JOURNAL
OF
RESEARCH
IN
PERSONALITY
76-82 (1976)
10,
There’s Method in Our Madness: Interpersonal Attraction as a Multidimensional Consttbct R. MICHAEL LATTA Iowa
State
University
The convergent validity and number of dimensions common to six measures of interpersonal attraction (three verbal and three nonverbal) were explored in this investigation. Convergent validity was found within verbal and nonverbal modes of measurement, but not between modes. No general factor was found to be common to the six measures of attraction. It was, therefore, concluded that method bias influences experimental results, threatening the comparability of results from research programs using different modes of measuring attraction. The results also question the assumption that social attitudes, such as attraction, are closely related to behavior, and imply that predictions derived from theories based on different modes of measuring attraction will have unequal utility.
. . . Our science is best advanced by each of us working very hard in our own sand boxes. . . We need not. . . be disconcerted by the fact that not everyone is doing the same thing in precisely the same way (Byrne, 1971, p. 15).
Most definitions of attitudes suggest that attitudes, as measured by verbal self-report scales, contribute to overt behaviors (Kiesler, Collins, & Miller, 1969). Since interpersonal attraction is a social attitude, it is assumed that the degree of attraction indicated by a subject on a verbal scale should be related to nonverbal behaviors which are considered to be influenced by attraction. However, there are no empirical data bearing upon this issue. It is, therefore, necessary to study the convergent validity of, and the number of dimensions measured by, the diverse procedures employed by social psychologists to study attraction. In general, the measures of attraction fall into two broad categories, verbal and nonverbal. Within the verbal category is Byrne’s Interpersonal Judgment Scale (IJS) and the semantic differential approach. The IJS consists of two seven-point scales on which evaluations of another’s likability and desirability as a work partner are made (Bryne, 1969), while the semantic differential measure of attraction consists of several bipolar adjectives which are evaluative in nature. Nonverbal measures of attraction include the distance between two individuals (Argyle & Dean, 1965), seating position at a rectangular table (Russo, 1967), and head orientation (Mehrabian & Friar, 1969). Address requests for reprints to the author: Department of Psychology, Iowa State University, Ames, IA 50010. 76 Copyrighhr @ 1976 by Academic Press. Inc All rights of reproduction in any form reserved.
CONVERGENT
VALIDITY
OF ATTRACTION
MEASURES
77
Although there are many different measures of attraction, most researchers tend to employ only one of these measures in their research programs. For example, Byrne (1969) has reported a tremendous body of experimental evidence showing that perceived similarity leads to attraction based on the IJS as an index of attraction. Lerner, in another research program, has gathered a great deal of evidence concerning attraction towards innocent victims based solely on bipolar adjectives (Lemer, 1971). Since comparability of findings from these two research programs depends upon the similarity of the attraction measures employed, it is necessary to determine if the various measuring procedures tap the same dimension. If these measures of attraction do not tap the same dimension or do not have convergent validity, comparison of the results from these two research programs has little meaning. Mehrabian (1972), on the other hand, has developed a comprehensive theory of communication based on nonverbal cues which indicate positiveness (attraction), potency (status), and responsiveness. Hence, the measurement of liking is considered to be best accomplished by using such indices as distance, seating position, head orientation, etc. Once again, however, convergent validation needs to be checked before results from studies employing verbal measures of attraction can be considered comparable to results from studies using nonverbal measures. An initial step in the right direction was taken by Byrne, Baskett, and Hodges (1971). These investigators used both the IJS and seating position as dual dependent measures of attraction. However, the two measures were not intercorrelated to see if they were properly related; and the hypothesized relationship between similarity and attraction as measured by both the IJS and seating position was confirmed for female but not male subjects in Experiment I, while the reverse was found in a second experiment. These findings suggest that verbal and nonverbal measures of attraction may involve method bias which precludes convergent validity. The present study was intended to systematically investigate six measures of attraction to determine if the pattern of experimental results observed in any given experiment depends on the type of dependent measure,employed and to see if the diverse measures of interpersonal attraction are measuring the same thing. METHOD
Design The data were gathered within the context of an attraction study testing the “just world” hypothesis (Lemer, 1965). A female confederate was employed as a stimulus person who suffered an injustice. Three independent variables were employed in a 23 factorial: role determination, or why the confederate suffered (fate or volunteer); intensity of suffering (high or low); and compensation for suffering ($0 or $2). For a more complete description of the procedure, see Latta, Bernhardt, Hildebrand, and Kahn (1974).
78
R. MICHAEL
LATTA
Subjects Subjects were 64 female introductory psychology students who were randomly assigned to one of the eight experimental conditions. All subjects were run individually with a female confederate.
Procedure After the experimental manipulations were carried out, the subject and confederate were asked to take a seat at one of the tables outside of the experimental room. The subject was detained by the experimenter, while the confederate left the experimental room and seated herself at the table in the position marked “C” depicted in Fig. 1.
FIG.
1. Available seating positions outside the experimental
room.
Measures After the subject chose her seat, the experimenter unobtrusively recorded the nonverbal measures of attraction, For each subject, a data sheet identical to that shown in Fig. 1 was used. The subject’s seating position and shoulder orientation was marked by a line drawn through the circle corresponding to the subject’s seating position. This line was parallel to an imaginary line through the subject’s shoulders, perpendicular to the subject’s line of sight. Head orientation was calculated by measuring the angle between the subject’s actual line of sight and the line of sight that would have occured had the subject attempted to maintain eye-contact with the confederate. Actual line of sight was taken as a line perpendicular to the recorded shoulder orientation since these two indices are highly related (Mehrabian & Friar, 1969). Seating positions were scored as follows: one-if the subject sat at the table directly across from the confederate, twc+if the subject sat diagonally across from the confederate, three-if the subject sat beside the confederate, and four-if the subject did not sit at the confederate’s table. The final nonverbal measure, that of distance, was taken as the distance from the confederate’s chair to the chair chosen by the subject. After unobtrusively recording the nonverbal measures of attraction, the experimenter asked the subject and confederate to fill out an impression formation questionnaire which consisted of five bipolar adjective scales and the IJS measure of desirability as a work partner. The subject used the scales to rate the confederate on the following five characteristics: goodness, likability, reputability, looks, and bravery. The goodness, likability, and IJS scores were considered verbal dependent measures of attraction because they seemed most related to attraction on face value.
CONVERGENT
VALIDITY
OF ATTRACTION
MEASURES
79
By way of summary, six measures of attraction were obtained: three verbal and three nonverbal. The three verbal measures were ratings of the confederate’s goodness, likability, and desirability as a work partner, while the three nonverbal measures were seating position, head orientation, and distance relative to the confederate.
RESULTS
To explore the latent dimensions of these six measures of attraction, the six variables were intercorrelated and a principle components factor analysis was performed. Two factors were extracted and rotated to a Varimax criterion of simple structure (Kaiser, 1958). The correlations matrix appears in Table 1 and the results of the factor analysis appear in Table 2. Inspection of Table 1 suggests no convergent validity outside of a mode of measurement. The three verbal measures of attraction are related, as are the three nonverbal measures; but verbal measures are not significantly related to nonverbal measures. Inspection of Table 2 indicates no general factor of attraction which is common to all six measures. Instead, the rotated factor pattern indicates two measurement factors, with the first representing nonverbal measures
OVERALL
CORRELATIONS
OF
TABLE 1 SIX MEASURES OF INTERPERSONAL
ATTRACTION
Variable
D
HO
G
L
IJS
Seating position (SP) Distance (0) Head orientation (HO) Goodness (G) Likability (L) IJS Question (IJS)
.65*
.79* .67*
.09 .08 .05
.22 .16 .20 .a3*
-.Ol -.08 .06 .57* .59*
*p < .Ol. TABLE ROTATED
Variable Seating position Distance Head orientation Goodness Likability IJS Question
FACTOR
MATRIX
2
OF ATTRACTION
DIMENSIONS
Rotated factor loadings I II 90889 .85704 .91207 .03890 .18797 - .07260
.06245 .0@452 .06270 .91196 .91084 .81079
Communalities .82998 .73454 .83580 .83319 .86495 .66265
80
R. MICHAEL
LATTA
of attraction and the second representing verbal measures. This indicates that these two measuring procedures are tapping different dimensions. It could be argued that the factor loadings derived from overall correlations are misleading since treatment effects play a part in determining the weights assigned to each variable (Caldwell, 1974).’ Since the factor loadings could be situation specific because they are affected by between cell differences, another principle components solution was obtained using within cell correlations (Weiner, 1962). The results using within cell correlations indicated a factor pattern similar to the original solution. This implies that the factor loadings derived from the overall correlations have some generality. DISCUSSION
The lack of convergent validity outside of a mode of measurement is disturbing when one considers continuity of research findings. Although the results of laboratory studies supporting Byrne’s theory of attraction and Lerner’s “just world” hypothesis may be comparable, the comparability of their results with studies employing nonverbal measures of attraction is limited, due to a lack of convergent validity for the two categories of attraction measures. This conclusion is consistent with results of Lerner and Becker (1962), Novak and Lerner (1968), and Lerner and Agar (1972) who report two dimensions of attraction, approach-avoidance (nonverbal) and attribution of positive characteristics (verbal), which gave conflicting patterns of results in those studies. Obviously, method bias has played a part in determining the results of past research. The lack of a general attraction dimension which is common to the various measures of attraction has implications for contemporary attitude theories. In general, social psychologists have accepted verbal measures of attraction while formulating theories of attraction that specify behavioral implications of attraction such as approach-avoidance. The present results suggest that the procedures used to measure the social attitude called interpersonal attraction are not closely related to behaviors which imply attraction, an assumption which is at the heart of attitude theories. The present results also suggest that a theory based on a verbal measure of attraction, such as Byrne’s, will not predict behavior in situations where a theory based on nonverbal measures operates best. Our task, then, becomes one of outlining the boundaries of these two kinds of attraction theories and specifying the relationships among the operational definitions of attraction employed in the two approaches. Multivariate studies such as 1 The author would like to express his thanks to Ed and Marcia Donnerstein for pointing out this possible criticism of the results of this investigation.
CONVERGENT
VALIDITY
OF ATTRACTION
MEASURES
81
the present one, with employment of factor scores as the dependent measure, should help accomplish this goal. In summary, the results of this study strongly suggest that verbal and nonverbal measures of attraction represent two nondependent dimensions; not necessarily independent, but uncorrelated in many instances. Thus, since the subject’s position on the verbal dimensions of attraction is not directly related to his/her position on the nonverbal dimension, studies conducted by Lerner and Byrne, who employ only verbal measures, may involve quite different phenomena than those conducted by Mehrabian and Argyle, who employ only nonverbal measures. Since, the situational context can affect how and when these two dimensions of attraction are related, it would be valuable for attraction theorists to integrate the various determinants of attraction on the verbal and nonverbal dimensions. A theoretical framework is needed to specify when and how these two nondependent dimensions of interpersonal attraction are related. Hopefully, paying attention to what happens in other people’s sandboxes will reduce the method in our madness, and ensure comparability of research findings. REFERENCES Argyle, M., & Dean, J. Eye-contact, distance, and tiliation. Sociometry, 1965,28,289-304. Byrne, D. Attitudes and attraction. In L. Berkowitz (Ed.), Advances in experimental social psychology. New York: Academic Press, 1969. Vol. 4., pp. 35-89. Byrne, D. Can Wright be wrong? Let me count the ways. Representative Research in Social Psychology, 1971, 2, 12-18. Byrne, D., Baskett, G. D., & Hodges, L. Behavioral indicators of interpersonal attraction. Journal of Applied Social Psychology, 1971, 1, 137-149. Caldwell, M. Within-cell vs. over-all correlations in the analysis of factor variance. Perceptual and Motor Skills, 1974, 38, 1191-1201. Kaiser, H. F. The Varimax criterion for analytic rotation in factor analysis. Psychometrika, 1958, 23, 187-200. Kiesler, C. A., Collins, B. E., & Miller, N. Artitrrdes change. New York: John Wiley and Sons, 1969. Latta, R. M., Bernhardt, V. L., Hildebrand, P., & Kahn, A.Attraction to a beneficent victim: Balance theory or “The Just World?” Paper presented at the annual meetings of the American Psychological Association, New Orleans, 1974. Lerner, M. J. Evaluation of performance as a function of performer’s reward and attractive. ness. Journal of Personality and Social Psychology, 1956, 1, 35>360. Lerner, M. J. Observer’s evaluation of a victim: Justice, guilt and veridical perceptim. Journal of Personality and Social Psychology, 1971, 20, 127-135. Lerner, M. J., & Agar, E. The consequences of perceived similarity: Attraction and rejectbn, approach and avoidance. Journal of Experimental Research in Personality, 197:, 6, 69-75. Lerner, M. J., & Becker, S. W. Interpersonal choice as a function of ascribed similarly and definition of the situation. Haman Relations, 1%2, 15, 27-34. Mehrabian, A., & Friar, J. Encoding of attitude by a seated communicator via pospre and position cues. Journal of Consulting and C:‘ ‘Cal Psychology, 1969, 33, 330-36.
82
R. MICHAEL
LATTA
Mehrabian, A. Nonverbal communication. Chicago: Aldine, 1972. Novak, D. S., and Lerner, M. J. Rejection as a consequence of perceived similarity. Journal of Personality and Social Psychology, 1968, 9, 147-152. Russo, N. Connotations of seating arrangements. Cornell Journal of Social Relations II, 1967, 37-44. Weiner, B. J. Statistical principles in experimental design. New York: McGraw-Hill, 1%2.