Communication Patterns in Laboratory Discussions of Safer Sex Between Dating Versus Nondating Partners.

Safer-sex negotiation is considered crucial for the implementation of safer sexual behavior in dating partners' sexual scripts (Catania, Coates, & Kegeles, 1994; Dolcini et al., 1995; Fisher, Fisher, Williams, & Malloy, 1994; Helwig & Collins, 1994; Kalichman, Rompa, & Coley, 1996; Winter & Goldy, 1993). Although condom use is the most recommended way to practice safer sex, individuals often prefer to rely on their perceived ability to select a safe partner and decide against condom use (Buysse, 1998; Maticka-Tyndale, 1991). In both stable and casual relationships, but especially in relationships perceived as close and intimate, partner selection is the most popular way to protect oneself against AIDS. Partner selection is found to be preferred in 70% to 80% of sexual encounters within stable relationships, and in about half of the sexual encounters within casual relationships (Buysse, 1998; Buysse & Van Oost, 1997).

Successful partner selection, however, requires an extensive evaluation of a potential partner's sexual risk history. All too often, individuals who have well-defined cognitions and intentions regarding safer sexual practices fail to communicate them to their sexual partners. Research findings suggest that about three quarters of young people talk about safer sex (Sprecher & McKinney, 1993). However, only 21% (Cline, Freeman, & Johnson, 1990) to 27% (Gray & Saracino, 1991) report having discussed monogamy, condom use, or their sexual history with their actual sexual partner(s) before having sex. In only about 40% of the sexual encounters in which young adults rely on their ability to select a safe partner did they actually know about the partner's prior sexual history, condom use, and serostatus before they had sex (Buysse, 1998). Effective safer-sex negotiation is, therefore, more the exception than the rule in dating couples.

The present study focuses on a critical component of safer-sex negotiations--conflict avoidance. We hypothesized that safer-sex negotiations evoke in stable dating partners a high level of goal conflict, which in turn results in an increased use of more conflict avoidance and in a decreased use of direct problem-solving strategies. Self-report data suggest that safer-sex discussions in intimate relationships are difficult to manage because of the high levels of perceived goal conflict they evoke (Lust, Celuch, & Showers, 1993; Metts & Fitzpatrick, 1992; Miller, Bettencourt, DeBro, & Hoffman, 1993).

What sets up this conflict are two opposing viewpoints that are both likely to become salient when the issue of safer sex is confronted. On the one hand is the romantic view which holds (a) that stable, intimate relationships are relatively close and exclusive (Snyder & Simpson, 1987); (b) that relationships are based on "trust, caring, ease in communication, general comfort and security in each other's presence, and openness" (Wilmot & Baxter, 1983, p. 10); and (c) that sex in stable dating relationships is person-centered and seen as an expression of love and affection, and is a way to increase the emotional intimacy of the relationship (Reiss, 1967). On the other hand, the more cynical view holds (a) that romance does not guarantee either fidelity or sexual health; and (b) that discussions of safer sex in intimate relationships are often associated with mistrust, psychological distance, and formality (Metts & Fitzpatrick, 1992).

Given these two opposing views, a plausible reason to decide not to discuss condom use or inquire about a partner's sexual risk history is the perceived incompatibility between (a) the salient goals in developing and sustaining a stable romantic relationship, and (b) the salient goals in discussing safer sex, which are likely to evoke embarrassing issues that have the potential to discredit one or both partners and the relationship itself (Fullilove, Fullilove, Haynes, & Gross, 1990; Mays & Cochran, 1988; Metts & Fitzpatrick, 1992; Miller et al., 1993). Because of the simultaneous presence of multiple and conflicting goals, safer-sex discussants have to deal with competing tasks that require an unusually high level of controlled and effortful cognitive processing (Ickes & Simpson, 1997). Because safer sex is a topic that dating partners tend to avoid discussing openly, couples who lack experience with safer-sex talk are likely to experience both increased cognitive conflict and increased stress during such a discussion. An effective strategy to minimize such stress would be to avoid the conflict-evoking topic, if possible, or to confront it in a more indirect manner. Topic avoidance--both direct and indirect---could therefore be used to minimize the stress that derives from the conflicting cognitions evoked in safer-sex discussions (see also Gilbert & Krull, 1988; Gilbert, Pelham, & Krull, 1988; Ickes & Simpson, 1997).

The seemingly paradoxical aspect of goal conflict during safer-sex negotiations is that relationship features such as closeness (Gold, Skinner & Ross, 1994), intimacy (Pilkington, 1994), and exclusivity (Herald & McNamee, 1982) actually tend to discourage safer-sex discussions and safer sexual practices (Abraham & Sheeran, 1993). They do so because the goals most pertinent to safer-sex communication (e.g., mutual honesty about one's sexual history and behavior in other relationships) are experienced as incompatible with highly valued relational goals such as exclusivity and emotional closeness. Relational goal conflict is therefore a likely and potentially distressing aspect of safer-sex discussions between intimates.

On the other hand, there is no compelling reason to believe that safer-sex conversations are likely to be as potentially distressing in less stable or more casual dating relationships. Conflict about the issue of fidelity/exclusivity does not enter into safer-sex discussions unless and until one of the partners raises this issue after a sexual relationship has been established. In these cases, jealousy and suspicion of infidelity cause the conflict and are expected to interfere with highly valued relational goals such as closeness, trust, intimacy, and exclusivity. The nature of the communication patterns in such cases are the focus of the current investigation.

There is some disagreement about the relative merits of confronting versus avoiding conflict (Sillars & Weisberg, 1987). In most circumstances, conflict avoidance is not considered adaptive within the context of marriage or dating relationships (Gottman, 1993). In particular, the interaction pattern involving avoidance--demand/withdrawal--is typically regarded as negative in its consequences. In this pattern, one partner pressures the other to change, whereupon the other partner withdraws (Christensen, 1987, 1988; Greenberg & Johnson, 1986; Napier, 1978; Watzlawick, Beavin, & Jackson, 1967; Wile, 1981). On the other hand, because attempts to resolve conflict may, in some circumstances, make the situation worse (Fitzpatrick & Winke, 1979; Rands, Levinger, & Mellinger, 1981), avoidance of Such conflicts can actually be constructive (Canary, Cupach, & Messman, 1995). In the case of AIDS prevention, however, it is clear that the failure of dating partners to confront the issue of safer sex can have devastating consequences later on.

In addition to a tendency to avoid safer-sex negotiation, we expect that during safer-sex discussions stable dating partners will exhibit some other communication patterns often observed in conflictive interactions between distressed spouses, such as a tendency to refer less to shared experiences (we-ness), exhibit less direct problem-solving behavior (Sabourin, Laporte, & Wright, 1990), more problem escalation (Revenstorf, Hahlweg, Schindler, & Vogel, 1984), less problem acceptance (Revenstorf et al., 1984), more coercive strategies (Noller, Feeney, Bonnell, & Callan, 1994; Patterson, 1982), more manipulation (Fitzpatrick & Winke, 1979; Noller et al., 1994), and more defensiveness or combativeness (Alberts & Driscoll, 1992; Canary et al., 1995; Gottman, 1993). Evidence that communication styles vary according to discussion themes can be drawn from several attempts to describe communication styles in thematic discussions of jealousy (Bennum, 1985; Epstein, Baucom, & Rankin, 1993), sexual intercourse and/or wanting to date others outside the primary relationship (Metts, 1994), autonomy/dependency (Bennum, 1985; Christensen, 1988; Epstein et al., 1993; Jacobson, 1989; Napier, 1978), complaints about the partner (Cupach & Metts, 1994), lies and betrayed confidence (Metts, 1994), and lack of support (Jones & Burdette, 1994).

Goal conflict has also been associated with particular communication behaviors. Pause phenomena are found to be related to the difficulty of generating plans for goals (Berger & Jordan, 1992), and to the pursuit of multiple goals (Greene, Lindsey, & Hawn, 1990) and multiple conflicting goals (Greene, 1995). The same logic predicting longer pauses would also predict decreased references to shared experiences (we-ness, see also above) and averted eye gaze (i.e., the more goals individuals are trying to juggle and the greater the incompatibility between these goals, the more the eyes are averted while utterances are being planned).

With regard to this last prediction, there is considerable evidence that speakers avert their gaze from another person's face during periods of increased cognitive activity in speech (Beattie, 1979, 1981; Kendon, 1967). One plausible interpretation is that individuals try to avoid cognitive overload or reduce the input of information by looking away (Argyle & Cook, 1976). On the other hand, eye contact has also been related to goal detection. Goal detection is only one of the many different functions of eye contact (Argyle, 1972; Phillips, Baron-Cohen, & Rutter, 1992; Rutter, 1984), but one that has potential relevance for the present study. Given these different and somewhat opposing functions of eye contact, it is difficult to predict the gaze patterns of intimate partners during discussions of safer sex. A relatively high level of gaze could result because of both the difficulty and the importance of diagnosing the other person's goals under relational stress, whereas a relatively low level could result from efforts to decrease one's cognitive load during discussions of safer sex between intimates.

Given this ambiguity about the gaze variable, the present study tested three predictions about specific interaction behaviors in safer-sex discussions--two predictions that are relatively unequivocal and a third that is necessarily more tentative. These predictions are that safer-sex discussions between dating partners will evoke relational goal conflict that is manifested in (a) increased pauses, (b) decreased use of personal pronouns expressing we-ness, and (c) a distinctive pattern of eye gaze (Hypothesis 1). In terms of more global patterns of interaction behavior, we further predict that safer-sex discussions between dating partners will result in negative communication behavior. More specifically, we expect dating partners to exhibit more withdrawal and less direct problem-solving during discussions of safer sex (Hypothesis 2).

Finally, we expect that the negative conflict management styles that are typically observed in distressed couples' discussions will be evident even in nondistressed couples' discussions when they are forced to confront safer sex--a topic that evokes relational goal conflict. In other words, negative communication patterns, including withdrawal and impaired problem solving, will be more evident in safer-sex discussions than in control-topic discussions, and this difference will be even more evident for intimates (actual dating partners) than for opposite-sex strangers (Hypothesis 3). As will be clear from the results reported below, a strong test of this hypothesis requires the use of sequential data analyses.

Our analysis suggests that relational goal conflict should occur only in the safer-sex discussions of actual dating partners, and not in the corresponding discussions of opposite-sex partners who have not yet established a personal relationship. For this reason, both the discussion topic (safer-sex discussions vs. control discussions) and the status of the discussants' relationship (dating partners vs. nondating partners) were independently varied by the random assignment of a discussion topic and a discussion partner to the respondents (either their own dating partner or an opposite-sex stranger). Our analysis further suggests that relational goal conflict will only occur when the issue of safer sex is raised after the partners have already established a sexual relationship, because for stable dating partners discussing this issue is an event that suddenly increases uncertainty and arouses suspicion. For this reason, transgression stimuli--stimuli that call into question the fundamental underpinnings of the relationship and evoke the goal conflict we discussed above--were used to elicit safer-sex talk in the present study.

METHOD

Sample

Heterosexual couples in a large Belgian city were approached by two members of the research team in cinemas and shopping areas. They were first asked whether they were dating and whether theirs was a sexual relationship; if they answered affirmatively, they were then asked to voluntarily participate in an extended investigation of conflict management in dating couples. The members of the research team identified themselves as researchers and warned the couples beforehand that they were going to ask some intimate questions. Those couples who were already married or were platonic cross-sex friends were not recruited to participate, but were instead debriefed about the specific goals of the study. The volunteer rate in this sample was 55% of those who met the criteria, closely approximating the volunteer rates in studies using comparable sampling techniques (Kipke, Palmer, LaFrance, & O'Connor, 1997; Liebman & Mulia, 1993; Rhodes & Malotte, 1996). An appointment for each couple to report to the research center was made within two weeks. Two couples previously unacquainted with each other were scheduled to come to the research center at the same time.

In total, 123 dating couples participated in the study, which took place in Ghent, a Belgian city of about 500,000 inhabitants, from July to August 1995. All participants were white, with a mean age of 22.7 years (SD = 3.82 years, range 18 - 30 years), and a mean length of relationship of 18 months (SD = 6.24 months, range 4 - 30 months). Almost all of the couples (98%) reported having a stable sexual relationship (see also Footnote 2), and 95% reported having ever discussed the issue of safer sex in their relationship. In the present study, which was the second part of a larger data collection involving these couples, the data for 120 couples are reported (the first part of this data collection is reported in Buysse & Ickes, 1999). Because the videotapes of three couples were not suitable for coding, the data for these three couples were not included in the analyses.

Research Design and Manipulations

The research design was a 2 (discussion topic: safer sex vs. joint leisure time) x 2 (discussion partner: intimate vs. stranger) factorial, with discussion topic and discussion partner treated as between-dyad variables. The dyads (dating couples) were randomly assigned to discuss one of two topics: either the experimental topic (safer-sex discussion), or the control topic (joint leisure time). Clearly, safer-sex discussions and joint leisure time discussions differ on a number of dimensions besides goal incompatibility, including the degree to which they are seen as intimate/nonintimate, consequential/inconsequential, or important/nonimportant. Because our primary goal was to contrast safer sex--a topic which has been shown to evoke high goal incompatibility (Buysse, 1998)--with a topic we could be reasonably sure would evoke significantly less goal incompatibility, we opted for a comparison that accomplished this goal without attempting to control all other dimensions of potential relevance to our results. For a detailed explanation of the control topic, see Buysse and Ickes (1999).

Because the focus of this paper is on safer-sex negotiation in intimate relationships, and because faith in the safety of the partner is the most frequently-used strategy to protect oneself against AIDS, the main focus of the safer-sex script was on communication about potential sexual experiences outside the relationship. The script was based on the results of a study of 1088 actual sexual encounters within both stable and casual relationships, in which detailed information about AIDS prophylaxis was obtained (Buysse, 1998). The mundane realism of both discussion scripts, and the relevance of the safer-sex script for both stable and casual dating partners, were evaluated in advance in a small sample (N = 5 couples). Based on this evaluation, some modifications to the initial safer-sex script were made. As a result of these modifications, the final safer-sex script (a) was judged to be realistic by the researchers, (b) included the most frequently used safer-sex strategy, and (c) enabled a better comparison with the control topic. The final scripts read as follows:

   The safer-sex conflict script.

   "Your relationship has recently been stressed (fill in events through which
   the relationship could have been stressed, e.g. exams, holiday ...). One of
   you is worried because of the possibility that your partner is sexually
   involved with someone else (choose who plays the unfairly treated person).
   The unfairly treated partner has the feeling that the other person is
   spending a lot of time with someone else, and is even sexually involved
   with another person (fill in what these activities could be, e.g., studying
   together, going out together). Last night was the first time in a long time
   you two saw each other. You and your partner had made plans, and then that
   other (i.e., third) person showed up (or made a phone call). Now you think
   it is really time to have a talk about the situation, especially within the
   context of the potential for contracting AIDS ..."

   The joint leisure-time activities script.

   "Your relationship has recently been stressed (fill in an event through
   which the relationship could have been stressed, e.g., exams, holiday). One
   of you is worried because the other one makes all the decisions about what
   to do during leisure time him/herself (choose who plays the unfairly
   treated person). The unfairly treated partner has the feeling that he/she
   has no input on how to spend joint leisure time (fill in what these
   activities could be, e.g., always watching football). Last night was the
   first time in a long time you two saw each other. You had made plans, and
   then your partner decided again to do something else. Now you think it is
   really time to have a talk about the situation ..."

Because two couples were scheduled to participate at the same time, it was possible to systematically vary the relationship status variable. Couples were randomly assigned to discuss the topic either with their own partner or with the opposite-sex partner from the other couple. This pairing procedure resulted in the four cells of the experimental design.

Procedure

As detailed elsewhere (Buysse & Ickes, 1999), this study is part of a larger data collection involving the same couples. The study was run in 2-hour sessions. Two couples, previously unacquainted with each other, were scheduled to come to the laboratory at the same time. The general procedure of the study can be broken down into three distinct phases. The data collection for this study concerns only the second phase.

Phase 1: Getting-acquainted exercises. In the first phase, the two couples were led to an experimental room where they met two assistant experimenters (E1 and E2). E1 greeted the participants and reminded them that, as they had been told during recruitment, the study was designed to investigate how dating couples discuss conflictive issues. The participants were informed that they would be asked to role-play a discussion of such an issue. As a warm-up designed to help the participants to prepare for these discussions, they were asked to participate in some brief getting-acquainted exercises. The last of these warm-up exercises was videotaped for analysis in a different aspect of this investigation (see Buysse & Ickes, 1999). E1 then introduced the Phase 2 conflict management discussion as follows.

Phase 2: Collection of the videotape data. E2 indicated that two opposite-sex participants (randomly paired) were to go to room 2 with E2, while the remaining participants were to stay in room 1 with E 1. These dyads (the actual dating couples in half of the sessions, and the two opposite-sex subjects from different couples in the other sessions) were asked to demonstrate, in a 6-minute interaction, how they would resolve the specific conflict that they were assigned to enact.(1) In the safer-sex discussion condition, this conflict involved a possible extradyadic sexual involvement and its potential consequences in terms of AIDS prevention. In the leisure activities condition, the conflict involved disagreement about joint leisure time activities.

The assistant experimenter (E1 or E2) read the script and asked the participants to personalize it and choose their respective roles. Neither the dating partners nor the opposite-sex strangers experienced difficulties choosing their respective roles. In most cases (81.6%), the women choose the role of the unfairly-treated partner, and agreement about this role selection was obtained within two minutes for all of the dyads. The assistant experimenter then explained to the participants that she would leave them alone while they role-played the conflict management situation, and would return 6 minutes later to terminate this phase of the study.

These conflict management discussions were videotaped. The video equipment was activated the moment the assistant experimenter left the room and all participants were informed beforehand that the videotaping would occur. After the conflict management discussions had been completed, E2 brought all four participants back to room 1 and partially debriefed them about the phase 1 getting-acquainted exercises. E2 also informed them that their written consent was required for the videotapes of their conflict management discussions to be used as data. None of the participants requested that any of these videotapes be erased, and all of them provided the requested written consent.

Phase 3: Collection of the self-report measures. In the part of Phase 3 that is relevant to the present study, the four participants were asked to individually complete a computer-administered post-test questionnaire in room 3. The questionnaire included a variety of measures about different aspects of the relationships of the actual dating partners. When all four participants had completed the post-test questionnaire, they were debriefed and thanked for their participation in the study. A small monetary reward ($7) was then given to each of the participants for their participation in the study.

Measures

Relationship satisfaction. To rule out the possibility that the discussion dyads would differ across conditions in their satisfaction with their actual dating relationships, all of the respondents were asked to report their satisfaction on a 7-point scale (1 = not at all satisfied, 7 = really satisfied). Because of the dyadic interdependence in these data due to the fact that half of the discussion dyads were composed of actual dating partners, the ratings for relationship satisfaction were aggregated (i.e., averaged) for the two members of each of the discussion dyads. In 46.4% of the discussion dyads, both individuals reported the maximum score on the scale.

The high percentage of extremely satisfied couples could be a result of a social desirability response bias. On the other hand, it is known that dating partners generally provide higher ratings of relationship satisfaction than do married couples (Noller et al., 1994).(2) We therefore transformed the relationship satisfaction measure into a categorical measure with two categories: dyads high in relationship satisfaction (14 on the aggregated 7-point scales), and dyads lower in relationship satisfaction (ranging from 8 to 13 on the aggregated 7-point scales). Dyads in the four conditions did not differ in their self-reports of relationship satisfaction, [chi square](1) = 0.17 for the interaction with discussion topic, and [chi square](1) = 0.62 for the interaction with type of discussion partner.

Dynamic behavior measures. Four independent judges who were kept unaware, insofar as possible, of the discussion partners' actual dating status, used the DYAD DYNAMICS software program (Ickes, Bissonnette, Garcia, & Stinson, 1990) to code the set of dynamic behavior measures relevant to this study. To further minimize the possibility of rater bias, the raters coded the behavior of only one dyad member at a time. The recorded behaviors included the total frequency and duration of both verbalizations and directed gazes. The interrater reliabilities (Cronbach's [Alpha]) were .68 for frequency of verbalizations, .62 for duration of verbalizations, .74 for frequency of directed gazes, and .64 for duration of gazing. The total duration of verbalizations was then divided by the total frequency of verbalizations for each dyad to compute the mean length of verbalizations (as opposed to mean length of pauses). Both the mean length of verbalizations and the total duration of gazing were used as dependent variables in subsequent analyses.

Personal pronoun usage. Personal pronoun usage was coded from the videotapes by a single judge who used tally marks to record the number of first-and second-person singular pronouns, and the number of third-person plural pronouns, occurring in each dyad's conversation. The number of first-person plural pronouns was divided by the total number of pronouns used to create a dyad-level index of we-ness.

Dyadic interaction coding scheme. The videotaped data were coded with a system designed to capture certain sequential interaction patterns. Because of the relatively short duration (6 min.) of the conflict management discussions, a limited coding system was used that was based on the summary codes of Sillars' Conflict Coding System (Revised Version), as described in Canary et al. (1995). In the Revised Conflict Coding System (Sillars, 1986), seven summary codes are used: Denial and Equivocation, Topic Management, Irreverent Remarks, Analytic Remarks, Conciliatory Remarks, Confrontative Remarks, and Noncommittal Remarks.

For the purposes of this study, the summary category Confrontative Remarks was divided into three subcodes: Demand (including Sillars' subcodes hostile imperatives and questions, and personal criticism), Sarcasm (including Sillars' hostile jokes and rejection), and Defensiveness (including Sillars' denial of responsibility and presumptive remarks). The summary category Noncommittal Remarks was also divided into three subcategories: Irrelevant Remarks (including all Sillars' noncommittal statements and questions, abstract remarks, and procedural remarks that are neutral in affective tone), Positive Affect (including all Sillars' noncommittal statements and questions, abstract remarks, and procedural remarks that are positive in affective tone), and Negative Affect (including all Sillars' noncommittal statements and questions, abstract remarks, and procedural remarks that are negative in affective tone). Finally, and consistent with Sillars' coding scheme, a code for uncodable sequences was provided. The coding system required the assignment of one of these 12 mutually exclusive codes to each spoken thought (i.e., a change in semantical content indicates a different message requiring a new code, Sillars, 1986).

For the final analyses, the 12 codes were aggregated into six summary codes in order to better address the research question and the demands of sequential analysis.(3) Final summary codes include Demand, Withdrawal (Denial/ Equivocation, Topic Management, Irreverent Remarks, and Neutral Noncommittal Remarks, with a Cronbach's [Alpha] = .41), Problem Solving (Analytic Remarks and Conciliatory Remarks, with an intercorrelation r = .35), Defensiveness (Defensiveness and Sarcasm, r = .41), Negative Affect (Negative Noncommittal Remarks), and Positive Affect (Positive Noncommittal Remarks). See the appendix for examples of actual transcripts, labeled by category.

Coding procedure. All of the tapes were coded by one observer. The coder received a detailed description of the code categories, including the description of the different subcodes from the Revised Conflict Coding System (Sillars, 1986), and was required to memorize the 12 code categories as well as the definitions of the different behaviors. Another independent observer, whose data served as a reliability check, coded a random tenth of the videotapes. The two observers practiced coding pilot tapes and were kept unaware, insofar as possible, of any predictor variables being studied. Kappa statistics for interrater reliability are based on a random sample of 12 tapes coded by the two observers. The overall kappa based on all the codes was k = .88. Kappas for the six summary codes ranged from .80 to .97.

A third independent judge viewed each discussion once, and for every tape rated the degree of conflict between the discussion partners, the degree to which the discussion partners reached a solution, and the degree of both positive and negative affect expressed by both partners. All of these ratings were made on 5-point Likert scales (1 = not at all, 5 = very much).

Validity of the coding scheme. The validity of the coding scheme was evaluated by correlating the relative frequency of the behavioral categories of the dyadic interaction coding scheme with the corresponding affect and conflict-resolution ratings just described. As expected, the problem-solving rating data were significantly and positively correlated with the relative frequency of problem-solving behavior, r = .52, p [is less than] .0005, whereas these problem-solving ratings were significantly negatively correlated with the relative frequencies of withdrawal behavior, r = -.29, p = .05, and defensiveness, r = -.40, p [is less than] .0005.

The ratings for positive affect were significantly and positively correlated with the relative frequency of positive affective behavior, r = .40, p [is less than] .0005, and significantly negatively correlated with the relative frequency of defensiveness, r = -.42, p [is less than] .0005. On the other hand, the ratings for negative affect were positively correlated with defensiveness, r = .34, p [is less than] .01, but were not significantly correlated with negative affective behavior.

Overall, these data indicate a general convergence between global ratings of conflict resolution by one observer and the micro-analytic coding of problem-solving and withdrawal behavior by another observer. Convergence between ratings and micro-analytic behavior codes was also found for positive affect, but not for negative affect. The observers' rating of negative affect seemed to be based on the degree of defensiveness expressed by the subjects, rather than on the degree of expressed distress.

RESULTS

Tests of the First Hypothesis: Evidence for Relational Goal Conflict

We predicted that safer-sex discussions between intimate partners would result in more relational goal conflict than would be observed in the three remaining conditions. Hypothesis 1 proposed that more relational goal conflict would result in shorter verbalizations (longer pauses), lower levels of we-ness, and a different pattern of directed gazes during the 6-minute discussions. This hypothesis was tested in a series of planned contrasts that compared the mean of the intimate partners/safer-sex discussion condition against the combined mean of the three remaining conditions.

These planned contrasts yielded the predicted pattern of results. Actual dating partners who discussed safer sex differed significantly from the remaining dyad types in the dyads' mean length of verbalizations, F(1,111) = 14.42, p [is less than] .05; proportion of first-person pronouns used in their conversation, F(1,111) = 19.91, p [is less than] .001; and the dyads' total duration of directed gazes, F(1,111) = 21.01, p [is less than] .001. As expected, dating partners who discussed safer sex displayed shorter verbalizations (longer pauses) and fewer we, our, and us references in their conversations relative to the other dyad types (means are displayed in Table 1). They also looked at each other longer, which could be interpreted as consistent with the goal-detection function of gazing. According to Scheffe post hoc tests, the means for directed gazes and verbalizations did not differ significantly among the remaining three dyad types. The mean use of first-person pronouns did, however, differ among the remaining dyads such that we, our, and us references were more frequently used during the joint leisure time discussion by both intimate partners and strangers.

Table 1. Means and Standard Deviations for the Dyad-Level Behavioral
Relational Goal-Conflict Measures

                                  Safer-Sex Discussions
                                Partner            Partner
                                  n=33               n=24
Behaviors                      M        SD        M        SD
Verbalizations (seconds)     3.56(a)    .32     3.81(b)    .38
Gazing (seconds)           230.76(a)   8.32   195.88(b)   9.75
We-ness (%)                  6.53(a)   1.21     6.41(b)   1.42

                                 Leisure Time Discussions
                                Partner           Stranger
                                  n=30              n=33
Behaviors                      M        SD        M        SD
Verbalizations (seconds)     4.19(b)    .34     5.13(b)     .32
Gazing (seconds)           193.63(b)   8.72   164.10(b)    8.32
We-ness (%)                 14.15(c)   1.27    15.32(c)    1.21

Note. Means with different superscripts differ significantly from each other.

If this constellation of behaviors is indeed indicative of greater relational goal conflict, then the three observational measures should load on one common factor. To test this hypothesis, we standardized each of the behavioral measures and then conducted a confirmatory factor analysis using LISREL 8. The results of the saturated model revealed that each of the three behavioral measures contained both a specific component and a significant contribution to one common latent variable, which we believe to be relational goal-conflict (significant standardized factor loadings are reported in Figure 1). Each of the behavioral measures was found to be a moderately valid indicator of this common, with mean duration of verbalizations being the most valid indicator, [R.sup.2] = .20, followed by use of first-person pronouns, [R.sup.2] =. 16, and duration of directed gazes, [R.sup.2] =. 14. A planned contrast that compared the mean of the intimate partners/safer-sex discussion condition against the combined mean of the three remaining conditions on the factor scores yielded the predicted pattern of results, F(1,111) = 30.78, p [is less than] .0005.

[Figure 1 ILLUSTRATION OMITTED]

Tests of Hypothesis 2: Negative Communication Behavior in the Four Dyad Types

Hypothesis 2 predicted that greater behavioral withdrawal and less direct problem-solving would be especially evident during the actual dating partners' discussions of safer sex. In a set of exploratory analyses, we examined the incidence of the different behavior codes in each of the four dyad types for further evidence that safer-sex discussions between intimate partners result in more goal conflict, less direct problem-solving behavior, and more withdrawal behavior. To avoid overlooking any possible interaction effects, we conducted a 2 (discussion topic: safer sex vs. joint leisure time) x 2 (discussion partners: intimates vs. strangers) multivariate analyses of variance (MANOVA) in which discussion partner and discussion topic were treated as between-dyad factors.

The dependent measures were the percentages for each of the summary behavior codes. The proportions of the summary codes were transformed using an arcsin transformation in order to avoid a nonsingular matrix in the multivariate analysis in which one of the six proportions is a linear combination of the other five proportions. The means and standard deviations of the percentages for each of the summary codes are presented in Table 2 (main effects) and Table 3 (interaction effects).

Table 2. Means and Standard Deviations for the Dyad-Level Mean
Behavioral Measures

                           Effect Topic
                   Safer Sex       Leisure
                     n=57           n=63
Behaviors (%)       M      SD      M      SD        F
Problem Solving   18.27   1.77   27.31   1.64   14.22(**)
Defensiveness     31.84   1.89   33.41   1.75    0.27
Avoidance         22.52   1.25   14.83   1.16   19.66(**)
Demand            14.53   1.14   12.93   1.06    1.58
Negative affect    5.63   0.80    5.09   0.74    0.13
Positive affect    7.21   0.98    6.42   0.91    0.19

                        Effect Partner
                     Partner      Stranger
                      n=63          n=57
Behaviors (%)       M      5D      M      SD        F
Problem Solving   18.06   1.66   27.52   1.74   12.46(**)
Defensiveness     36.95   1.78   28.62   1.86    9.01(*)
Avoidance         18.57   1.18   18.78   1.23    0.03
Demand            15.22   1.08   12.24   1.13    2.23
Negative affect    5.72   0.75    5.00   0.79    0.43
Positive affect    5.79   0.92    7.84   0.97    2.21

(*) p<.01. (**) p<.001.

Table 3. Means and Standard Deviations for the Dyad-Level Mean
Behavioral Measures

                     Safer-Sex Discussions
                     Partner       Stranger
                      n=33           n=24
Behaviors (%)       M      SD      M      SD
Problem Solving   14.35   2.38   21.20   2.71
Defensiveness     37.55   2.47   26.01   2.90
Avoidance         21.22   1.65   24.57   1.94
Demand            16.09   1.51   13.37   1.77
Negative Affect    5.45   1.07    5.97   1.25
Positive Affect    5.33   1.29    8.87   1.51

                    Leisure Time Discussions
                    Partner       Stranger      Interaction
                     n=30           n=33           Effect
Behaviors (%)       M      SD      M      SD
Problem Solving   21.89   2.42   32.40   2.31       0.54
Defensiveness     35.29   2.59   31.02   2.47       1.93
Avoidance         16.49   1.74   13.78   1.65       2.90(*)
Demand            13.78   1.58   11.73   1.51       0.04
Negative Affect    6.30   1.12    4.30   1.07       1.23
Positive Affect    6.25   1.35    6.77   1.29       1.83

(*) p< .10.

The multivariate analysis yielded no significant interactions, but two significant main effects: one for discussion partner (F(6,111) = 2.99, p [is less than] .01) and one for discussion topic (F(6,111)= 4.68, p [is less than] .001). Univariate analyses revealed that discussions with intimate partners were characterized by less direct problem solving and more defensiveness than discussions with strangers (see Table 2). Safer-sex discussions were characterized by less direct problem solving and more avoidance than discussions of joint leisure time (see Table 2). The effects for direct problem-solving behavior were additive (the probability of a nonsignificant interaction effect is p [is greater than] .25), such that the safer-sex discussions between intimate partners were characterized by the fewest problem-solving behaviors (see Table 3).

Consistent with the goal-conflict behavioral data, a follow-up contrast comparing the mean for the first dyad type against the remaining three revealed that intimate partners who discussed safer sex displayed significantly less direct problem solving than did partners in the remaining three conditions, F(1,111) = 13.44, p [is less than] .001. On the other hand, the means for direct problem solving differed significantly among the remaining three dyad types such that problem solving was more frequently used during joint leisure time discussions between strangers than in similar discussions between intimates, and problem solving was more frequently used when strangers discussed joint leisure time than when they discussed safer sex.

Test of Hypothesis 3: Sequential Analyses

Hypothesis 3 proposed distinctive ways in which one partner responds to the other partner during dating partners' discussions of safer sex. Of particular importance were the predicted negative communication patterns, the predictors of withdrawal and problem solving in safer-sex discussions compared to joint leisure time discussions, and discussions between intimates compared to discussions between strangers (see Hypothesis 3). In order to test whether or not the response of a dating partner is conditionally dependent on the behavior of the other partner, a matrix was computed for each cell of the 2 (discussion topic: safer sex vs. joint leisure time) x 2 (discussion partners: intimates vs. strangers) design with the six codes as Antecedents and the six codes as Consequents. The sequential analyses aim at investigating whether a particular communication sequence occurs more or less likely than one would expect from the base rate obtained in the test of Hypothesis 2.

Because there were six code categories and relatively short streams of data for each dyad, there were insufficient data to do sequential analyses on a dyad-by-dyad basis. The solution to this problem was to pool the data across the dyads in each of the four dyad types (Gottman & Roy, 1990). Pooled over the four cells of the design, the data contained 5453 codes. The problem with pooling data is the implicit assumption that the parameters of the sequential connection are stable across dyads and over time. These assumptions were tested for the four cells of the design. Dyads within each cell were randomly divided into two groups to test stability across dyads, and the test for stability over time was based on dividing data streams into halves. Loglinear models with and without these factors were compared using backward elimination. Based on the results of these analyses, it can be concluded that the first-order Markov chain is stable across the cells of the design, but cannot be considered stable over time. The latter reflects the time-dependence of the occurrence of communication behavior during the conflict management role-plays (see also Margolin, Burman & John, 1989). Analyses on the pooled data, however, will not allow us to investigate these fluctuations in the development of the dyadic interactions.

Log-linear analysis. To fit a hierarchical log-linear model to the data, backward elimination was used. We added .5 to every cell in the design that was not a logical zero prior to analysis, a procedure recommended by Goodman (1971). The strategy for finding the best-fitting model is illustrated in Table 4. First, the saturated model, including all possible covariations between the variables, is tested. This model fits the data perfectly, but uses all the degrees of freedom and might involve too good a fit, that is, it might include unnecessary parameters (Bishop & Holland, 1975). Then, gradually, covariations are set to zero, and as long as a subsequent model does not differ significantly from the former model, the model is considered acceptable (Knoke & Burke, 1980).

Table 4. Multivariate Log-Linear Analysis

Fitted marginals(a)                      [chi square]   df
1 Antecedents x consequences                  0         0
   x topic x partner
2 Antecedents x consequences x topic
  Antecedents x consequences x partner
  Antecedents x topic x partner
  Consequences x topic x partner            27.97       25

                                                 [chi square]
Fitted marginals(a)                        p        [Alpha]
1 Antecedents x consequences              1.00
   x topic x partner
2 Antecedents x consequences x topic
  Antecedents x consequences x partner
  Antecedents x topic x partner
  Consequences x topic x partner          0.31       27.97

Fitted marginals(a)                      df [Alpha]    p
1 Antecedents x consequences
   x topic x partner
2 Antecedents x consequences x topic
  Antecedents x consequences x partner
  Antecedents x topic x partner
  Consequences x topic x partner             25       0.31

(a) A general test of the models specified in column 2, is presented in column 3 (chi-square value), column 4 (degrees of freedom), and column 5 (significance level). The testing strategy reported in this table is the opposite of the one taught in conjunction with traditional chi-square tests since a low chi-square value relative to the degrees of freedom is expected. The changes in chi-square value and in the degrees of freedom when comparing the model specified in column 2 with the former model, are reported in column 6 and 7. Finally, the significance of these changes is reported in column 8.

The hierarchical model selected by this procedure includes the terms Topic x Partner x Antecedent, Topic x Partner x Consequent, Topic x Antecedent x Consequent, and Partner x Antecedent x Consequent, to produce the desired nonsignificant [LRX.sup.2] (having a value of 27.97 with 25 degrees of freedom, p = .31). Because the probability of a Type I error lies between .10 and .35, the model can be accepted as one fitting the data (Knoke & Burke, 1980).

Two important results emerged from this analysis. First, two of the terms--Topic x Partner x Antecedents and Topic x Partner x Consequences--indicated that the Antecedents and Consequences changed with the interaction of discussion topic and discussion partner. However, both terms were essentially the same term because the Antecedent at time t + 1 is the Consequent at time 1. These effects are comparable to those obtained in our ANOVAs of the nonpooled data (see Table 2 and Table 3). In that analysis, the three-way interaction Topic x Partner x Behavioral Codes was significant for only one of the six codes, and resulted in a nonsignificant multivariate effect.

Second, two other terms indicate that the 6 x 6 matrix itself was associated significantly with both discussion topic and discussion partner. These terms are theoretically important in revealing that the truly interactive nature of conflict (reflected by the matrix) was affected by the variation in both the discussion topic and the discussion partner, but not by the interaction of both.

The variation of discussion topic for the matrix, collapsing over discussion partner. The transitional probabilities needed for sequential analyses and the corresponding z-scores were calculated using Bakeman and Quera's SDIS and GSEQ program. Lag 1 z-scores represent the degree to which the target code is immediately contingent on the occurrence of the given code. Between-group comparisons were made by comparison of the z-scores.

As mentioned earlier, safer-sex discussions were characterized by more withdrawal and less direct problem solving compared to joint leisure time discussions. Of interest in the sequential analyses were the predicted negative communication patterns and the predictors of withdrawal in safer-sex discussions compared with those in joint leisure time discussions. Table 5 presents the significant between-group differences in z-scores.

Table 5. Lag 1 z-Scores and Between Group (Safer Sex Discussions vs. Leisure Time Discussions) Comparisons

                                               Discussion Topic
                                          Safer Sex    Leisure Time
Sequences                                 z - scores    z - scores
Negative Reciprocity - Distancing
 Negative Affect - Negative Affect          -2.36          0.45
Defensiveness - Defensiveness                1.27          4.33
Problem Escalation
 Problem Solving - Negative Affect           0.43         -3.22
 Problem Solving - Defensiveness            -2.85         -7.79
Open Conflict
 Demand - Defensiveness                      5.83          7.13
Demand/Withdrawal
 Demand - Withdrawal                         7.10          2.96
Attraction
 Positive Affect - Positive Affect          10.26          3.99
Problem Acceptance

 Problem Solving - Problem Solving           8.60         12.16
 Problem Solving - Positive Affect           6.33         11.36
 Positive Affect - Problem Solving           4.83          8.07
Specific Antecedents of Avoidance
 Positive Affect - Withdrawal               -4.33         -0.43
Specific Antecedents of Problem Solving
Withdrawal - Problem Solving                -1.86         -5.44
Defensiveness - Problem Solving             -3.62         -6.57

Sequences                                 z - comparison
Negative Reciprocity - Distancing
 Negative Affect - Negative Affect            -1.72
Defensiveness - Defensiveness                 -2.16(*)
Problem Escalation
 Problem Solving - Negative Affect             2.58(**)
 Problem Solving - Defensiveness               3.49(**)
Open Conflict
 Demand - Defensiveness                       -0.92
Demand/Withdrawal
 Demand - Withdrawal                           2.93(**)
Attraction
 Positive Affect - Positive Affect             4.34(**)
Problem Acceptance
 Problem Solving - Problem Solving            -2.52(*)
 Problem Solving - Positive Affect            -3.56(**)
 Positive Affect - Problem Solving            -2.29(*)
Specific Antecedents of Avoidance
 Positive Affect - Withdrawal                 -2.76(**)
Specific Antecedents of Problem Solving
Withdrawal - Problem Solving                   2.53(*)
Defensiveness - Problem Solving                2.09(*)

Note. A positive z-score indicates that the consequent behavior, following the antecedent, was more likely than one would expect from the base rate. Similarly, a negative z-score indicates that the chances of the consequent behavior, following the antecedent, were smaller than the chances based on the overall base rates. The formula for comparing two z-scores is z = ([z.sub.1] - [z.sub.2]) / x [square root of] (Rosentahl & Rosnow, 1991).

(*) p < .05. (**) p < .01.

Table 5 reveals that the typical initial response to a partner's problem solving, in all dyads, was the reciprocity of problem solving. Interaction in the control dyads was characterized not only by more problem-solving interventions, but also by a number of contingent behavior patterns implying problem solving--such as the initial induction of problem solving in response to a partner's positive affect, and the initial reduction of problem solving in response to a partner's defensiveness and withdrawal. Other contingent behavior patterns implying problem solving were problem acceptance (positive affect in response to a partner's problem solving) and problem escalation (negative affect or defense in response to a partner's problem solving).

All of these behavior patterns were significantly less contingent during the safer-sex discussions than during the joint leisure time discussions. In other words, the occurrence of problem solving by either partner was significantly more likely to result in problem solving or positive affect and significantly less likely to result in negative affect or defense during joint leisure time discussions than during safer-sex discussions. Similarly, the occurrence of positive affect was significantly more likely to result in problem solving, whereas it was significantly less likely to result in negative affect or defense, during the joint leisure time discussions.

The occurrence of expressions of both negative and positive affect did not vary as a function of discussion topic. Similarly, Table 5 shows that no negative communication patterns regarding expression of affect--increased negative reciprocity and decreased attraction--were evident in these discussions. On the contrary, although attraction was a contingent sequence in both discussions, the contingency of positive affect expressed by both partners was stronger in the safer-sex discussions than in the joint leisure time discussions. Moreover, no differences between the two conditions were found in contingencies of negative affect. Based on the ratings used in the validation of the coding scheme, however, one might conclude that the rating of negative affect was based more on the occurrence of defense than on the occurrence of negative affect per se. Reciprocity of defense was found to be contingent only during the joint leisure time discussions.

The predicted negative communication patterns, demand-withdrawal and demand-defense, were exhibited during both discussions. The occurrence of demand by either partner was, however, more likely to result in withdrawal during the safer-sex discussions than during the leisure time discussions. Withdrawal was found to be evoked by demandingness, and not by positive affect, in both conditions. Again, this contingency was found to be stronger during the safer-sex discussions.

The variation of relationship status for the matrix, collapsing over discussion topic. As noted earlier, discussions between the intimate partners were characterized by more defense and less direct problem solving. Defensiveness during discussions between intimates was found to be evoked by withdrawal, whereas this same contingency was not significant in the discussions between strangers (see Table 6).

Table 6. Lag 1 z - scores and Between Group (Intimate Partners vs. Strangers) Comparisons

                                                Discussion Partner
                                              Intimate     Stranger
Sequences                                     z-scores     z-scores
Negative Reciprocity - Distancing
 Negative Affect - Negative Affect             -2.57        -1.42
 Defensiveness - Defensiveness                  1.20         4.31
Problem Escalation
 Problem Solving - Negative Affect             -0.80        -2.29
 Problem Solving - Defensiveness               -5.30        -4.77
Open Conflict
 Demand - Defensiveness                         6.83         5.94
Demand/Withdrawal
 Demand - Withdrawal                            4.27         6.64
Avoidance Reciprocity
 Withdrawal - Withdrawal                       -3.05         0.49
Attraction
 Positive Affect - Positive Affect              8.66         5.26
Problem Acceptance
 Problem Solving - Problem Solving             10.09        10.66
 Problem Solving - Positive Affect              9.10         8.20
 Problem Solving - Avoidance                   -2.67        -5.66
Antecedents/Consequences of Defensiveness
 Withdrawal - Defensiveness                     3.11        -0.65
 Defensiveness - Positive Affect               -8.36        -5.38
 Defensiveness - Demand                         7.67         3.94

Sequences                                   z-comparison
Negative Reciprocity - Distancing
 Negative Affect -Negative Affect              -0.81
 Defensiveness - Defensiveness                 -2.20(*)
Problem Escalation
 Problem Solving - Negative Affect              1.05
 Problem Solving - Defensiveness               -0.37
Open Conflict
 Demand - Defensiveness                         0.63
Demand/Withdrawal
 Demand - Withdrawal                           -1.68
Avoidance Reciprocity
 Withdrawal - Withdrawal                       -2.50(*)
Attraction
 Positive Affect - Positive Affect              2.40(*)
Problem Acceptance
 Problem Solving - Problem Solving             -0.40
 Problem Solving - Positive Affect              0.63
 Problem Solving - Avoidance                    2.11(*)
Antecedents/Consequences of Defensiveness
 Withdrawal - Defensiveness                     2.66(**)
 Defensiveness - Positive Affect               -2.11(*)
 Defensiveness - Demand                         2.64(**)

(*) p < .05. (**) p < .01.

Besides the negative reciprocity with regard to defensiveness, the occurrence of defensiveness by either partner was found to result in demandingness, and not in positive affect, in both conditions. Reciprocity of defensiveness was, however, significantly more likely to occur during discussions between strangers, whereas defensiveness was more likely to result in demandingness and less likely to result in positive affect during discussions between intimates.

There were no significant differences in the contingencies of antecedents of problem solving during discussions between intimates and strangers, and there was only one difference in the contingency of the consequences of problem solving. Although problem solving by either partner was not likely to result in withdrawal in both conditions, this was significantly more the case during discussions between strangers than during discussions between intimates.

The predicted sequences demand-withdrawal and demand-defensiveness were found to be contingent in both conditions. Some differences between the conditions in the antecedents of both withdrawal and demandingness were found, however. As already noted, withdrawal was not likely to be preceded by withdrawal during discussions between intimates, whereas withdrawal was not likely to be preceded by problem solving in both conditions, especially in discussions between strangers. Demandingness was likely to be preceded by withdrawal and by defensiveness in both conditions, but the sequence withdrawal-demandingness was significantly more contingent during discussions between strangers, whereas the sequence defensiveness-demandingness was significantly more contingent during discussions between intimates. Particularly during discussions between intimates, demandingness was found to be preceded by and to result in negative affect. These contingencies were not significant during discussions between strangers.

Although no differences were found in the occurrences of both negative and positive affect during discussions between strangers and intimates, negative reciprocity was found to be more frequent during discussions between strangers, whereas attraction was found to be more frequent during discussions between intimates.

DISCUSSION

This study examined conflict interactions with the goal of identifying distinctive behavior patterns in stable dating couples who discuss the implications of potential outside sexual involvements for practicing safer sex in their relationship. Safer-sex discussions in both dating and nondating couples were compared to discussions of a control topic--a conflict about joint leisure time activities--which pretesting had revealed to be a topic less threatening to a relationship. Our predictions were based on the assumption that, relative to the leisure activities discussion, the safer-sex discussion would evoke a high level of goal conflict--but only in actual dating partners, as opposed to opposite-sex strangers.

Three hypotheses were tested. Our first hypothesis proposed that safer-sex discussions between stable dating partners would be characterized by relatively longer pauses, less references to shared experiences (we-ness), and a distinctive pattern of directed gazes that all suggest the presence of high levels of goal incompatibility. Consistent with Hypothesis 1 at the most general level, the results of a confirmatory factor analysis indeed confirmed that each of the three behavioral measures--total length of directed gazes; mean length of verbalizations; and we, our, and us references in conversations-had a unique contribution to a latent construct that appears to represent relational goal conflict.

Consistent with Hypothesis 1 at a more specific level, dating couples who discussed safer sex displayed shorter verbalizations (longer pauses)--a finding which has previously been interpreted as being a consequence of the simultaneous presence of multiple and incompatible goals (Greene, 1995; Greene, Lindsey, & Hawn, 1990). In addition, the couples who discussed safer sex referred less to shared experiences (we-ness), a finding which, in the context of conflictive interactions, indicates threat to the relationship (Gottman, 1993). Finally, consistent with the goal-detection function of eye contact but inconsistent with research showing that eye contact decreases during cognitive load, dating couples who discussed safer sex displayed longer directed gazes, suggesting that they had an increased need for goal detection in this ambiguous situation (Phillips et al., 1992). Given the nature of the conflict situations that the partners were asked to discuss, the relatively high level of eye contact displayed by the dating couples in the safer-sex discussions may have reflected both the difficulty and the importance of diagnosing the partner's goals under high relational stress.

Our second hypothesis proposed that the safer-sex discussions would evoke some dysfunctional communication behaviors that would be significantly less evident in the control (leisure activity) discussions. Consistent with Hypothesis 2, direct problem-solving interventions were found to occur less often in the safer-sex discussions of actual dating partners than in the other three conditions of the study. In addition to this predicted effect, safer-sex discussions were found to engender more withdrawal in both intimates and strangers, and the discussions involving intimates were found to engender more defensive reactions than those involving strangers.

Our third hypothesis led us to expect greater evidence of negative communication patterns in the safer-sex discussions of intimate partners than in the other types of discussions we studied. These negative communication patterns were investigated through sequential analyses conducted along the three most salient conflict dimensions: (a) constructive versus destructive behavioral contingencies, (b) interaction patterns referring to engagement versus avoidance, and (c) expression of negative affect versus positive affect. With regard to two of these dimensions--constructive/destructive and engagement/avoidance--negative patterns of the types identified in distressed couples were also detected in the safer-sex discussions. On the other hand, typical patterns of reciprocated negative affect in distressed couples were not replicated in these discussions.

Interaction in the control topic condition was characterized not only by more problem-solving interventions, but also by a number of contingent behavior patterns implying problem solving. A major difference between dyads involved in safer-sex discussions and dyads involved in joint leisure time discussions was the relative absence of behavioral dependencies related to problem-solving in the former dyads. Specifically, both the positive contingencies related to problem acceptance and the negative contingencies related to problem escalation were weaker. Discussions between strangers were also characterized by more problem solving than discussions between intimates, but none of the contingencies related to problem solving varied as a function of relationship status.

Dyads involved in safer-sex discussions also exhibited more avoidance, in addition to increased problem escalation and decreased problem acceptance. The relative merits of avoidance versus engagement have been questioned (Gottman, 1993; Noller et al., 1994), but the demand/withdrawal pattern is still typically seen as negative during conflictive interactions (cf. Christensen & Shenk, 1991; Gottman, 1993; Heavey, Christensen, & Malamuth, 1995). The contingencies of both the demand/withdrawal sequence and the demand/defense sequence were found to be stronger during the safer-sex discussions. Reciprocal defensiveness was also found to be more contingent during these discussions, so that both open conflict (sequences of reciprocal defensiveness after a partner's demand) and uninvolvement (of one partner in response to the other's demand) were more likely to occur during discussions implying high goal conflict than during control discussions.

Surprisingly, negative reciprocity was not found in the safer-sex discussions, despite their presumably high goal conflict. Moreover, mutual attraction was more evident in these discussions than in the control discussions. The most obvious explanation for these seemingly anomalous

findings is the topic itself. Discussing safer sex with an opposite--sex partner--either one's own or someone else's--probably created a sexually-charged atmosphere in which the discussion partners' attraction to each other could more easily be acknowledged.

Our findings further revealed that, in the thematic discussions of potential extradyadic affairs and their relevance to practicing safer sex, both destructive and avoidant communication styles and patterns were evident. The relative merits of avoidance versus engagement have been questioned, but because these couples were generally satisfied with their actual dating relationships, avoidance was probably the best way for them to cope with the goal conflict that was induced during the safer-sex discussion. On the other hand, the relative absence of problem solving and problem acceptance, and the relatively frequent occurrence of problem escalation, must clearly be seen as instances of negative communication. It seems evident that these patterns of behavior are--at least to some extent--evoked by the induced relational goal-conflict.

The study has a couple of noteworthy limitations: the use of role-played laboratory discussions and some potential sampling biases deriving from the fact that 45% of the couples who were approached declined to participate in the study. Although these limitations might not compromise the internal validity of the findings, it is clear that their external validity (i.e., generalizability) has not yet been demonstrated and cannot be assumed. The present study is, in essence, an analogue study that is suggestive rather than conclusive in its theoretical implications.

Although the use of transgression scenarios was of particular relevance for the current study, it undoubtedly restricts the generality of the results. Beyond the issue of external validity, there is an alternative explanation for many of the findings that cannot be ruled out. Whereas some of the analyses supported our contention that discussing safer sex is more difficult for actual dating couples than for opposite-sex strangers, other analyses suggested that discussing safer sex was more difficult than discussing leisure activities regardless of couple type. This pattern of differences could reflect a demand effect of the research. We asked individuals to discuss intimate sexual details, albeit imaginary, in front of a camera and--in half of the cases--with a complete stranger. It is possible that the same communicative patterns would emerge under these conditions regardless of which sexual topic the couples discussed.

On the other hand, if the sheer difficulty of talking about sexuality induces goal conflict and relational strain in itself, there is still merit in our argument that discussing safer sex with an established dating partner should create especially high levels of goal conflict and relational strain. If talking about safer sex is necessary for practicing safer sex--as the positive correlation for these variables in previous studies indicates (see e.g. Barthlow, Horan, DiClemente, & Lanier, 1995; Catania et al., 1994)--intimate dating partners are in a decidedly paradoxical position. On the one hand, they should be encouraged to talk about safer sex in order to practice safer sex, but on the other hand they may need to protect themselves from experiencing the distressingly high levels of goal conflict and relational strain that such discussions induce. Their basic dilemma is that although effective safer-sex talk can be good for the partners' sexual health, it can also have negative implications for the health of their relationship (i.e., their perceptions of mutual trust and commitment).

Resolving this paradox may not be easy, as it may require the partners to suddenly and dramatically redefine their relationship, seeing it perhaps as less unique, less exclusive, less romantic, and less committed. Because many, if not most, couples will resist such redefinition, the corresponding dilemma for health educators should be clear. How can health educators reframe the emotional and relational significance of safer-sex talk in a way that is less threatening to relationship partners? For there can be little doubt that discussions of previous sexual partners, history of condom use, and so on can be profoundly threatening in just the ways we have described.

The topic of how to cope effectively with such a relationship threat is one that is still largely unexplored. However, several writers have made useful suggestions that include (a) talking about safer sex in the context of birth control or in the context of stories about the experiences of friends and acquaintances, (b) making safer sex an issue in the very beginning of a relationship, and (c) using humor to discuss safer-sex issues in a less threatening way (see Adelman, 1992; Buysse, 1998; Buysse & Van Oost, 1997; Metts & Fitzpatrick, 1992). Empirical evidence regarding the effectiveness of these techniques is still lacking, however. Although the present study provides evidence that stable dating partners display a wide range of behaviors believed to be indicative of goal conflict during safer-sex discussions, we still need clearer answers to the question of exactly how partners can discuss this important issue in a way that does not threaten the very relationship the discussion is intended to preserve.

This study is part of the first author's doctoral dissertation at the University of Ghent under the supervision of Paulette Van Oost, to whom the authors extend their thanks. Thanks also to the students for their assistance in coding the data. The study was supported by a doctoral fellowship awarded to the first author by the Belgian National Fund for Scientific Research.

Address correspondence to Ann Buysse, University of Ghent, Research Group Health and Behavior, H. Dunantlaan 2, 9000 Ghent, Belgium; e-mail: Ann. [e-mail]

(1) The time limit of 6 minutes was derived from an earlier pilot study with 13 couples, in which the average length of the conflicts ranged from 4 to 6 minutes. Burman, Margolin, and John (1993) observed two conflict interactions without time limits as they normally evolved at home. In their study, the interactions ranged from approximately 5 to 20 minutes, with an average exchange lasting about 5 minutes, averaged across the two conflicts.

(2) Individuals who rated their relationship as highly satisfactory differed from individuals with lower relational satisfaction on a variety of relational measures, F(12,217) = 12.17, p [is less than] .0005. More specifically, they reported higher ratings of relational closeness, similarity, trust, and uniqueness. They also reported fewer discussions, more sexual behavior, and more specific future plans, and they were older at relationship onset.

(3) It was necessary to reduce the number of separate codes for the sequential analysis that examines each possible combination of a two-code sequence. The six summary codes refer to the three most studied dimensions in conflict interactions (see a.o. Canary et al., 1995): (a) engagement (demand) versus avoidance (withdrawal), (b) negative affect versus positive affect, and (c) constructive (problem solving) versus destructive (defensiveness).

REFERENCES

Abraham, C. S., & Sheeran, P. (1993). In search of a psychology of safer-sex promotion: Beyond beliefs and texts. Health Education Research, 8, 245-254.

Adelman, M. B. (1992). Healthy passions: Safer sex as play. In T. Edgar, M. A. Fitzpatrick, & V. S. Freimuth (Eds.), AIDS, a communication perspective (pp. 69-90). Hillsdale, NJ: Lawrence Erlbaum Associates.

Alberts, J. K., & Driscoll, G. (1992). Containment versus escalation: The trajectory of couples' conversation complaints. Western Journal of Communication, 56, 394--412.

Argyle, M. (1972). The psychology of interpersonal behaviour. Harmondsworth, UK: Penguin.

Argyle, M., & Cook, M. (1976). Gaze and mutual gaze. Cambridge, UK: Cambridge University Press.

Barthlow, D. J., Horan, P. F., DiClemente, R. J., & Lanier, M. (1995). Correlates of condom use among incarcerated adolescents in a rural state. Criminal Justice and Behavior, 22, 295-306.

Beattie, G. W. (1979). Planning units in spontaneous speech: Some evidence from hesitation in speech and speaker gaze direction in conversation. Linguistics, 17, 61-78.

Beattie, G. W. (1981). A further investigation of the cognitive inferences hypothesis of gaze patterns during conversations. British Journal of Social Psychology, 20, 243-248.

Bennum, I. (1985). Prediction and responsiveness in behavioral martial therapy. Behavioural Psychotherapy, 13, 186-201.

Berger, C. R., & Jordan, J. M. (1992). Planning sources, planning difficulty and verbal fluency. Communication Monographs, 59, 130-149.

Bishop, Y. M. M., & Holland, P. W. (1975). Discrete multivariate analysis: Theory and practice. Cambridge, UK: MIT Press.

Burman, B., Margolin, G., & John, R. S. (1993). America's angriest home videos: Behavioral contingencies observed in home reenactments of marital conflict. Journal of Consulting and Clinical Psychology, 61, 28-39.

Buysse, A. (1998). Safer sexual decision making in stable and casual relationships: A prototype approach. Psychology and Health, 13, 55-66.

Buysse, A., & Ickes, W. (1999). Topic-relevant cognition and empathic accuracy in laboratory discussions of safer sex. Psychology and Health, 14, 351-366.

Buysse, A., & Van Oost, P. (1997). Appropriate male and female safer sexual behavior in heterosexual relationships. AIDS Care, 9, 549-561.

Canary, D. J., Cupach, W. R., & Messman, S. J. (1995). Relationship conflict: Conflict in parent-child, friendship, and romantic relationships. Thousand Oaks, CA: Sage.

Catania, J. A., Coates, T J., & Kegeles, S. (1994). A test of the AIDS Risk Reduction Model: Psychosocial correlates of condom use in the AMEN cohort survey. Health Psychology, 13, 548-555.

Christensen, A. (1987). Assessment of behavior. In K. D. O'Leary (Ed.), Assessment of marital discord (pp. 13-57). Hillsdale, NJ: Erlbaum.

Christensen, A. (1988). Dysfunctional interaction patterns in couples. In P. Noller & M. A. Fitzpatrick (Eds.), Perspectives on martial interaction (pp. 31-52). Clevedon, Avon, UK: Multilingual Matters.

Christensen, A., & Shenk, J. L. (1991). Communication, conflict, and psychological distance in nondistressed, clinic, and divorcing couples. Journal of Consulting and Clinical Psychology, 59, 458-463.

Cline, R. J., Freeman, K. E., & Johnson, S. J. (1990). Talk among sexual partners about AIDS: Factors differentiating those who talk from those who do not. Communication Research, 17, 792-808.

Cupach, W. R., & Metts, S. (1994). Facework. Thousand Oaks, CA: Sage.

Dolcini, M. M., Coates, T. J., Catania, J. A., & Kegeles, S. (1995). Multiple sexual partners and their psychosocial correlates: The population-based AIDS in multiethnic neighborhoods (AMEN) study. Health Psychology, 14, 22-31.

Epstein, N., Baucom, D. H., & Rankin, L. A. (1993). Treatment of marital conflict: A cognitive-behavioral approach. Clinical Psychology Review, 13, 45-57.

Fisher, J. D., Fisher, W. A., Williams, S. S., & Malloy, T. E. (1994). Empirical tests of an information-motivation-behavioral skills model of AIDS-preventive behavior with gay men and heterosexual university students. Health Psychology, 13, 238-250.

Fitzpatrick, M. A., & Winke, J. (1979). You always hurt the one you love: Strategies and tactics in interpersonal conflict, Communication Quarterly, 27, 3-11.

Fullilove, M. T., Fullilove, R. E., Haynes, K., & Gross, S. (1990). Black women and AIDS prevention: A view towards understanding the gender rules. The Journal of Sex Research, 27, 47-64.

Gilbert, D. T., & Krull, D. S. (1988). Seeing less and knowing more: The benefits of perceptual ignorance. Journal of Personality and Social Psychology, 54, 193-202.

Gilbert, D. T., Pelham, B. W., Krull, D. S. (1988). Cognitive busyness: When person perceivers meet persons perceived. Journal of Personality and Social Psychology, 54, 733-740.

Gold, R. S., Skineer, M. J., & Ross, M. W. (1994). Unprotected anal intercourse in HIV-infected and non-HIV-infected gay men. The Journal of Sex Research, 31, 59-77.

Goodman, L. A. (1971). The analysis of multidimensional contingency tables: Stepwise procedures and direct estimation methods for building models for multiple classifications. Technometrics, 13, 33-61.

Gottman, J. M. (1993). The roles of conflict engagement, escalation, and avoidance in marital interaction: A longitudinal view of five types of couples. Journal of Consulting and Clinical Psychology, 61, 6-15.

Gottman, J. M., & Roy, A. K. (1990). Sequential analysis: A guide for behavioral researchers. Cambridge, UK: Cambridge University Press.

Gray, L. A., & Saracino, M. (1991). College students' attitudes, beliefs, and behaviors about AIDS: Implications for family life educators. Family Relations, 40, 258-263.

Greenberg, L. S., & Johnson, S. M. (1986). Affect in marital therapy. Journal of Marital and Family Therapy, 12, 1-10.

Greene, J. O. (1995). An action-assembly perspective on verbal and nonverbal message production: A dancer's message unveiled. In D. E. Hewes (Ed.), The cognitive bases of interpersonal communication (pp. 51-86). Hillsdale, NJ: Lawrence Erlbaum Associates.

Greene, J. O., Lindsey, A. E., & Hawn, J. J. (1990). Social goals and speech production: Effects of multiple goals on pausal phenomena. Journal of Language and Social Psychology, 9, 119-134.

Heavy, C. L., Christensen, A., & Malamuth, N. M. (1995). The longitudinal impact of demand and withdrawal during marital conflict. Journal of Consulting and Clinical Psychology, 63, 797-801.

Helwig, L. M., & Collins, B. E. (1994). The UCLA multidimensional condom attitudes scale: Documenting the complex determinants of condom use in college students. Health Psychology, 13, 224-237.

Herald, E. S., & McNamee, J. E. (1982). An explanatory model of contraceptive use among young women. The Journal of Sex Research, 18, 289-304.

Ickes, W., & Simpson, J. A. (1997). Managing empathic accuracy in close relationships. In W. Ickes (Ed.), Empathic accuracy (pp. 218-250). New York: The Guilford Press.

Ickes, W., Stinson, L., Bissonnette, V., & Garcia, S. (1990). Naturalistic social cognition: Empathic accuracy in mixed-sex dyads. Journal of Personality and Social Psychology, 59, 730-742.

Jacobson, N. S. (1989). The politics of intimacy. Behaviour Therapist, 12, 29-32.

Jones, W. H., & Burdette, M. P. (1994). Betrayal in relationships. In A. L. Weber & J. H. Harvey (Eds.), Perspectives on close relationships (pp. 243-262). Boston: Allyn & Bacon.

Kalichman, S. C., Rompa, D. C., Coley, B. (1996). Experimental component analysis of a behavioral HIV-AIDS prevention intervention for inner-city women. Journal of Consulting and Clinical Psychology, 64, 687-693.

Kendon, A. (1967). Some functions of gaze direction in social interaction. Acta Psychologica, 26, 22-63.

Kipke, M. D., Palmer, R. E, LaFrance, S., & O'Connor, S. (1997). Homeless youths' descriptions of their parents' child-rearing practices. Youth and Society, 28, 415-431.

Knoke, D., & Burke, P. J. (1980). Log-linear models. London: Sage Publications.

Liebman, J., & Mulia, N. (1993). An office-based AIDS prevention program for high risk drug users. Drugs and Society, 7, 205-223.

Lust, J. A., Celuch, K. G., & Showers, L. S. (1993). A note on issues concerning the measurement of self-efficacy. Journal of Applied Social Psychology, 23, 1426-1434.

Margolin, G., Burman, B., & John, R. S. (1989). Home observations of marfled couples reenacting naturalistic conflicts. Behavioral Assessment, 11, 101-118.

Maticka-Tyndale, E. (1991). Sexual scripts and AIDS prevention: Variations in adherence to safer-sex guidelines by heterosexual adolescents. The Journal of Sex Research, 28, 45-66.

Mays, V. M., & Cochran, S. D. (1988). Issues in the perception of AIDS risk and risk reduction activities by Black and Hispanic/Latina women. American Psychologist, 43, 949-957.

Metts, S. (1994). Relational transgressions. In W. R. Cupach & B. H. Spitzberg (Eds.), The dark side of interpersonal communication (pp. 217-239). Hillsdale, NJ: Lawrence Erlbaum.

Metts, S., & Fitzpartick, M. A. (1992). Thinking about safer sex: The risky business of "know your partner" advice. In T. Edgar, M. A. Fitzpatrick, & V. S. Freimuth (Eds.), AIDS, a communication perspective. Hillsdale, NJ: Lawrence Erlbaum Associates.

Miller, L. C., Bettencourt, B. A., DeBro, S. C., & Hoffman, V. (1993). Negotiating safer sex: Interpersonal dynamics. In J. B. Pryor & G. D. Reeder (Eds.), The social psychology of HIV infection (pp. 85-119). Hillsdale, NJ: Lawrence Erlbaum Associates.

Napier, A. Y. (1978). The rejection-intrusion pattern: A central family dynamic. Journal of Marriage and Family Counseling, 4, 5-12.

Noller, P., Feeney, J. A., Bonnell, D., & Callan, V. J. (1994). A longitudinal study of conflict in early marriage. Journal of Social and Personal Relationships, 11, 233-252.

Patterson, G. R. (1982). Coercive family process: A social learning approach. Eugene, OR: Castalia.

Phillips, W., Baron-Cohen, S., & Rutter, M. (1992). The role of eye-contact in goal-detection: Evidence from normal toddlers and children with autism or mental handicap. Development and Psychopathology, 4, 375-384.

Pilkington, C. J. (1994). Is safer sex necessary with a "safe" partner? Condom use and romantic feelings. The Journal of Sex Research, 31, 203-210.

Rands, M., Levinger, G., & Mellinger, G. (1981). Patterns of conflict resolution and marital satisfaction, Journal of Family Issues, 2, 297-321.

Reiss, I. L. (1967). The social context of premarital sexual permissiveness. New York: Holt, Rinehart, & Winston.

Revenstorf, D., Hahlweg, K., Schindler, L., & Vogel, B. (1984). Interaction analysis of marital conflict. In K. Hahlweg & N. S. Jacobson (Eds.), Marital interaction: Analysis and modification (pp. 159-181). New York: Guilford.

Rhodes, F., & Malotte, C. (1996). Using stages of change to assess intervention readiness and outcome in modifying drug-related and sexual HIV risk behaviors of IDUs and crack users. Drugs and Society, 9, 109-136.

Rutter, D. (1984). Looking and seeing: The role of visual communication in social interaction. Chichester, UK: John Wiley and Sons.

Sabourin, S., Laporte, L., & Wright, J. (1990). Problem solving self-appraisal and coping efforts in distressed and nondistressed couples. Journal of Marital and Family Therapy, 16, 89-97.

Sillars, A. L. (1986). Interpersonal perception in relationships. In W. Ickes (Ed.), Compatible and incompatible relationships (pp. 277-305). New York: Springer Verlag.

Sillars, A. L., & Weisberg, J. (1987). Conflict as a social skill. In M. E. Roloff & G. R. Miller (Eds.), Interpersonal processes: New directions in communication research (pp. 140-171). Newbury Park, CA: Sage.

Snyder, M., & Simpson, J. A. (1987). Orientations toward romantic relationships. In D. Perlman & S. Duck (Eds.), Intimate relationships: Development, dynamics and deterioration (pp. 45-62). Newbury Park, CA: Sage.

Sprecher, S., & McKinney, K. (1993). Sexuality. In C. Hendrick & S. S. Hendrick (Series Eds.), Sage series on close relationships. Newbury Park, CA: Sage.

Watzlawick, P., Beavin, J. H., & Jackson, D. D. (1967). Pragmatics of human communication: A study of interactional patterns, pathologies, and paradoxes. New York: Norton.

Wile, D. B. (1981). Couples therapy: A non-traditional approach. New York: Wiley.

Wilmot, W. W., & Baxter, L. A. (1983). Relationship definitions and episodic interaction. Western Journal of Speech Communication, 47, 205-217.

Winter, L., & Goldy, A. S. (1993). Effects of prebehavioral cognitive work on adolescents' acceptance of condoms. Health Psychology, 15, 308-312.

Manuscript accepted June 18, 1998

APPENDIX

ACTUAL SAMPLE STATEMENTS FROM THE SAFER-SEX CONVERSATIONS, LABELED BY CATEGORY

Category          Sample Statements
Demand            * You should tell me, I insist, please, why
                      (you want me to use a condom)?
                      Oh no, I can't stand it if you keep silent
                      like this, you should tell me, really.
                  * I want you to tell me about that girlfriend
                      you had two years ago, you know, how the sex
                      was, you know like nice and safe and so,
                      you know like everything.
Withdrawal        * I don't belief that AIDS stuff is that bad,
                      I haven't heard anyone I know considering
                      safer sex.
                  * Nice place, this room, although they should
                      consider redecorating the walls.
Problem Solving   * I don't know, but maybe for our own safety it
                      isn't really such a bad idea, I mean it is
                      just a little talk for our own safety, if
                      you understand what I mean.
                  * They sometimes say it is a matter of
                      responsibility, you know, if you love
                      someone, you want the best for him, you want
                      it to be safe for him.
Defensiveness     * Don't be ridiculous, I really can't see what
                      I have to do with that (safer sex).
                  * It is that friend of yours who told you to
                      mistrust me, isn't it?
Negative Affect   * If you want an AIDS test, fine, get you one,
                      take your mother with you.
                  * This is a trial, but you'll get nowhere,
                      my dear, nowhere.
Positive Affect   * I love you, I want it to be safe for you.
                  * You are so beautiful the way you sit there,
                      and then talking about safer sex, how lucky
                      am I to know you, you are really beautiful.