Results and Discussion
Means on the closeness composite were 4.06 for the closeness condition and 3.25 for the small-talk condition. This difference corresponds to an effect size (d) of .88 standard deviations (.8 is considered large in relation to typical effect sizes in the psychological literature; Cohen, 1988). This difference was also clearly significant. We evaluated significance in the context of a 2 (Task) X 2 (Cross-sex vs. All-women) x 4 (Attachment- style Pairing) ANOVA with the pair as the unit of analysis (a conservative procedure; Kenny, 1988). The effect for task condition (closeness vs. small talk) was ^(1, 37) = 5.68, p< .05. There were no significant or near-significant interactions of task condition with any other variable; nor did the main effect for cross-sex versus all-women reach or approach significance. (Attachment-style findings are presented at the end of the Study 2 Results and Discussion section in the context of the larger, combined samples of Studies 1 and 2.) Also, in the analysis of the cross-sex pairs only, there were no significant or near- significant within-pair main effects for sex or any sex interactions involving task condition.
These data support the importance of task type in developing closeness through our procedure. The contents of the tasks—whether they required self-disclosure and other intimacy-associated behaviors—made a considerable difference. Thus any effect of this procedure is not simply a matter of putting two people together in any kind of structured interaction for 45 min.
STUDY2
This study focused on the importance for generating closeness of two additional aspects of our procedure: (a) matching within a pair for nondisagreement on attitudes and (b) leading subjects to expect mutual liking between self and partner. That is, we manipulated these two variables as crossed experimental factors. In addition, subjects were paired by attachment style as in Study
1 so that data from the two studies could be combined for an illustration of the application of our procedure to attachment issues.
Method
Procedures were the same as in the Study 1 closeness condition, except for this study's two manipulated variables and two additional items on the postinteraction questionnaire. There were 52 cross-sex and 19 all-women pairs. The study was conducted 8 weeks into the term.
Experimental manipulations. All potential pairings (that is, within Attachment-style Combination x Cross-sex vs. All-women categories) were made so that half did not disagree on any important issue (the procedure used for all subjects in Study 1) and half clearly did disagree on one or more important issues. Crossed with this division, subjects were randomly assigned to expect or not expect mutual liking.
The disagreement/nondisagreement manipulation used the 17 attitude items on the initial questionnaire. Those pairs in the disagreement condition had either (a) one strong disagreement (one rated the item a 1 or
1 and the other a 6 or 7 on the 7-point scale) on an issue rated as highly important (5 or higher on this 7-point scale), (b) two strong disagreements on issues rated as moderately important (4 or higher), or (c) three moderate disagreements (the two people's ratings are on opposite sides of the midpoint, or one is at the midpoint and the other is a 1 or 7) on issues of very high importance (6 or higher). Those pairs in the nondisagreement condition were matched so that there were no disagreements of any of the kinds listed above (this is the same rule as used for all subjects in Study 1).
Regarding the manipulation of expectation of mutual liking, in this study nothing was said on the initial questionnaire or in any oral instructions of any expectation that the pair would like each other or that any special matching had occurred. However, in the expectation-of- mutual-liking condition the instruction sheet included an explicit and prominent section noting that the pair had been carefully matched:
We have taken great care in matching partners. Based on our experience in previous research we expect that you and your partner will like one another—that is, you have been matched with someone we expect you will like and who will like you.
In the no-expectation-of-mutual-liking condition, this section instead read as follows:
Partners in this study have been put together in ways that pair different categories of individuals. We are investigating the effect of different kinds of pairings. We have no special reason in your case to assume that you and your partner will like each other.
Additional items on the postinteraction questionnaire. In this study only, we included a version of Byrne's (1971) two-item Interpersonal Judgment Scale (IJS), which asked, "How much would you like to work with your partner on a project?" and "How much do you like your partner?" We included the IJS in this study because the disagreement/nondisagreement manipulation seemed directly relevant to issues Byrne and others have explored with this measure.
Results and Discussion: Instruction Conditions (Attitude Disagreement and Expected Mutual Liking)
Overall mean closeness was 4.02, a figure comparable to that obtained for the closeness-condition subjects in Study 1. There were no significant or near-significant differences on the closeness composite or the IJS for either of the instruction condition variables (disagreement vs. nondisagreement or expectation of mutual liking vs. no expectation of mutual liking) or their interaction; in all cases, Fs < 1. (There were also no significant or near-significant interactions of the instruction condition variables with cross-sex vs. all-women pairing or with the attachment-style pairings.)
Overall, these data suggest that matching in terms of not disagreeing on important attitudes or leading subjects to believe that they and their partners will like each other probably has little impact on the overall closeness subjects achieve through this procedure, or even on their mutual attraction. There was about 90% power in this study for achieving significant effects (or interactions) for the two manipulated variables if in fact there were a large effect of this kind (d= .8). Indeed, the power is about 90% for finding at least a near-significant (p < .10) medium-sized effect (d= .5). Thus it seems unlikely that we would have obtained the present results if in fact there is more than a small effect for either of these variables.
In light of extensive research showing the importance of similarity (e.g., Byrne, 1971) and expected liking (Aron et al., 1989) in attraction, it is surprising we did not find any effects for these variables on either closeness or attraction. Perhaps the self-disclosure and relation- ship-building process has such an impact that agreement or expectation of liking is not relevant (Aron & Aron, 1986). Or perhaps this result is due to the specific conditions of the study: In terms of disagreement/non- disagreement, the tasks employed may not bring up topics that allowed subjects to discover any attitude dissimilarity; and in terms of expectation of mutual liking, the general expectation of closeness as a purpose/result of the study may already imply mutual liking, so that the explicit mention of expecting them to like each other made little difference. It is also possible that any attitude similarity effect was minimized in this study because subjects were aware that their partners were in the same class (and may have actually noticed them in the classroom), so that they expected they were similar, thus reducing the range on this variable. However, Byrne (1971) regularly found attitude similarity effects using a somewhat similar situation (subjects in most of his studies believed that they were evaluating partners from another psychology class at the same university).
Results and Discussion: Illustration of Application to Theoretical Issues (Closeness of Attachment-Style Pairings)— Combined Data From Studies 1 and 2
These analyses included pairs from Study 1 in the closeness condition and all pairs from Study 2, for a total of 97 pairs—37 secure, 11 avoidant/dismissive, 23 avoidant/fearful, and 26 with a preoccupied partner.3 Three results stood out. First, the avoidant/dismissive pairs reported less closeness than other pairings (Ms = 4.10 for secure pairs; 3.59, avoidant/dismissive; 4.09, avoidant/fearful; 4.07, preoccupied with other; contrast p < .05). Correlational research (e.g., Hazan 8c Shaver, 1987) has found that having an avoidant attachment style is associated with reporting poorer quality relationships. Our findings support an underlying causal direction for this association from attachment style to poorer relationship quality, illustrating the potential of the present method for sorting out causality.4 (Of some theoretical interest in its own right is the finding that the avoidant-poor-relationship link may be limited to avoidant/dismissive individuals; most previous research has used the three-category typology in which the two types of avoidant individuals are not distinguished.)