, 273-285. Rosenthal, R., Hall, J. A., DiMatteo, M. R., Rogers, P. L, & Archer, D. (1979). Sensitivity to nonverbal communication: The PONS test. Baltimore: Johns Hopkins University Press. Rosenthal, R., Hall, J. A., Archer, D., DiMatteo, M. R., &Rogers, P. L. (1977). The PONS Tests: Measuring sensitivity to nonverbal cues. In P. McReynolds (Ed.), Advances in psychological assessment (Vol. 4, pp. 179-221). San Francisco: Jossey-Bass. Rotter, J. B (1990). Internal versus external control of reinforcement. American Psychologist, 45, 489-493. Rowe, E. (1992). The relationship among nonverbal processing abilities, intention cue performance, and popularity. Unpublished masters thesis, Department of Psychology, Emory University, Atlanta. Snodgrass, S. E., & Rosenthal, R. (1985). Interpersonal sensitivity and skills in decoding nonverbal channels: The value of face value. Basic and Applied Social Psychology, 6, 243-255. Stone, W. L. (1980). Comprehension of nonverbal communication in learning disabled and nonlearning disabled children. Unpublished doctoral dissertation, University of Miami, Miami, OH. Stone, W. L., & la Greca, A. M. (1984). Comprehension of nonverbal communication: A reexamination of the social competencies of learning-disabled children. Journal of Abnormal Child Psychology, 12, 505-518. Wolitzky, D (1973). Cognitive control and person perception. Perceptual and Motor Skills, 36, 619-623. Zanier, D. J. (1991). The decoding of facial expressions and its relationship to compliance behavior in boys characterized as extemalizers. Unpublished doctoral dissertation, Department of Psychology, Windsor University, Windsor, Canada. Zenker, C. L. (1992). The association between children's nonverbal processing accuracy and ratings of emotional disturbance and social maladjustment. Unpublished manuscript, Department of Psychology, Emory University, Atlanta. Zuckerman, M., Blanck, P. D., DePaulo, B. M., & Rosenthal, R. (1980). Developmental changes in decoding discrepant and nondiscrepant nonverbal cues. Developmental Psycholofo~/, 16, 220-228.