Method
Procedures were generally the same as in Studies 1 and 2 except that the manipulated variable was whether closeness was an explicit task and subjects were randomly paired on the basis of introversion/extraversion. The study was conducted 3 weeks into the term.
Subjects. There were 45 cross-sex and 24 all-women pairs. Of these, 36 cross-sex and 18 all-women pairs were matched in advance on introversion/extraversion. (We did not have introversion/extraversion data on the remaining subjects.) There were 14 extravert pairs, 23 introvert pairs, and 17 mixed pairs.
Initial questionnaire. The initial questionnaire was the same as that used in previous studies except that instead of an attachment-style measure, subjects indicated their introversion/extraversion type, based on the Myers- Briggs Type Indicator (Myers 8c McCaulley, 1985), which they had taken and self-scored in a recent class session.
The Myers-Briggs Type Indicator is a widely used measure, employed here because it was used in the Thorne (1987) study.
Experimental manipulation and procedures. We took great care that the students in this class would not be aware of previous studies or have any basis for thinking this study focused on closeness. In class announcements and in all materials, the research was described as a study of social interaction in which subjects would participate in pairs. The experimental manipulation consisted of whether the subjects received one of two sets of instructions. In the closeness-as-a-task condition, the written instructions began in the way they had in the previous studies—'This is a study of interpersonal closeness, and your task ... is to get close to your partner"—and proceeded with several more lines about getting close (see appendix). Also, in this condition, the reminders interspersed among the slips repeated this emphasis on the task being one of getting close. In the no-mention-of-closeness condition, the written instructions began, 'This is a study of dyadic interaction, and your task is simply to follow instructions, doing the tasks with your partner"; reminders in this condition similarly did not mention closeness.
Results and Discussion: Closeness as a Task Versus No Mention of Closeness
Overall mean closeness was 3.76, a figure in the same range as the closeness condition in Study 1 and overall in Study 2. There was no difference between instruction conditions (F < 1), and the interaction with cross-sex versus all-women pairing did not reach or approach significance. The lack of difference suggests that, overall, making closeness a task versus not mentioning it does not much affect the level of closeness achieved. There was about 90% power in this study for achieving a significant result if in fact there were a large effect for this variable and also about 90% for a near-significant result with a true medium effect. Further, the lack of a direct overall effect for this manipulation cannot easily be attributed to the manipulation not having an impact on the subjects at all, because (as described below) there were significant interactions of this manipulation with introversion/extraversion. Thus these findings suggest that, overall, making closeness an explicit task is not very important for producing closeness using this paradigm. They also suggest that, on the average, the closeness produced by the procedure is probably not due to making closeness an explicit goal serving as a demand characteristic.
Results and Discussion: Illustration of Application to Theoretical Issues (Introversion/Extraversion and Its Interaction With Closeness as a Goal)
Our analysis strategy, based on Thome's (1987) approach, separately compared (a) extravert-extravert pairs with introvert-introvert pairs (a between-subjects analysis) and (b) within mixed-type pairs, the extra- verted person's ratings to the introverted person's ratings (a within-subjects analysis). Separating the data into these two analyses is a quite conservative procedure («s for each analysis are small). Thus near-significant results should probably be taken a bit more seriously than usual. Also, because cross-sex versus all-women pairings did not significantly interact with any introversion/extraversion effect in any of these analyses, we collapsed over this variable. (Nor were there any sex differences or interactions within the cross-sex pairs.)
Considering the extravert pairs versus the introvert pairs, there was a marginal overall main effect of extravert pairs reporting greater closeness (p = . 11). However, any such tendency was qualified by a strong interaction with instructional condition,/^!, 33) = 5.61, p< .05, such that extravert pairs become much closer than introvert pairs in the no-mention-of-closeness condition, but there was relatively little difference between extravert and introvert pairs in the closeness-as-a-task condition. The analysis comparing extraverted versus introverted individuals who were paired with each other revealed findings strikingly similar to the analysis comparing all- extraverted to all-introverted pairs. This result approached significance (main effect, p = .09; interaction effect,/? =.10).
The overall tendency for introverted individuals to report somewhat lower closeness (and the quite clear tendency for them to do so when no special instructions about closeness as a task are given) is consistent with the general understanding of introversion as a discomfort with social interactions with strangers. Because the method for getting close is presumably provided by the task, these data shed doubt on the view that introverts are less social because they are less skilled at getting close. Indeed, when getting close is made an explicit task, introverts became as close as extraverts. This may be because to the extent introverts do engage in social interaction, this task was precisely the kind of self-disclosure that is typical of their conversations (Thorne, 1987). Thus, when the goal of closeness was made salient, introverts may have perceived themselves as especially effective at these tasks (or perhaps as a result they were less aroused than they would have been without the instructions). In any case, these tentative findings (both the main and interaction effects) are important here because they demonstrate the potential of our procedure for shedding light on theoretical issues. In the present study, they demonstrate the possibilities arising from (a) being able to control the pairings (as was also done in Studies 1 and 2 with attachment style) and (b) being able to manipulate the circumstances of interaction, in this case the goal of closeness, and evaluate
its moderating effect in the context of theory (which was illustrated here for the first time in Study 3).
GENERAL DISCUSSION
This series of studies explored a closeness-generating experimental paradigm, simultaneously examining various aspects of the procedure and illustrating its potential for addressing theoretical issues in the study of close relationships. Below, we consider the level of closeness produced by the procedure, implications of the findings for various aspects of the paradigm, implications of the data from the illustrative theoretical issues regarding the usefulness of these procedures, and how this paradigm could be used to advance knowledge in the social psychology of close relationships.
Overall Level and Type of Closeness Produced by the Procedure
Over the three studies (excluding the small-talk condition in Study 1), the mean IOS Scale score was 3.82. In an independent sample of 296 students (Aron et al., 1992) at the same level at the same university and collected at about the same time as this study, subjects used the IOS Scale to rate their "closest, deepest, most involved, and most intimate relationship" (these instructions were taken originally from Berscheid et al., 1989, p. 806). In that sample, the mean IOS Scale score was 4.65 (SD= 1.50) and was approximately normally distributed. (In yet another similar sample of 88, reported in the same article, means, SDs, and distributions were about the same as for the larger sample.) Using those data as a standard, the mean scores for the subjects who participated in our closeness-generating procedure were at about the 30th percentile (Z= -.55) of ratings of one's closest, deepest, most involved, and most intimate relationship. That is, immediately after about 45 min of interaction, this relationship is rated as closer than the closest relationship in the lives of 30% of similar students. Of course, we cannot rule out the possibility that subjects used the scale differently in the experimental context than they would when rating actual ongoing relationships.