a behavioral game methodology for the...

A BEHAVIORAL GAME METHODOLOGY FOR THE STUDY OF PROXEMIC BEHAVIOR

1


A Dissertation Presented to the Faculty of the Graduate School

of Yale University

in Candidacy for the Degree of Doctor of Philosophy

by

William J. Ickinger

May 1982 @ Copyright William J. Ickinger, Jr. 1982 ALL RIGHTS RESERVED


2

ABSTRACT


William J. Ickinger

Yale University 1982 "Proxemics," a conceptual framework developed by the anthropologist E. T. Hall for the study of the way people use physical space in interpersonal interaction is described. Some related research on the nonverbal immediacy behaviors is briefly discussed. Methodologies used in proxemics research are discussed and representative laboratory, field, and simulation studies briefly described. Major findings in proxemic research, including results of studies relating interpersonal relationships and individual characteristics to interpersonal distance, and studies of "spatial invasion" and other behaviors are summarized. Some criticisms of proxemic research and suggestions for future research from the literature are reviewed. A "behavioral game" methodology devised by the author for the study of proxemic behavior in order to meet some of the objections in the literature to the existing research is discussed. A "device," a special room which provides a physical realization of the methodology, is described as are some specialized statistical measures that can be used in conjunction with the device to analyze proxemic behavior acted out within it. Computer programs, in fortran, to perform the necessary computations are available from the author. Two experiments performed using the methodology are described. In a validation experiment, proxemic behavior in a behavioral game played by pairs of experimental subjects within the device is shown to be non-random. Significant differences in proxemic behavior are observed in conjunction with differences in pretest scores on Rotter's I-E scale, Christie's Mach IV scale, and Schutz's FIRO eI and eA scales. Experimental subjects who engaged in easy mutual conversation during the game are shown to have used significantly closer interpersonal distances than those who did not. No significant differences in interpersonal distance are observed in conjunction with sex differences within the interacting dyads, their degree of similarity in physical appearance, or a weak acquaintance differential produced by having some dyads complete a pretest in the same room and some in separate rooms. A second experiment was performed to explore the interaction of interpersonal distance, eye contact, and verbal behavior and to continue the exploration of correlations between proxemic behavior and pretest scores on personality variables. Confederates varied the amount of talking and eye contact they engaged in with subjects during the experiment. In terms of interpersonal immediacy or intimacy, a compensatory effect was observed between the amount of eye contact and the interpersonal distance used by subjects. There was a clear reciprocity effect in the amount of talking between members of 'experimental dyads, and also weaker reciprocity effects in other immediacy behaviors. An analytical framework for the interpersonal immediacy behaviors based on a hypothesized interactive balance between compensatory behavior produced by information overload and reciprocal behavior in response to social norms of reciprocity is proposed.


3

TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES CHAPTER 1: AN INTRODUCTION TO PROXEMIC BEHAVIOR E. T. Hall's "Proxemics" Framework Intimate Distance Personal Distance Social Distance Public Distance The Nonverbal Immediacy Behaviors Methodologies Used in Proxemics Research A Representative Laboratory Study Two Representative Field Studies A Simulation Study Findings in Proxemics Research Studies of Spatial Intrusion or Invasion Interpersonal Relationships and Spatial Distance Individual Characteristics and Spatial Distance Interpersonal Distance and Other Behaviors Suggestions for Future Research CHAPTER 2: A BEHAVIORAL GAME METHODOLOGY Characteristics of One Such Methodology A physical Realization Some Specialized Statistics The Pattern on the Floor of the Device CHAPTER 3: A VALIDATION EXPERIMENT Procedure Hypotheses A Primary Null Hypothesis Other Hypotheses Experimental Subjects


4

Results Issues in Data Analysis Analytical Technique Attitudes Detailed Analysis of Selected Comparisons Sex Differences Verbal Interaction Acquaintance Similarity in Appearance Instrument Effects CHAPTER 4: AN EXPERIMENT Experimental Design Independent Variables Intervening Variables Dependent Variables Subjects and Confederates Description of the Experiment Hypotheses and Analysis Factorial Design Pretest variables--Chi-square Analysis Pretest variables--Regression Analysis CHAPTER 5: DISCUSSION Some General Impressions An Approach-Avoidance Paradigm Methodological Issues in Approach-Avoidance Analysis An Application of Approach-Avoidance Analysis Toward an Integrated Model of the Immediacy Behaviors1 A Review of Existing Theories An Integrated Model Research Findings in the Integrated Model Concluding Discussion APPENDICES: (Available from the author) Appendix A: Computer Programs Appendix B: Instruments-Validation Experiment Appendix C: Problems in the Development of a Single Criterion of Individual Behavior Appendix D: Instruments - An Experiment (Chapter 4) BIBLIOGRAPHY


5

LIST OF TABLES Chapter 3 1. Null hypotheses expected and observed frequencies 2. Probabilities that proxemic behavioral differences are significant (subjects categorized by attitude scores) 3. Comparison of subjects' proxemic behavior (subjects categorized by I E score) 4. Comparison of subjects' proxemic behavior (subjects categorized by Mach IV score) 5. Comparison of subjects' proxemic behavior (subjects categorized by FIRO eA and eI scores) 6. Comparison of mean interpersonal separations (subjects categorized by sex) 7. Comparison of subjects' proxemic behavior (subjects categorized by sex) 8. Similarities in proxemic behavior of subjects categorized by sex and F score 9. Comparison of mean interpersonal separations (subjects categorized by level of verbal interaction) 10. Comparison of mean interpersonal separations (subjects categorized by acquaintance) 11. Comparisons of mean interpersonal separations (subjects categorized by similarity in appearance) Chapter 4 1. Pattern of significant relationships observed in the factorial design 2. Interpersonal separations maintained by subjects with the two confederates 3. Interpersonal separations maintained by subjects under low and high eye contact conditions 4. Significant effects due to confederates' behavioral manipulations 5. Comparison of subjects' proxemic behavior (subjects categorized by I E score Phase 3 of

the experiments reported in Chapters 3 and 4)


6

6. Comparison of subjects' proxemic behavior (subjects categorized by I E score Phase 1 of

the experiments reported in Chapters 3 and 4) 7. A derivation of the expected behavior of subjects in Phase 1 8. Comparison of subjects' proxemic behavior (subjects categorized by I E and Emotionality

scores) 9. Significant results regression analysis on pretest variables LIST OF FIGURES Chapter 1 1. Arousal model of interpersonal intimacy 2 . Observation room Chapter 2 1. Floor plan of experimental apparatus 2. Probability distributions 3. Probability of separations 4. Comparison of "distance errors" Chapter 3 1. Floor plan of experimental apparatus 2. Expected frequencies of "attraction" and "avoidance" 3. Behavioral choices of subjects Chapter 4 1. A representation of conceptualized relationships among variables


7

Chapter 5 1. A comparison of the proxemic behavior of subjects labeled X and those labeled Y in the validation experiment 2. A tabular presentation of Figure 1 3. Illustration of gradients of approach, avoidance, and indifference in the validation experiment 4. Comparison of subjects' approach avoidance behavior in the validation experiment and the experiment reported in Chapter 4 5. A comparison of the approach avoidance behavior of subjects categorized by I E score 6. A minor revision of Patterson's arousal model of interpersonal intimacy 7. An analytical framework 8. Selected experimental results in an analytical framework


8

Chapter 1

AN INTRODUCTION TO PROXEMIC BEHAVIOR Studies of how people use physical space in interpersonal interaction most commonly use E. T. Hall's conceptual framework, which he calls 'proxemics." Hall has defined proxemics in various ways over the years, and a sample of definitions arranged in chronological order from early to more recent is provided below. . . . the study of how man unconsciously structures microspace - the distance between men in the conduct of daily transactions, the organization of space in his houses and buildings, and ultimately the layout of his towns. (Hall, 1963, p. 1003) . . . the study of the ways in which man gains knowledge of the content of other men's minds through judgements of behavior patterns associated with varying degrees of [spatial] proximity to them. (Hall, 1964, p.41) . . . the interrelated observations and theories of man's use of space as a specialized elaboration of culture. (Hall, 1966, P. 1) . . . the study of man's transactions as he perceives and uses intimate, personal, social and public space in various settings while following out of awareness dictates of cultural paradigms. (Hall, 1974, p. 2) The last definition above is, perhaps, closest to the specific concept of proxemics utilized in this research. This definition has three key components. First, proxemics involves the study of transactions, in the present research interpersonal interactions. Second, these interactions are viewed in a spatial context, which Hall has defined by four "zones" that he has termed intimate, personal, social, and public. Third, this behavior is considered to be largely learned or culturally determined rather than entirely dictated by innate biological or physiological processes. Although the purely proxemic aspect of interpersonal interaction the physical separation or "interpersonal distance" between people has been shown to vary significantly across cultures and in different situational contexts, proxemic behavior cannot, in general, be studied apart from other behavior in interpersonal interaction. Personal experience and common sense, supported by a growing body of research, suggests that information is transmitted between people in face to face interaction through a combination of verbal and non verbal behaviors, and that it is conditioned and filtered by the predispositions and attitudes that make up a person's internal state before meaning is attributed to it. The totality of the behaviors involved in interpersonal interaction may be bewilderingly complex, but a relatively small subset of them has been shown to be exceptionally significant. In one theory, Argyle and Dean (1965) proposed that interpersonal interaction involves a conflict between approach forces such as a desire for closeness, love, security, etc. and avoidance forces such as fear of dominance or rejection. They hypothesized that people will vary the amount of eye contact, physical proximity, and the topic of conversation, along with other behaviors such as smiling, etc. in an interactive fashion. Their theory, which has received considerable empirical support (Patterson, 1973), states that interacting individuals will vary these factors in a compensatory manner until approach and avoidance factors balance at a level of mutual comfort. This compensation process has been likened to a homeostatic or regulatory mechanism where the sum of the intimacy components of an interaction remains constant through a process of reducing certain of these "immediacy behaviors" when others are increased. Other theories have been proposed, and will be discussed later, but the three categories of


9

interpersonal behavior stressed by Argyle and Dean verbal, eye contact, and proxemic remain the most studied and seemingly the most significant. Another figure that should be mentioned in attempting to "set the stage" for a review of proxemic research is Erving Goffman. Although Goffman's work tends to resist categorization, observation of the spatial context in which interpersonal activities occur plays a significant part in his analysis. He utilizes a theatrical paradigm, distinguishing between "front" (public or onstage) settings where one performs for significant audiences and "back" (or offstage) settings where one rests and prepares these performances. Although the proxemic and kinesic aspects of various performances are regularly described by Goffman, there is little effort made to deal with them systematically. The most explicit consideration is given in Behavior in Public Places (1963) where he defines "body idiom" which includes "bodily appearance and personal acts: dress, bearing, movement and position, sound level, physical gestures such as waving or saluting, facial decorations, and broad emotional expression." One interesting, explicit contribution that Goffman makes to the body of thought about such behaviors is his observation of their "unfocussed" nature, and the potential for using them instrumentally rather than expressively. In every society these communication possibilities are institutionalized. While many such usable events may be neglected, at least some are likely to be regularized and accorded a common meaning. Half aware that a certain aspect of his activity is available for all present to perceive, the individual tends to modify this activity, employing it with its public character in mind. Sometimes, in fact, he may employ these signs solely because they can be witnessed . . . . Further, while these signs seem ill suited for extended discursive messages, in contrast to speech, they do seem well designed to convey information about the actor's social attributes and about his conception of himself, of the others present, and of the setting. These signs, then, form the basis of unfocused interaction, even though they can also play a role in the focused kind. (Goffman, 1963, pp. 33 34) Finally, it should be mentioned that proxemics, along with other related research, tends to concentrate on the effects of interpersonal spacing in the horizontal plane, more or less ignoring the significance of position in the vertical dimension. This omission has been noted particularly by Ball (1973). Identification of the vertical dimension with authority, leadership, social status, etc. permeates our language upper and lower classes, higher rank, superior/subordinate, etc., all have spatial connotations. This identification of height and dominance is said by Ball to be thought to have derived from the advantage that height affords in battle or personal combat, forcing the opponent to fight gravity as well as the person. Anecdotal evidence and examples of this perceived significance of height in popular literature are too numerous to mention, and Ball (p. 26) indicates that it is shared by most, if not all, cultures. E. T. Hall's "Proxemics" Framework Apart from his pioneering work in bringing the spatial characteristics of human action and interaction to the attention of social scientific researchers, Hall's major conceptual contribution to the study of behavior is his division of personal space into four zones. These zones are labeled, in order of distance proceeding outward from the person, intimate, personal, social, and public. Hall began his study of social space with a perspective derived from cultural anthropology, and more particularly from observation of how territoriality is elaborated and differentiated from culture to culture. (Hall, 1959) This, in turn, was stimulated by observation of territoriality in animals. However, in his later work he has taken an approach which emphasizes the nonverbal sensory


10

mechanisms involved in person perception, including visual cues, odors, perception of body heat and other cues which are often processed below the level of conscious awareness. Originally Hall identified eight social distances by observing variations in loudness and tone of voice associated with changes in distance. However, "further observation of human beings in social situations convinced [Hall] that these eight distances were overly complex. Four were sufficient. . . ." (Hall, 1966, p. 108) However, he does retain a vestige of his eightfold categorization by dividing each of his four zones into a "close" and "far" phase. (Hall, 1966, pp. 110 120) The characteristics of each zone are summarized below.

Intimate Distance This zone extends from zero to eighteen inches from the person, including a close phase (zero to six inches) and a far phase (six to eighteen inches). At intimate distance the presence of the other person is unmistakable and may at times be overwhelming because of the greatly stepped up sensory inputs. Sight (often distorted), olfaction heat from the other person's body, sound, smell, and feel of the breath all combine to signal unmistakable involvement with another body. (Hall, 1966, p. 110) Strangers are not invited into the intimate zone by adult, middle class Americans. Hall describes the typical response to such an "invasion" in a crowded bus or subway as "to withdraw if possible," or, failing this, to be rigid, tense, and immobile, and to fix the eyes on infinity and not to look at anyone except perhaps in a passing glance. Many of Hall's anecdotal comments about the intimate zone indicate that he believes that, for members of non contact cultures this sensory overload produces a threat to the perceiver's sense of individuality.

Personal Distance The personal zone extends from one and one half to four feet, with a near phase extending to two and one half feet and a far phase beyond that. "Personal distance" is the term originally used by

Hediger to designate the distance consistently separating the members of non contact species. It might be thought of as a small protective sphere or bubble that an organism maintains between itself and others. (Hall, 1966, p. 112)

Hall indicates his belief that this zone is largely defined by the possibility of physical domination by means of grappling with the arms. "Beyond it, a person cannot easily 'get his hands on' someone else." (Hall, 1966, p. 113) Elsewhere this zone has been described as one of "transition." The personal zone permits a range of contact between people, from relatively intimate to more formal. "It is a zone that people [i.e., Americans] use in public (perhaps less so in the close phase), and it seems to be a 'normal' contact distance that enables people to remain in reasonable proximity or to move toward more or-less personal communication." (Altman and Vinsel, 1977, p. 185) In the personal zone, although the number and range of potential communicative cues is still large, they do not seem to threaten the individuality of members of non-contact cultures as in the intimate zone. Thermal cues are not usually evident and most natural body odors are not easily detectable, except perhaps in the closest parts of the zone. Perfumes, colognes or deodorants may be detectable, if used, and fine details of the face and clothing can be detected visually with little perceived distortion.


11

Social Distance

This zone spans from four to twelve feet, the close phase extending from four to seven feet and the far phase from seven to twelve feet. The boundary line between the far phase of personal distance and the close phase of social distance marks, in the words of one subject, the "limit of domination." Intimate visual detail in the face is not perceived, and nobody touches or expects to touch another person unless there is some special effort. Voice level is normal for Americans. There is little change between the far and close phases, and conversations can be overheard at a distance of up to twenty feet. (Hall, 1966, P. 114) Social distance is typically used in business settings, among people who are working together, and in more formal social situations. Beyond this distance, communication requires noticeably more effort. The far phase is characterized as a distance in which the maintenance of eye contact becomes critical. At around twelve feet, feedback from the eye

muscles used to hold the eyes inward on a single spot falls off rapidly. . . . During conversations of any significant length it is more important to maintain visual contact at this distance than it is at closer distances.

Proxemic behavior of this sort is culturally conditioned and entirely arbitrary. It is also binding on all concerned. To fail to hold the other person's eye is to shut him out and bring conversation to a halt . . . . (Hall, 1966, p. 114)

Public Distance

The public distance zone extends from twelve feet on out, with a close phase from twelve to twenty five feet and a far phase beyond that. Hall defines the close phase in terms of self defense, stating that "at twelve feet an alert subject can take evasive or defensive action if threatened" (Hall, 1966, p. 116), and also in terms of sensory shifts that occur at a separation of about twelve feet. In discussing the defensive implications of this distance he speculates that it "may even cue a vestigal but subliminal form of flight reaction." (Hall, 1966, p. 116) The sensory shifts associated with this distance are speech which becomes loud although not shouting, along with a more careful choice of words and phrasing of sentences, and also changes in visual perception such that depth perception diminishes and fine details of the skin and eyes are no longer visible. The far phase is truly public. At this distance people tend to be perceived "in a setting" rather than the person filling the entire stage of attention. Hall states that "thirty feet is the distance automatically set around important public figures." (Hall, 1966, p. 117) Also subtle shades of meaning conveyed by the normal voice and details of facial expression are lost at this distance, and voice, gestures, and expressive movements must be exaggerated. The fourfold classification system described above (as opposed to six or some other number of zones) was chosen by Hall because, first, there is a need for some classification system in order to apply the mechanisms of scientific enquiry to any subject, and second, because it seems consistent with the observed phenomena and the basic theory of territorial behavior in humans and animals.


12

The . . . descriptions of the four distance zones have been compiled from observations and interviews with non contact, middle class, healthy adults, mainly natives of the northeastern seaboard of the United States. A high percentage of the subjects were men and women from business and professions; many could be classified as intellectuals. The interviews were affectively neutral; that is, the subjects were not noticeably excited, depressed, or angry. There were no unusual environmental factors, such as extremes of temperature or noise. These descriptions represent only a first approximation. They will doubtless seem crude when more is known about proxemic observation and how people distinguish one distance from another. (Hall, 1966, pp. 109 110)

The Nonverbal Immediacy Behaviors

A subset of the personal behaviors having communicative value has received considerable attention from empirical researchers. These are the "nonverbal immediacy behaviors," including physical distance, eye contact, body orientation, and body lean. As mentioned earlier, the first theory to specifically deal with the interaction of these behaviors was proposed by Argyle and Dean (1965) who suggested that, once an equilibrium level of intimacy has been reached in interpersonal interaction, an increase in one immediacy behavior will result in a compensatory decrease in others so that the original level of intimacy is maintained. However, a minority of reported studies do not support the compensation hypothesis. Breed (1972), upon finding that increases in a confederate's "search for eye contact" and forward lean were met by reciprocal increases from experimental subjects, proposed an alternative model. Basing his explanation on social exchange theory and the observation that "approving responses are normatively reciprocated in our society," he speculated that an increase in a confederate's immediacy behavior might put increased pressure on subjects "to return the approval (eye contact, forward lean) of the confederate." (p. 141) The compensation and reciprocity hypothesis were then reconciled by Patterson (1976) who proposed that responses to a change in the intimacy of interpersonal interaction by one person (person A) are mediated by changes in the arousal level of the other person (person B) and their emotional labeling of this change. The arousal model of intimacy is relatively simple, but it seems to have some advantages in terms of explaining and predicting diverse patterns of intimacy expression. By proposing arousal change as the necessary mediator in initiating adjustments in intimacy, it becomes possible to reconcile diverse patterns of results in this area of research. Instances in which manipulations of intimacy produce no changes in the target person may be due to the failure of those manipulations to effect arousal change. In suggesting that compensatory and reciprocal changes in intimacy share a necessary common mediator of arousal change, the specific course of the behavioral reaction becomes dependent on the nature of the labeled arousal or emotional state. (Patterson, 1976, p. 241) (Patterson’s model is reproduced as Figure 1 below.)


13

Methodologies Used in Proxemics Research

At the time of this writing (1981) there are well over 300 quantitative, empirical studies of human spatial behavior reported in the social scientific literature. In this section a few of these studies are briefly reviewed and summarized in order to acquaint the reader with some of the various methodological approaches used. This research is notable for its variety. Interpersonal distance has been used as either a dependent or an independent variable. Standing and seated interaction has been observed. Laboratory and field studies of actual behavior have been performed, as well as a large number of simulation studies where subjects are asked to position dolls or symbolic figures. In the simulation studies, the most common methodology has been Kuethe's (1962) "felt figure" technique reviewed below. The variables used in studies of proxemic behavior have been categorized by Altman (1975) into three classes: individual factors, interpersonal factors, and situational factors. These three classes are further specified as follows: Individual factors deal with properties of specific persons; biographical variables such as age and sex, cultural variables such as creativity and intelligence, personality factors such as need achievement and need affiliation, personal disorders and abnormality, and personal handicaps such as speech and physical disorders. . . Interpersonal factors refer to social relationships among people in regard to attraction, cohesion and linking, interpersonal influence, group composition (with attention to sex and ethnic variables), and group structural factors such as size and status . . . .


14

Situational factors deal with the general setting within which people or groups function, including physical factors such as seating arrangements, the formality of situations, and public versus private setting. (Altman, 1975, p. 66) Although Hall's hypothesized spatial zones are almost invariably cited in studies of proxemic behavior, a recent detailed review of the literature (Altman and Vinsel, 1977, p. 192) was unable to find even one empirical study that directly tests their validity. His hypothesized cultural differences in proxemic behavior have, however, been confirmed in quantitative studies. A Representative Laboratory Study One of the earliest of these was a laboratory study performed by Watson and Graves (1966) using Arab and American college students. Hall, in his impressionistic accounts of proxemic behavior, had often used examples showing the differences between Arabs and Americans. On the basis of these observations, Watson and Graves formulated three hypotheses: 1. Arabs will exhibit significant differences in proxemic behavior from Americans with Arabs

being closer and more direct in their proxemic behavior than Americans. 2. Within the group of Arabs, persons from any particular Arab culture will be more similar to

persons from any other Arab country in proxemic behavior than to Americans from any particular region of the United States, the direction of this difference being the same as for overall Arab American differences.

3. Similarly, Americans from any particular region of the United States will be more similar to Americans from any other region of the United States in proxemic behavior than to persons from any Arab culture, the direction of difference again being the same as overall Arab American differences. (Watson and Graves, 1966, P. 972)

Sixteen Arab and sixteen American male students from the University of Colorado were chosen as experimental subjects. The Arab students were chosen so that there were four each from Saudi Arabia, the United Arab Republic, Iran, and Kuwait; and the Americans were similarly chosen to represent four regions, New York New Jersey, Colorado, California, and the Midwest (Michigan, Illinois, and Wisconsin). Subjects were divided into pairs, always consisting of two Arabs from the same country or two Americans from the same geographical region, and told only to talk about anything they wanted (the Arabs in Arabic), and that they were going to be observed. Each pair was then directed to an observation room (Figure 2) where they were given one to two minutes to "warm up" before being observed over a period of five minutes. This procedure was followed until all possible pairs of subjects (six possible pairs from each group of four subjects from the same ethnic group and geographical area) had been observed and their behavior recorded using a system described below.


15

Hall (1963, 1974) has developed a system of notation for recording proxemic behavior based on eight categories:

(1) postural sex identifiers Identifies person's sex and whether he or she is sitting, standing, or prone. Since all subjects were seated and male, this category was not used.

(2) sociofugal sociopetal axis Scores the relation of the axis of one person's shoulders to that

of the other. Arabs were expected to face each other more directly.

(3) kinesthetic factors Scores interpersonal distance in terms of the potential for holding, grasping or touching each other. Arabs were expected to interact more closely.

(4) touch code Scores the amount of physical contact. Arabs were expected to touch each

other more.

(5) visual code Scores the degree of "visual contact." This includes all "looking at" the other person with eye contact scored as a 1, looking at the other person's head and face scored as 2, and so on. Arabs expected to score lower (have "more direct" visual contact) than Americans.

(6) thermal code Provides for the detection of one person's body heat by another. No such

detection was reported by subjects so this variable was not used.

(7) olfaction code Scores whether the odor of one person is reported by the other. None of the subjects detected any odors so this category was discarded.

(8) voice loudness scale Self explanatory. In this

a. experiment speech was recorded by a tape recorder b. equipped with a decimeter (for measuring loudness). c. Arabs were expected to speak more loudly than Americans.


16

Subjects were observed and their behavior in categories 2, 3, 4, 5, and 8 was recorded using Hall's notation. The scores were compared using standard statistical techniques and results were that each of the three original hypotheses was confirmed to a statistically significant degree in each of the five categories of behavior observed. This provides strong support for Hall's impressionistic observations. There were also significant regional differences among the American groups which appeared "worthy of further investigation." (Watson and Graves, 1966, p.984) Two Representative Field Studies

In addition to laboratory studies such as the one described above, a number of field studies of proxemic behavior have been reported. In one such study (Baxter, 1970), the distance at which people interacted was observed in natural settings. The settings chosen were four locations in the Houston zoo. Pairs of subjects (N=859 pairs) were studied, typically standing side by side looking at an animal. Observed dyads were categorized by ethnic group (White, Black, or Mexican American), sex (male male, male female, and female female), and age (children, adolescents, and adults). Relatively small but consistent and statistically significant differences were found in interpersonal interaction distances in each category. Mexican American dyads were found to stand closest together. (NOTE x's SHOULD BE "x-bars" IN THIS PARAGRAPH) (x=1.78 ft.), Whites were intermediate (x=2.29 ft.) and Blacks most distant (x=2.66 ft.). This effect accounted for about 32 percent of the sample variance. The second most powerful effect was due to age (about 10 percent of the sample variance). Children interacted most proximally (x=1.99 ft.), adolescents were intermediate (x=2.28 ft.), and adults at the greatest distance (x=2.46 ft.). Sex grouping also produced a significant, though relatively small (3 percent of variance) effect. Male female pairs interacted most proximally (x=2.23 ft.), and male male groups were most distant (x=2.39 ft.). All of these main effects were significant at a p<.001 level. Another field study, one of the earliest studies of proxemic behavior (even before Hall coined the term) was performed by John T. Gullahorn (1952) in an office where 29 women and 8 men worked for a large corporation. One section of the office where 12 women worked at individual non cooperative clerical tasks was observed. The women were seated at desks, arranged in three rows with four desks in each row, in one corner of the room. Each row was separated by file cabinets which the women could see over to speak with those on the other side, although this was more difficult than interacting with others within the same row. Subjects had been brought together as a single work group about 1 1/2 years before the study, prior to which they had worked as two groups in two separate offices. "This meant, of course, that some firm friendships had formed among girls in each of the two separate offices before the merger." (p. 124) Two weeks were spent observing interaction in this work group. Approximately every fifteen minutes during the working day an observation was made of which persons were engaging in conversation. Gross results were that "of 1558 interactions, 78.18 percent took place within the row in which the girl sat." (p. 126) The results of the gross interaction study are summarized by Gullahorn as follows:

(1) Interaction within the row is far greater than is interaction of each row with the other two.

(2) Interaction is more frequent between adjacent rows than between those separated by another row.

(3) Within each row, the girls interact more frequently with those seated nearest them

than with any others.


17

(4) With few exceptions, the frequency of interaction diminishes as the distance within the row increases and also as the distance by row increases. (p. 128)

In addition, friendship choices were determined in interviews with each worker. When the women were segmented by age, "the younger group (supported) the hypothesis that those who interact frequently tend to develop sentiments of friendship." (p. 132) However, friendship accounted for only a small part of the variance in the interaction data. Gullahorn's findings are in no way startling or contrary to common sense, but they do underscore the relative importance of proximity in human interaction. It is not safe to ignore the sheer fact of contiguity as a factor in interaction, despite the fact that many investigators might consider it an uninteresting variable in comparison with personality, friendship, and other such factors . . . . It was discovered that even where the conditions of work do not require cooperative effort the gross interaction rate among the employees was largely determined by distance. Friendship and business necessity were additional factors. (p. 123) A Simulation Study

Another stream of proxemic research involves the movement or placement of symbols rather than actual physical movement or body positioning on the part of subjects. "Simulation techniques have been the most popular procedure and account for almost half of the studies in the literature, perhaps because they are easy to administer and because they permit experimental control over situations." (Altman and Vinsel, 1977, p. 191) The most common methodology used in these experiments was developed by Kuethe (1962). Kuethe describes it as follows: A piece of blue felt, 2 yards X 2.5 yards was stretched on a wall of the experimental room. On each trial the subject was given two or more objects cut from yellow felt and was told to place them on the blue felt field in any manner he wished. The nap of the felt permitted the objects to cling wherever they were placed. The main advantage of the technique is that the objects may be placed anywhere on the field and with any orientation. When the objects are removed there is no mark left on the field that could influence future trials. (Kuethe, 1962, p. 32) Kuethe analyzed subjects' behavior in positioning objects on the felt field in terms of what he called "social schemas." Unit forming principles in social perception can be regarded as social schemas or response sets to the extent that they function to structure ambiguous situations involving human objects. When a person indicates that two objects "belong together" he has employed some schema or plan. If these objects are people or people symbols, the schema employed may be considered, by definition, a social schema. (Kuethe, 1962, p. 31)


18

He found that subjects given a set of felt "objects" which could typically be representations of a man, woman, child and dog or rectangles of differing heights, positioned them on the field in an organized manner. For example, subjects showed a strong tendency to order the placement of rectangles with the tallest at the left and in decreasing order of height as one moved to the right. Perhaps more interesting from the standpoint of proxemics was a second experiment. The second investigation was concerned with whether or not social schemas function in reconstruction situations. The two rectangles were placed 30 inches apart on the stimulus field. The first subject looked at this display for five seconds from a distance of 12 feet. The experimenter then took the objects down and gave them to the subject whose task it was to replace them exactly where they had been. Next, the experimenter measured the separation of the objects as placed by the subject. The man and woman figures were then placed on the field by the experimenter with a 30 inch separation and the procedure was repeated. The next subject started with the placement of the figures produced by the last subject. This sequence was continued until 30 subjects had reconstructed the displays produced by the subject that preceded them. For every other subject the order of presentation was reversed so that 15 subjects reconstructed the rectangle display first and 15 subjects reconstructed the man woman display first. (Kuethe, 1962, p. 36) Results were that when rectangles were used the separation remained fairly constant at about 30 inches. However, when the man and woman figures were used, successive subjects consistently placed them closer together until, after about the sixteenth subject, they were separated by about two inches. This spacing was consistently reproduced by the next fourteen subjects. Kuethe explained this result by postulating that the schema "man and woman 'belong together' induces errors of reconstruction." (1962, p. 36) Kuethe also speculated about the significance of atypical responses. Although these investigations were designed to study high commonality schemas, the clinical significance of idiosyncratic responses should not be overlooked. A person's organization of the human figures may be identified as idiosyncratic once the typical responses of other people are known. for example, a subject used the rectangles, one on top the other, to construct a high "wall" between the man and woman. This behavior attains significance because it contrasts with the strong tendency of most people to put the man and woman figures together. A response of this type might reflect a disturbance of the normal concept of the relation between male and female figures. Other responses made by subjects that could reflect disturbances of normal social thinking were grouping the non human objects to a greater extent than human figures and placing the child next to the man on the side away from the woman. (Kuethe, 1962, p. 38)


19

This speculation was confirmed by Higgins, Peterson and Dolby (1969), who attempted to account for varying familial schemata. It was predicted that normal, well adjusted males would place a figure of a son closer to that of a mother, while less well adjusted males would place the son figure closer to a figure of a father. Male undergraduates coming from intact families were subjects. Each was given a questionnaire to measure social adjustment and Kuethe's felt figure technique to determine familial schema. The stimuli figures were those representing a father, a mother, a son, and a sister. Subjects were asked to place them in any manner they wished. It was found that subjects with poor social adjustment placed the son closer to the father, while those with good social adjustment placed the son closer to the mother. The combined groups placed the sister closer to the mother than to the father. Additionally, poor social adjustment subjects placed a greater number of intervening figures between the son and mother than between the son and father, whereas the good social adjustment subjects more frequently placed intervening figures between the boy and father than the boy and mother.

Findings in Proxemics Research

The recent publication of an extensive, thorough, and generally well done review of proxemic research (Altman and Vinsel, 1977) is of great value to those interested in this area, but it presents a problem to anyone attempting to write on the subject in that it invites reliance on it to the possible detriment of other perspectives. Be this as it may, its thoroughness and superior organization of the material make it the almost inescapable basis for other reviews and commentary at this time. This section of the present work closely follows Altman and Vinsel. Their review was consciously limited to studies which measured "actual physical distance between real people, in either an independent or a dependent variable role." (p. 193) The most significant omission resulting from this decision was the simulation studies. However, although agreement between simulation studies and actual proxemic studies tends to be very good in many areas, the decision to omit them is justifiable. Given the relative newness of proxemic research and the early stage of development of both theory and methodology, it is possible that simulations tap other variables in addition to those relevant to proxemic behavior, and their inclusion could serve to further confuse the already complex behavioral issues. This concern is reinforced by findings demonstrating the importance of non verbal immediacy behaviors, which are not tapped in simulation studies. One recent paper specifies three requirements for methodological adequacy in the study of nonverbal immediacy behaviors, "1) that actual interactions be studied; 2) that each subject experience a change in immediacy; and 3) that special care be used (because of the difficulty of changing or programming these behaviors) if confederates are used to introduce differences in immediacy." (Bakken, 1978, p. 301) Studies included in the Altman and Vinsel review were organized in five categories, which have been mirrored in the four major headings below, the difference being that cultural factors have been included in "individual characteristics" here. This follows Altman and Vinsel's categorization scheme, but cultural factors were considered separately in their review because of the special significance attributed to them by Hall.

Studies of Spatial Intrusion or Invasion

Altman and Vinsel selected 30 "invasion" studies for review on the basis of methodological adequacy. All were conducted in field situations rather than in the laboratory. Twenty of the studies involved standing subjects with the remainder using seated subjects. Twenty one studied responses to being intruded upon with the rest dealing with a person's reactions to intruding on others. In terms of Hall's spatial zones, twenty of the twenty-one studies of intrusion involved invasions of the intimate zone.


20

The typical response to invasion in the intimate zone was movement away (including higher frequency and shorter latency of flight). Other studies showed decrements in performance, nonverbal compensatory reactions, or expressions of dissatisfaction or tension. Of the 13 studies that used some measure of moving away, all but one results in more or faster movement when intrusions occurred in the intimate zone compared with other zones or control conditions. And there is some evidence that females exhibit more negative reactions to intrusion than males, regardless of the sex of the intruder. Finally, people who were approached in the social zone typically did not respond with discomfort or annoyance. In fact, social distance conditions were generally used as control or non intrusion conditions. (Altman and Vinsel, 1977, pp. 209, 210) The studies of reactions to intruding on others were summarized as follows: First, when given a choice of going around or penetrating a group, or intruding on an individual, people avoided intruding on one person and bypassed the group, especially mixed sex groups, interacting groups, small groups, and high status groups. Individual characteristics also seemed to play a role in these studies. For example, mixed sex groups were not approached as closely as homogeneous sex groups, some times males were approached less closely than females, and at other times the opposite was true; conservatively dressed people were approached more closely than flashily dressed people. But even these variables played only a minor role in affecting distances; in most cases variations in approach distances were only in the neighborhood of a few inches. Distances between pair members also were important. For example, people were reluctant to pass between two or more people when they were at a distance no greater than approximately four feet (the edge of the social zone). Beyond four feet, intrusion rates increased. Finally, when forced to invade a dyad, people were quite uncomfortable, as evidenced by a lowered gaze, fidgeting and other nonverbal behaviors, and feelings of discomfort. (Altman and Vinsel, 1977, pp. 210, 211) In both "invader" and "invadee" studies, Altman and Vinsel found a need for "parametric studies that chart a range of reactions- flight, nonverbal behavior, attitudes, etc. at each zone and within zones in order to provide profiles or patterns of behavior. . . ." (p. 210) They point out that this would allow both a more systematic assessment of Hall's hypotheses about spatial zones and also make it possible to define zones inductively based upon behavioral profiles exhibited at various distances.


21

Interpersonal Relationships and Spatial Distance

Thirteen studies were included, of which eight used interpersonal distance as a dependent variable and six studied it as an independent variable (one studied both). Other variables were feelings of attraction and liking, closeness of relationship (friends, acquaintances, etc.), similarity of attitudes, sensitivity to rejection, and approval seeking. Eleven studies were done in laboratory settings, and ten used seated positions as a distance measure. In studies where subjects could choose their own seating position, almost two thirds of the average interaction distances were in Hall's personal zone (one and one half to four feet). The sample suggests the importance of two major types of variables: similarity and acquaintance/attraction. First, people use closer distances when interacting with others who are similar rather than dissimilar in age, sex, status, etc. Second, and most frequently researched, people use less space when interacting with people whom they like, with whom they have had positive experiences, or with whom they are better acquainted, such as friends versus acquaintances versus strangers. (Altman and Vinsel, 1977, p. 213) Interaction distances for close friends were near the outer edge of Hall's intimate zone, particularly for females, whereas subjects with more "distant" relationships tended to use the personal zone. "Negative affect" dyads used interpersonal spacings in the personal zone. Females tended to differentiate among friends, acquaintances and strangers more than males, who generally used somewhat greater distances. Females reported more positive feelings toward new acquaintances when forced to sit closer to them, while males reported more positive feelings at greater distances (i.e., the reverse effect). More formal settings produced greater interpersonal distances for both sitting and standing interaction regardless of the nature of the interpersonal relationship. People with more positive affect for each other tended to face each other more directly, and, when seating positions were forced, face to face orientations resulted in more positive affect than angled seating. These findings agree well with the results of simulation studies that were not included in the Altman and Vinsel review. The experimental data tends to fit fairly well in Hall's framework but further research is called for. The fact of differences as a function of seating standing, sex and sex composition, and other variables emphasizes the need for more systematic research to assess the contribution of such variables to distancing behavior. Also, as in the case of invasion studies, there is a dearth of research on the patterns or profiles of behavior that people exhibit under different conditions of acquaintance and attraction. We know little about use of verbal, nonverbal, kinesthetic, aural, olfactory, or thermal cues as interaction occurs with friends, acquaintances, etc. Also . . . the process of communication within various spatial zones is what is ultimately important, not the use of distances considered in isolation. At this time, studies of attraction/acquaintance and distance have only tapped the surface and are largely of a "demonstration" variety. We now need more parametric studies that weave together distancing


22

behavior (1) with related factors, such as standing seating, sex, and other relevant variables, and (2) with other levels of behavior; for example, verbal, nonverbal, and kinesthetic channels). (Altman and Vinsel, 1977, p. 218)

Individual Characteristics and Spatial Distance

Twenty eight studies of this topic were examined by Altman and Vinsel. All but one used interpersonal distance as a dependent variable, and all used individual factors (personality characteristics, biographical or demographic variables, and social roles, such as status or position in an organization) as independent variables. Twenty of the twenty eight were laboratory studies, and twenty three used standing interpersonal distance rather than seated distance. Results indicated that "normative" interpersonal distances for people when standing were at the outer limit of Hall's intimate zone extending to the near phase (one and one half to two and one half feet) of the personal zone. The variation produced by individual factors tended to be small, less than one foot. Altman and Vinsel argue fairly convincingly that the dominant personal factor in interpersonal spacing is the perceived relative power of those involved in the interaction. In summary, research on individual characteristics and personal space suggests a framework based on the power, confidence, and control properties of the subject and the other participant. Low power in the subject is associated with increased distance from others, while confidence or power is associated with a willingness to be physically closer to others. Also, larger distances are maintained from a more threatening target person with whom one is interacting. These data correlate nicely with attraction and intrusion studies. That is, a positively valued person is approached more closely than a negatively valued person in the case of attraction studies, and a stranger who comes overly close is presumably threatening and is avoided, as shown in intrusion studies. (Altman and Vinsel, 1977, p. 223) Cultural factors are categorized as individual factors by Altman and Vinsel, but they are singled out for special attention because of Hall's emphasis on cultural differences. Eleven studies were selected for review as dealing specifically with cultural variables. Culture was an independent variable in all studies and nine studies used standing rather than seated subjects. Five used laboratory rather than field settings. Based on their review, they were not as persuaded of the importance of cultural factors as Hall has been. Although based on limited data, our sample showed some support for the idea that so called contact cultures for example, Mediterranean, Latin American, and Arabic societies use closer spatial distancing than other cultures. On the other hand, there is inconsistant evidence regarding the distances used by ethnic groups in the United States. If anything, the data suggest that socioeconomic and related variables are more important determinants of proxemic behavior than ethnicity. (Altman and Vinsel, 1977, p. 251)


23

Interpersonal Distance and Other Behaviors

This research is almost entirely informed by Argyle and Dean's affiliative conflict theory described earlier in this paper. Twenty-six studies from twenty three articles were selected for review by Altman and Vinsel. Twenty four of these were laboratory studies and twenty one involved seated interaction. Interpersonal distance was treated as an independent variable in twenty two of the studies. Hall (1966), Argyle and Dean (1965), Patterson (1976), and Altman (1975) have all described the "systemlike" interaction of the various modes of sensory perception, verbal behavior, and non verbal behavior in interpersonal interaction. Seventeen of the studies examined the relationship between "looking" behavior and distance. Sixteen confirmed the Argyle and Dean (1965) hypothesis that the amount of eye contact (and other looking behaviors) should increase with distance. Seven studies examined other facets of nonverbal behavior in relation to distance and generally found that the closer the distance between people, the more they exhibited behaviors reflecting discomfort and/or compensatory reactions. In this sense, these studies are congruent with both the intrusion research described earlier and the equilibrium hypothesis. For example, several studies found that close distances yielded less direct body orientations and more signs of discomfort, such as leaning backward and scratching the head. But these results were often complicated by interactions with other factors, such as friendship and sex composition. Four studies examined the relationship between distance and psychological states: feelings and attitudes about the self or the situation. Surprisingly, three studies found positive feelings with greater distances between participants. These studies are difficult to evaluate because of their unusual characteristics. For example, the single positive relationship between distance and comfort involved a difficult performance situation, and people were most comfortable 32 feet from the experimenter. On the other hand, two of the three studies that found negative feelings at increased distances involved groups sitting on the floor, again an atypical setting. (Altman and Vinsel, 1977, p. 238)


24

Suggestions for Future Research Criticism of proxemic research has focused both on theory and methodology. Baldassare and Feller (1975) have criticized Hall's conceptualization of the process by which patterns of individual proxemic behavior develop, and they suggest that "the process by which proxemic behaviors are acquired, extinguished, and maintained within cultures should be thoroughly examined throughout the (individual) life cycle." (p. 495) In addition they see a need for increased understanding of how situational features (both physical and social) affect the use of personal space. They criticize the "one shot hit and run" (p. 496) approach to the study of space use, and suggest that larger, more elaborate, multivariate studies are necessary. These and other suggestions from within the Hall tradition have been summarized and elaborated by Altman and Vinsel. Some of their more specific suggestions have already been incorporated into the previous section of this paper, and their general comments are as follows: 1. The most critical requirement is for parametric studies of distancing in relation to other behaviors. That is, the idea that distance serves as a medium for various communication channels should be researched in a systematic manner. We must better understand which combinations or profiles of behavior are employed at various distances and how these shift over distances. Such parametric studies of a multivariate, profile like nature should not only identify patterns of behavior, but they should assess the relative contribution of different channels at different distances . . . . 2. Another important direction of future research concerns systematic analyses of standing seated arrangements of participants. Existing data show gross differences in distance as a function of standing sitting, but there is little information available about (a) differences in patterns of behavior for the standing and seated positions and (b) the relationship between other variables and distance, such as attraction and intrusion, in standing and seated arrangements . . . . As a related issue, our analysis revealed the noncomparability of information across topical areas. For example, the intrusion/invasion studies primarily used standing arrangements, attraction studies involved seated participants, and so on, making it difficult to compare across or within variables classes. 3. Another topic requiring more systematic study is sex and sex composition . . . . 4. Several lines of work are also necessary on culture in relation to spatial behavior. First, a greater volume of studies is necessary, since there are only a handful of studies tapping Hall's ideas. Second, the range of cultures to be studied should be extended to include more than Arabic, northern European, Latin American, and ethnic groups in the United States. Third, additional research is needed on


25

sex factors in relation to culture and distance. Fourth, it is crucial to understand profiles of behavior in different cultures. That is, what similar and different patterns of verbal, nonverbal, distancing, visual, auditory, etc., channels do various cultures use at different distances? 5. Implicit in several of the proceding points was the idea that future research on spatial distancing would profit by analyses of interactions across categories for example, sex X culture, personality X attraction, and personality X intrusion in relation to spatial distancing. That is, having demonstrated the relationship of single variables to spatial behavior, we should now study important variables in combination with one another. (Altman and Vinsel, 1977, pp. 252, 253) Other criticism and suggestions have been directed toward proxemic research from researchers operating outside the Hall framework. Undoubtedly the most persistent and prolific of these critics has been Gary W. Evans (Evans, 1972, 1973, and 1978; Evans and Eichelman, 1976; Evans and Howard, 1972 and 1973). Evans laments the relatively large number of non supported findings and suggests greater reliance on objective rather than self report measures in proxemic research. Evans (1972), Evans and Howard (1972), and McBride, King, and James (1965) have reported consistent changes in subjects' GSR related to approach by another person.


26

Chapter 2

A BEHAVIORAL GAME METHODOLOGY

The suggestions for further research reviewed in the last section of Chapter 1, when taken together, indicate that it might be useful to have a standardized methodology for proxemics research. Ideally such a methodology would permit researchers to conveniently perform a variety of meaningful experiments that could be standardized and administered to diverse populations of subjects from various cultures, at different age levels, etc. so that the proxemic behavior of various groups could be more readily compared and differences could be interpreted with less ambiguity. It should permit researchers to transcend the "one shot hit and run" approach that, according to Baldassare and Feller (1975), has characterized previous research. It should also encourage a larger volume of studies by facilitating large systematic research projects and/or permitting the results of smaller projects to be integrated into a standardized body of data on proxemic behavior. An analogy with the use of the Skinner box in animal research has some unfortunate connotations, but may be apt. The Skinner box, while not appropriate for all types of such experiments, has permitted the relatively rapid accumulation of an extensive data base in animal research that is readily verifiable and that provides a basis for further research in many areas. If a similar methodological approach could be devised for research with humans, particularly one that did not present subjects with a completely sterile and dehumanized experience, then there is reason to believe that it might prove valuable.

Characteristics of One Such Methodology Given the nature of the problem, one approach is to consider the possibility of creating a readily reproducible physical setting having features that would facilitate proxemic research, much as a Skinner box is a reproducible physical setting having features that facilitate certain types of animal research. Such a setting must have certain characteristics, a few of the most critical of which are listed below.

1. It must be fairly large physically. Subjects should have enough space available to maintain any desired interpersonal distance. In general one would expect his to encompass a range from Hall's intimate zone (zero to 18 inches) at least through the social zone, which extends to 12 feet.

2. it should permit all sensory modalities of subjects to operate without obstruction. (See Hall, 1966, pp.118, 119)

3. Environmental and situational factors should be standardized or controllable in as much as this is possible.

4. There should be enough potential variability possible in the proxemic behavior of subjects to allow individual differences to manifest.

5. It should lend itself to analysis. In addition, the large physical size of the required setting necessitates, as a practical matter, that it cannot be a permanent structure since the cost of constructing and maintaining it would then be prohibitive.


27

A Physical Realization

These requirements led the author to consider whether a suitable setting could be devised in which people could play "behavioral games" that could be analyzed to gain insight into their proxemic behavior. The result of these (ACTUALLY TO USE PROXEMIC BEHAVIOR TO GAIN INSIGHT INTO OTHER FACETS OF THE PERSON) considerations could be termed an "instrument" for the study of proxemic behavior. This instrument is essentially a portable room with a pattern inscribed on the floor which has spatial and statistical properties that permit precise analysis of individual or interpersonal proxemic behavior acted out upon it according to rules determined by an experimenter. The rules, in general, specify a "game" that can be played by one or more experimental subjects and/or confederates within the room. The room has a floor plan in the shape of a regular hexagon measuring 7.5 meters (approximately 24.6 feet) across its largest dimension. The walls of the room are made of opaque white plastic sheeting and are supported by a collapsible framework. Each wall (i.e., side of the hexagon) is a separate plastic sheet which overlaps the adjacent wall by about two feet. Entrance to, or exit from, the room is gained at any of the six corners by pulling the curtain walls to the side. Once experimental subjects are in the room, the curtain walls fall together to create a symmetrical and featureless barrier to extraneous ambient visual stimuli. The walls serve two functions. They eliminate doors or other environmental features that might influence behavior within the device, and they create a setting that can be assumed to be equally unfamiliar to all experimental subjects. The floor of the room is a circular, blue, reinforced polyethylene sheet approximately 27 feet in diameter. The pattern on the floor is laid out with white cellophane tape 19 millimeters (3/4 inch) wide.1 The entire room is portable and can be set up or taken down by two people in about an hour. It is designed to be set up indoors in a larger room as shown in Figure 1. Portability is an advantage because it allows an essentially identical setting to be recreated in a variety of locations, the only requirement being a room large enough to accommodate it. When not in use the disassembled device can be stored in a medium sized closet (about 15 square feet). _____________________ 1 A patent for the device described has been granted to the author. Anyone wishing to use this or a similar system should contact the author to obtain permission or to insure that there is no infringement of patent rights. ____________________


28


29

The first three of the requirements specified earlier are met by the overall size and shape of the hexagonal room. The last two are met by the pattern of lines on the floor of the device. The points on this pattern where two or more lines intersect are referred to as "nodes" for convenience. An experimental run within the device could be described as a "game." In general, a game within the device would be "played" by one or more experimental subjects initially positioned by the experimenter so that they are standing on certain nodes; for example, those labeled SX and SY in Figure 1. The experimenter would then describe the rules of the game, which are the rules that subjects must follow in moving from node to node. In the simple game that was used in the experiments described in this report, subjects can only move when instructed to do so by the experimenter, and can only move from the node on which they are standing to any adjacent node. An adjacent node is any node connected to the node a subject is occupying by a line, but not separated from it by another node. The distance between adjacent nodes is always 0.5 meter (approximately 19.7 inches). It can be seen by inspection of Figure 1 that, unless the subject is standing on a node at the edge of the floor pattern, in which case the number of possible moves is reduced, he or she will have six possible moves to choose from each time they are instructed to move. To provide a simple example and avoid the complications introduced by "edge effects," suppose that a subject is initially positioned on the node at the center of the pattern and makes five moves as described above. The total number of possible five move paths is then 65 or 7776, indicating that even for relatively short games enough potential variety is available to satisfy requirement 4.

Some Specialized Statistics

Requirement 5 provides the primary justification for using the game format described above for the study of proxemic behavior. A number of useful specialized probability distributions can be generated specifically for use with the device, and standard statistical techniques can be readily employed. Both these specialized distributions and standard statistical techniques are used in the analysis of the experiments that are reported later in this dissertation, and an example may aid the reader to see how the specialized distributions are generated. If, as a null hypothesis, it is assumed that subjects are making moves in a purely random manner, it is possible to determine the probability that a subject will reside on any node after any number of moves. This probability distribution, in the basic game described above, depends solely upon the subject's initial position and the number of moves. Two sample distributions for two subjects beginning the game in the initial positions shown in Figure 1, for a two move game, are illustrated in Figure 2. A computer program, developed by the author, that generates probability distributions similar to those shown in Figure 2 is described in Appendix A to this dissertation ("Program 1"). Further, since the probability of residing on any node is independent of the probability of residing on any other node, the probability of a subject residing on any set or combination of nodes is simply the summed probabilities associated with each individual node in the set.


30

FIGURE 2. Probability distributions after two moves for two subjects who started on the nodes marked "X" and "Y". Probabilities are computed by noting that there are six equally probable moves from any node to the adjacent nodes, making the probability 1/6 that any path will be chosen at random. For example, since there is only one path of two moves by which subject Y can reach node B, the probability is 1/6 X 1/6 = .028. Probabilities are assigned to the other nodes in a similar manner.


31

Second, ignoring boundary effects occurring at the edge of the pattern (which mildly complicate the computations, but do not affect the basic analysis), each node is surrounded by a "ring" of Six adjacent nodes arranged in a hexagonal pattern. Furthermore, this hexagonal pattern of nodes is surrounded by another concentric hexagon of 12 nodes, each being one move "outward" from the nodes forming the first hexagon. This second concentric hexagon is surrounded by a third hexagon of 18 nodes, and so on until the boundary of the pattern is reached. These concentric hexagons are significant because a distinct minimum number of moves is required to reach each one from the node chosen as the initial position. This minimum number of moves is also a fairly good approximation of the actual physical distance between the initial position and any node on the appropriate hexagon.2 ____________________ 2 Although each node on any concentric hexagon is the same number of moves from the node chosen as the initial position, those nodes near the center of one of the sides of the hexagon are physically closer to the initial node than those at the corners of the hexagon. This "distance error" is significant, but is minimized by the geometry chosen as will be shown later. ____________________ Combining these two observations, one can see how the game described above can be analyzed to determine whether the subjects' physical behavior indicates attraction to (or avoidance of) any node on the floor. Suppose that, after two moves, subject X is standing on the node marked A in Figure 2, and subject Y is on node B. Further suppose that we want to know whether the subjects' behavior indicates that they are "attracted" to the center of the device. Node A is four moves from the center. The probability that subject X will be less than or equal to four moves from the center, assuming random movement, is the sum of the probabilities attached to all nodes four moves or closer to the center. The sum of these probabilities is p = .811. This is interpreted as a probability of .811 that subject X's "attraction" to the center is due to chance.3 However, subject Y is only one move from the center. There is only one possible node with this property on which this subject could reside, and the probability associated with this node is p = .028. 3 The majority of probabilities possible in a two move game are not statistically significant at p<.05 As the number of moves is increased, significant outcomes also increase,as does the computational complexity. The two move game is illustrated only as an example. Actual games would have more moves, and computer programs have been written to generate the required probability distributions. Also, typical experiments combine many experimental runs so that even minor departures from random movement, if repeated systematically and frequently enough will, when properly analyzed, be statistically significant. _____________________ Perhaps more interesting, it is possible to determine the probability associated with the two subjects' separation from each other. Since this separation is a measure of their behavioral attraction to or avoidance of each other, the associated probability is a measure of whether any observed attraction or avoidance is due to chance. The required computations are somewhat more difficult than for the statistical measures described above, since they involve determining the joint distribution for the distance between subjects. To see how this is done, think for a moment about the most extreme possible case of attraction, which would occur if subject X and subject Y in Figure 2 were two moves apart. This can only occur when each subject is on one unique node, that closest to the center.4 (Note that for subject Y this is the node labeled B.) The probability of residing on this node is .028 for each subject so the joint probability is .028X.028 = .001. ____________________ 4 This also illustrates a potential difficulty in interpreting behavior in the device. In this particular instance the behavior could be interpreted either as mutual "attraction" between the two subjects or as both subjects being "attracted" to the center of the instrument. ____________________


32

Similarly, one can compute the probability that the two subjects will be separated by three moves. There are eight possible combinations of positions for subjects X and Y that are separated by three moves in Figure 2. These are indicated by the curved lines in Figure 3. It can then be seen that the associated probability is (6 X .028 X .056) + (2 X .0562) = .016, and the probability of a separation of three moves or less is .016 + .001 = .017.


33

FIGURE 3. The eight possible ways that the two subjects in the example discussed in the text could be separated by three moves after completing two moves are illustrated by the curved lines. The summed probabilities are p= .0155.


34

The probabilities associated with other separations can be computed although the computations are difficult even for a two subject, two move game. However, the computational rules are relatively simple and the problem has been computerized. The computer program developed to perform this analysis is described in Appendix A (“Program 3”).

The Pattern on the Floor of the Device

A final explanation is, perhaps, in order before describing the experiments performed in the device. Why was the particular pattern on the floor of the device chosen? The regular pattern of equilateral triangles that covers the floor is known as a tesselation. Tesselations have been described as follows:

Imagine that you have an infinite supply of jigsaw puzzle pieces, all identical. If it is possible to fit them together without gaps or overlaps to cover the entire plane, the piece is said to tile the plane, and the resulting pattern is called a tessellation. (Gardner, 1975, p. 112)

The question that must be answered is, "What is the best tessellation for our purposes?" The problem seems formidable at first blush, since there are an infinite number of shapes that tile the plane. For example, with a little fiddling with pencil, paper, and scissors, one can convince one's self that any triangle or quadrilateral will tile the plane. However, a closer study of the requirements of a behavioral instrument simplifies the problem dramatically. Two requirements can be identified. First, the distance between all adjacent nodes should be equal. Among other things this justifies the assumption, used in the statistical analyses described above, that the probability of moving to any adjacent node is the same as that of moving to any other adjacent node. If the distances between nodes were not equal, the problem of assigning a probability to any move between adjacent nodes would be complicated by possible preferences of subjects for "long" or "short" moves. Since not all subjects would be likely to share the same preference ordering, the assignment of probabilities to various longer or shorter moves would, itself, involve a probability distribution, thereby greatly complicating the interpretation of behavior in the instrument. A second requirement is that the "game distance" measured in "moves" between any two nodes should approximate the straight line distance (measured in arbitrary units one "move" long) as nearly as possible. The first condition, equal distances between all adjacent nodes, immediately and by definition limits the shapes of the "tiles" or pieces that can be considered to the regular polygons. Fortunately, the ancient Greeks have proved that only three of the regular polygons, the equilateral triangle, the square, and the regular hexagon, tile the plane. Since a tesselation of regular hexagons can be obtained from the pattern of equilateral triangles that was chosen for use by eliminating certain nodes, this alternative can be rejected immediately as offering no advantages. In this manner, by applying the first condition, the possibilities for suitable tesselations have been reduced from an infinite number to two, the familiar checkerboard pattern of squares or the pattern of equilateral triangles chosen for use. The problem is now reduced to deciding between these two. To do this the second condition, that "game distance" in moves between any two nodes approximate the straight line distance as nearly as possible, is invoked. It is convenient to define a term which can be called "distance error" as a measure of how closely distance measured by the number of moves separating two nodes approximates the physical distance between the two nodes under "worst case" conditions. The method used to determine distance error is illustrated in Figure 4. (NOTE THAT IN THIS PARAGRAPH THE "DEGREES" SIGN IS REPRESENTED BY A ") It can be seen by inspection that the maximum distance error for either tesselation occurs in the conditions illustrated in Figure 4. For both tesselations, movement between the two nodes connected by the solid line and the two nodes connected by the broken line each requires two moves. For the tesselation consisting of equilateral triangles, if the distance between adjacent nodes is considered one unit, then the solid line is two units long and the broken line is 2 X cosine


35

30" units in length. The distance error is then computed as 2/ (2 X cos. 30") = 1.155. For the checkerboard tesselation the length of the solid line is still two units, but the length of the broken line is now secant 45" units. The distance error for the checkerboard tesselation is 2/ secant 45", or, since the secant equals 1/ cosine, 2 X cos. 45" = 1.414. Since the distance error inherent in the checkerboard tesselation is substantially greater than that in the tesselation chosen for use, it has been shown that the tesselation chosen for use is the best possible, given the requirements that were established.5 Although distance error could be a serious problem in some types of experiments, its effects are somewhat mitigated because it exists only at separations greater than one move. At separations greater than one move, the actual physical distance can vary between 1.0 and .866 times the length of a move (.5 meter) times the separation in moves. Because of the potential difficulties associated with distance error, and also to make the initial reduction of the data obtained with the instrument described above less laborious, a computer program was developed. This program displays the positions of experimental subjects (and/or confederates) during the game, computes their interpersonal distance in both moves and meters (i.e., actual distance), and categorizes their moves. It is described in Appendix A ("Program 4").

FIGURE 2. Method used to compare "distance errors" in the checkerboard tesselation and the tesselation chosen for use. Note that, for clarity of presentation, the lines forming the nodes in the actual device have been omitted, and the nodes are represented by dots. Note particularly that the nodes joined by the broken lines are not "adjacent" as would be apparent if the lines forming the checkerboard of the pattern of equilateral triangles were present. ___________________ 5 For the sake of brevity the above discussion of tesselations omits some of the finer points. Interested readers are referred to Martin Gardner's (1975) article. ___________________


36

Chapter 3

A VALIDATION EXPERIMENT

The review of findings in proxemic research presented in Chapter 1 suggests that an exploratory study of proxemic behavior using the methodology described in the previous chapter should attempt to assess the significance of the factors listed below.

1. Personality characteristics of experimental subjects related to power, confidence, and control.

2. Acquaintance. 3. Similarities or differences apparent in the physical appearance of subjects, such as their

sex, age, and style of clothing. 4. The relationship among proxemic behavior, verbal behavior, and eye contact.1

In addition, the behavior of people in the device should be analyzed to assess any confounding effects that might be produced by the physical characteristics of the apparatus. Since this methodology presents subjects with an environment that is quite different from the naturalistic settings used in most prior proxemic research, a very simple experiment seemed to be the most appropriate strategy for an initial exploration of its characteristics. ______________________ 1 Eye contact was not observed in this experiment, but was incorporated in the design of the experiment reported in the next chapter. _____________________ Accordingly, it was decided that, for each experimental run, two naive subjects would be positioned in the device on the nodes marked SX and SY in Figure 1 in Chapter 2 (reproduced here for convenience). Each would then be asked to make twelve moves, the two subjects alternating moves, with no constraints placed on their choice of how to move.


37

FIGURE 1. A floor plan of the experimental apparatus as set up for the experiment described in this report.


38

Procedure

The following experimental procedure was used. Subjects were first given a pretest. This consisted of Rotter's (1966) I E (internal vs. external locus of control of reinforcement) scale with the six "filler" items deleted, a forced choice version of the F (authoritarian personality) scale (Christie, Havel, and Seidenberg, 1958), the Mach IV (Machiavellianism) scale (Christie, 1973), and Schutz's (1958) FIRO B scales. A copy of this instrument and the other forms completed by participants in this experiment is presented as Appendix B of this dissertation. Subjects were given 30 minutes to complete the pretest, with a few subjects given an extra 5 or 10 minutes. Half of the dyads (3 male male, 4 male female, and 3 female female) completed the pretest together, sitting at the same table in a small room, and the other half completed the pretest in separate rooms. This created a slight acquaintance differential. After completing the pretest, subjects were led to another room where the "game" was set up as shown in Figure 1. Once inside the device, one subject was designated "Subject X" and the other "Subject Y", and they were positioned on the nodes labelled SX and SY in Figure 1. (The nodes were not labelled in the actual device.) The experimenter then moved to the center of the device, and after demonstrating which nodes were adjacent, read the following instructions. We are interested in how people use physical space in interpersonal interaction, and to help us study this we want you to play a simple game. There is no 'point' or 'goal' to the game other than this. You are now standing on a point where a number of white lines on the floor come together. We refer to these points as nodes. After reading these instructions and answering any questions you may have, the experimenter will leave the apparatus and you will hear a number of recorded statements as follows. "Subject X, move 1 please," followed by a fifteen second pause, then "Subject Y, move 1 please," followed by another fifteen second pause, then "Subject X, move 2 please," and so on. Each time you are requested to move, you can move to any node adjacent to the one you are on. You must move. You cannot remain where you are. Unless you are at the edge of the pattern on the floor you will always have six adjacent nodes to which you can move. If you are at the edge of the pattern you will have less than six, but the principle is the same. That is, when requested to move you can move to any adjacent node. You may both occupy the same node if you wish. You can do this by standing close together so that you are physically touching while you are both standing as close to the "target" node as possible. The total number of moves that you will be asked to make is unknown, but will be 25 or less.2 The experimenter will observe and record your moves from outside the apparatus. You may talk to each other during the game if you wish. Your conversation will be analyzed as part of the experiment. You may talk about anything you wish, the game in general, how you feel, etc., except you are forbidden to discuss your actual moves or those of the other person. You may not tell the other person where you are going to move or discuss or attempt to


39

influence his or her moves in any way. After the last move, the end of the game will be announced. At this point remain where you are until the experimenter returns to lead you out of the apparatus. Do you have any questions? ____________________ 2 The actual game always consisted of twelve moves for each subject. Subjects were not told the actual number of moves in order to avoid the possibility of a "horizon effect" where subjects might modify their behavior in some way in expectation of the end of the game. ____________________ After answering any questions, the experimenter left the device and switched on a recorder to play the prerecorded "move" instructions. These instructions were played through three speakers positioned as shown in Figure 1 to minimize any possibility that subjects might be influenced by the direction from which the instructions were heard to come. Three 2 x 2 ft. plastic one way mirrors were located in the walls above the speaker locations, and the experimenter recorded the physical position of each subject after each move, as well as their verbal behavior (i.e., whether each subject spoke during each fifteen second interval between moves) by observing through one of these windows from outside the device. After the experiment was completed, subjects were debriefed and paid. None of the subjects reported any unusual anxiety during the game, although a number of them were quite curious about it.

Hypotheses

A total of five hypotheses were generated to determine whether non random proxemic behavior does indeed occur when people are asked to play a game in the apparatus described, and to assess the effects of the factors listed in the introduction to this chapter.

A Primary Null Hypothesis

Probably the most important single requirement to demonstrate the potential usefulness of the methodology is to show that the proxemic behavior of people in the device is non random. A null hypothesis, which must be disproved if the methodology is to be useful, is formally stated as follows.

1. In a game of any number of moves, each subject will make each move in a random fashion.

This hypothesis can be tested using a chi square test. This method allows a table of expected (theoretical) frequencies of events to be compared with their observed frequencies, and allows a determination of the statistical significance of any differences. It is desirable to analyze the 12 move game at three move intervals, that is, as a 3 move game (first three moves), a six move game, etc. Allowing both experimental subjects to make two moves between observations provides assurance that the observed interpersonal separations are not constrained by the mechanics of the game in such a way that they are dependent upon each other, since independent observations are required. If only the final positions after a twelve move game were analyzed, the subjects may have, for example, moved close together and then separated during the game. By analyzing the game at three move intervals it should be possible to pick up these "dynamic" interaction effects. This multiple analysis can be justified because experimental subjects are not told how many moves the game will consist of. Therefore there is no reason for them to assume that any particular move will or will not be the last and to plan their moves accordingly (i.e., the "horizon effect").


40

The most serious constraint on the use of this technique is a requirement that, for chi square tables larger than 1 x 2 (one degree of freedom) a minimum expected frequency of 5 in the least possible cell is desirable. This precludes the use of the probability distributions generated by Program 3 (see Appendix A) "as is" each distribution being a 1 x 13 table where each probability for each separation in moves constitutes a cell in the table. To obtain an expected frequency of 5 in the least probable cell would, it turns out, require a prohibitive number of experimental runs (233,645 to be exact). Therefore, to reduce the required number of runs to a practical level, the theoretical probability distributions must be "compressed" or "collapsed", by summing probabilities, into the four 1 x 3 tables shown in Figure 2. Happily, the cells in these tables are meaningful as well as practical. That is, "Attraction", "Neutral/Minimal Avoidance", and "Avoidance" represent fairly closely the actual behaviors we wish to observe. The complete set of four 1 x 3 tables, giving probabilities and expected frequencies (in parentheses) assuming 20 experimental runs,3 is presented in Figure 2. FIGURE 2

Other Hypotheses Five other hypotheses can be stated as follows, based upon prior findings in proxemic research.

2. Personality characteristics. Subjects scoring high on pretest scales reflecting power, confidence, and control will attempt to use closer interaction distances (i.e., will make a larger number of moves reducing the interpersonal separation), than subjects scoring low on these scales.

3. Sex differences. Female female pairs (n=6) of experimental subjects will use closer

interpersonal separations than will female male pairs (n=8), who will use closer separations than male male pairs (n=6).

4. Verbal interaction. Pairs of subjects who engage in a high level of verbal interaction will

use closer interpersonal separations than pairs who engage in a low level of verbal interaction.

5. Acquaintance. Pairs of subjects who have completed the pretest questionnaire in the

same room (acquaintance condition) will use closer interpersonal separations than pairs who completed the pretest in separate rooms (no acquaintance).

6. Physical similarity. Pairs of subjects rated as being globally more similar in physical

appearance will use closer interpersonal separations than those rated less similar.


41

____________________ 3 Note from Figure 2 that the number of runs required for a minimum expected frequency of 5 is 5 divided by .2439 = 20.5 or approximately 20. ____________________ Two statistical techniques are used to test these hypotheses. The analysis, along with a rationale for preferring one technique over another in specific instances, is presented below in the Results section of this chapter.

Experimental Subjects

Twenty three pairs of university students were used as subjects. Subjects were obtained through signup sheets posted on bulletin boards around the campus. Data for three pairs of subjects were not analyzed, two because one member of the dyad did not show up, and one because the pair was acquainted. All twenty dyads for which data was analyzed were unacquainted before the experiment. Sex composition of the dyads analyzed was eight male female, six female female, and six male male.

Results

As was stated earlier, a logical first step in analyzing the results of this experiment is to determine whether the proxemic behavior of the experimental subjects was non random. The results of the chi-square test (see Figure 2 and related text) applied to the null hypothesis of random movement are shown in Table 1. hypothesis 1 In a game of any number of moves, each subject will make each move in a random fashion.

TABLE 1 A test of the null hypothesis of random movement. After 3 moves after 6 moves after 9 moves after 12 moves

Exp. Freq.

Obs. Freq.

Exp. Freq.

Obs. Freq.

Exp. Freq.

Obs. Freq.

Exp. Freq.

Obs. Freq.

Separation 0-5 moves 5 20 6 20 6 20 6 20 6-7 moves 8 0 6 0 6 0 6 0 8-12 moves 7 0 8 0 8 0 8 0

Statistics Pearson Chi-Square 24.000 21.538 21.538 21.538

P 0.000 0.000 0.000 0.000 As indicated in Table 1, there was a highly significant tendency for subjects to first approach each other and then to maintain an interpersonal distance between 0 and 5 moves. Therefore the null hypothesis of random movement can be rejected and interpersonal proxemic interaction of a non-random nature is shown to have occurred. The overall tendency for subjects to first approach each other and then to maintain a relatively constant separation is detailed in Figure 3. It can be seen that the first three moves of the game


42

were characterized by a high number of moves decreasing the interpersonal separation and very few increasing it, with slightly more than a third of the moves leaving the separation unchanged. After the third move, and continuing throughout the remainder of the game, a different pattern is seen in which slightly more than half of the moves leave the separation unchanged, and the remaining moves are divided more or less equally between those increasing and decreasing the interpersonal separation. This led the author to divide the game into two "phases" for purposes of analysis, the first three moves being characterized as the "approach phase" of the game and the remaining moves as the "interaction phase."

FIGURE 3. Behavioral choices of subjects (moves decreasing, increasing or leaving the interpersonal separation unchanged) plotted as a function of the sequence of moves from move 1 to move 12. Once non random proxemic behavior is shown to have occurred, it becomes of interest to determine what, if any, individual or interpersonal factors are associated with differences in behavior. Five additional hypotheses, based on prior findings in proxemic research, were tested. The first of these, hypothesis 2, was stated as follows. hypothesis 2 Subjects scoring high on pretest scales reflecting power, confidence, and control will attempt to use closer interaction distances than subjects scoring low on these scales.

Issues in Data Analysis

Hypotheses 2 and 3 (regarding attitudes and sex) can be viewed as being concerned with individual characteristics, whereas subsequent hypotheses relate to interpersonal level variables (verbal interaction, acquaintance, and physical similarity) that would be expected to manifest themselves directly in the interpersonal separation of the two subjects. Individual characteristics would not necessarily manifest themselves directly in the interpersonal separation of subjects because of the reciprocal and compensatory responses of subjects to each other's behavior. It seems more appropriate to look for the manifestation of individual characteristics in each subject's attempts to influence the interpersonal separation. The construction of an effective measure of individual efforts to influence interpersonal separation, in such a fashion as to extract


43

as much useful information as possible from the available data, is a problem for which the optimal solution may not yet have been found. The difficulty results from the fact that each time an experimental subject makes a move in the game, that move can affect the interpersonal separation in one of three ways either reducing it, leaving it unchanged, or increasing it. Therefore, to take full advantage of this information, one would ideally wish to have some means (a) to divide subjects into two or more different but internally homogeneous groups of roughly equal size, (b) to count the number of moves in each category (i.e., separation reduced, unchanged, or increased) made by members of each group, and (c) to compare the number of moves made in each category by members of each group to determine whether any differences are statistically significant. However, the psychological tests commonly used to measure attitudes usually produce a range of scores, not a categorization into homogeneous groups. One can use the scores to form categories, the most obvious being high scorers (those scoring above the median score) and low scorers, but this results in a loss of some of the information contained in the original range of test scores. Extremely high scores, for example, would not be differentiated from scores only slightly above the median since both would be categorized as high scores. A second approach is to use only one of the three types of moves (for example the number of moves reducing the interpersonal separation) as a dependent variable and to consider the test scores as independent variables, which would preserve the full range of the test scores and permit a conventional regression analysis to be performed. However, this results in the loss of two thirds of the behavioral data (represented in this example by the number of moves increasing the interpersonal separation or leaving it unchanged) from the game. At first thought it appears that the situation can be improved by mathematically combining the three categories of moves in some way. As an example, suppose that one decided to use the number of moves increasing the interpersonal separation (Ni) less the number of moves reducing the separation (Nr) as a dependent variable, ignoring the number of moves leaving the separation unchanged (Nu) as being irrelevant. This new dependent variable (call it C1 ) can now be expressed as C1=Ni-Nr. This seems promising until we note that many combinations of Ni-Nr produce the same C1. For example is the person who, out of a total of twelve moves in the game, makes moves Ni = 0, Nr = 2, Nu = 10 the same as a person who makes moves Ni = 5, Nr = 7, Nu = 0? Both have the same C1 = 2 but they have obviously behaved differently. One might feel that this problem can be overcome if Ni, Nr, and Nu are combined properly. For example, suppose we define C2=(Ni-Nr)/Nu. Notice that this number goes to infinity if Nu = 0, certainly an undesirable characteristic, and in any case how does one interpret C2? What does it mean? In short, although there are any number of possible Cn that can be constructed, it is impossible to preserve all of the information contained in three different numbers (assuming that they are, at most, only weakly dependent on each other as is the case here) if they are combined into one number.4 As a result, it appears that there is an unavoidable dilemma that must be faced in analyzing the attempts of individual experimental subjects to change the interpersonal separation. Either one can use the first approach described above, taking full advantage of the behavioral information contained in the moves made by subjects but losing some of the information in the test scores, or one can use the second approach which retains all of the information in the test scores but loses some of the information in the behavior of the subjects. In general, since the purpose of this experiment is principally to explore the characteristics and applications of the "behavioral game" methodology, the first approach is favored in the analyses that follow. ____________________ 4 In some possible experiments using the behavioral game methodology one might wish to have a means of assessing whether certain patterns of behavior on the part of subjects are "purposeful." One approach to constructing such a measure is discussed in Appendix C. ____________________


44

Analytical Technique

The analytical technique used to assess the effects of attitude differences has the following features. First, the basic unit of analysis or observed "event" is one two person experimental run. Within each run, usually the two subjects had different scores on most, if not all, of the nine attitudinal measures used. Two methods were used to select runs to be included in the analysis. In one analysis, every run where the two subjects had different scores on the attitude being tested was included. The disadvantage of this method is that both subjects could, for example, score below the overall mean score (for all forty subjects) for the attitude in question, but if one subject scored one point higher than the other, he or she would be categorized as the "higher" scorer while the other subject would be categorized as the "lower" scorer. The corresponding advantage is that the total number of runs analyzed is maximized, with only those runs where both subjects had the same score being omitted. In the other analysis, only those runs where one subject scored above the overall mean and the other subject scored below the overall mean were used. Since only "high" scorers vs. "low" scorers were compared, the magnitude of any observed effects would be expected to be greater in this analysis. The disadvantage of this method is that the total number of runs analyzed is reduced. A second major feature of the analysis is that once runs to be analyzed are selected, a contingency table technique is used with the contingency tables constructed as follows. As mentioned above, at each move in the game, the subject can influence the interpersonal separation in only one of three possible ways. He can make a move that increases the interpersonal separation (in "moves"), leaves it unchanged, or reduces it. After subjects are categorized by one of the two techniques described above, the total number of moves of each of the three types are totaled for the subjects in each category. A 2 x 3 contingency table is then constructed having the form shown in Tables 3 and 4. The two "distributions" of moves (one for "high" or "higher" scorers and one for "low" or "lower" scorers respectively) are then compared to determine whether any differences are significant, using the standard chi square methodology. A third feature of the total analysis is that separate contingency tables are constructed for the first three moves of the game and the last nine moves of the game, as well as for the entire run of twelve moves. This is done because of the observed tendency for experimental runs to consist of two distinct phases, as illustrated in Figure 3.

Attitudes

Table 2 provides a tabulation of the probabilities that observed differences in behavior between groups categorized according to scores on the various attitude measures and analyzed using the technique described above are due to chance. It can be seen that significant behavioral differences are associated with differences in Mach IV scores, I E scores, and FIRO eI and eA scores. The overall pattern of results seems reasonably consistent given the relatively small sample sizes. Within the FIRO subscales, significant results are obtained for two of the three "expressed behavior" subscales but for none of the three "wanted behavior" subscales. This would be expected if self reports of expressed behavior corresponded with actual expressed behavior, and if there were no opportunities to significantly influence the behavior of the other subject, as seems to have been the case. The pattern of results over the two phases of the game (moves 1 3, the "approach" phase, and moves 4 12, the "interaction" phase) is also consistent. The only significant probability observed in the approach phase of the game was associated with differences in eI score ("I initiate interaction with people"). The significant probability associated with eA ("I act close and personal toward people") was found in the "interaction" phase of the game, also as would be expected. No significant probabilities were observed in association with the FIRO scales related to control (eC and wC ), or with the F scale (authoritarian personality), which is closely related conceptually.


45

TABLE 2

Summary of probabilities that observed differences in behavior between subjects categorized by attitude scores are due to chance.


46

Although the overall pattern of results for individual characteristics was not entirely consistent, the primary anomolies being the more highly significant results on eI for the "higher" vs. "lower" scorers rather than for the "above mean" vs. "below mean" scorers, and the disappointingly high probability associated with eA in the interaction phase of the game in the "higher" vs. "lower" scorers comparison, the results seem promising. The relatively small sample sizes, ranging from 6 to 20 runs for each variable, might be expected to produce some instability in the results.

Detailed Analysis of Selected Comparisons

The most consistent and highly significant behavioral differences were observed between subjects categorized according to I E score. As can be seen by inspection of Table 3, the major difference between the high and low scorers is a tendency for the high scorers (externals) to make fewer moves reducing the interpersonal separation and more moves increasing it. Since subjects generally orient themselves to face each other, moves increasing the separation are steps backward (or, more commonly, "backward and to the side" due to the geometry of the device). Except for the case when subjects are categorized by FIRO eI score, discussed later, differences in the number of moves increasing the separation are prominent in all the tables where significant differences appear. The precise significance of this tendency is unclear to the author at this point. However, one could speculate that these moves tend to be more difficult, and perhaps more "calculated", than others because of the necessity to turn one's head to see where one is going, among other things. This pattern of behavioral differences is repeated consistently in all cases where subjects were categorized by I E score. There were also behavioral differences in the game that were associated with Mach IV scores. In addition to the two significant (P<.05) probabilities shown in Table 2, there were two other probabilities that approached significance (p<.10). Since the pattern of behavioral differences was similar in all cases, the contingency table for "higher" vs. "lower" scorers for the entire game (moves 1 12) is presented in Table 4 to facilitate comparison with the analogous table for I E scores (Table 3). Significant, though less consistent, behavioral differences were also observed when subjects were categorized by their scores on the FIRO B eA and eI subscales. These are illustrated in Table 5. An examination of Table 5 reveals interesting differences in behavioral patterns. When subjects are categorized by eA score the most significant behavioral differences are observed in the interaction phase of the game, whereas when subjects are categorized by eI score, differences are more significant in the approach phase. In addition, examination of the cells within the contingency tables shows that behavioral differences between high and low scorers on the eA scale are concentrated in the moves increasing the interpersonal separation while differences between subjects classified by eI score are manifested primarily in moves reducing the interpersonal separation. In all cases the observed behavioral patterns can be interpreted in a way that is consistent with the rationale of the attitudinal scales and with prior research findings. For subjects categorized by I E score, the smaller number of moves reducing the interpersonal separation and the larger number increasing it exhibited by higher scorers ("externals") when compared to internals could be characterized as "defensive" behavior. The similar, though less pronounced, behavioral pattern for high Machs as opposed to low Machs could be interpreted as indicating that high Machs are more "cautious" in the situation presented by the game. The smaller number of moves increasing the interpersonal separation in the interaction phase of the game by those who scored higher on the FIRO eA subscale could be interpreted as "friendly" behavior, while the larger number of moves reducing the separation in both the approach and interaction phases of the game seem to indicate that high eI scorers are more interpersonally "aggressive".


47

TABLE 3 A comparison of the number of moves reducing, leaving unchanged, or increasing the interpersonal separation (moves 1 12) for subjects scoring higher vs. lower on the I E scale. Separation Separation Separation (N = 17 runs)a Increased Reduced Unchanged Higher Scorers 55 107 42 (M=14.5) Lower Scorers 69 114 21 (M=8.9) Pearson Chi Square = 8.80 p = 0.012

TABLE 4 A comparison of the number of moves reducing, leaving unchanged, or increasing the interpersonal separation (moves 1-12) for subjects scoring higher vs. lower on the Mach IV scale. Separation Separation Separation (N = 20 runs) Reduced Unchanged Increased Higher Scorers 67 127 46 (M=94) Lower Scorers 79 134 27 (M=84) Pearson Chi Square = 6.119 p = 0.047 a Three of the twenty runs were not included because both subjects had the same scores.


48

TABLE 5 Contingency tables showing the observed behavior of subjects differentiated by FIRO eA score (left side) and FIRO eI score (right side) during the "approach phase" of the game (top) and the "interaction phase" (bottom).


49

Sex Differences

First it should be noted that a case could be made for conceptualizing sex differences as interpersonal rather than individual level variables since the behavior of males vis a vis males, males vis a vis females, and females vis a vis females is being compared. Since interpersonal behavior depends on the actions of both subjects, it seems that it might be more appropriate to observe the interpersonal separation directly, rather than observing the attempts of each subject to change the separation as was done in analyzing the attitudinal variables in the prior section. The hypothesis concerning sex differences was formulated as follows.

hypothesis 3 Female female pairs of experimental subjects will use closer interpersonal separations than will male-female pairs, who will use closer separations than will male male pairs.

Average separations in each dyad were observed during the interaction phase of the game (moves 4 12). The interpersonal separation was recorded after each subject had made each move, and the mean separations were compared using T tests. The results are shown in Table 6.

TABLE 6 T-test comparisons of mean interpersonal separations (moves 4 12) for pairs of subjects categorized by sex. Female Male Male (Female- Female Female Male Female) N (dyads) 6 8 6 Mean Separation (moves) 2.51 2.54 2.65 Std. Deviation (moves) 0.52 0.70 0.60 T (separate variances) 0.13 0.43 0.60 p 0.89 0.67 0.55 Mean Separation (meters) 1.15 1.17 1.21 Std. Deviation (meters) 0.23 0.32 0.27 T (separate variances) 0.13 0.40 0.58 p 0.90 0.69 0.57 As can be seen from Table 6, the differences in interpersonal separation maintained by male male, male female, and female female pairs of subjects were very small (less than 3 inches) and not statistically significant, although the predicted pattern of differences was observed. This result is not altogether surprising, since sex differences reported in the literature have tended to be small in magnitude and the number of subjects run in this experiment is probably too small for such differences to become statistically significant. Since sex differences can alternately be conceptualized as individual level differences, a contingency table analysis similar to that utilized previously to study behavioral differences associated with attitudes can also be performed.5 The results are shown in Table 7.


50

6

TABLE 7 Contingency tables showing the observed behavior of subjects differentiated by sex (moves 4-12) Males (vis a vis Females) vs. Females (vis a vis Males) Separation Separation Separation (N = 8 runs) Reduced Unchanged Increased Males 29 51 16 (vis a vis (30%) (53%) (17%) Females) Females (vis a vis 29 56 11 Males) (30%) (58%) (12%) Pearson Chi Square = 1.16 p = 0.56 _________________________________________________________ Males (vis a vis Males) vs. Females (vis a vis Females) Separation Separation Separation (N = 12 runs) Reduced Unchanged Increased Males 47 69 28 (vis a vis (33%) (48%) (19%) Males) Females (vis a vis 44 81 19 Females) (31%) (56%) (13%) Pearson Chi Square = 2.78 p = 0.25 ___________________________ 5 One technical difference should, perhaps, be noted. In the attitude analysis (Tables 3, 4, and 5) the higher scoring subject was always compared with the lower scoring subject within a single run on the attitude measure in question. That is not the case in the analysis presented in the lower half of Table 7 (Males (vis a vis Males) vs. Females (vis a vis Females)) or in the top half of Table 8. 6 To facilitate comparison of the two tables, because of the different N's, the percentages of the total number of moves in each category for each group are included in parentheses below the actual numbers of moves in each cell of the tables.


51

Table 7 confirms the results of the analysis of mean interpersonal separations presented in Table 6, since no significant differences were observed in either case. As shown in Table 7, there is a consistent tendency for males to make about the same number of moves reducing the interpersonal separation, somewhat fewer leaving it unchanged, and somewhat more increasing it than do females. This pattern remains essentially the same for same sex or mixed sex dyads. Although this observed pattern of differences is small, interestingly, it is almost identical to the differences observed when the behavior of subjects with the higher F score in a dyad is compared with that of the subjects with the lower F score. This rather striking similarity is shown in Table 8.

TABLE 8 Similarities in the behavior of subjects categorized by sex and by F score (moves 4 -12) Separation Separation Separation Reduced Unchanged Increased All Males 76 120 44 All Females 73 137 30 Higher F score 76 122 42 Lower F score 76 139 31 Although the similarity of the two tables shown in Table 8 certainly cannot be said to "prove" anything, it might, perhaps, suggest that the observed minor behavioral differences between males and females could be related to corresponding differences in authoritarian attitudes. It is interesting to note that the observed pattern is consistent whether males are interacting with males or females (see the two tables presented in Table 7) and also for females interacting with males or females, and there is no indication that males behaved in a more "oppressive" manner with females than they did with each other. Verbal Interaction The hypothesis concerning verbal interaction was stated as follows. hypothesis 4 Pairs of subjects who engage in a high level of verbal interaction will use closer interpersonal separations than pairs who engage in a low level of verbal interaction. In observing the experimental runs it became apparent that there were three somewhat overlapping basic patterns of verbal interaction between subjects during the approximately six minutes of the game. These were labelled low, intermediate, and high verbal interaction and can be characterized as below. LOW VERBAL This condition was characterized by a very low level of verbal activity throughout the entire experimental run. To the experimenter, the "atmosphere" of the interpersonal interaction seemed noticeably tense during these runs. INTERMEDIATE VERBAL These runs were characterized by a relatively low level of verbal interaction in the beginning of the run, with a tendency toward an increasing amount of verbal interaction as the run progressed. The experimenter felt that these subjects were attempting to be polite, but that there was some awkwardness in the interaction. Most often, one subject would take the lead in the conversation with the other tending to respond. Topics of conversation tended to be impersonal.


52

HIGH VERBAL In these runs there tended to be a high level of verbal activity from the beginning of the run and continuing throughout the entire run. The flow of conversation was easy with both subjects tending to contribute. There were a wide range of topics of conversation and the atmosphere seemed easy and open. Ratings were made by the experimenter during and after each run. In addition to the three categories described above, two intermediate categories, one between low and intermediate and one between intermediate and high, were used for questionable runs. In the analysis illustrated in Table 9, three categories corresponding to those described above were used. The "low" category (N=6) included questionable cases between the low and intermediate categories described above. "Intermediate" (N=7) included questionable cases between intermediate and high verbal. "High" verbal (N=7) included only those cases that clearly were in the "high" category described above. Results of this analysis indicated that there were no significant differences in interpersonal separation between the low and intermediate groups, but that both differed significantly from the high verbal interaction group.

TABLE 9 T test comparisons of mean interpersonal separations (moves 4 -12) for pairs of subjects categorized by level of verbal interaction. Low Intermediate High (Low) N (dyads) 6 7 7 Mean Separation (moves) 2.66 2.84 2.20 Std. Deviation (moves) 0.61 0.72 0.23 T (separate variances) -0.70 3.15 2.44 p 0.50 0.006* 0.03* Mean Separation (meters) 1.22 1.30 1.01 Std. Deviation (meters) 0.27 0.32 0.10 T (separate variances) -0.69 3.25 2.57 p 0.50 0.003* 0.02* *Significant at p>.05 Although the author is somewhat reluctant to interpret these results, given the small sample size and the preliminary nature of this experiment, it would appear that perhaps the "high verbal" condition is indicative of a sense of ease in the experimental situation on the part of both subjects, which is also manifested in a reduced interpersonal separation. When neither subject (i.e., the low verbal condition) or only one subject (the intermediate verbal condition) attempts to converse, a significantly greater interpersonal separation was maintained. It should be noted that the experiment does not directly bear upon the relationship among the non verbal immediacy behaviors discussed in Chapter 1 (Argyle and Dean, 1965; Breed, 1972; Patterson, 1976), since these theories are concerned with people's response to changes in the various immediacy behaviors. This experiment did not present subjects with a change in immediacy behaviors, and only demonstrates that subjects who conversed easily used a closer "steady state" interpersonal separation than those who did not.


53

Acquaintance

A weak acquaintance differential was created by having half of the experimental dyads complete the pretest questionnaire in the same room and the other half in separate rooms.7 Completion of the pretest took about thirty minutes. The hypothesis concerning acquaintance was stated as follows.

hypothesis 5 Pairs of subjects who have completed the pretest questionnaire in the same room (acquaintance condition) will use closer interpersonal separations than pairs who completed the pretest in separate rooms (no acquaintance).

As was the case with the prior hypothesis, a T test on the mean separations was the most appropriate analytical technique. As shown in Table 10, no significant differences were found between subjects who completed the pretest in the same room and those who completed it in separate rooms. However, the overall mean separation was slightly (approximately 2 inches) less for the acquaintance condition, which was the expected direction of the difference.

TABLE 10

T-test comparisons of mean interpersonal separations (moves 4 -12) for pairs of subjects categorized by acquaintance. Acquaintance Non Acquaintance Condition Condition N (dyads) 10 10 Mean Separation (moves) 2.52 2.61 Standard Deviation (moves) 0.54 0.69 T (separate variances) 0.48 p 0.64 Mean Separation (meters) 1.15 1.20 Standard Deviation (meters) 0.24 0.31 T (separate variances) 0.51 p 0.61 ______________________________ 7Half of each of the male male, male female, and female female dyads completed the pretest under each condition. Individual dyads were assigned to one condition or another primarily on the basis of which rooms were available at the time, creating an essentially random assignment. ______________________________ In all likelihood, the acquaintance differential in this experiment was too small to produce significant effects, and a more powerful induction of differences in acquaintance, or a series of increasingly strong inductions, would be of interest in future research. Similarity in Appearance The hypothesis concerning the appearance of subjects was stated as follows.

hypothesis 6 Pairs of subjects rated as being globally more similar in physical appearance will use closer interpersonal separations than those rated less similar.

Two ratings were made, one on the basis of the subjects' clothing, and one on the basis of physiological factors. Both ratings were made primarily on the basis of the experimenter's global


54

impressions. Major factors used in rating similarity of clothing were "style" (ranging from "sloppy" or "informal" through "casual" to "snappy" or "formal") and color. Most subjects wore bluejeans or slacks and a shirt and sweater. Major factors used in rating physiological similarity were race, height, weight, and facial characteristics. As part of the pretest (see Appendix B), subjects were asked to answer some questions about their clothing and physique, and these answers were compared later as a check on the experimenter's subjective ratings. Both clothing and physiological similarity were rated on a three category (low, intermediate, high) scale. An overall similarity measure was constructed by averaging the clothing and physiological similarity ratings of each dyad. As with the three prior hypotheses, testing was done using T tests of differences in mean interpersonal separations. No significant differences were observed for any of the measures of similarity. The results for the overall similarity measure are shown in Table 11.

TABLE 11 T test comparisons of mean interpersonal separations (moves 4 12) for pairs of subjects categorized by their overall similarity in appearance. Low Intermediate High (Low) N (dyads) 6 8 6 Mean Separation (moves) 2.54 2.58 2.56 Std. Deviation (moves) 0.66 0.71 0.44 T (separate variances) 0.17 0.08 0.12 p 0.87 0.94 0.91 Mean Separation (meters) 1.17 1.18 1.17 Std. Deviation (meters) 0.29 0.32 0.19 T (separate variances 0.08 0.12 0.02 p 0.94 0.91 0.98 Instrument Effects As mentioned earlier, the geometry of the floor pattern makes it possible, given the rules of the game used in this particular experiment, to determine whether subjects are being attracted to (or are avoiding) any particular node on the floor to a greater extent than would be predicted by a null hypothesis of random movement. Inspection of the data resulting from this experiment showed no discernible evidence for such "instrument effects" with the possible exception of a tendency for experimental subjects to remain near the center of the device. Comparison of observed and expected frequencies for subjects three moves or less from the center vs. subjects more than three moves from the center, using a chi-square methodology similar to that illustrated in Table 1, provided conclusive (p=0.000) evidence that subjects did tend to remain near the center of the device. However the starting positions chosen for subjects in this experiment were such that this tendency could be due to proxemic effects between subjects rather than any real tendency for subjects to remain near the center of the device (or vice versa, although this seems much less likely). However, the persistence of the "centrality" effect after the approach phase of the game suggests that it may be due to some situational factor inherent in the device. Further experiments will be necessary to clarify this effect. A more general discussion of the results of this experiment and the experiment described in the next chapter is presented in the final chapter of this dissertation.


55

Chapter 4

AN EXPERIMENT

Results of the validation experiment described in Chapter 3 were sufficiently encouraging to justify attempting a more ambitious experiment. The experiment was planned to attempt to serve two purposes simultaneously. First to continue to explore and validate the basic "behavioral game" methodology described in Chapter 2, and the observation of proxemic behavior in general, as a means of increasing the scientific understanding of interpersonal behavior. Second, to specifically test, clarify, and explore the sometimes conflicting hypothesized relationships among some of the more prominent immediacy behaviors postulated by Breed (1972), Argyle and Dean (1965), and Patterson (1976). Although the majority of the studies reviewed by Patterson (1973) support Argyle and Dean's "compensation" hypothesis, other studies "indicate that changes in nonverbal intimacy may produce reciprocal as well as compensatory adjustments." (Bakken, 1978, p. 300) That is, attempts to increase intimacy by one person may also result in increased attempts by the other person. Other studies have failed to find either reciprocal or compensatory changes (Coutts and Schneider, 1976). All of these results can be explained by recourse to Patterson's (1976) "arousal" model (described in the section of Chapter 1 titled "The Nonverbal Immediacy Behaviors"). Although the compensation hypothesis seems dominant at the present time, it seems fair to say that neither the effect of the immediacy behaviors on person perception nor the nature of the mutual interaction of these behaviors in interpersonal interaction is truly well understood.

Experimental Design

Because of the many unresolved issues in the theoretical frameworks described above, and the questions still to be answered about the characteristics of the methodology described in this dissertation, it seemed best to perform a rather general experiment rather than one that is very closely tied to the specific predictions of one or another of the theories. Therefore, the experiment described below is conceptualized by the author as a reasonably straightforward embodiment of the classical paradigm of psychological research, which has been stated as follows At a certain moment the organism makes a response. The response [R] depends on the stimuli [S] acting at that moment and on factors present in the organism [O] at that moment. This general statement can be put into the form of an equation, R = f (S,O). (Woodworth and Schlosberg, 1954, p. 3) In the experimental design chosen, the stimuli are considered to be the controlled behaviors of a confederate. The specific behaviors varied (verbal behavior, eye contact, and proxemic behavior) are relevant to the compensation and other hypothesis and have been shown, at a minimum, to influence interpersonal immediacy in some way. The "organismic" factors are considered mediating variables, and have been chosen based on the conclusions of Altman and Vinsel (1975) and the results of the validation experiment described in the prior chapter. The two factors that seem both most relevant and most amenable to measurement are (a) a subset of attitudes or personality characteristics of the confederate related to power, confidence, and control, and (b) the global impression (i.e., from "liked" to "disliked") made by the confederate on the subject. The latter has particular relevance to Patterson's (1976) theoretical model which predicts that subjects will reciprocate changes in the confederate's immediacy (intimacy) behaviors if they experience positive emotion, but will respond in a compensatory fashion if they experience negative emotion toward the confederate.


56

The responses or dependent variables observed in the subjects are the same as the independent variables manipulated by the confederate, that is the subject's verbal behavior, eye contact, and proxemic behavior. The experimental design can be diagrammed as follows. INDEPENDENT INTERVENING DEPENDENT VARIABLES VARIABLES VARIABLES (S) (0) (R) Physically Observ- Non-Observable Physically Observable able Behaviors of Characteristics Behaviors of the Confederate: of the Subject: of the subject: ---------------------------►►►►- ----------------------------►►►►------------------------------------------- (a) verbal a) attitudes or (a) verbal (b) eye contact personality (b) eye contact (c) proxemic characterstics (c) proxemic ▼ (b) perception of ▲ ▼ confederate ▲ ▼ ▲ ►►►►►►►►►►►►►►►►►►►►►►►►►►►►► (The arrow connecting the independent variables directly to the dependent variables is meant to indicate that the particular set of intervening variables chosen for study, while expected to be significant, is not expected to directly account for a large percentage of the variance in the effects produced by the behaviors of the confederates upon the behaviors of the subjects.) FIGURE 1. A generalized representation of the conceptualized relationships among the variables chosen for observation. Independent Variables As indicated in Figure 1 above, three variables, the verbal behavior of the confederate, his "looking behavior" (eye contact), and his proxemic behavior during the game were manipulated. The confederates were coached in two conditions for each of the first two variables as follows. High verbal

The confederate initiates (if necessary) and maintains a casual, friendly conversation with the subject, asking a number of questions if necessary to maintain the verbal interaction. Low verbal

The confederate speaks only when spoken to, and gives only brief responses.


57

High eye contact The confederate maintains his gaze in the

vicinity of the subject's head, without staring, so that he can meet the subject's eyes when the subject initiates eye contact. Low eye contact The confederate does not look at the sub ject's head, looking either down or to the right or left, and does not meet the subject's eyes when the subject initiates eye contact. Since both confederates were generally friendly and outgoing in their interpersonal behavior, most of the coaching effort was devoted to the low verbal and low eye contact conditions. The proxemic behavior of both confederates was programmed to follow the same pattern, described later in the "Procedure" section of this chapter, for all experimental runs. Intervening variables Two uncontrolled variables, personality characteristics or attitudes of the subject and the subject's perception of the confederate, were considered intervening variables, that is, variables that mediated between the independent and dependent variables or moderated the effect of the independent variables. A pretest was used to assess certain personality characteristics or attitudes of the subjects. The following scales were included:

(a) I E scale (Rotter, 1966)

(b) F scale forced choice version (Christie, Havel,and Seidenberg, 1958)

(c) Mach IV (Christie, 1973)

(d) FIRO eI,eA, and eC subscales (Schutz, 1958)

(e) The Pittsburgh social extraversion introversion (SEI) and emotionality (Em) scales (Bendig, 1962).

A post-test was used to rate the subject's perception of the confederate with a "likableness" scale, a copy of which is included, along with the other instruments and forms completed by participants in this experiment, as Appendix D of this dissertation. The likableness scale was constructed from twelve items presented in a semantic differential format. The anchor words for each item were taken from a list of words that have been previously rated high and low in likableness (Anderson, 1968). Since the subject's perception of the likableness of the confederate must result from the interaction during the experiment, it could be at least equally valid to categorize this factor as a dependent variable rather than as an intervening variable as the author has done. However, the author is philosophically predisposed to consider physical "observables", such as the behaviors of the subject and confederate, as dependent and independent variables, and to conceptualize all non observable internal states of the subject as intervening variables. This conceptualization is also in agreement with that expressed by Patterson (1976), quoted at length in the section of Chapter 1 titled "The Nonverbal Immediacy Behaviors". Viewed in this sense, the attitudes measured by the pretest are regarded as relatively enduring and generalized internal states, while the "likableness" rating from the post test is considered a relatively short term and situationally specific internal state.


58

The procedure used, measuring the confederate's "likableness" (as perceived by the subject) after the experiment, does present some difficulties since presumably the subject could "like" the confederate until a certain point in the "game" say move 13 when the confederate begins to attempt to "close in" on the subject (see the "Description of the Experiment" section later in this chapter) and "dislike" him based on the particular behaviors manifested late in the game. Although this potential problem is recognized, it is not viewed as being unacceptable because: (a) since the proxemic behavior of the confederate follows the same patterns for all subjects, differing individual responses to various proxemic behaviors that are not correlated with the specific variables analyzed in the study should be randomly distributed and therefore should not produce systemmatic effects in the analysis; (b) there is no practicable way to measure the internal state of the subject continuously during the game; and (c) in any case, inclusion of this variable is one of the more exploratory features of the experimental design. Dependent Variables As mentioned previously, three aspects of the subject's behavior, his spatial behavior, verbal behavior, and looking behavior were recorded and analyzed as dependent variables. a) Spatial behavior the subject's spatial behavior is recorded and analyzed in the same way as in the validation experiment described in Chapter 3. b) Only the amount of verbalization by the subject is measured, not its content. Two types of verbalizations, "short"- approximately ten words or less, and "long" statements of more than ten words, were distinguished. The number of each type made by the subject during each fifteen second interval between moves was recorded. d) Two types of "eye contact" (i.e., when the subject looks at the confederate's face, regardless of whether actual eye contact is established) were distinguished and recorded. "Glances" were defined as looks of about one second or less, and "looks" were defined as extended looks lasting longer than one second. The number of each were recorded in the same manner as the verbalizations described above. Subjects and Confederates In addition to the author, two male confederates, a research assistant and forty male subjects participated in the experiment. Confederates were paid. Each confederate spent about twenty hours on the experiment, four hours for coaching and training and sixteen hours observing or participating in the actual experimental runs. Confederates alternated being in the device with the subject or being outside recording the subject's verbal behavior for each run. Subject's proxemic and "looking" behavior were recorded by the experimenter who observed all runs. The research assistant was needed to "direct traffic" and control the various forms as the subjects took the pretest, participated in the game, and took the post test. Subjects were white American males recruited from the Yale student body and spent about one hour completing the experimental procedure. White American males were selected because it was hoped that the experiment can be replicated in the future using females and subjects from other cultures or subcultures to assess the extent of any differences in proxemic behavior due to these effects. Subjects were paid for their participation.


59

Description of the Experiment As subjects arrived, they were first asked to read and sign the "Informed Consent Form" (Appendix D), then given the pretest (described above under "Intervening Variables.") Two small rooms were available so that most subjects completed the pretest alone. Completing the pretest took about one half hour. After the subject completed the pretest the experimenter, accompanied by a confederate who was introduced as the subject's "partner" in the experiment, led the subject to a large room in which the experimental apparatus had been set up as shown in Figure 1 in Chapter 3. The subject and confederate were positioned on the nodes labeled SX and SY respectively in Figure 1 in Chapter 3, exactly as in the validation experiment. Exactly the same instructions were read to the confederate and subject as were used in the experiment described in Chapter 3 (see the "Procedure" section of that chapter). The game consisted of 15 "moves" by both the confederate and the subject in alternating order at 15 second intervals, exactly as described in the validation experiment except for the total number of moves. The confederate always moved first, and the pattern of moves made by the confederate was identical for all runs. This pattern of moves made by the confederates was designed to break the game into four distinct "phases," which could then be separately analyzed to determine whether different proxemic behaviors on the part of the confederates evoked different behaviors from the subjects. The four phases are described below. Phase 1 "Approach" (Moves 1- 3) The subject and confederate approach each other until a

comfortable interpersonal distance is found. Relatively minor differences in proxemic behavior would be expected during this phase. (In the validation experiment the only significant differences observed were when subjects were differentiated by FIRO eI score.)

Confederate's moves move 1 reduce interpersonal separation move 2 separation unchanged (move to the right) move 3 reduce separation

(This approximates the typical pattern for the first three moves observed in the validation experiment see Figure 3 in Chapter 3.) Phase 2 "Distance Measurement" (Moves 4 - 6) The subject is allowed to control the interpersonal distance, which will be the primary (proxemic) variable of interest. Confederate's moves move 4 - separation unchanged move 5 - separation unchanged move 6 - separation unchanged (The purpose of these three moves was to allow the subject to establish a comfortable "baseline" interpersonal distance, since the confederate's behavior did not influence the interpersonal separation.) Phase 3 "Influence Attempts" (Moves 7 12) The confederate controls the interpersonal distance

by maintaining it at the separation established by the subject after move 6. The primary proxemic variable of interest was the subjects' attempts to modify this distance (i.e., number of moves in which the subject attempted to reduce, increase, or maintain the same interpersonal distance).


60

Confederate's moves move 7 During this sequence of six moves, the move 8 confederate "reverse mirrors" the behavior move 9 of the subject, moving back when the move 10 subject moves forward and vice versa, and move 11 moving sideways when the subject moves move 12 sideways.

Phase 4 "Spatial Invasion" (Moves 13 -15) The confederate makes three successive moves

reducing the interpersonal separation. Whereas the manipulation of the confederates' verbal and looking behavior (described in section "Independent variables" above) presents different groups of subjects with different levels of immediacy, this presents each individual subject with a distinct change in immediacy to which he must respond.The primary proxemic variable of interest will be the same as in Phase 3, the number of moves the subject makes reducing, increasing, or maintaining the same interpersonal distance in response to the intervention by the confederate.

Confederate's moves move 13 reduce separation move 14 reduce separation move 15 reduce separation

After the game, the subject and confederate were asked to complete the post test out of sight of each other. They were told that they would not be told how they were evaluated on the post test. Actually the confederate does not complete the post test, but sets up to record the next subject's verbal behavior as described earlier in this section.

Hypotheses and Analysis

Factorial Design Three of the independent variables the two confederates and the eye contact and verbal behavior of each confederate form a 2 x 2 x 2 factorial design (confederate 1/confederate 2 x high/low verbal x high/low eye contact) with five subjects in each cell. As noted above, the third independent variable, the proxemic behavior of the confederate, was not varied across subjects, but varied over time breaking the game into the four phases as described. The results of the analysis of the factorial design are presented in Table 1. The null hypothesis in each case was that there were no differences in the subjects' behavior that would result in significant main or interaction effects. As can be seen from Table 1, the results tended to be consistent across all phases of the game. It should be noted that the method used to measure and record the subjects' verbal and looking behavior (see the section "Subjects and Confederates" above) was too coarse to provide meaningful qualitative data, but the consistency, and the high level of statistical significance of the results in the case of verbal behavior, provides some assurance of the reliability and validity of the rather crude quantitative measures used. The behavioral game methodology did provide a means to check for errors in the confederate's proxemic behavior. Any departure from the programmed series of moves on the part of the confederates was apparent in the computer printout from Program 4 (see Appendix A) for that game. Out of a total of 600 moves made by the confederates during the experiment, only 7, or 1.2% were in error. These errors could have resulted from either the confederates' behavior or mistakes on the part of the experimenter in recording their behavior. Examination of Table 1 yields a number of interesting observations. First it can be seen from the first column "CONFEDERATE" that the subjects consistently used a greater interpersonal


61

separation with one confederate than they did with the other. The actual separations are detailed in Table 2.


62

The entries for the main effects in the table are in the form X(Y) where X is the probability (1 tail) of chance occurrence of the observed relationship, and Y is given as either (+) indicating a positive relationship or ( ) indicating a negative relationship. For clarity, only significant (p<.05) probabilities are shown. aThe direction of the relationship is arbitrary in the case of the CONFEDERATE variable, and is shown only to indicate that the direction of the differences in interpersonal separation maintained by subjects with the two confederates was consistent. bThe interpersonal separation measured in "moves" was used for this analysis. c The abbreviations used to identify the interactions have the following meanings. CL = confederate X looking behavior CV = confederate X verbal behavior LV = looking behavior X verbal behavior CLV = confederate X looking behavior X verbal behavior


63

TABLE 2

Interpersonal separations maintained by

subjects with the two confederates.

MEAN INTERPERSONAL SEPARATIONS (IN METERS) CONFEDERATE 1 CONFEDERATE 2 DIFFERENCE Phase 1 1.62 1.81 0.19* (moves 1 - 3) Phase 2 1.07 1.37 0.30* (moves 4 - 6) Phase 3 1.06 1.46 0.40* (moves 7 - 12) Phase 4 0.66 0.90 0.24* (moves 13 - 15) ALL PHASES 1.09 1.40 0.31* (moves 1 - 15) *P<.05 The mean difference for the game as a whole, 0.31 meters, or slightly more than one foot, is larger than would have been expected. There seems to be no completely satisfactory explanation for it. The two confederates were fairly similar overall, having been selected to be close to average in height, weight, and general appearance. Both were personable and there were no gross differences in their behavior as they interacted with subjects in the experiment, that were apparent to the author as he observed the experiment, that would account for the differences. The significant interaction effects among confederate, looking behavior, and verbal behavior (CLV, the right hand column in Table 1) probably reflect individual differences in the way the two confederates expressed the high and low conditions of verbal and looking behavior, but these differences were seemingly in relatively subtle behaviors (topic of conversation, smiling, gestures, etc.) that were not recorded in this experiment. Probably the most reasonable approach is to note the differences apparent in Table 2 but to leave them unexplained for the present. They serve as a reminder that the rather coarse analysis of interpersonal behavior performed in this experiment only scratches the surface, and leaves many questions unanswered. Taken at face value, the effects of the confederates' modifications of their looking behavior (column 2 in Table 1) seem to be the weakest of any of the three factors in the design. However, it appears that this manipulation did significantly affect the interpersonal distance maintained by subjects. The separations maintained by subjects under the low and high eye contact conditions are displayed in Table 3.


64

TABLE 3

Interpersonal separations maintained by subjects under the low and high eye contact conditions. HIGH EYE CONTACT LOW EYE CONTACT DIFFERENCE CONDITION CONDITION Phase 1 1.77 1.66 0.11 (moves 1- 3) Phase 2 1.33 1.11 0.22 (moves 4 - 6) Phase 3 1.38 1.14 0.24* (moves 7 -12) Phase 4 0.86 0.69 0.17 (moves 13 -15) ALL PHASES 1.35 1.15 0.20* (moves 1 - 15) *P<.05 (< should be “less than or equal”) The differences in interpersonal separation are consistent in direction, although smaller in magnitude than those observed in association with the different confederates (Table 2). The overall mean difference of 0.20 meters or slightly less than 8 inches is statistically significant. The closer separation associated with the low eye contact condition supports a compensatory immediacy relationship between these two variables as would be predicted by Argyle and Dean (1965). The manipulation of the confederates' verbal behavior produced no significant effect on the interpersonal separation maintained by the subjects. However, it did result in the only significant difference in their evaluation of the confederates' "likability" as measured by the post test, and it also resulted in highly significant differences in their verbal and looking behavior. It can be seen from Table 1 that the subjects' immediacy responses to changes in the verbal behavior of the confederates were reciprocal in nature rather than compensatory, contrary to the results produced by the confederates' manipulation of eye contact discussed above and shown in Table 3. Increases in the amount of talking by the confederates resulted in increases, rather than decreases, in the amount of talking and also in the amount of eye contact on the part of the subjects. As noted above, the impression that the confederates made on the subjects varied significantly only with the amount of talking by the confederates. A neutral "likability" score on the post test would be 48, a "four" on each of the twelve items (see Appendix D). The overall mean score recorded for all subjects was 57.5, showing the usual positive evaluation bias. The mean scores for subjects exposed to the low verbal condition was 54.2, compared to a mean score of 60.8 for subjects exposed to the high verbal condition. Two other significant effects, in addition to those discussed above, were observed. Both were associated with the subjects' looking behavior and both occurred during Phase 4 ("spatial invasion") of the game. Subjects tended to make significantly more eye contact with one of the confederates during this phase of the game, and there was also a significant interaction between their looking and verbal behavior. The most probable explanation for these effects is an increase in tension produced by the confederates' abrupt and unilateral reduction of the interpersonal distance. The subjects' looking behavior possibly would seem to be the most sensitive indicator of increased arousal or tension. The effects discussed above are summarized in Table 4.


65

TABLE 4

Significant effects due to the confederates' behavioral manipulations. CONFEDERATE BEHAVIOR SUBJECT RESPONSE NATURE OF EFFECT Reduce eye contact Reduce Indicates a compen- interpersonal satory relationship distance in interpersonal immediacy between these variables Reduce talking (a) Reduce Indicates a recip- talking rocal effect in

interpersonal immediacy.

(b) Reduce Eye contact appears eye to vary directly contact with the amount of talking. (c) Reduce "Liking" varies perceived directly with the "likability" amount of talking. of confederate Spatial invasion Increase eye Possible evidence of contact increased tension in looking behavior. Pretest Variables Chi Square Analysis Significant results of the analysis of the pretest personality measures (locus of control; authoritarianism; extroversion; emotionality; Machiavellianism; and need for inclusion, affection, and control) using the chi square methodology discussed in Chapter 3 were sparse but interesting. Tables similar to those shown in Tables 3 through 5 in Chapter 3 (some differences will be discussed later in attempting to account for the results) were constructed for low (below the median) vs. high scorers on each of the pretest scales for each "phase" of the game. This resulted in 8 (pretest variables) x 4 (phases) = 32 such tables. Eight additional tables summing the behaviors for the entire game (all four phases) were also analyzed, giving a total of 40 tables. Only two of these tables showed significantly (p=.05) different behaviors for the two groups of subjects compared. It was anticipated that the most significant results in this analysis would occur in Phase 3, when the confederate was manipulating the interpersonal distance by maintaining a constant separation (in moves) from the subject, and this expectation was confirmed as both of the significant comparisons were observed in Phase 3.1 Not surprisingly, given the results of the experiment reported in Chapter 3, the most highly significant differences in behavior were observed when the subjects were categorized by I E score. This tends to confirm the significance of locus of control in interpersonal behavior that was apparent in the results of the validation experiment summarized in Table 2 in Chapter 3.


66

___________________ 1 The only other comparison that approached significance (p=.08) was also observed in Phase 3 for subjects differentiated by F score. _______________________ The interesting thing about these behavioral differences is that the observed pattern of behavior for high scorers (externals) and low scorers (internals) was reversed in the two experiments. The difference between the two patterns of behavior is illustrated in Table 5

TABLE 5 Patterns of behavior for subjects categorized by I E scorea in the experiment reported in Chapter 3 (moves 4 - 12, the "interaction" phase only) and the experiment reported in this chapter (moves 7 12, Phase 3 only)

EXPERIMENT REPORTED IN CHAPTER 3 (Moves 4 -12, the "interaction" phase only.)

N(runs)=17 Separation Separation Separation N(subjects)=34 Reduced Unchanged Increased Lower Scorers 36 96 21 (24%) (62%) (14%) Higher Scorers 29 84 40 (19%) (55%) (26%) Pearson Chi square = 7.47 p = 0.024

EXPERIMENT REPORTED IN CHAPTER 4 (Moves 7 -12, Phase 3 only.)

N(runs)=40 Separation Separation Separation N(subjects)=40 Reduced Unchanged Increased Lower Scorers 4 88 28

(3%) (73%) (23%) Higher Scorers 16 85 19 (13%) (71%) (16%) Pearson Chi square = 8.98 p = 0.011 a Note that there are differences in the procedure used to categorize subjects in the two tables. These are explained in 3 below.


67

Comparison of the two tables presented in Table 5 shows that the pattern of behavior shown in each table is quite different from that shown in the other. The differences can be resolved into two components. First, in the experiment reported in Chapter 3 there was a larger percentage of the total moves (for all subjects) in the "Separation Reduced" column and a smaller percentage in the "Separation Unchanged" column than in the experiment reported in this chapter. Second, in the "Separation Reduced" and "Separation Increased" columns in the two tables, the group (Low or High Scorers) with the largest percentage of moves in each column is reversed from table to table. Such large differences, unless they can be explained, bring the reliability and validity of the methodology into question. Some possible explanations of the observed differences are listed below. 1. The methodology is inherently unstable. 2. The arrangement and order of the items in the I E scale used in the two experiments was different (see the two pretests in Appendices B and D) and this resulted in subjects being categorized differently in the two experiments. 3. In the experiment reported in Chapter 3, the unit of analysis was always a single experimental run. The higher scoring subject within a run would always be categorized differently from the lower scoring subject within that run regardless of their absolute scores. In the experiment reported in Chapter 4, all subjects were run with a confederate whose pattern of moves was the same for all subjects. So the unit of analysis could not, because of the design of the experiment, be a single run. Instead, the twenty subjects (i.e., 20 experimental runs) who scored above the median on the I E scale were compared with the twenty who scored below the median. 4. The behavioral manipulations of the confederates during the game reported in this chapter may have evoked different responses in the subjects than the more naturalistic conditions in the experiment reported in Chapter 3 where naive subjects were run with other naive subjects. Although there is no certain way to distinguish among these possible explanations or others not listed, given the available data, a thoughtful examination of the data does provide indications as to the possible source of the apparent discrepancies in behavior in the two experiments. It could be argued that the most similar phases of the two games should have been Phase 1, the "approach" phase. During the first three moves of the game reported in this chapter the subjects have had minimal exposure to the behavioral manipulations of the confederates, and would be expected to behave more similarly to the subjects in the validation experiment than during later phases of the game. Therefore it seems logical to compare the behaviors of the subjects in the two experiments during the first three moves of each experiment to see if the differences apparent in Table 5 manifest themselves early in the games. If so, it would tend to lend support to explanations 1 or 2 above, or possibly to explanation 3. If explanation 4 is valid, then the behavioral differences should be minimized. The two relevant tables are shown in Table 6.


68

TABLE 6

Patterns of behavior for subjects categorized by I E score in the experiment reported in Chapter 3 (moves 1- 3, the "approach" phase only) and the experiment reported in this chapter (moves 1 - 3, Phase 1 only). EXPERIMENT REPORTED IN CHAPTER 3 (Moves 1- 3, the "approach" phase only.) N(runs)=17 Separation Separation Separation N(subjects)=34 Reduced Unchanged Increased Lower Scorers 33 18 0 (65%) (35%) (0%) Higher Scorers 26 23 2 (51%) (45%) (4%) EXPERIMENT REPORTED IN CHAPTER 4 (Moves 1 - 3, Phase 1 only.) N(runs)=40 Separation Separation Separation N(subjects)=40 Reduced Unchanged Increased Lower Scorers 31 29 0 (52%) (48%) (0%) Higher Scorers 25 32 3 (42%) (53%) (5%) A comparison of the two tables shown in Table 5 with the two tables shown in Table 6 reveals that, of the two "components" of the differences between the tables presented in Table 5 that were discussed above, only one is present in the differences between the tables presented in Table 6. Only the first component, a larger percentage of the total moves in the "Separation Reduced" column and a smaller percentage in the "Separation Unchanged" column in the experiment reported in Chapter 3, is present. The second component, a reversal of the group (Low or High Scorers) with the largest percentage of moves in the "Separation Reduced" and "Separation Increased" columns, is not observed. This lends some weak support to the fourth possible explanation of the observed differences proposed above, but it would be more convincing if some further rationale based on the behavior of the confederates could be invoked to account for the remaining differences in the tables shown in Table 6. In fact, an obvious rationale does exist. In the experiment reported in Chapter 3, the actual percentages of moves made by all subjects that reduced, left unchanged, or increased the interpersonal separation were 58%, 40% and 2% respectively. (These figures are obtained by summing the moves in each category made by all subjects as shown in the upper table in Table 6 and representing each sum as a percentage of the overall total of moves.) These percentages were approximated in the experiment reported in this chapter by having the confederate make two of the first three moves reducing the separation and one leaving it unchanged. This was the best approximation possible to the natural behavior observed in the validation experiment if the same pattern of moves by the confederates was to be presented to all subjects. However, the


69

approximation was not perfect 67% moves reducing the separation vs. 58%, 33% moves leaving the separation unchanged vs. 40%, and 0% increasing the separation vs. 2%. If participants in the experiments were approaching each other during this phase of the experiment until a comfortable interpersonal distance was achieved, as Hall's (1966) formulation of proxemics would suggest, then to achieve this distance, assumed to be the same in both experiments, the subjects in the experiment reported in this chapter would have to compensate for the imperfect approximation of the natural pattern of moves made by the confederates. Using this line of reasoning, the expected behavior of the subjects in the experiment reported in this chapter can be derived as shown in Table 7.

TABLE 7

A derivation of the expected behavior of subjects in Phase 1 of the experiment reported in this chapter.

PERCENTAGE OF MOVES

Separation Separation Separation Reduced Unchanged Increased A - The actual behavior 58% 40% 2% of subjects in the experiment reported in Chapter 3. (From Table 6) B - The behavior of the 67% 33% 0% confederates in the experiment reported in this chapter. C - Compensation required -9% 7% 2% by subjects in the experiment reported in this chapter to maintain the overall percentages of moves in each category observed in the experiment reported in Chapter 3.(A - B) D - Expected behavior 49% 47% 4% by subjects in the experiment reported in this chapter. (A+C) E - Actual behavior by 46% 51% 3% subjects in the experiment reported in this chapter. While the agreement between the expected and observed behavior (D and E in Table 7) is not precise, it can be seen that it does produce a clear reduction in the discrepancy between the behavior of subjects in the first phase of the two experiments. Unfortunately the same procedure does not produce good results when applied to the data displayed in Table 5. Here, the unusual manipulation of the proxemic behavior of the confederates in the experiment reported in this


70

chapter combined, perhaps, with the effects of the manipulation of their verbal and looking behavior of the subjects. (Recall that in this "phase" of the game the confederates "reverse mirrored" the proxemic behavior of the subjects refer to the earlier section of this chapter "Description of the Experiment" for a summary of the various manipulations.) Alternatively, it is possible that some non behavioral explanation for the discrepancies between the two tables shown in Table 5 is the correct one. Further experiments would be necessary to definitely resolve the issue, but the results of the analysis of the simpler situation earlier in the two games, as presented in Table 7, provides reason for some optimism that a behavioral explanation will ultimately account for the observed differences. The other pretest variable that was associated with significant behavioral effects was Bendig's (1962) "Emotionality" scale. Since this scale was not used in the validation experiment, and given the rather dramatic changes in the behavior of subjects categorized by I E score that were apparently produced by the behavior of the confederates in the experiment reported in this chapter, any attempt to characterize the meaning of the pattern of moves exhibited by subjects categorized by their "emotionality" score would be speculative. However, the pattern of behavior for subjects categorized by their Emotionality score was somewhat similar to that of subjects classified by I E score so, for purposes of comparison, the two tables are shown in Table 8.

TABLE 8 Behavioral similarities between subjects categorized by I -E score and Emotionality score. SUBJECTS CATEGORIZED BY EMOTIONALITY SCORE (Moves 7 - 12, Phase 3) N(subjects)=40 Separation Separation Separation Reduced Unchanged Increased Low Scorers 4 93 23 High Scorers 16 80 24 Pearson Chi square = 8.20 P = (not on orig.) _________________________________________________________ SUBJECTS CATEGORIZED BY I - E SCORE (Moves 7 - 12, Phase 3) N(subjects)=40 Separation Separation Separation Reduced Unchanged Increased Low Scorers 4 88 28 High Scorers 16 85 19 Pearson Chi square = 8.98 p = 0.011 Note that the tables presented in Table 7 are identical in the numbers of moves reducing the separation, somewhat similar in the moves leaving the separation unchanged, but dissimilar in the moves increasing the interpersonal separation. For the reasons discussed above, the


71

significance of this is unclear and can only be clarified by further research and experience with the methodology. Pretest Variables - Regression Analysis In the section "Issues in data analysis" in Chapter 3 a second analytical approach that can be used to assess the effects of personality characteristics was discussed. This consists of using the number of moves made by a subject reducing the interpersonal separation (or alternatively leaving the separation unchanged or increasing it or some mathematically defined combination of the three categories of moves) as a dependent variable and the pretest scores as independent variables in a regression analysis. Four such independent variables were tested, (a) the number of moves reducing the interpersonal separation, (b) the number increasing it, (c) the number reducing it minus the number increasing it, and (d) a more complex measure that weights (c) by the number of moves leaving the separation unchanged in such a fashion as to reduce the value of the variable as more moves are made leaving the separation unchanged (see the formula in Table 9). The results of the regression analysis tend to confirm the results of the Chi square tests reported in the prior section of this chapter, showing significant relationships for I E and Emotionality scores in Phase 3. In addition there was another significant relationship for I E score in Phase 4, three significant results associated with the FIRO scales one with eA in Phase 1, and one with eI and one with eC in Phase 4 and also one significant relationship associated with Machiavellianism in Phase 4.

TABLE 9 Significant results in a regression analysis on the pretest variables. The entries in the table are in the form X(Y) where X is the probability (1 tail) of chance occurrence and Y is either (+) indicating a positive relationship or ( - ) indicating a negative relationship. R = Number of moves reducing the interpersonal separation. U = Number of moves leaving the interpersonal separation unchanged. I = Number of moves increasing the interpersonal separation. PRETEST DEPENDENT VARIABLE VARIABLE ____________________________________________________________ R I R-I (R - I) * (R+U+I) U Phase 1 (Moves 1 3) FIRO eA - - - 0.03(+) Phase 3 (Moves 7 12) I E 0.05(+) - 0.05(+) - Emotionality 0.03(+) - - - Phase 4 (Moves 13 15) I E - - 0.05( - ) - FIRO eI 0.03(-) FIRO eC 0.02(+) Machiavellianism - - 0.03(+) - ____________________________________________________________


72

The major difficulty associated with this type of regression analysis springs from the difficulty of unambiguously assigning a meaningful behavioral interpretation to the dependent variables. The results of a regression analysis such as those presented in Table 9 can be used to obtain a "quick and dirty" indication of where significant differences in behavior might be found, but (for reasons discussed in the section "Issues in data analysis" in Chapter 3) it seems risky at best to attempt to interpret the results beyond the purely mathematical description of the dependent variable. The Chi square methodology discussed in the previous section or, better yet for large experiments, an expansion of it that will be presented in the next chapter which the author has termed "approach avoidance analysis", should be used to gain a more detailed insight into the actual behavior of groups of subjects. The greater number of statistically significant results obtained with this method of analysis, as opposed to the Chi square technique is interesting. It probably results from two factors. First, of course, the ability to rather arbitrarily define a number of dependent variables almost assures that significant relationships with some of them will be found through random statistical processes. Second, where the variance of the independent variables is small there is a greater chance of misclassifying cases using median splits as is done in the Chi square technique. The regression analysis makes full use of the variance in these cases, avoiding the inherent conservative bias of the Chi square analysis. This effect is probably responsible for the three significant results obtained on the FIRO scales shown in Table 9 which were not found in the Chi square analysis. The relatively small variance observed in the FIRO scores (4.4 for eI, 3.9 for eA and 7.3 for eC compared to variances of 12.5 for the I-E scores and 36.8 for the emotionality scores) probably significantly limits the ability of the Chi-square analysis to detect significant relationships associated with them. The results of this experiment and the experiment reported in Chapter 3 are discussed in a more general way in the next chapter. Two paradigms, an approach avoidance analysis of the proxemic behavior of subjects in the two experiments and a more general model relating the non verbal immediacy behaviors, verbal interaction and the impression one person makes on another in interpersonal interaction, are developed.


73

Chapter 5

DISCUSSION Some General Impressions This is, perhaps, an appropriate place to recapitulate, in a general way, a few of the more salient impressions made on the author by the process of running and analyzing the two experiments reported in this dissertation. Most of these impressions are supported by statistically significant results in one or both experiments but a few are not. Some are incorporated in the explanatory model developed later in this chapter and some are not. So this section provides an opportunity to at least mention some observations that are not discussed in greater detail elsewhere as well as providing a kind of "grab bag" out of which many of the results discussed later in this chapter are drawn. First, proxemic behavior, in the apparatus described in Chapter 2, of dyads engaged in casual interpersonal interaction has been clearly shown to be non random (Chapter 3). The mean interpersonal separation maintained by unacquainted, naive, subjects in the "Interaction Phase" (moves 4 12) of the experiment reported in Chapter 3 was 1.17 meters or 3 feet 10 inches. This is just inside the outer boundary of Hall's "Personal Distance Zone" described in Chapter 1 (4 feet). When the behavior of one member of the dyad was manipulated in a reasonably subtle but "unnatural" way (moves 4 12 of the experiment reported in Chapter 4) the mean interpersonal separation was increased slightly to 1.20 meters or 4 feet 1 inch, just beyond the inner boundary of Hall's "Social Distance Zone." This rather narrow bracketing of one of Hall's zone boundaries seems curious, but it does lend credence to the idea that an interpersonal distance of four feet has some exceptional significance in interpersonal interaction in our culture. At a minimum it seems to represent an equilibrium point for strangers engaged in standing interaction in an unfamiliar setting. Second, and somewhat to the author's surprise, certain measurable psychological needs and personality characteristics seem to be more highly correlated with differences in proxemic behavior than such readily apparent physical factors as the sex composition of the dyad or their similarity in physiology and dress (Chapter 3). Of the scales used, Rotter's (1966) Internal External locus of control score is the most consistently and significantly associated with such differences. Differences that were almost equally significant were associated with scores on Bendig's (1962) Emotionality scale in the experiment described in Chapter 4, but unfortunately this scale was not used in the experiment reported in Chapter 3. Third, a reciprocal effect on the total immediacy of interpersonal interaction was observed between the amount of talking and the interpersonal separation in the experiment reported in Chapter 3 (i.e., dyads where both members conversed freely used significantly closer interpersonal separations than those where neither member or only one member talked), but in the experiment reported in Chapter 4, where the verbal behavior of one member of the dyad was intentionally manipulated, no effect on the interpersonal separation was observed. The explanation favored by the author is that stress has a moderating effect on interpersonal separation and that the "natural" high verbal condition indicated a lack of stress in the experiment reported in Chapter 3, but not in the more contrived or manipulated situation presented to subjects in the experiment reported in Chapter 4. This idea is developed in greater detail later in this chapter. In general there were rather large differences in the behavior of all subjects, and groups of subjects categorized by their scores on the various pretest scales, between the two experiments. This would seem to indicate that subjects were quite sensitive to the behavioral manipulation of the confederates in the experiment reported in Chapter 4, although, with the exception of the manipulation of their verbal behavior, these seemed relatively unobtrusive to the author as he observed the experiment. A fifth observation is that results of the analysis of the factorial design of the experiment reported in Chapter 4 (Tables 1 and 4) are more supportive of an overall reciprocity model of the effects of the major interpersonal immediacy behaviors (proxemic, looking and verbal behavior) than a compensation model, with the exception of a relatively weak compensatory effect between proxemic and looking behavior that seems to be perceptual in nature. That is, the major


74

perceptual cues about interpersonal separation at the distances observed in these experiments are visual, and an increase in eye contact with the other person seems to make interpersonal separation more salient perceptually, resulting in a tendency to increase the separation to compensate for the greater perceived immediacy. Another significant result was that the global impression of the other person that is retained in the memory of the focal person is influenced more by how much the other person talked than by differences in their use of eye contact or the way they expressed the experimental conditions described in Chapter 4. The latter is particularly noteworthy because of the large differences in the interpersonal separation maintained by subjects with the two confederates. Finally it should be noted that the differences in the interpersonal separation that subjects maintained with the two confederates (Table 2 in Ch. 4) indicate that other, presumably more subtle, differences in their behavior which were not captured in this experiment were important determinants of the immediacy of the interpersonal interaction. It would, of course, be quite wrong to think that other variables, not observed in this research, do not have significant effects. An Approach Avoidance Paradigm Before proceeding to develop a model of the interaction of proxemic and other immediacy behaviors, it is useful to have a method of analyzing the purely proxemic behavior of subjects that is more revealing than the Chi square tests of gross differences in the number of moves made reducing, leaving unchanged, or increasing the interpersonal separation that were used in Chapters 3 and 4. One such method is what the author has termed approach avoidance analysis. The approach avoidance paradigm is a classical conceptualization of spatial behavior that has been used for many years. One particularly clear and concise statement of this paradigm is contained in the following four assumptions.

1. The tendency to approach a goal is stronger the nearer the subject is to it. This will be called the gradient of approach.

2. The tendency to avoid a feared stimulus is stronger the nearer the subject is to it. This will be called the gradient of avoidance.\

3. The strength of avoidance increases more rapidly with nearness than does that of approach. In other words, the gradient of avoidance is steeper than that of approach.

4. The strength o? the tendencies to approach or avoid varies with the strength of the drive upon which they are based. In other words, an increase in drive raises the height of the entire gradient. (Dollard and Miller, 1950, pp. 352 353)

With one addition and one minor modification discussed later, these assumptions and definitions can be directly applied to the analysis of proxemic behavior in the apparatus described in Chapter 2. Methodological issues in Approach Avoidance Analysis The methodology used in this dissertation permits a simple derivation of the gradients of approach and avoidance for groups of subjects engaged in interpersonal interaction. The derivation is illustrated in two steps below to illustrate how the Chi square test can be applied at an intermediate stage in the analysis (e.g., the first step) to determine whether the behavior of different groups of subjects differs in a statistically significant fashion. The first step is to count the number of moves made by subjects in the group of interest at each possible interpersonal separation (in moves) and to further categorize the moves at each separation as either reducing, leaving unchanged, or increasing the interpersonal separation. Recall that each possible move must fall into one of these categories. This is done in Figure 1 below for two groups of subjects in the experiment reported in Chapter 3, those who were arbitrarily labelled "Subject X" and those who were labelled "Subject Y" during the experimental runs.


75

It can be seen from Figure 1 that the distributions of moves of each type for both groups of subjects are fairly similar. This would be expected since the two groups of subjects were drawn from the same population, each subject being arbitrarily designated as X or Y at the time the experiment was run. However, it would be desirable to have a means of determining whether the observed differences are statistically significant. This can be done using the same Chi square test used in Chapters 3 and 4. Since the numbers of moves leaving the interpersonal separation reduced, unchanged, or increased at each possible interpersonal separation are independent of each other, there is no reason why the patterns of moves for the two groups of subjects shown in Figure 1 cannot be expressed as two 18 x 1 tables as shown in Figure 2, which permits them to be compared directly.


76

FIGURE 2

SEPARATION

REDUCED (R)

SEPARATION UNCHANGED (U)

SEPARATION REDUCED (I)

INTERPERSONAL SEPARATION

(MOVES) 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

SUBJECT X 0 0 8 12 10 2 0 5 50 28 14 4 1 11 18 14 3 0SUBJECT Y 0 1 7 17 14 4 0 8 60 33 9 3 0 5 12 6 1 0

Pearson Chi-Shi-Square = 15.15

P = 0.37 An alternative presentation of the moves of subjects "X" and "Y" shown in Figure 1. As can be seen from Figure 2, when the analysis is performed the resulting probability that the two distributions are different is non significant (i.e., p>.05). This is encouraging since it is the expected outcome, but, unfortunately, the result must be treated with caution. As noted in the section "A Primary Null Hypothesis" in Chapter 3, it is desirable to have a minimum expected frequency of 5 in each cell of the table if one is to be able to treat the results of the Chi square test with confidence. However, this requirement can be relaxed under certain circumstances. "In particular, if the number of degrees of freedom is large, then it is fairly safe to use the [Chi square] test for association even if the minimum expected frequency is as small as 1, provided that there are only a few cells with small expected frequencies (such as one out of five or fewer)." (Hays, 1973, p. 736) The data presented in Figure 2 is marginal in terms of these requirements. Since the three columns of data consisting of two zeros in Figure 2 are not analyzed, the number of degrees of freedom is the number of remaining columns minus one, or fourteen. This qualifies as "large." Also if "small expected frequencies" are defined as frequencies of two or less, a not unreasonable definition, then only one in five cells have small expected frequencies. The data presented in Figure 2 is typical of the data analyzed in deriving Figures 4 and 5 later in this section, so the technique is probably valid for this data, although rather marginally so. This difficulty with approach avoidance analysis can be eliminated in future experiments by using a larger number of subjects. Once the moves have been categorized as shown in Figure 2, the gradients of approach and avoidance can be obtained by plotting the percentages of the total number of moves made at each interpersonal separation that reduce or increase the separation respectively. In addition, another useful gradient, which the author has termed the "gradient of indifference" can be obtained by similarly plotting the percentage of moves leaving the interpersonal separation unchanged. The moves of all subjects (combining the moves of subjects X and Y shown separately in Figure 1) are plotted in this fashion in Figure 3.


77

FIGURE 3 The percentages of moves made by subjects reducing the interpersonal separation (gradient of approach), leaving it unchanged (gradient of indifference), and increasing it (gradient of avoidance) as a function of the interpersonal separation during the interaction phase (moves 4 12) of the experiment reported in Chapter 3. Note that the percentages of moves in the three categories shown in Figure 3 can also be interpreted as probabilities. For example, at an interpersonal separation of one move, 53% of the moves increased the interpersonal separation. This could be expressed, alternatively, as a probability of .53 that subjects will make a move increasing the interpersonal separation at a separation of one move. Examination of Figure 3 reveals that the gradients of approach and avoidance as defined in that figure meet all of the four "assumptions" of Dollard and Miller that were presented earlier in this section as defining the approach avoidance paradigm except for a trivial departure from the first one. In this case the gradient of approach decreases instead of increasing as one subject nears the other. The gradient of avoidance satisfies the second assumption, as would be expected. Taken together, the gradients of approach and avoidance as defined in Figure 3 also satisfy the third assumption since the strength of avoidance does increase more rapidly with nearness than does that of approach (which actually decreases as noted above). Finally, as required by the fourth assumption, an increase in drive would raise the height of an entire gradient. "Slack" in the system of analysis that would permit this to occur is provided by those moves leaving the separation unchanged, or the "gradient of indifference." As the height of, for example, the gradient of approach increases, given the same level of the gradient of avoidance,


78

the height of the gradient of indifference would have to decrease. It can be seen that the height of the gradient of indifference is an inverse measure of the strength of the other two gradients added together. The results of this approach avoidance analysis are also consistent (as they must be mathematically) with results reported in Chapter 3. The gradients of approach and avoidance cross at an interpersonal separation of about 2.5 moves, which agrees with the mean separation of 2.56 moves maintained by subjects during the interaction phase of the experiment. Also, the gradient of indifference reaches a peak at an interpersonal separation of 2 moves, about as would be expected since the most "comfortable" separation was 2.56 moves. It is, perhaps, noteworthy that the gradient of indifference declines so much between separations of two and three moves, since one would anticipate a more "rounded" peak occurring at 2 and 3 moves. The shift toward the closer separation may be due to subjects experiencing greater tension or stress at the closer separation and making more moves leaving the separation unchanged because of stress induced indecision. If so, the term "gradient of indifference" may be somewhat misleading as a description of this behavior on the part of subjects. Other names, including "gradient of uncertainty" and "gradient of satisfaction" were considered, but "gradient of indifference" seems to be somewhat more general and neutral in tone. It should probably be kept in mind, however, that moves leaving the interpersonal separation unchanged do not necessarily indicate indifference. An Application of Approach Avoidance Analysis Of the attitude scales included in the pretests used in the experiments reported in Chapters 3 and 4, scores on Rotter's (1966) I E scale were the most consistently associated with significant behavioral differences. However, as noted in the section headed "Pretest Variables Chi Square Analysis" in Chapter 4, these differences were manifested in different patterns of behavior in the two experiments. To attempt to explain these differences, approach-avoidance tables illustrating behavior in the two experiments were constructed and are shown in Figures 4 and 5. In Figure 4 the behavior of all subjects in the interaction phase (moves 4 12) of the experiment reported in Chapter 3 is compared with the behavior of all subjects in phase 3 (moves 7 12) of the experiment reported in Chapter 4.


79

As can be seen from inspection of Figure 4, the differences in approach avoidance behavior between subjects in the two experiments were quite pronounced. Whereas the gradients of approach and avoidance for the subjects in the experiment reported in Chapter 3 were monotonically increasing and decreasing (respectively) in slope over the entire range of interpersonal separations except for a slight reduction in the gradient of approach at a separation of five moves, these gradients in the experiment reported in Chapter 4 exhibit quite different characteristics. The gradient of approach has been substantially reduced in mean height (14% vs. 28% in the experiment reported in Chapter 3), and its slope is essentially flat as opposed to the positive slope of the gradient of approach observed in the experiment reported in Chapter 3. The gradient of avoidance is also somewhat different, exhibiting a "U" shape as opposed to the negative slope observed in the earlier experiment, although the mean heights of the two curves are about the same. The shapes of the gradients of indifference observed in the two experiments exhibit the same general "inverted U" shape, but the peak of the gradient observed in the experiment reported in Chapter 4 occurs at a greater interpersonal separation and its mean height is also somewhat higher. The most likely explanation for these differences is that it is an effect of the behavioral manipulations of the confederates in the experiment reported in Chapter 4. Overall it appears that subjects in Phase 3 of the experiment reported in Chapter 4 were somewhat unnerved by the behavior of the confederates, exhibiting a much reduced tendency or "drive" to approach them, maintaining a greater interpersonal separation, and apparently exhibiting a slight tendency to "flee" from the interpersonal interaction at separations greater than three moves. The approach avoidance behavior of subjects categorized by I E score in the two experiments is detailed in Figure 5 below.


80


81

When tested using the Chi square method illustrated in Figures 1 and 2, the approach avoidance behavior of the subjects in the experiment reported in Chapter 3 who had low I E scores was not significantly different from that of subjects in that experiment with high I E scores (P=.406).1 All other horizontal and vertical comparisons in Figure 5 are significantly (p<.Ol) different. Overall, a comparison of the four graphs in Figure 5 yields the following observations. 1. The approach avoidance behavior of all subjects was significantly altered by the conditions present in the experiment described in Chapter 4. 2. The differences in approach avoidance behavior between externals and internals were more pronounced in the experiment reported in Chapter 4 (Phase 3) than in the experiment reported in Chapter 3 (the "interaction phase"). 3. The approach avoidance behavior of internals was affected more drastically in the experiment reported in Chapter 4 than that of externals, particularly in the percentage of moves reducing the interpersonal separation (a difference of 3% vs. 24% for internals compared to 14% vs. 28% for externals). If, as seems likely, the behavioral differences are the result of stress produced by the unusual behavior of the confederates, then these results seem to support the conclusions in a review of research on locus of control presented below. _____________________ 1 This does not contradict the results of the Chi square tests reported in Chapter 3, since these involved only cases where the lower and higher scores within a single experimental run were compared. In the tables in Figure 5, the behavior of low (below the median) scorers is compared with that of high scorers regardless of the score of the other subject in the experiment. _____________________ All this suggests that, at least in some ways, externals are less disturbed. Put another way, they may be better able to handle immediately threatening material because their externality provides a convenient mode of anxiety reduction (denial). This is quite compatible with . . . evidence that suggests sensitizers show less evidence of physiological arousal to stress than repressors. Thus, repressors or internals may deny verbally what they fail to hide physiologically. (Phares, 1976, pp. 131 132) The explanation based on differential responses to stress is supported by the heights and slopes of the gradients of approach, avoidance and indifference for internals shown in the upper right hand graph in Figure 5. The reversal of the usually observed slope of the gradient of avoidance, combined with the very high gradient of indifference at close separations indicates that internals may be strongly inhibited at close interpersonal separations under stressful conditions, and tend to "flee" from the stressful situation at greater interpersonal separations.


82

Toward an Integrated Model of the Immediacy Behaviors

A Review of Existing Theories The three most prominent descriptive models of the interaction of the nonverbal immediacy behaviors, Argyle and Dean's (1965) "compensation" model, Breed's (1972) "reciprocity" model, and Patterson's (1976) "arousal" model were briefly reviewed in the section "The Nonverbal Immediacy Behaviors" in Chapter 1. These models have generally been used in connection with E. T. Hall's "proxemics" framework, described in some detail in Chapter 1. A relatively small number of writers have attempted to go beyond Hall's conceptual framework to a broader perspective on the causes and analysis of human spatial behavior. Two such articles (Evans and Eichelman, 1976; and Pederson and Shears, 1973) have been selected for review. Evans and Eichelman present an "outsiders"' characterization of the Hall framework and two others that they identify as currently relevant. Pedersen and Shears present a reinterpretation of proxemic terminology in the framework of system theory. Evans and Eichelman characterize the Hall framework as an "information overload model ." Probably the most widespread theoretical construct underlying most proxemic work is the notion that when we are in close proximity to another individual, we are forced to process more information than we normally do. This "overloads" the system causing confusion and stress. For example, Hall cites the increase in visual detail, odors, and temperature changes that result from an invasion of personal space (Hall, 1966). There are two aspects of this hypothesis which should be kept separate. One is that there: is some quantitative change in the stimulation impinging upon the organism so that in order to access the situation the individual must process more information and make decisions more rapidly. The other is that a qualitative change occurs such that the individual is unsure of the consequences of his decisions because he is not used to dealing with the new information. Actually it is the quantitative aspect that has dominated research strategies. (Evans and Eichelman, 1976, pp. 96 97) It could be debated whether this is a completely accurate description, but, having set it up, they proceed to criticize it on several grounds. First, they believe that it is an oversimplification. They argue that humans are not stimulus bound, but respond to what the stimulus symbolizes, a process which is affected by the entire history of the individual. They then cite a number of findings showing differing effects of personal space invasion (or crowding) and information overload or overstimulation) on the rate of maturation of organisms and task perfomance, among other things, to argue that they are different phenomena. Another framework is referred to as the "stress model." Although stress is a component of each of the models they describe, this one gives it a central explanatory position. It is argued that responses to crowding or personal space invasion are mediated by stress stress being defined in accordance with Selye's (1956) conceptualization.


83

Stress is a bodily state manifested by the general adaptation syndrome (GAS). The GAS is a tripartite, non specifically induced reaction consisting of alarm, resistance, and exhaustion. The physiological chain of events concurrent with the GAS is manifested by several indices centered around the endocrine system. The most common indicators include. . . increased heart rate, blood pressure, and skin conductance. (Evans and Eichelman, 1976, p. 88) These physiological signs are often produced by, and linked to, interpretations of external events rather than the direct physiological impingement of events on the organism. Behavioral indicators of stress "include increased reaction time, erratic performance, malcoordination, increased error and fatigue, self reports of stress, nervousness, and anxiety." (p. 88) In support of the stress model they cite evidence that spatial zones are generally enlarged under stressful conditions, a finding that is weakly supported by the research reported in this dissertation (see the second paragraph at the beginning of this chapter), and the presence of physiological and behavioral indicators of stress when personal space is invaded. The stress framework is quite useful for researching proxemic effects, providing, for example, a theoretical basis for using physiological measures in proxemic research. However, Evans and Eichelman do not find it entirely satisfactory. While there may be some overlap between indices of stress and unmet spatial needs, some differences exist. First, spatial restrictions by definition refer to a unique component of stressors related to unmet spatial needs. Second, there are data which indicate that the effects of stressors on developing organisms are quite distinct from those of crowding. Finally, there is considerable controversy and ambiguity over definitions of stress and consequently over indices of stress. (Evans and Eichelman, 1976, p. 88) Another current school of thought is identified as the "micro macro" or "individual social" perspective. Robert Sommer has thoughtfully argued that a fundamental issue in the design of space is the conflict between individual and societal priorities of function (Sommer, 1969, 1972). For example, society's needs for limiting costs in providing office space and classrooms may conflict with the personal needs of individual faculty and students . . . . A similar dichotomy has been suggested by Esser (1973). Building upon Paul MacClean's theory of the triune brain, he has suggested that a fundamental conflict exists in the operation of our old brain (the brain stem and limbic system) with the new brain (neocortex). This clash of primary functions and operational modes produces stress. It is of interest that the old brain is critically involved with emotional communication. His analysis suggested that emotional response is most often involved in individual, personal interactions, particularly with the small intimate group . . . .


84

Perhaps the interface between personal space and crowding can be viewed within this context. Under spatial invasion conditions, spatial impositions which may be more primitive are going to be more important than primarily informational factors such as social variables. (Evans and Eichelman, 1976, p. 103) Although Evans and Eichelman prefer this model to the information overload or stress perspectives, it seems more applicable to research on crowding or population density than that on interpersonal distance. Therefore it will not be elaborated further, its primary value being seen by the present author as conceptual rather than practical, and its implications for proxemic research are unclear. Perhaps the most interesting alternative theoretical formulation for proxemic research is the system theory approach of Pedersen and Shears (1973). They begin their discussion with a brief definition of basic system teminology. A system performs three distinct functions: input (sensing changes in the environment), "throughput" (evaluating sensed changes against internal criteria that yield a decision concerning their acceptability), and output (acting to alter the external state if it is unacceptable). Unacceptability is defined in terms of threat to the maintenance of an internal steady state of energy at a different level from the surroundings. An open system has a more complex organization than its surroundings and has the ability to maintain itself. It uses the three mechanisms to (a) process input over pathways, (b) process input against criteria within the system, and (c) send output to the environment to control environmental threats to its internal state. (Pedersen and Shears, 1973, p. 367) The three characteristics, throughput, input, and output are then characterized in terminology more familiar in the study of proxemic behavior. Throughput is identified with Kuethe's (1962) social schemas described in the section of Chapter 1 of this dissertation titled "A Simulation Study." Throughput has been illustrated by social schemata, and they in turn have been conceptualized as response sets. Thus, the throughput operation is akin to a response set. This function is an intervening variable and is not observable; it can be described according to its influence upon the action (output) of the person system. Some dimensions have been related consistently to actions. Demographic, cultural, and personality variables can be used to verify the presence of a framework of schemata. (Pedersen and Shears, 1973, pp. 370 371) Within the system framework, input to the person system is identified with the perception of events. Perception may be inferred or identified by observing expressive and intensive behavior in at least four ways. The first way is to ask the person if he is aware of an event. . . . The second way to identify perception is to take physiological readings from the


85

person system. . . . The third way of inferring input is by noting the bodily acts involved in the "orienting response". . . .2, The fourth way of inferring input is to note the responses the person makes in coping with changes in his surroundings. (Pedersen and Shears, 1973, pp. 371 372) Output is identified with instrumental behavior. Output consists of completing the feedback circuit with a message to influence the environment. The spacing of people at distances that are mutually satisfactory is conducive to a steady state within each of them. The use of eye contact brings out the interrelatedness of subsystems within the person system. Eye contact can be used to shorten or lengthen the comfortable distance between two people. (Pedersen and Shears, 1973, p. 372) This distinction between expressive behavior as a measure of input (perception) and instrumental behavior as output points up a major difficulty in proxemic research. The issue is whether the person system is engaging in instrumental or expressive behavior. In discussing this issue, it is important to distinguish between spatial behavior with and without awareness. Since awareness is likely to be the concomitant of discomfort with the existing spacing location, a shift from expressive to intentional behavior may occur. If awareness produces sufficient discomfort, the individual either changes his spacing or leaves the situation. (Pedersen and Shears, 1973, P. 377) ____________________ 2 Gibson and Peck (1963) discussed evidence that may be used to infer that stimuli have been received: "The classical senses in normal use require not only receptors but also muscles for adjusting them. . . Focused, stabilized, and appropriately centered retinal images, for instance, are the product of a whole complex of ocular responses which have been recorded and are fairly well understood. The sense organ adjustments, in short, are a form of observable behavior. The act of visual attention in another person can be perceived simply by watching his face, and we do in fact notice where a person is looking (p. 386)." (Pedersen and Shears, 1973, p. 371) ____________________


86

FIGURE 6

A minor revision of Patterson's (1976) arousal model of interpersonal intimacy, emphasizing its "systems" components.

It remains to be seen whether the systems perspective will advance proxemic studies. However, its application has been so fruitful in other areas of research in both the physical and behavioral sciences that it seems worthwhile to utilize a model that is compatible with systems terminology if this can be accomplished in a "natural" way that does not require awkward manipulations of the data. Fortunately, the most general of the available models of the interaction of the immediacy behaviors, Patterson's (1976) arousal model, is presented in a systems format, although Patterson does not make this explicit. This model is shown in a slightly modified form to emphasize its systems components in Figure 6. Here the change in A's intimacy is the input to the system (i.e., person B), person B's internalized behaviors of arousal change and emotional labeling are equated with the throughput of the system, and the overt behavioral adjustments represent the system output which is fed back to person A. An Integrated Model A notable feature of the three major existing models of the interaction of the interpersonal immediacy behaviors (Argyle and Dean, 1965; Breed, 1972; and Patterson, 1976) is their simplicity. Argyle and Dean postulate one primary mechanism, compensation, to account for the most commonly observed pattern of results while Breed suggests another mechanism, reciprocity, to account for certain discrepant findings. Patterson's model, reproduced in Figure 6 above, is the most elaborate of the three but it essentially combines the other two, proposing that differences in the "emotional labelling" of a perceived change in intimacy determines whether the response will be reciprocal or compensatory.


87

The use of such simplistic models at the present time may be justified by the relative newness of the immediacy behaviors as a field of study and the relative scarcity of unambiguous and generally accepted findings in the related research. However, these models are unsatisfying in at least two respects. First, they are not as helpful as one might wish in suggesting theoretical issues of interest or directions for further research. Patterson's model is better than the other two in this regard since it does make concrete, testable predictions about relationships among relevant and measurable qualities that seem intuitively meaningful in interpersonal interaction, such as arousal change and emotional labelling. however, even Patterson's model seems unacceptably sparse when the range of human immediacy behaviors is considered. The second source of the author's dissatisfaction with existing theories is related to the first, but somewhat more basic and philosophical. Interpersonal immediacy behavior is a very significant and fundamental aspect of the experience of being human. Any theory that purports to explain such behavior would ideally be able to accommodate and interpret a broad range of human experience. Logically, any conceptual model of interpersonal behavior presupposes a conceptual model of the individuals who are interacting. On the other hand there is some question whether the current "state of the art" in proxemic research is sufficient to justify advancing more detailed models. Certainly no theoretical approach to interpersonal immediacy can currently be presented with any assurance that it will ultimately be proven correct, given the present state of the relevant empirical research. Nonetheless, the author feels that it is useful to develop a more elaborate theoretical framework within which the findings of the present research can be interpreted. The term "theoretical framework" is carefully chosen. The model illustrated in Figure 7 is presented only as the author's personal view, not as being "true" or proved by the research described in this dissertation. It represents a statement of the author's philosophical speculations and subjective experience more than it represents an analytical interpretation of the research findings. However, these findings can be interpreted rather straightforwardly within the model, as will be demonstrated by Figure 8 and the associated text. Hopefully, viewing the model as a theoretical framework will emphasize its provisional nature and focus the attention of the reader on its primary purpose, which is to suggest areas of research on interpersonal immediacy that would be useful in either confirming and elucidating specific linkages proposed in the model or in disconfirming this model and pointing the way to one that is More satisfactory. The primary justification for the general model shown in Figure 7 is to utilize a terminology that is both humanistic and compatible with some basic findings in neurophysiology about how stimuli are processed in the nervous system. The term humanistic is used in a very broad sense and is meant to indicate that the model does not force one to deny the significance of any facet of the subjective experience of being human, although, depending upon how the individual evaluates and conceptualizes his or her own subjective experience he or she may find it more or less difficult to frame that experience in the categories used in the model. The best example of a theoretical framework that is "non humanistic" in the sense that the author uses this term is pure behaviorism, which denies the significance and in extreme presentations even the existence of cognition and affect, or at best considers them epiphenomena playing no causal role in human behavior. Figure 7 consists primarily of a schematic representation of a person, labelled "B" to facilitate comparison with Patterson's model shown in Figure 6. Person B is shown as existing in a social and physical environment where the dominant behavioral norm is reciprocity in social and physical exchange, an environment where, as the expression goes "there is no free lunch." B is represented as composed of three concentric circles labelled behavior, affect, and cognition. For purposes of conceptual clarity these terms are defined as follows. Behavior is any external activity of person B that can be observed or measured or recorded using physical instruments. Affect and cognition are defined as internal, subjective experiences that cannot be directly observed by physical devices. Specifically, affect is an experienced state of subjective arousal which may be labelled (happy, sad, love, hate, etc.) or unlabelled. The model requires no physiological definition of affect, since it is concerned only with the subjective experience, but it is compatible with findings that affect seems to be physiologically associated with chemical changes in the nervous system and the body. The definition of cognition, again a subjective one, is an internal verbal or pictorial representation.


88


89


90

It is the author's view that these three categories, behavior, affect, and cognition, as defined above, constitute both a necessary and a sufficient set with which the entire range of individual, subjective human experience can be described. As a test of the completeness of this set, the reader is invited to think of a personal experience that cannot be completely described to another person by describing the physical activity (behavior), the feelings (affect) that were experienced, and the way you represented the experience to yourself either verbally or pictorially (cognition). In the author's experience with this exercise different people may disagree in certain cases about whether a certain component of an experience was cognitive or affective (hence the precautionary label "boundary is indistinct" between affect and cognition in Figure 7), but no fundamentally new category seems to be needed. Of course it should also be noted that the use of this simple theoretical framework by no means reduces the actual complexity of human experience, which reappears in the great variety possible within each of the three categories. The value of the framework lies in the possibility that there are fundamental causal relationships among subjective experiences of an event on the three different levels. There is nothing new about this idea. Research on cognitive dissonance, to give only the most obvious example, has established certain relationships between cognition and behavior. However, this type of analysis has not, to the best of the author's knowledge, been applied to the study of interpersonal immediacy. The arrangement of the components of the model, with behavior and cognition separated by affect, reflects a fundamentally dualist philosophical perspective. That is, as has long been argued by that school of philosophy, we as human beings seem to deal simultaneously with two representations of reality. One existing outside ourselves in the environment in which we behave, and one existing within ourselves in our cognitive representations. Affect seems, as subjectively experienced, to mediate between these two representations. When the cognitive representation agrees or "fits" well with external reality (and, to get a bit ahead of the description of the theoretical framework, when the level of stimuli is moderate) the subjective experience of affect is minimal. There is a general feeling of well being perhaps, but it is usually out of awareness. Affect seems to become more prominent when there is a conflict or contradiction between the internal cognitive representation, "how we expect things to be" or "how things should be," and the external reality "how things are." Another feature of the model represented in Figure 7 is the series of arrows beginning in the upper left of the diagram labelled "increase in A's intimacy" (again to facilitate comparision with Figure 6) and splitting into two parts labeled "reflex pathway" and "conscious pathway." These arrows represent the hypothesized "route" of incoming stimuli and suggest the primary mechanism proposed by the author for compensatory and reciprocal immediacy behaviors. As with other significant terms in the model, the words "reflex" and "conscious" are used in a special sense. These particular terms are used because their definitions in common usage are close to what the author intends to convey, and also because they are used in a similar way by House, Pansky, and Siegel (1979) in their book on neuroanatomy. In general the findings of research in neuroanatomy show that incoming stimuli are transmitted to the brain through the spinal cord and through afferent nerves that terminate in the brainstem. From the brainstem and cerebeller cortex, nerve impulses are routed both directly to efferent nerves, producing "reflex" or unconscious behavioral responses, and to the associative areas of the cerebrum, which are associated with cognition. nerve impulses from the cerebrum are then routed back through the brainstem and outward through efferent nerves to be translated into behavior through the muscles. This latter pathway, involving the cerebrum, corresponds to the author's "conscious" pathway.3 It should be noted that we can become conscious of our reflex pathway responses by observing the resulting behavior, and conversely that we are not always "conscious" of our conscious pathway responses, which often ____________________ 3 This is, of course, an oversimplification of the actual neurological process, which is only partially understood. Interested readers are referred to House, Pansky, and Siegel , 1979 or to Stratton, 1981 . In particular there is a more complex, but recognizably similar, diagram on page 266 in Stratton that is itself described as "simplistic." ____________________


91

proceed outside of awareness. However, it appears that we can be directly conscious of the internal representational aspects of conscious pathway responses if we attend to them, while reflex pathway responses are not directly associated with internal representations, although secondary internal representations can be formed by observing reflex pathway produced behaviors. The terms "conscious pathway" and "reflex pathway" are used in this sense. The author's basic hypothesis is that reciprocal immediacy responses generally result from the action of the conscious pathway, but that the information processing capacity of this channel is limited. Therefore, to avoid overloading the conscious pathway (Argyle and Dean's "information overload"), in high immediacy situations we engage in compensatory immediacy behaviors, mediated largely by the reflex pathway, to reduce the total amount of incoming stimuli to a level that we can comfortably process. Having now defined the most significant terms, the hypothesized mechanism producing the interpersonal immediacy behaviors can be described as follows. In the absence of information overload, the primary mechanism is the "conscious pathway" and the dominant (but not necessarily exclusive) mode of behavior is reciprocity. As stimuli cross the subjective "person environment boundary" (see Figure 7) they first result in unconscious affective "arousal." This prepares the system to respond. The stimuli are then processed cognitively, being compared with internal representations of past experience that can be subjectively experienced, if attended to, as internal verbalizations or images. Depending on the nature of these internal representations, individual differences in immediacy behavior will be manifested. Past experience, the amount of processing that must be done to interpret the experience, etc. will all play a part in determining the resulting behavior. However, it is suggested that social exchange reciprocity is the dominant organizing principle of our conscious pathway responses. Our physical survival ultimately depends upon our ability to engage in productive exchanges with our environment although some opportunities for exchange that are presented to us may be rejected, on the basis of our cognitive representation of them, as unproductive. The next step in this conscious pathway processing is the routing of the resulting stimuli back through the brainstem where, it might be hypothesized, the cognitively processed experience acquires a "higher" affective label such as love, hate, friendship, etc. Finally, the resulting subjective cognitive and affective representation is translated into behavior as indicated in Figure 7. However, if the amount of stimuli initially entering the system is too great to be comfortably processed by the conscious pathway, the reflex pathway will be actuated. The function of this pathway is to reduce the amount of stimuli entering the conscious pathway to a manageable level, so the resulting behavioral responses are primarily compensatory to reduce the amount of information entering the system. As shown in Figure 7, these responses would normally be out of awareness and not accompanied by an internal cognitive representation although they are probably associated with some rudimentary affective labelling (annoyance, unpleasant, etc.). The totality of person B's immediacy behaviors is therefore hypothesized to result from the interaction of these two primary mechanisms. One natural consequence of the hypothesized limited capacity of the conscious pathway is a ready explanation of the need to severely restrict the amount of incoming stimuli when we are engaged in activities that require a large amount of cognitive processing, such as reading. Pursuing this line of thought, it may be that people who habitually engage in a large amount of cognitive processing would tend toward more compensatory immediacy behaviors, giving the appearance of "shyness" in their immediacy behavior. One possible interpretation of behavioral correlates of such cognitive differences as internal vs. external locus of control, which will be discussed later, could be based on this line of reasoning. Of course there are also, no doubt, organic individual differences in information processing capacity. The primary deficiency of the model presented in Figure 7, if indeed it is valid at all, is probably that it might give the impression that immediacy behavior is the result of a single pass by stimuli through the system. In actuality, of course, it would be a continuous, self adjusting process, and actual responses are probably built up from a multitude of inputs involving moment by moment adjustments and feedback between the two pathways.


92

Research Findings in the Integrated Model Some of the most consistent and significant results of the current research can be embedded in the theoretical framework presented in Figure 7 as shown in Figure 8. Since Figure 8 may at first appear to be forbiddingly complex, it may be helpful to discuss the format of the figure itself. A good place to start is with the heading "Location in Figure 7" in the upper left corner of the figure. This refers to the row of headings extending to the right. Each of these headings identifies the column extending below it and, as one moves from left to right, the columns correspond to the path of the arrows in Figure 7. This is further indicated by the top row of headings in Figure 8 which identifies the conscious and reflex pathways. Moving down the column at the far left of Figure 8, directly under the top heading, is a long space containing the heading "Hypothesized effects." The wide row extending to the right of this contains the research findings and the arrows representing the hypothesized relationships among them. This is the main body of the figure. Notice that along with the heading "Hypothesized effects" there are two other headings in parentheses. Person B's sensory cues about A's proxemic and looking behaviors are primarily visual. Therefore these behaviors are combined for purposes of the model as shown in the second column. B's cues about A's talking behaviors are primarily aural, so talking is diagrammed separately. This separation of visual and aural sensory pathways is a further elaboration of the basic framework presented in Figure 7, and it is discussed further below.


93

FIG. 8


94

The bottom two rows in Figure 8 contain supplementary information providing references to the text of this dissertation and to the primary source in the literature describing the effect, if one exists. The first feature of the model that should be noticed, is that both the visual and aural sensory pathways are hypothesized to operate in the same fashion. That is, higher stress or arousal produced by an excess of incoming stimuli activates the reflex pathway producing compensatory immediacy behaviors on the part of person B. If the stimuli do not produce excessive arousal or stress, the conscious pathway predominates and immediacy responses are primarily reciprocal. The rationale for considering the sensory pathways separately is that, while significant compensatory responses were observed in subjects' proxemic and looking behaviors in the two experiments, only reciprocal effects were observed in response to increases in talking. The experiments themselves give no clue as to why this occurred, but two possible explanations could be as follows. First, possibly the two sensory pathway do operate somewhat independently, and the experimental conditions may have been such that the conscious visual pathway was overloaded but not the conscious aural pathway. Alternatively, the level of arousal or stress required to activate the reflex pathway for verbal behavior may be higher than for proxemic and looking behavior. A second possibility is that increased talking may, in itself be a mechanism for reducing stress or arousal. Further experiments would be required to explain this phenomena and, of course, it could also indicate a basic error or incompleteness in the proposed theoretical framework. For the sake of clarity, only two of the findings from the pretest and posttest psychological scales are included in Figure 8. The first of these, locus of control, which is shown in the visual sensory pathway, is included because behavioral differences were so consistently observed in association with different scores on this variable in both experiments. Of the variables included on the pretests in both experiments, locus of control score clearly seems, more than the others, to tap a significant dimension in interpersonal interaction.4 However, the precise nature of the underlying mechanism producing these behavioral differences is unclear. As shown in Figure 8, locus of control is represented as a cognitive variable whose behavioral effects are potentiated by stress or arousal (see Figure 5 and the associated text in this chapter). One possible explanation of the observed differences in behavior might be constructed as follows. Both internals and externals experienced stress in the experiment because the experimental conditions overloaded the visual conscious pathway. Under conditions of stress the person with a more external locus of control, because of their propensity to accept events as externally caused and to reject "ownership" of them, will respond in much the same way as if they were not under stress because they are predisposed to believe that they have no ability to control the situation in any case. On the other hand, a person with a more internal locus of control will tend to attribute the stressful nature of the situation to a loss of control and to exhibit a more proactive pattern of behavior in an attempt to bring the situation under control, as would be required for consistency with their "internal" orientation. An explanation of this type would be consistent with the observed facts, but this should not be allowed to blind us to other possible interpretations. Another possibility is that stress and locus of control are not separate and distinct entities, as is implied in the first explanation presented above, but are clues to a more fundamental causal reality which is the characteristics, state, and condition of the conscious pathway itself. An interpretation of the observed behavioral differences based on this conceptualization might be as follows. Actually, internals and externals respond to experienced stress in the same way, but internals experience stress, and therefore actuate the reflex pathway, at lower levels of sensory immediacy input than externals. This reduced ability of internals to handle immediacy stimuli without stress is caused by the reduction in the capacity of the conscious pathway produced by ____________________ 4 There were behavioral differences of similar magnitude associated with Bendig's (1962) "Emotionality" scale, in the experiment reported in Chapter 4, but, unfortunately, this scale was not included in the pretest used in the experiment reported in Chapter 3. ____________________


95

the additional cognitive processing which is inherent in the internal orientation. It seems reasonable to hypothesize that an internal does more cognitive processing or "explaining" of events than an external because, among other things, an internal needs to rationalize events to maintain the experience of being in control to a much greater extent than an external, who operates under no such assumption and therefore is able to accept events and respond to them "in the moment" without as extensive a burden of cognitive processing. According to this view, the behavioral differences between internals and externals are not caused by the cognitive differences reflected in the locus of control score per se, but by the reduced amount of information that can be processed in the conscious pathway by internals. Although the experimental designs and data collection methods used in the two experiments described in this dissertation do not permit a meaningful analytical test that would distinguish between the two possible explanations discussed above, these explanations do provide an example of the potential usefulness of the analytical framework described in Figure 7 for generating hypotheses for future research. In fact, it seems feasible to design an experiment that would support one over another. The key to making such a distinction would be to determine whether internals begin to experience stress or arousal in interpersonal interaction at lower levels of sensory immediacy than do externals. Self reports as well as behavioral and/or physiological indicators of stress could be monitored under conditions where the amount of stimuli was controlled, in order to provide an answer to this question. A negative result would be more supportive of the first explanation and would indicate that observed behavioral differences result directly from associated differences in the cognitive representation of the experimental situation, as is suggested by Figure 8. A positive result would tend to support the second alternative, that it is the additional cognitive processing required to maintain the internal perspective that produces the behavioral effects by constricting the conscious pathway, rather than the content of the cognitive processing itself. If greater "emotionality" as measured by Bendig's (1962) scale (see Appendix D) is equated with a propensity to engage in less cognitive processing, then the global similarity between the behavior of externals and high "emotionality" scorers, as shown in Chapter 4, Table 8, may provide some indirect support for this hypothesis. The only feature of Figure 8 that has not been discussed is the increase in B's liking of A associated with increased talking by A. This result (see Table 1 in Chapter 4) has particular significance with respect to Patterson's model, reproduced as Figure 6, because this model postulates "emotional labelling" as the factor determining whether B's immediacy responses will be reciprocal or compensatory. In general, the results of the research reported in this dissertation do not support this hypothesized mechanism. The fact that no significant differences in the subjects' "liking" of the confederates were associated with the confederates' manipulation of their looking behavior could be due to methodological problems connected with the way looking behavior was manipulated or to the use of the post test "liking" score as a measure of "emotional labelling," but it does weaken the case for emotional labeling as the primary determinant of the way immediacy responses are differentiated. In fact, the observed tendency toward reciprocal responses in verbal interaction was so strong (p=0.000 in every phase of the game, see Table 1 in Chapter 4) that the existence of the compensatory "reflex pathway" in the aural sensory pathway shown in Figure 8 is only hypothetical and not based on the observed results of the experiment. The observation of compensatory effects between subjects looking and distancing behaviors, but not in their verbal behavior, complicates the theoretical framework somewhat. It would appear that if the framework is correct in postulating stress or arousal as the primary factor determining whether immediacy responses will be predominantly reciprocal or predominantly compensatory, then a given level of generalized stress can have different effects on verbal and non-verbal immediacy behaviors. If the visual and aural sensory pathways are, in fact, processed at least somewhat separately, then one possible explanation is that the experimental conditions were such that the confederates' behavioral manipulations may have overloaded the visual sensory pathway but not the aural pathway. This seems reasonable, since in the "high verbal" condition the confederates were told to maintain a "casual, friendly conversation," a "natural" condition, whereas in the high eye contact condition they maintained their gaze constantly in the direction of the subjects' faces, which is more than the usual "natural" amount of eye contact. An alternative explanation for the


96

failure to observe compensatory responses in the subject's verbal behavior is that social norms of reciprocity in verbal interaction may be stronger than those for the nonverbal behaviors. It is unclear how an experiment could be designed to distinguish between these two explanations, but the theoretical framework does suggest an experiment that would be expected to elicit compensatory verbal responses. In such an experiment one might set up a situation where subjects are engaged in a "cognitive task" requiring one of a number of possible different levels of cognitive processing, such as adding numbers presented at one of a number of different frequencies. A confederate would talk to "distract" subjects while they were engaged in the task, and the subjects' verbal responses would be observed. At some level of cognitive effort on the task, the conscious pathway should become overloaded and compensatory verbal responses should begin to occur. An instrument similar to, but perhaps more elaborate than, the posttest used in the experiment reported in Chapter 4 could be administered to subjects following the experiment to determine whether negative "emotional labelling" is associated with compensatory verbal immediacy behaviors, as would be predicted by both Patterson's model (Figure 6) and the author's framework.

Concluding Discussion

This dissertation represents what seems to the author at this point to be perhaps a rather overly ambitious attempt to (a) develop a new methodological approach to the study of proxemic behavior, (b) develop the necessary analytical techniques, including computer programs, needed to interpret the results, (c) validate the methodology, and (d) apply it to a meaningful test of some basic theories of interpersonal interaction. Of necessity, many questions have been left unanswered along the way. The author feels that the process of preparing this dissertation has improved his own understanding of interpersonal behavior. It is even possible that the rather speculative and incomplete analytical framework for interpersonal immediacy presented in this chapter may after considerably more experimentation and confirmation or revision eventually advance the general scientific understanding of the relationships among some basic interpersonal behaviors. This framework is, in a fundamental sense, the result of a tension or conflict in the mind of the author between the "scientific perspective" that wishes to discover general principles and laws of human behavior that will allow predictions to be made and situations to be structured in such a way as to produce desired behaviors, and the "humanist perspective" that wishes to preserve the richness, diversity, and apparent unpredictability of human behavior. The solution to this dilemma that is incorporated into the model is essentially to define subjective human experience as being composed of combinations of three elements, behavior, affect, and cognition, among which certain more or less mechanical "scientific" relationships are postulated. In addition an essentially mechanical mechanism, information overload in the "conscious pathway" is proposed to account for gross differences between compensatory and reciprocal immediacy behaviors. The humanist perspective re emerges in a recognition of the great diversity of possible human experience in thought, feelings, and behavior. Since, at least in our culture, we do not usually conceptualize our experience as being composed of combinations of cognition, affect, and behavior, but in categories that are more socially oriented (friendly/ hostile, pleasant/unpleasant, introverted/ extroverted, etc.) this may tend to conceal the degree to which our experience, in its totality, forms a lawfully related whole. Similarly, since the major proposed mechanism for compensatory immediacy behavior, the "reflex pathway," normally operates outside of our conscious awareness, this may conceal the extent to which our subjective experience is shaped and limited by our finite capacity to process information. If this view of human behavior is valid, it does have implications affecting the way we interpret certain psychological variables. An example of this was given in the prior section in the discussion of the behavioral correlates of locus of control and emotionality scores. To some extent these behavioral correlates may be caused by underlying predispositions toward a greater or lesser degree of cognitive processing which is indicated by the score on these variables, rather than by a behavioral expression of the different cognitive representations that are tapped by these variables.


97

To the extent that our subjective experience is ordered by a more or less hidden underlying dynamic and this dynamic can be articulated and brought into the awareness of people, and to the extent that our limited capacity to process information controls our experience outside of our awareness, and people can be made aware of this, then it is the author's contention that research directed along these lines can only enrich the human experience and make it more productive. Because of this the author feels no compunction about describing himself as a humanist, despite the rather "mechanistic" nature of the research described in this dissertation and the proposed analytical framework.


98

BIBLIOGRAPHY

Altman, I. Environment and Social Behavior: Privacy, personal space, territory and crowding. Monterey, California: Brooks Cole, 1975. Altman, I., and A. M. Vinsel. "Personal Space: an Analysis of E. T. Hall's Proxemic Framework." In Human Behavior and Environment, ed. I. Altman, J. Wohlwill, Vol. 2, New York: Plenum, 1977. Anderson, N. H. "Likableness Ratings of 555 Personality Trait Words." Journal of Personality and Social Psychology, 9, 272 279, 1968. Argyle, M. and J. Dean. "Eye Contact, Distance, and Affiliation." Sociometry, 28, 289 304, 1965. Bakken, D. "Behavioral Adjustment in Nonverbal Immediacy: A Methodological Note." Personality and Social Psychology Bulletin, 4, 301 303, 1978. Baldassare, M. and S. Feller. "Cultural Variations in Personal Space." Ethos, 3, 481 503, 1975. Ball, Donald W. Microecology: Social Situations and Intimate Space. New York: Bobbs Merrill, 1973. Baxter, J. C. "Interpersonal Spacing in Natural Settings." Sociometry, 33, 444 456, 1970. Bendig, A. W. "The Pittsburgh Scales of Social Extraversion, Introversion and Emotionality." The Journal of Psychology, 53, 199 209, 1962. Breed, G. "The Effect of Intimacy: Reciprocity or Retreat." British Journal of Social and Clinical Psychology, 11, 135 142, 1972. Christie, R. "Mach IV" in Measures of Social Psychological Attitudes, Revised Edition by J. Robinson and P. Shaver. Ann Arbor: University of Michigan Institute for Social Research, 1973, pp. 593 594.


99

Christie, R., J. Havel, and B. Seidenberg. "Is the F Scale Irreversible?" Journal of Abnormal and Social Psychology, 56, 143-159, 1958 Coutts, L. T., and F. W. Schneider. "Affiliative Conflict Theory: An Investigation of the Intimacy Equilibrium and Compensation Hypothesis." Journal of Personality and Social Psychology, 34, 1135 1142, 1976. Dollard, J. and N. Miller. Personality and Psychotherapy. New York: McGraw Hill, 1950. Esser, A. H. "Experiences of Crowding." Representative Research in Social Psychology., 4, 207 218, 1973. Evans, G. W. "Personal Space: The Experimental Approach." Man Environment Systems, 2, 3, May 1972. Evans, G. W. "Personal Space: Research Review and Bibliography." Man Environment Systems, 3, 4, July 1973. Evans, Gary W. "Human Spatial Behavior: The Arousal Model" in Baum, A. and Epstein, Y. (Eds.) Human Response to Crowding. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 1978. Evans, Gary W. and William Eichelman. "Preliminary Models of Conceptual Linkages Among Proxemic Variables." Environment and Behaviors, 8, 87 116, March 1976. Evans, Gary W. and Roger B. Howard. "A Methodological Investigation of Personal Space." EDRA III, Proceedings of the Third Environmental Design Research Association Conference, Los Angeles, California, January 1972 (Prepublication copy reviewed). Evans, G. W. and R. B. Howard. "Personal Space." Psychological Bulletin, 80, 4, 334 344, 1973. Gardner, Martin. "On Tessellating the Plane with Convex Polygon Tiles". Scientific American, Vol. 233, No. 1, July 1975, pp. 112 117. Gibson, J. J., and A. D. Peck. "Perception of Another Person's Looking Behavior." American Journal of Psychology, 76, 368 394, 1963. Goffman, Erving. Behavior in Public Places. New York: The Free Press of Glencoe, 1963. Gullahorn, J. T. "Distance and Friendship in the Gross Interaction Matrix." Sociometry, 15, 123 134, 1952.


100

Hall, Edward T. The Silent Language. Garden City, New York: Doubleday, 1959. Hall, Edward T. "A System for the Notation of Proxemic Behavior." American Anthropologist, Vol. 65, No. 1, Feb. 1963, pp. 1003 1026. Hall, Edward T. "Silent Assumptions in Social Communication." Disorders of Communication, Research Publications, Association for Research and Mental Disease, Vol. 17. Baltimore: Williams and Wilkins, 1964. Hall, Edward T. The Hidden Dimension. Garden City, New York: Doubleday & Co., 1966. Hall, Edward T. Handbook for Proxemic Research. Washington, D. C.: Society for the Anthropology of Visual Communication, 1974. Hays, W. Statistics for the Social Sciences Second Edition. New York: Holt, Rinehart & Winston, 1973. Higgins, J., J. C. Peterson and L. L. Dolby. "Social Adjustment and Familial Schema." Journal of Abnormal Psychology, 74, 3, 296 299, 1969. House, E., B. Pansky, and A. Siegel. A Systematic Approach to Neuroscience Third Edition. New York: McGraw Hill, 1979. Kuethe, J. L. "Social Schemas." Journal of Abnormal and Social Psychology, 64, 31 38, 1962. McBride, G., M. G. King and J. W. James. "Social Proximity Effects on Galvanic Skin Responses in Adult Humans." Journal of Psychology, 61 , 153 157, 1965. Patterson, M. "Compensation in Nonverbal Immediacy Behaviors: A Review." Sociometry, 36, 237 252, 1973. Patterson, M. "An Arousal Model of Interpersonal Intimacy." Psychological Review, 83, 235 245, 1976. Pedersen, Darhl M. and Loyola M. Shears. "A Review of Personal Space Research in the Framework of General System Theory." Psychological Bulletin, 80, 367 388, 1973. Phares, E. J. Locus of Control in Personality. Morristown, N. J.: General Learning Press, 1976.


101

Rotter, J. B. "Generalized expectancies for internal versus external control of reinforcement." Psychological Monographs, 80 (Whole No. 609), 1966. Schutz, W. FIRO: A Three Dimensional Theory of Interpersonal Behavior. New York: Holt, Rinehart & Winston, 1958. Selye, H. The Stress of Life. New York: McGraw Hill, 1956. Sommer, Robert. Personal Space: The Behavioral Basis of Design. Englewood Cliffs, New Jersey: Prentice Hall, 1969. Sommer, R. Design Awareness. San Francisco: Rinehart, 1972. Stratton, D. Neurophysiology. New York: McGraw Hill, 1981. Watson, 0. Michael, and Theodore D. Graves. "Quantitative Research in Proxemic Behavior." American Anthropologist, 68, August, 971 985, 1966. Woodworth, R. S. and H. Schlosberg. Experimental Psychology: Revised Edition, New York: Holt and Co., 1954.


102

Appendices

To obtain appendices A, B, D, and E, please contact the author at:

[email protected] APPENDIX C is included in this document below

Appendix C

PROBLEMS IN THE DEVELOPMENT OF

A SINGLE CRITERION OF INDIVIDUAL BEHAVIOR


103

Analysis of the problem to date indicates that there are two potentially useful approaches to constructing a criterion of individual behavior discussed in the section "Issues in Data Analysis" in Chapter 3. The first and simplest, discussed in Chapter 3, involves the construction of a number which contains information about the behavior of the individual, but which does not fully utilize the intrinsic statistical properties of the floor pattern. This number is then used in a conventional statistical analysis such as a multiple regression analysis. Such a number could be constructed in a variety of ways, depending upon the information about the behavior of the person that one wishes to incorporate in it. Assuming that we wish to detemine the extent to which the person's moves affect the interpersonal separation in the device by either (a) reducing it, (b) leaving it unchanged, or (c) increasing it, the following method yields three relatively simple and potentially useful criteria.

Then numbers quantifying the subject's efforts to reduce the separation (Cr ), leave it unchanged (Cu), or increase it (Ci) can be constructed as follows.

Advantages of this approach include its simplicity and versatility (many other numbers could be constructed to represent the behaviors summarized by the three criteria defined above). The disadvantage is that, while these numbers allow the experimentor to compare the behavior of different groups of subjects, they do not contain information that would allow the experimentor to make statistical inferences about the purposefulness of such behavior on the part of any particular subject. To obtain this information, a more complex procedure is required which utilizes more of the total information contained in the pattern of the subject's moves. An approach to developing such a procedure is outlined below. Suppose that we are interested in determining the probability that a subject's moves reducing the interpersonal separation of the two subjects are due to chance. If two simplifying assumptions are made the problem is relatively simple. The first assumption is that the probabability of a move reducing the interpersonal separation is always the same (for example


104

1/3). The second assumption is that we can dichotomize all possible moves into two categories, those that reduce the interpersonal separation and those that don't. Notice that the second assumption implies that, for purposes of this analysis, a U move is equivalent to an I move. Given these two assumptions, the problem is easily solved by recourse to the formula for the binomial distribution. This formula gives the probability of k "successes" in N "attempts" where the probability of success in any single attempt is p (and where q = 1 p) as follows.

In the experiments described in this paper, for example, each move by a subject is an "attempt." Since there are a total of 12 moves per subject per experiment, N = 12. A "success" occurs when the subject makes a move reducing the interpersonal separation, and, according to our first assumption, the probability of success is 1/3. Using these assumptions, the following table has been constructed.

Since the actual criteria desired is the probability of k or more successes, this indicates that 7 "R" moves in this experiment would have a probability only slightly higher than p = .05 of chance occurrence, while 8 "R" moves would indicate p<.05 that this behavior was not purposeful. Now the question becomes, "What happens if we relax our first simplifying assumption that all probabilities of success are the same? Examination of the Equilab floor pattern shows that there are five different possible probabilities that a move by one subject in a two subject game will reduce the interpersonal separation. The actual probability (i.e., which of the five possible probabilities) is a function of the positions of the two subjects. If a subject is on one of the six "corner" nodes at the end of the pattern, the only possible probabilities are 1/3 or 2/3 (which one depends on the position of the other subject). If a subject is on a node at the edge of the pattern that is not a "corner," the possible probabilities are 1/4 or 1/2. At any other node, the probabilities are 1/3 or 1/6. Since 1/3 is repeated, there are only five different probabilities for a move that reduces the interpersonal distance 1/3, 1/6, 1/2, 1/4, or 2/3. A preliminary analysis indicates that the frequency of occurrence should decrease according to the above ordering, with about 2/3 of the actual probabilities having the value 1/3. For a twelve move game, the series of probabilities that each move of a subject will result in a reduction of the interpersonal distance might typically be similar to the example that follows, with the "successes" (actual moves reducing the distance) circled.


105

In this example there are six successes. The first step in determining the probability of this occurring by chance is to enumerate all the possible ways that six successes could occur. Summing all the moves with the same probabilities shows that there are 8 where p = 1/3, 2 where p = 1/6, 1 where p = 1/4, and 1 where p = 1/2. The following table summarizes all the possible ways that six successes could occur given these probabilities.

and the total probability is the sum of the twelve possible binomial probabilities, one for each row in the table, computed as above. Since the amount of computation required is large, obtaining the actual probabilities is best done with a suitable computer program. Finally, we can ask what happens if we also relax our second simplifying assumption that we will only categorize moves as "R" or "not R", ignoring the difference between "U" and "I" moves? The author has not been able to find any literature on this problem, but it seems reasonable to


106

assume that it could be solved by a procedure similar to that above, only substituting a "trinomial" distribution for the binomial used above. This would involve a more complex enumeration of the possible ways that combinations of R, U, and I moves could occur, and use of the formula for the trinomial distribution given below.1

At present, for the exploratory experiments described in this paper, this exact solution does not seem necessary. However, it seems feasible to develop a computer program to perform the required analysis if the results of the experiment indicate that it would be worthwhile. ____________________ 1 William Feller, An Introduction to Probability Theory and its Applications, Volume 1, third edition, New York: Wiley, 1950 is a good reference work pertaining to the trinomial distribution and the problem covered in this appendix in general. ____________________


107

a behavioral game methodology for the...

Documents