Nonverbal behavior and client state changes during psychotherapy

NONVERBAL BEHAVIOR AND CLIENT STATE CHANGES DURING PSYCHOTHERAPY

MARTHA DAVIS

New York State Psychiatric Institute

DEAN HADIKS

The Methodist Hospital Brooklyn, New York

In an intensive videotape analysis of 10 psychotherapy sessions, the body positions and gesticulation patterns of the client were examined in relation to changes in her verbal behavior. Verbal ratings were obtained on the Ex- periencing Scale (Klein, Mathieu, Gendlin, & Kiesler, 1970), and nonverbal ratings were obtained on the Davis Nonverbal States Scales (Davis, 1986). Results revealed that the position “accessibility” ratings were related significantly to levels of self-disclosure and involvement as determined by the verbal ratings. The client’s gesticulation ratings were not related significantly to the Experiencing Scale ratings, but clinically interesting relationships between gesticulation patterns and verbal content were noted. While body movement long has been recognized as an important source of clinical information, replicable coding of complex patterns of position and gesture has been very difficult to develop. This study presents data that support a promising, replicable method for coding nonverbal behavior in psychotherapy process research.

Recently, efforts to assess the process of psychotherapy have shifted to include intensive designs and structural methodologies (Elliott, 1984; Frieswyk et al., 1986; Greenberg, 1984; Luborsky, Singer, Hartke, Crits-Christoph, & Cohen, 1984; Mahrer, 1988; Safran, Greenberg, & Rice, 1988). Such approaches have the potential for generating clinically meaningful data that have a direct effect on theory and practice. Intensive analyses of client and therapist communicative behavior are particularly valuable for monitoring and understanding the subtle fluctuations in client psychological states that are characteristic of the change process (Horowitz, 1987; Marmar, Wilner, & Horowitz, 1984; Safran et al., 1988). Rice and Greenberg (1984) have argued that therapists continually assess these fluctuations and use this information to guide their interventions. Jones, Cumming, and Horowitz (1988) have demonstrated how experi- enced clinicians tailor their interventions to meet the needs of their clients. Such research efforts are beginning to make more explicit the means by which effective clinicians achieve their results.

It long has been recognized that evaluation of nonverbal behavior is an integral part of clinicians’ assessments of clients’ psychological states (Freud, 1893; Greenberg, 1984; Horowitz, 1987; Reich, 1949; Rice & Saperia, 1984). Of particular interest are Mahl’s (1968, 1977) clinical studies, which indicate that nonverbal behavior may anticipate verbal content. Mahl refers to this as an “A + B phenomenon’’ in which the

We are deeply grateful for the support of Dr. Israel Zwerling, former chair of the Department of Mental Health Sciences, Hahnemann University, Philadelphia, PA. This therapy research project was sponsored by Hahnemann University Department of Mental Health Sciences for more than 10 years. A major part of the present study was completed at Fairleigh Dickinson University, Teaneck, NJ.

We especially thank the client for her consent. We wish to express our appreciation to Dr. Robert McGrath for his generous contribution to the statistical analysis and to the coders on the project. Finally, we are in- debted to Dr. Philippa Matheiu-Coughlan and Dr. Marjorie Klein for their vital contributions to the verbal analysis.

Requests for reprints should be addressed to Dr. Martha Davis, Psychiatric Institute, Division of Clinical Psychology, P.O. Box 80, 722 West 168th Street, New York, NY 10032.

340

Nonverbal Behavior and Client State Changes 341

emergence of new material is preceded by a new inadvertent action or body position of the patient. These writings have anticipated the current interest in the role of motoric processes in the development of mental structures (Izard, Kagan, & Zajonc, 1984; Stern, 1985; Zajonc & Marcus, 1984); the interrelationship of emotion and cognition (Basch, 1988; Izard, 1984; Lewis, Wolan-Sullivan, & Michalson, 1984; Zajonc, 1980); and at- tention to nonverbal behavior, particularly facial expression, in the empirical study of emotion (Ekman, Friesen, & Ellsworth, 1972; Ekman & Oster, 1979).

Empirical research on nonverbal behavior in psychotherapy, however, has been problematic. Some studies have utilized actors rather than real therapists (Fretz, Corn, Tuemmler, & Bellet, 1979; La Cross, 1975; Smith, 1972; Trout & Rosenfeld, 1980) and have relied on college students to evaluate complex constructs such as rapport (Dinges & Oetting, 1972; Karger, 1974; La Cross, 1975; Lassen, 1973). These “impression” studies have been criticized for the weakness of their external validity (Bakken, 1978; Davis, 1985; Gladstein, 1974; Myers & Myers, 1972). Wallbott (1985) has demonstrated that there are serious limitations to data based on judgments of patients’ movements made by untrained observers. Other studies have measured single variables, such as amount of eye contact (Fretz et al., 1979) or interpersonal distance (Boucher, 1972; Lassen, 1973), in ways that are so simplistic that they generate naive, clinically trivial information (Davis, 1985).

However, more sophisticated projects that involved clinically promising, multivariable analyses of actual therapy sessions have tended to be limited to single sessions. For example, Scheflen (1973) conducted a 4-year film analysis of one family therapy session. Verbal transcripts and highly detailed logs of participants’ nonverbal behavior were analyzed in a procedure that identified distinct phases of recurrent sequences of interaction in what appeared to be an anatomy of a symbiotic relationship. Charny (1966) discovered clinically important verbahonverbal relationships in his analysis of posture sharing in a single psychotherapy session. Davis (1974) analyzed a session that occurred in the tenth month of the therapy of a chronic schizophrenic man. She found relationships between the relative degree of organization of the client’s behavior and the degree of coherence and appropriateness of his verbal productions. In addition, she found a “fit” between the client’s and therapist’s positions and movement dynamics, which sug- gested an exquisite attunement achieved through a progression of distinct phases. She discovered later that the therapist considered this their best session.

Davis has adapted methods of coding body movement from dance and movement notation, in particular, aspects of Laban movement analysis (Bartenieff & Lewis, 1980) that describe dynamics and intensity. These “parakinesic” features of movement largely are ignored in nonverbal research despite the fact that the clinical literature is replete with indications that affect is expressed through movement intensity and quality as well as through facial expression and paralinguistic patterns.

These single-session analyses demonstrate the potential of nonverbal analysis for empirically identifying brief clinical states in clientltherapist interactions. Although they are provocative in their implications for theory and practice, they are difficult to replicate and have not generated data that can be generalized to other clients.

The present study is an intensive analysis of one client’s verbal and nonverbal behavior during the course of her psychotherapy. While it, too, is limited to one patient, it represents a considerable methodological advance over these earlier studies because it involves the first reliable multivariable coding system for gesticulation and position patterns of both client and therapist (Davis, 1986; Davis & Hadiks, 1987).’

‘This study of patient nonverbal behavior is the first part of a project whose other parts involve the therapist’s nonverbal patterns in relation to his verbal interventions and analysis of the ways in which the client’s and therapist’s movements and positions are related.

342 Journal of Clinical Psychology, May 1990, Vol. 46, No. 3

The study reported here was conceived as a rigorous investigation of two key questions in the nonverbal behavior literature related to psychotherapy; whether gesticulations and positions accurately reflect clinically significant cognitive and emotional processes of clients during psychotherapy and, if they do, whether they anticipate verbal expression of these processes.

METHOD Procedure

Videotapes of the 62 sessions of one complete outpatient psychotherapy were the source of the data. Initially, these sessions were recorded in the office of the therapist for a seminar for psychiatry residents. Permission to use them for research purposes was obtained midway in the therapy. No camera operator was present. The therapist set up the equipment in his office, focused the camera on both chairs for a total view of both himself and the client, and turned on the camera before each session. The client was a woman in her mid-twenties who sought therapy for anxiety attacks. Therapy sessions were held once weekly for 1 ‘/z years, with occasional breaks for vacations or the therapist’s professional travel.

Ten of these sessions were selected randomly from the 62 so that each quarter of the therapy was represented (i.e., 2 sessions from the first 15, etc.). Each session was divided into six 7-minute segments. Forty-two minutes were subtracted from the point at which the therapist signalled the end of the session to mark the beginning of the first segment. This yielded sixty 7-minute segments for a total of 7 hours of data. The entire image of the therapist was blocked out with a screen to eliminate any interference of this information on the coding of the client’s nonverbal behavior. The time in minutes and seconds was superimposed on the bottom of the screen to help coders record onset time. Raters

The six raters used for the nonverbal ratings were Certified Movement Analysts, trained in movement observation at the LabadBartenieff Institute of Movement Studies in New York City. For this project, the raters were trained further on the Davis Nonverbal States Scales (DNSS; Davis, 1986). This is a system that seeks to balance replicability and efficiency in coding of positions and gestures with a complexity that enhances clinical usefulness. It is based on a wide range of clinical and research observations with regard to how orientation, distance, and the dynamics of gesticulation and position reflect psychological states (Davis, 1985). A detailed presentation of the system, criteria for item selection, and reliability are discussed in Davis (1986), Davis and Hadiks (1987), and Hadiks (1987). Training involved repeated viewings of non-research videotapes to acquaint the raters with the behaviors that were targeted in the study and to obtain preliminary tests of interrater agreement before the research coding began.

Working independently, raters could view any portion of each 7-minute segment as frequently as they felt was necessary. They did not hear the sound, and they did not know about the aims of the project; nor were they able to infer accurately what the goals of the project were after data collection had been completed. Nonverbal Variables

Nonverbal ratings were completed separately for the position and gesture variables. That is, one team of three raters coded positions, then another team of three coded gesticulations. The following behaviors comprised the position variable: trunk lean (direc- tion of upper body lean), lower body width (relative separation of knees), and leg tuck. Furthermore, some behaviors were rated in terms of their relationship to the therapist. Specifically, the upper and lower body facing variables were rated in terms of their being toward the therapist, a “neutral” or deflected orientation, and away from the


therapist .' Additionally, the patient could be coded as swiveling her chair markedly away from the therapist.

The descriptive categories for gesticulation were: laterality (unilateral, bilateral or some sequence of these); body part articulation (fingers, hands, forearm, or whole arm); dynamics (quality of movement, e.g., sudden, light, etc.); gesticulation length (short, medium, etc.); and spatial complexity (low, medium, etc.). Each behavior is operational- ized rigorously in a DNSS guide (Davis, 1986). Three observers coded each segment and generally needed about three viewings of a segment to complete the gesticulations or position coding. The range of gesture qualities perceptible within a 30-second interval was the unit for the gesticulation coding rather than frequency of occurrence of a quality. For positions, however, the initial position was coded, then the onset time and nature of each subsequent change of po~ i t ion .~ Verbal Analysis

Verbal ratings were made from audiotapes using the Experiencing Scale (EXP; Klein et al., 1970), a widely used 7-point scale that measures the level of client involvement in the psychotherapy process, from superficial discourse to affectively involved passages in which the client is judged to be working actively at clarifying and understanding. Previous research on interrater reliabilities for this scale is reported in the manual and shows that Ebel's (1951) coefficients range from .76 to .91. Typed transcripts were made to facilitate the analyses. Audiotapes supplemented by these transcripts were rated independently by the two senior authors of the Experiencing Scale. The ratings were in- serted in the margins of the transcript next to the client's statements, after the number of the statement in the segment. Each speaking turn of the client was numbered and received an EXP rating from 1 to 7. Speaking turns, particularly long ones, sometimes contained multiple ratings when the coders determined that the client changed level of experiencing in the course of the speaking turn. When there was an Experiencing Scale rating change within a monologue, the point at which this shift occurred was noted on the transcript.

RESULTS In terrater Reliabilities

Interrater reliabilities were calculated separately for each of the items of the DNSS coding (e.g., trunk lean, spatial complexity, lower body orientation, laterality pattern). Because of its rigor, Cohen's (1960) kappa was the statistic used to compute these reliabilities. Landis and Koch (1977) have proposed that accepted levels of kappa reliability are as follows: less than .40, poor; .40 to .75, fair to good; greater than .75, excellent. Kappas for gesture unit, gesture length, and spatial complexity were performed after collapsing the lower two levels (very low/low) and the upper two levels (medium and high) because this degree of detail was sufficient for data analysis.

Some judgments were considerably more difficult to make. The degree of postural lean and facing was particularly difficult to judge. The two interactants oriented to each other in a somewhat deflected vis A vis facing, and the camera provided a view of the interactants almost in profile; consequently, the degree of lean and shift away from or toward the other was difficult to judge. The coefficients for these judgments were weighted to account for this difficulty according to the methods proposed by Fleiss (1981). Specifically, adjacent cells (those next to ones that contained perfect agreement) were weighted by S O , a conservative approach according to Fleiss. This procedure results

2"Where" the therapist was had to be inferred from the client's head orientation as she addressed him. 'The DNSS includes a third area of coding, specific actions, that has not been included in the present

analysis, but is important for a more specific analysis of defenses.


in higher reliabilities and reflects the degree of difficulty of the ratings. The weighting did not extend beyond adjacent cells.

The reliabilities were calculated on codings of 20 of the 60 segments, chosen randomly with the restriction that all quarters of the therapy be represented. Measurements were taken for every 30-second section of each segment because this was the shortest interval in which both aspects of movement were coded. The reliabilities for each of the nonverbal variables can be seen in Table 1.

Table 1 Reliability Coefficients and Percentages of Agreement for Each of the Nonverbal Variables

Average Percentage of percentage 2 out of 3

Variable agreement Kappaa agreement

Position categories Lean 72% .40a 97 To

Facing upper 71% .59a 98 qo

Facing lower Lower Leg tuck Swivel

61% 51" 95 Vo

88% .82 98%

92 qo .85 97% b 84% 94%

Gesticulation categories Laterality 88% .81 98 qo

Gesture unit 83% .71 98 %

Gesture length 81% .70 96 %

Spatial complexity 87 '70 .73 98%

Dynamics 78% .65 94 qo

aWeighted. bBecause of the large number of non-occurrences in this variable, kappas were not calculated.

As can be seen in Table 1, the coding of these behaviors proved to be quite reliable. Both kappas and percentage agreement are reported in order to provide a basis for com- parison. The nonverbal literature has been inconsistent with respect to reliability statistics, with percentage agreement reported far more frequently than chance-corrected coefficients. In addition, the percentage of time that 2 out of 3 raters agreed on a judgment was between 94% and 98%. Subsequent analyses were performed on all judgments in which 2 out of 3 raters agreed. Less than 4% of the nonverbal data were lost to lack of agreement among raters using this criterion.

Because the verbal ratings were completed by the two senior authors of the Ex- periencing Scale, the reliabilities for the verbal ratings were considerably higher than the .76 - .91 Ebel's (1951) coefficients reported for previous research with this scale. The ratings were completed independently. There were so few differences that formal calcula- tion of a reliability statistic was unnecessary. The few instances of disagreement were discussed after the ratings were completed and consensus was reached on each one such that all of the EXP coding was usable for subsequent analyses. Determination of Nonverbal States

Gestural and postural nonverbal states then were constructed for this client according to criteria defined in detail in Davis (1986). The position states were formed by order-


ing the various position codes along an ordinal 12-point scale that represented the degree of “accessibility” to the therapist in terms of orientation and openness of position from level 1 (leaning with upper and lower body facing markedly away from the therapist) to level 12 (sitting upright or forward with upper and lower body facing in full vis A vis the therapist). An ordinal 9-point scale for gestures also was constructed that clustered the level of complexity in spatial organization, intensity of movement dynamics, and quality and extent of body part activation from level 1 (a baseline of short, single forms of laterality of hand and forearm having no intensity or spatial complexity) to level 9 (long, intricate gestures in complex laterality sequences with high degree of spatial complexity and dynamic intensity).

A factor analysis was run on all of the coded variables. If positions and gestures are related to different aspects of psychological experience (Davis, 1985; Scheflen, 1964, 1973), it would be important to assess whether they are in fact independently occurring variables. It was, therefore, important to have an empirical check on the coding system and the nonverbal states scales determination method. Factor A nalysis

The 26 variables that comprise the position and gesture variables (i.e., spatial complexity, lower body neutral, lean forward, etc.) were factor analyzed utilizing a Varimax rotation. The 12-factor solution, which represented the best compromise between detailed and inclusive solutions, is presented in Table 2. The criterion for inclusions of a factor was set at a minimum eigenvalue of 1.31. Variables with factor loadings of .40 or less were not included.

Six factors were not included in Table 2 because they only described what amounted to physically possible or “logical” clusters. In one of these factors, for example, leg tuck loads positively with knees slightly apart and negatively with knees close together, a physical inevitability. The six factors described in Table 2 reveal that position codings were distinct from the gesture codings and that they closely approximate the position and gesture scales constructed for this study from an a priori method based on research and clinical experience (Davis, 1986; Hadiks, 1987). Specifically, the a priori constructed states were the most frequently co-occurring clusters of the position and gesture variables. Consequently, they were used as landmarks along a continuum of increased involvement as defined for each scale. The rest of the points along each continuum were constructed so that all of the data could be analyzed, not just the most frequently occurring clusters of this patient. For example, Factor 1 of Table 2 exactly parallels what are described in the states analysis procedure as the indicators of level 1, a “pro forma” state in which gesticulations are short, small, limited to distal body parts, low in spatial complexity, simple in laterality pattern, and without dynamic intensity. Factor 2 parallels what is articulated in the states analysis as gesticulation features of level 6, the highest degree of an “intensity/excitation” state that is composed of medium length phrases and spatial complexity and strong and/or sudden movement qualities. Factor 3 cor- responds to level 9, the highest degree of what is called “elaboration,” that is, high spatial and laterality complexity along with indirect, light, or slow movement qualities. In other words, the states labeled “pro forma,” “excitation,” and “elaboration” are arranged in a progression, and each is refined further into three levels (low, medium, and high). Thus, arranged sequentially, they form the 9-point gesticulation scale. Factors 4, 5 , and 6 on Table 2 are very close to points, 11, 12, and 3, respectively, on the position accessibility scale.4

41t should be noted that one variable, trunk involvement, is treated differently (although not incompatibly) in the DNSS procedure for determining nonverbal states from specific clusters of coding.


Table 2 Nonverbal Variable Factors: Factor Loadings for Varimax Rotated Factor Solution

Factors 1 2 3 4 5 6

Finger, hands, forearm only .91 Trunk involvement .52

Very low, vague, or low spatial complexity Medium spatial complexity .53 High spatial complexity .67

Very short, or short gesture length Medium gesture length .45

Single level laterality .91 All three laterality types .55

No dynamics .81

Sudden dynamics .61

Direct and/or strong dynamics .65

.95

.9 1

Indirect, light or slow dynamics .55

Lean right - .82 Lean back .80 Facing upper - toward .75

Facing upper - neutral

Facing lower - toward .77

- .73 Facing upper - away .69

Facing lower - neutral Facing lower - away Lower - neutral .74 Lower - closed - .63

- .60 - .58

Swivel markedly away .63

Both legs tucked .70 Leg tuck release .49

Eigenvalues 4.75 1.90 1.65 3.13 2.96 1.82

070 of variance 15 6 5 10 10 6

VerbaVNon verbal Correlations The ratings for the Experiencing Scale and the position and gesture variables were

collapsed into groups of four 30-second segments (2 minutes) each, and the mean value of each such group was calculated. This manipulation was done to test the hypothesis that concerned the temporal relationship between the EXP verbal and DNSS nonverbal variables.

A table of correlations between the Experiencing Scale and each nonverbal variable appears as Table 3. Experiencing Scale scores were used as the criterion for all subsequent analyses. These scores were correlated with the 2 minutes before, during, and after the experiencing period on the position and gesture scales.


Table 3 Correlations Between Each Nonverbal Variable and the Experiencing Scale

EXP Gesture Before Gesture Simult. Gesture After

EXP 1 .Ooo .095 .lo7 .049 Gesture Before 1 .OOo .350 .225 Gesture Simult. 1 .oo .380 Gesture After 1 .Ooo

EXP Position Before Position Simult. Position After

EXP 1 .Ooo .182 .265* .312* Position Before 1 .ooo 327 - . lo7 Position Simult. 1 .Ooo - .146 Position After 1 .Ooo

* p < .05.

As can be seen in Table 3, the position simultaneous (r = .27; p < .05) and position after (r = .31; p < .05) variables were correlated significantly with the Experiencing Scale. Regression analyses were run to test whether each variable and each combination of variables could predict the Experiencing Scale ratings. These analyses were not significant. Time series analyses also were run in order to test the strength of the patterns among the variables. These also were not significant.

DISCUSSION The data of the present study yielded significant correlations between the Experienc-

ing Scale and two of the position variables. This means that as the client shifted from superficial discussion to actively exploring her internal reactions, her bodily positions became increasingly more accessible, open, and oriented toward the therapist. These results support previous observations by Scheflen (1973), Davis (1974), and Charny (1966) on the potential of positions as a valid and replicable measure of subtle fluctuations in level of defensiveness and rapport in psychotherapy. The present study is a considerable advance on these previous single-session analyses, based as it is on a replicable coding method for positions and gestures that is theoretically applicable to any patient or therapist.’ The coding of discrete, unique positions with categories that then are scaled according to degree of “accessibility” has both qualitative and quantitative components. That is, much of the original detail of the position is recorded in a way that allows for a 12-point scaling useful for quantitative assessment. Position analysis heretofore has been largely descriptive and qualitative, so the evidence that this 12-point scaling is related to clinically meaningful changes is encouraging.

Although changes in positions were related to changes in verbal involvement, the position ratings did not anticipate this change in ways similar to the A - B phenomenon posited by Mahl(l977). It is possible, however, that this unit of measurement may have been too restrictive, and this hypothesis might be tested in more naturalistic units, such

’The DNSS coding method was developed through study of videotapes of a number of patients and therapists with the aim of its being applicable to diverse subjects. DNSS coding of the therapist’s movement in another study (Davis & Hadiks, 1987) indicates that the system works for someone who has a very different style than the patient. Of course, research is needed with large samples to test this assumption.


as the actual point of change. It is possible, in other words, that the nonverbal shifts preceded the verbal shifts within the 2-minute segments used in this study. In the present study, however, dynamics of gesticulations were not related significantly to the Experienc- ing Scale ratings.

A number of hypotheses can account for this finding. The most apparent appears to be an interaction of two factors: the relatively subtle variability in the gesture ratings and the often unvarying ratings of the Experiencing Scale. It should be noted that the client was not a hospitalized psychiatric patient and that her movement was not disorganized or disturbed in ways such as can be seen in hospitalized patients (Davis, 1974). More importantly, she was not particularly animated, and although over the course of the 10-session sample her range of variation was appreciable, within sessions she tended to display subtle variations, not dramatic peaks of intensity. In the Experiencing Scale there were large segments of some sessions, as well as one entire session, in which her ratings stayed the same. Given the subtle fluctuations of the gesticulation intensity and the low variation of the Experiencing Scale ratings, the likelihood of a high degree of covariation was compromised.

The lack of a significant relationship between the Experiencing Scale and the client’s gestures also may reflect the differences between the Experiencing Scale and the Davis Nonverbal States Scales. Although the Experiencing Scale measures the extent of involvement in psychotherapy, it appears to stress its cognitive dimension. This presented a problem in the present study in that there were numerous instances during which the client appeared more actively engaged in the psychotherapy process, but apparently not in such a way as to earn higher Experiencing Scale ratings. During all of session 6, for example, the client earned an EXP rating of 2 in spite of the fact that she spent a considerable amount of time discussing important events of her early childhood. In fact, the session evolved to a point at which the client spent considerable time describing get- ting her first menstrual period. All of this material, however, was rated as the client being at level 2 of experiencing, a level at which the person largely is giving information. The nonverbal ratings for gesticulation intensity were more variable, however, and were highest during the time she was talking about her first period.

The subject of her first period and her early friendships with boys again came up in session 23, again with the corresponding low experiencing and higher gesticulation dynamics ratings. Similarly, in discussing a dream during session 52 she again earned low scores (2 and 3) even though in the dream she is being attacked by someone with a meat cleaver. The gesture codings during this time suggest emotional arousal. Although this hypothesis needs to be tested more rigorously, there are indications in the present study that the client’s nonverbal behavior indicated emotional involvement in the psychotherapy process in ways that were not detectable by the Experiencing Scale. The possibility that each variable also is related to the therapist’s activity currently is being examined. Positions, in particular, have been shown to be highly context sensitive. The coordinated and inticately woven patterns of echoing, mirroring, and posture sharing have been related to periods of more personal, emotionally loaded material for clients.

Some of the most significant data generated by the present study were the interrater reliabilities for this unique and relatively complex analysis of qualitative aspects of nonverbal behavior. Although the kappas for the position variables fell within the “fair to good” category, when one considers the subtlety of these judgments, the angle of the camera, and the less than optimal clarity of the tapes, they are still notable. The percentage agreements for these ratings are also high, and the two out of three agreements are very high. Overall, the data presented here do not support the notion that DNSS coding is redundant information that could be obtained more efficiently simply by count- ing the frequency or duration of gestures or positions. A client may be gesturing frequently and continually, but in a cursory, unelaborated manner. Focus on the presence


or absence of specific qualitative features -in particular, combinations that give the gesticulations character and intensity - seems more clinically useful.

Perhaps most intriguing, these data indicate that position shifts and gesticulation patterns as coded here are independent and related to different intrapsychic phenomena. DNSS coding of positions may be tapping into fluctuations in “cognitive set” and at- titude toward what is transpiring at the moment. When Scheflen (1973) used the term “position” to describe units of communicative behavior bounded by position shifts, he considered these to be physical manifestations of psychological phases or brief attitudinal stances. Changes in gesticulation quality and intensity, on the other hand, may reflect subtle fluctuations in affect and arousal level. These subtle and elusive aspects of nonverbal behavior are now quantifiable and, along with the study of the “categorical” emotions (Ekman & Oster, 1979; Frijda, 1986; Tomkins, 1984), may allow greater insight into emotional processes during psychotherapy.

Is such a complex and time-consuming coding justifiable? Approximately 40 hours of observation training and 80 hours of data collection per observer for 10 psychotherapy sessions of one patient seem unwarranted. Arguably, much of the training and coding time can be reduced with more experience. Three coders for all sessions may not be necessary, although in this study this ensured that virtually all of the data collected were usable. And, of course, only part of the coding method might be used for specific prob- lems. But whatever might increase its efficiency, the DNSS remains a multivariable coding of continuous behavior to a 2-second level of accuracy. It can generate an enormous amount of clinically rich data. As was stated earlier, a great deal of research in nonverbal communication indicates that if observations are very simple and untrained judges are used to note the most obvious details of nonverbal behavior, the results will be clinically trivial (Davis, 1985). But if it is possible to reliably code subtle changes in clinical states from visible behavior, then it is possible that one is identifying the nonverbal bases of clinical intuition and empirically monitoring essential aspects of psychotherapy process. This study and others in progress on the relationship between the therapist’s and patient’s nonverbal behavior indicate how much clinically salient information can be obtained from this coding. And while DNSS is complex, it is designed to capture details of motor behavior sufficiently gross that clinicians may perceive them either peripherally and subliminally or explicitly through training.

REFERENCES BAKKEN, D. G. (1978). Behavioral adjustment in nonverbal intimacy: A methodological note. Personality

BARTENIEFF, I . , I LEWIS, D. (1980). Body movement: Coping with the environment. New York: Gordon

BASCH, M. F. (1988). Understandingpsychotherapy: The science behind the art. New York: Basic Books. BOUCHER, M. L. (1972). Effects of seating distance on interpersonal attraction in an interview situation.

CHARNY, E. .I. (1 966). Psychosomatic manifestations of rapport in psychotherapy. Psychosomatic Medicine,

COHEN, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measure- ment, I , 31-46.

DAVIS, M. (1974). Analysis of movement styles and interaction in psychotherapy. Paper presented at the First International Symposium on Nonverbal Aspects and Techniques of Psychotherapy, University of British Columbia.

Nonverbal behavior research and psychotherapy. In G. Streicker & R. H. Kiesner (Eds.), From research to clinical practice. New York: Plenum Press.

Guide to the Davis Nonverbal States Scales coding method. Available from the author, Psychiatric Institute, Division of Clinical Psychology, Box 80,722 West 168th Street, New York, NY 10032.

and Social Psychology Bulletin, 4, 300-303.

& Breach.

Journal of Consulting and Clinical Psychology, 38, 15-19.

28. 305-315.

DAVIS, M. (1985).

DAVIS, M. (1986).

3 50 Journal of Clinical Psychology, May 1990, Vol. 46, No. 3

DAVIS, M., HADKS, D. (1987). The Davis Nonverbal States Scales for psychotherapy research: Reliability of LMA-based coding. Movement Studies, 2, 29-34.

DINGES, N. G., & OETTING, E. R. (1972). Interaction distance anxiety in the counseling dyad. Journal of Counseling Psychology, 19, 146-149.

EBEL, R. L. (1951). Estimation of the reliability of ratings. Psychometrika, 16, 407-424. EKMAN, P., FRIESEN, W. V., & ELLSWORTH, P. (1972). Emotion in the human face. New York: Pergamon

Press. EKMAN, P., & OSTER, H. (1979) Facial expression of emotion. In M. R. Rosensweig & L. W. Porter (Eds.),

Annual Review of Psychology, 30, 527-554. ELLIOTT, R. (1984). A discovery-oriented approach to events in psychotherapy: interpersonal process recall

and comprehensive process analysis. In L. N. Rice & L. S . Greenberg (Eds.), Patterns of change (pp. 249-286). New York: Guilford Press.

FLEISS, J. C. (1981). FRETZ, B. R., CORN, R., TUEMMLER, J. M., & BELLET, W. (1979). Counselor nonverbal behavior and client

FREUD, S. (1893). FRIESWYIC, S. H., ALLEN, J. G., COLSON, D. B., COYNE, L., GABBARD, G. O., HORWITZ, L., & NEWSON,

G . (1986). Therapeutic alliance: Its place as a process and outcome variable in dynamic psychotherapy research. Journal of Consulting and Clinical Psychology, 54, 32-38.

Statistical methods for raters and proportions (2nd ed.). New York: John Wiley.

evaluation. Journal of Counseling Psychology, 26, 304-31 1. Studies in hysteria. London: Hogarth Press.

FRUDA, N. H. (1986). The emotions. Cambridge: Cambridge University Press. GLADSTEIN, G. A. (1974). Nonverbal communication and counseling/psychotherapy: A review. Counseling

Psychologist, 4. 34-57. GREENBERG, L. S. (1984). Task analysis: The general approach. In L. N. Rice & L. S . Greenberg (Eds.),

Patterns of change (pp. 124-148). New York: Guilford Press. HADKS, D. (1987). A structural analysis of verbal and nonverbal behavior in psychotherapy: Detecting cross

dimensional patterns. Unpublished doctoral dissertation, Fairleigh Dickinson University. HOROWITZ, M. (1987). States of mind: Analysis of change in psychotherapy (2nd ed.). New York: Plenum

Press. IZARD, C. E. (1984). On the ontogenesis of emotions and emotion-cognition relationships in infancy. In

C. E. Izard, J . Kagan, & R. B. Zajonc (Eds.), Emotion, cognition, and behavior (pp. 389-413). Cam- bridge: Cambridge University Press.

IZARD, C. E., KAGAN, J., & ZAJONC, R. B. (1984). Introduction. In C. E. Izard, J. Kagan, & R. B. Zajonc (Eds.), Emotion, cognition. and behavior (pp. 1-14). Cambridge: Cambridge University Press.

JONES, E. E., CUMMING, J. D., a HOROWITZ, M. J. (1988). Another look at the nonspecific hypothesis of therapeutic effectiveness. Journal of Consulting and Clinical Psychology, 56, 48-55.

KARGER, K. J. (1974). The relationship of nonverbal counselor behavior to client and rater perception of empathy. Unpublished doctoral dissertation, University of Wisconsin.

KLEIN, M., MATHEIU, P., GENDLIN, E., & KIESLER, D. J. (1970). The Experiencing Scale: A research and training manual. Madison: Wisconsin Psychiatric Institute.

LA CROSS, M. B. (1975). Nonverbal behavior and perceived counselor attractiveness and persuasiveness. Journal of Counseling Psychology, 22, 563-566.

LANDIS, J. R., & KOCH, G. G. (1977). A one-way components of variance model for categorical data. Biomdrics, 33, 671-679.

LASSEN, C. I. (1973). Effect of proximity on anxiety and communication in the initial psychiatric interview. Journal of Abnormal Psychology, 81, 226-232.

LEWIS, M., WOLAN-SULLIVAN, M. W., & MICHALSON, L. (1984). The cognitive-emotional fugue. In C. E. Izard, J. Kagan, & R. B. Zajonc (Eds.), Emotion, cognition. and behavior (pp. 264288). Cambridge: Cambridge University Press.

LUBORSKY, L., SINGER, B., HARTKE, J., CRITS-CHRISTOPH, P., & COHEN, M. (1984). Shifts in depressive state during psychotherapy: Which concepts of depression fit the context of Mr. Q's shift? In L. N. Rice & L. S. Greenberg (Eds.), Patterns of change (pp. 157-193). New York: Plenum Press.

MAHL, G. F. (1968). Gestures and body movements in interviews. In J. M. Schlien (Ed.), Research in psychotherapy (pp. 295-346). Washington: American Psychological Association.

MAHL, G. F. (1977). Body movement, ideation, and verbalization during psychoanalysis. In N. Freedman & S . Grand (Eds.), Communicative structures andpsychic structures (pp. 291-310). New York: Plenum Press.

MAHRER, A. R. (1 988). Discovery-oriented psychotherapy research: Rationale, aims, and methods. American Psychologist, 43, 694-702.

Nonverbal Behavior and Client State Changes 35 1

MARMAR, C. R., WILNER, W., & HOROWITZ, M. J. (1984). Recurrent client states in psychotherapy. In L.

MYERS, A. E., & MYERS, G. E. (1972). Discussion on papers on nonverbal communication. Comparative

REICH, W. (1949). RICE, L. N., & GREENBERG, L. S. (1984). Patterns of change. New York: Guilford Press. RICE, L. N., & SAPERIA, E. P. (1984). Task analysis of the resolution of problematic situations. In L. N.

SAFRAN, J. D., GREENBERG, L. S . , & RICE, L. N. (1988). Integrating psychotherapy research and practice:

SCHEFLEN, A. E. (1964). The significance of posture in communication systems. Psychiatry, 27, 316-331. SCHEFLBN, A. E. (1973). Communicational structure: Analysis of apsychotherapy transaction. Bloomington:

SMITH, E. W. L. (1972). Postural and gestural communication of A and B therapist types during dyadic

STERN, D. (1985). TOMKINS, S. S. (1984). meet theory. In K. R. Scherer & P. Ekman (Eds.), Approaches to emotion (pp.

TROUT, D. L., & ROSENFELD, H. M. (1980). The effect of postural lean and body congruence on the judg-

WALLBOTT, H. G. (1985). Hand movement quality: A neglected aspect of nonverbal behavior in clinical

ZAJONC, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35,

ZAJONC, R. B., & MARCUS, H. (1984). Affect and cognition: The hard interface. In C. E. Izard, J. Kagan, & R. B. Zajonc (Eds.), Emotion, cognition, and behavior (pp. 73-102). Cambridge: Cambridge Univer- sity Press.

N. Rice & L. S. Greenberg (Eds.), Patterns of change (pp. 194-212). New York: Guilford Press.

Group Studies, 3, 487-496. Character analysis. New York: Farrar, Straus, & Giroux.

Rice and L. S. Greenberg (Eds.), Patterns of change (pp. 29-66). New York: Guilford Press.

Modeling the change process. Psychotherapy: Theory, Research, and Practice, 25, 1-1 7.

Indiana University Press.

interviews. Journal of Consulting and Clinical Psychology, 39, 29-36. The interpersonal world of the infant: New York: Basic Books.

163-195). New York: Erlbaum.

ment of psychotherapeutic rapport. Journal of Nonverbal Behavior, 4. 176-190.

judgment and person perception. Journal of Clinical Psychology, 41, 345-359.

151-175.

CAN PSYCHOTHERAPISTS TRANSCEND THE SHACKLES OF THEIR TRAINING AND SUPERSTITIONS?

ARNOLD A. LAZARUS

Rutgers - The State University of New Jersey

Formal courses and training in psychological diagnosis and treatment often undermine the talents and skills with which many trainees are endowed naturally. Too many seem to acquire a stylized professionalism replete with general labels, questionable theories, and unfortunate proscriptions. Despite lip service paid to prescriptive eclecticism, most graduate programs socialize their students into delimited schools of thought. Meanwhile, hundreds of psychotherapeutic systems have proliferated. Unsubstantiated claims abound. Nevertheless, there are data that show that specific treatments of choice exist for many disorders. It is argued that a multimodal, systematic, technically eclectic model can point the way to a genuinely scientific aproach to psychotherapy.

In 1960, shortly after obtaining my Ph.D., I was introduced to a man a few years my senior with whom I soon became close friends. He was a remarkable fellow-

This paper is based on an Invited Address delivered to the Annual Convention of the American

Correspondence that concerns this article should be addressed to Arnold A. Lazarus, 56 Herrontown Psychological Association in New Orleans, August 1989.

Circle, Princeton, NJ 08540.

Nonverbal behavior and client state changes during psychotherapy

Documents

Transcript of Nonverbal behavior and client state changes during psychotherapy