10.1177/0146167202250920 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN Cross, Morris / RELATIONAL SELF-CONSTRUAL

ARTICLE

Getting to Know You: The Relational Self-Construal, Relational Cognition, and Well-Being Susan E. Cross Michael L. Morris Iowa State University necessary for self-fulfillment and satisfaction. Consequently, thoughts about relationships and relationship partners will be less important or frequent for these individuals. Likewise, consequences of close relationships for well-being may differ depending on the selfconstrual. The research reported here focuses on the relational self-construal, relational cognition, and wellbeing in the context of a new roommate relationship.

Individuals with a highly relational self-construal define the self in terms of their close relationships with others. Consequently, they seek to nurture and develop new relationships. These studies examine individual differences in the self-construal in the context of a new roommate relationship, with a focus on cognitive aspects of relationship development. Study 1 revealed that persons with a highly relational self-construal were better able than others to predict a new roommate’s values and beliefs. Study 2 showed that highly relational individuals tended to think optimistically about a new roommate’s feelings about the relationship. The relational self-construal was more strongly related to these measures of relationship cognition in distant relationships than in very close relationships. Participants’ self-construals and their perceptions of the closeness of the roommate relationship interacted in predicting well-being, revealing an unexpected negative relation between closeness and well-being for participants with a low relational self-construal.

THE RELATIONAL-INTERDEPENDENT SELF-CONSTRUAL

In most of the history of social psychological research on the self, psychologists assumed a very individualistic self-structure, in which the self was defined by abstract characteristics, unique abilities, and personal preferences and was represented as separate from others and social contexts (Markus & Kitayama, 1991). The assumption of the universality of this independent representation of the self was challenged by cross-cultural researchers (Markus & Kitayama, 1991) and by gender researchers (Cross & Madson, 1997; Surrey, 1991). For many North American women (and some men), the self is defined, at least in part, in terms of close relationships with other people. In this type of self-representation, termed the relational-interdependent self-construal (Cross, Bacon, & Morris, 2000; for simplicity, we use rela-

Keywords: relational self-construal; social cognition; roommates I count myself in nothing else so happy as . . . remembering my good friends. —Shakespeare, Richard II, II, iii

As the quote from Richard II indicates, many people find great happiness in thinking about close friends. Others spend relatively little time mulling their relationships. Individual differences in the self-construal may be one contributor to such divergent orientations to relationships. For some people, the self is defined in terms of close relationships, and the development and maintenance of close relationships is a central goal. Thus, these individuals will attend to and think about important relationships. Other people define themselves as independent of relationships and view relationships as much less

Authors’ Note: We thank Dan Russell, Carolyn Cutrona, and Jonathan S. Gore for their comments on this article and the members of the Social Psychology seminar for their comments on this research. Carrie Driscoll, Anne M. Koenig, Tangela Duncan, Christina Schwab, Carolyn J. Myers, Violetta T. Fung, and Marietta Dyer assisted in the data collection. Correspondence concerning this article should be addressed to Susan E. Cross, Department of Psychology, W112 Lagomarcino Hall, Iowa State University, Ames, IA 50011; e-mail: [email protected]. PSPB, Vol. 29 No. 4, April 2003 512-523 DOI: 10.1177/0146167202250920 © 2003 by the Society for Personality and Social Psychology, Inc.

512

Cross, Morris / RELATIONAL SELF-CONSTRUAL tional self-construal), the self-space includes representations of oneself in important relationships. For individuals with a highly relational self-construal, relationships are essential for self-definition, self-expression, and selfenhancement; thus, developing and nurturing close relationships are primary goals. Individuals who have defined themselves in terms of close relationships will therefore tend to think and behave so as to develop, enhance, and maintain harmonious and close relationships with important others (Cross et al., 2000). Until recently, most of the research on the role of the self in psychological processes focused only on the independent form of self-representation. Much has been learned about role of the self in cognition, motivation, self-regulation, and social behavior using this model of the self (Markus & Wurf, 1987; for a review, see Baumeister, 1998). Given a very different self-structure, in which the relational dimension dominates, the psychological processes associated with the self may vary as well (Brewer & Gardner, 1996). A series of studies focused on the implications of the relational selfconstrual for social cognition revealed that individuals who have defined themselves primarily in terms of close relationships were more likely than others to have dense cognitive networks for relational constructs (Cross, Morris, & Gore, 2002). The highly relational persons were more likely than others to remember relational information about others and to organize information about others in terms of relationships. High relationals described a friend and themselves more similarly than did lows (Cross, Morris, & Gore, 2002). Whereas previous studies focused on rather abstract cognitive tasks that were not necessarily relationship-specific, the studies described here focus on relationship cognition in the context of a specific new relationship with a roommate. In addition, these studies use this roommate relationship to examine the association between the relational self-construal and well-being. THE RELATIONAL SELF-CONSTRUAL AND RELATIONSHIP COGNITION

The starting point for this research is the assumption that individuals who have developed a highly relational self-construal will think and behave in ways that nurture and enhance important or close relationships. One way to nurture a new relationship is to self-disclose and to attend to one’s partner’s self-disclosures (Harvey & Omarzu, 1997; Reis & Shaver, 1988). When sensitive or personal information is shared within a dyad, it communicates trust, liking, and openness to increased closeness in the relationship (Collins & Miller, 1994). The listener’s response to this disclosure also influences the development of the relationship. For example, when Thelma describes how she misses her family to her new

513

roommate, Louise, Louise’s communication of understanding and caring signals her interest in a closer relationship (Reis & Patrick, 1996). Previous research using unacquainted pairs of participants who were asked to engage in a get-acquainted task in the lab revealed that participants with highly relational self-construals were more likely to self-disclose than were those with low relational self-construals and were perceived by their partners as more responsive than lows (Cross et al., 2000). Although this laboratory study supported the hypothesized influence of the relational self-construal in the development of relationships, it represented only a very temporary, uncommitted relationship. Commitment to an ongoing relationship must be reflected in more than just warm and reassuring responses to another’s disclosures. Imagine that Louise repeatedly failed to relay messages to homesick Thelma that her parents had called. Even if she had responded with understanding and caring at the time of Thelma’s disclosure, her failure to either remember or demonstrate concern for Thelma’s feelings could undermine the relationship. Therefore, if a person wants to build a harmonious relationship, it is important not only to respond sensitively to one’s partner’s disclosures but also to remember what has been disclosed and to take this information into account in future interactions. This knowledge allows the individual to predict the partner’s behavior and to prevent conflict in the relationship. A person may steer away from sensitive topics to avoid disagreements or change his or her behavior to maintain harmony in the relationship. Because relationship building is important to individuals with a highly relational self-construal, we hypothesize that they would be more likely than others to attend to and remember important information revealed by a new roommate. Thus, Study 1 goes beyond the findings that the relational self-construal is associated with self-disclosure and responsiveness to examine a potential cognitive consequence of these behaviors: memory for information that is likely to be shared by a new roommate. The tendency to pay close attention to and remember the disclosures of a relationship partner, particularly a new roommate, may depend on both the nature of the self-construal and the closeness of the relationship. In very close relationships, almost everyone will attend to roommates’ disclosures of their values and beliefs because doing so supports a valued relationship. But in more distant relationships, this behavior may be a function of the self-construal. For individuals with a highly relational self-construal, a cold or difficult relationship with one’s roommate may be seen as a failure (given that the roommate is not extremely odd or deviant in some way). Even if a relationship is not very close, the highly

514

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

relational person may pay close attention to the beliefs and values of his or her roommate to prevent uncomfortable situations, to anticipate the roommate’s behavior, or in an attempt to improve the relationship. In contrast, low relationals will not be as motivated to work on a difficult or distant relationship by attending to and remembering the partner’s disclosures. Therefore, we hypothesize that the closeness of the relationship moderates the association between judges’ self-construals and the extent to which they are able to predict their roommate’s responses to a variety of issues. Study 2 addresses motivational aspects of relational cognition. One way to nurture a developing relationship is to think positively about it and about the other person’s commitment to it. Relatively positive estimates of one’s partner’s feelings about the relationship reveals optimism and may contribute to increased effort to make the relationship work. This optimistic view can buffer the individual from disappointments that arise in the relationship and promote relationship-serving attributions for the partner’s behavior (Murray, Holmes, & Griffin, 1996). Because high relationals are presumed to be more concerned about the success of this relationship than lows, we hypothesize that they are more likely to be optimistic in their estimates of their roommate’s feelings about the relationship. THE IMPORTANCE OF ROOMMATE RELATIONSHIPS FOR WELL-BEING

Relationships are one of the most important aspects of individuals’ lives and a significant contributor to wellbeing (Baumeister & Leary, 1995; Myers & Diener, 1995). Why focus on relationships with roommates rather than the oft-studied romantic relationship? Roommate relationships, especially among college students, represent a unique form of potentially close or interdependent relationships. Unlike other types of nonromantic friendships, roommates share the same space and must coordinate their activities to live together peacefully. Because the roommates’ behaviors impact each other strongly, they may be described as very interdependent, in Kelley’s (1979) terms, even if they are not emotionally close. In addition, roommate relationships are sometimes involuntary and relatively difficult to exit. Students in residence halls are often assigned to live with strangers; others may choose to live with a stranger for the sake of convenience or financial savings. When other friendships go sour or become dissatisfying, they are usually fairly easy to end—the partners can simply quit seeing each other. In contrast, students are often forced by economic or other considerations to persist in a dissatisfying roommate situation at least until a semester break. Even when students

do not expect to be emotionally close to a new roommate, they may seek to have positive, harmonious interactions with the person so that the situation is as livable as possible. Research on students’ views of their roommates shows that good relationships are associated with physical health and psychological well-being (Joiner, Vohs, & Schmidt, 2000; Snydersmith & Cacioppo, 1992; Waldo, 1984). Because the roommate relationship is generally unavoidable and somewhat difficult to leave, we assume it will be important to highly relational individuals to develop a good relationship (even if it is not intimate) with their roommate. This, in turn, will be perceived as progress toward an important goal, which is related to well-being (Diener, Suh, Lucas, & Smith, 1999). In addition, a good relationship can affirm these persons’ relational self-views. A close roommate relationship may permit the expression of the high relational person’s authentic self, which also has been associated with wellbeing (Sheldon, Ryan, Rawsthorne, & Ilardi, 1997). Consequently, for participants with highly relational selfconstruals, we expected to find a strong positive association between their ratings of the closeness of their roommate relationship and well-being. The consequences of roommate relationships for students with a low relational self-construal are more difficult to predict. As mentioned above, research generally reveals a positive association between close relationships, including roommate relationships, and wellbeing. Individuals with a low relational self-construal, however, may prefer roommate relationships that are somewhat distant and that make few demands. In a roommate situation, in which a degree of interdependence may be obligatory, a very close relationship may be experienced as restricting one’s independence and lead to reduced well-being. Nevertheless, given that research overwhelmingly shows a positive association between relationship closeness and well-being, we expected to find a small positive relation between the closeness of this relationship and well-being for the low relationals and a stronger association between closeness and wellbeing for high relationals. HYPOTHESES

In summary, these studies examined the association of the relational self-construal with relationship-enhancing cognitions and well-being in the context of new roommate relationships. Specifically, we propose the following hypotheses: Hypothesis 1: Participants’ scores on a measure of the relational self-construal will be positively associated with the extent to which they can predict their roommate’s values and beliefs (Study 1).

Cross, Morris / RELATIONAL SELF-CONSTRUAL Hypothesis 2: The association between the judges’ relational self-construal scores and their prediction of their roommate’s responses will be moderated by the judge’s closeness to the roommate. We expect to find little relation between the self-construal and agreement between the roommates in very close relationships and a strong relation in less close relationships. Hypothesis 3: In Study 2, the participants’ relational selfconstrual scale scores will be positively associated with their estimates of their roommate’s feelings about the relationship. Hypothesis 4: The participants’ relational self-construal scale scores will interact with their evaluations of the closeness of the relationship to predict well-being. We expect to find a positive association between closeness and wellbeing for both high and low relationals, but this association will be stronger for highs than for lows. Study 1 uses a cross-sectional design to examine this hypothesis, whereas Study 2 examines change in well-being over a 1month period. STUDY 1

We expect that individuals with a highly relational self-construal will strive to enhance and develop a relationship with a new roommate by attending to and remembering their roommate’s disclosures. Most other research on agreement among partners has focused on personality (Bernieri, Zuckerman, Koestner, & Rosenthal, 1994; Funder & Colvin, 1988). Judges can fairly accurately infer many aspects of a target’s personality after exposure to only a brief sample of the person’s behavior (Ambady & Rosenthal, 1992; Funder & Colvin, 1988). Because we were interested in the processes that facilitate the development of a close relationship, such as self-disclosure, we focused on aspects of the target that are most likely to be revealed in conversation but are not as readily available through simple observation. Research on roommate relationships reveals that the initial stages of the relationship are marked by small talk and the disclosure of relatively superficial information about the self (Hays, 1985). As the relationship deepens, the partners are more likely to share information about their values and beliefs (Andersen & Ross, 1984; Berg, 1984). We hypothesized that individuals who defined themselves in terms of relationships would have sought out and remembered their roommate’s statements about their beliefs and values and so would be fairly accurate when predicting their roommate’s responses on this dimension.1 This study also examined the association between the participants’ self-construals, closeness to their roommate, and well-being. We hypothesized that among high relationals, closeness would be positively related to well-

515

being but that this relation would be weaker among low relationals. METHOD

Participants and Procedure Research participants were asked to sign up for the study if they were living with someone for the first time that they did not know well before the semester began. Only same-sex roommates were allowed to participate. The first phase of the study was conducted in university classrooms in groups of 2 to 10. Participants provided informed consent and completed a questionnaire packet, which included the measures described below. A total of 176 introductory psychology students (129 women, 47 men) completed the questionnaire in exchange for extra course credit. Twenty participants (12 women and 8 men) who indicated they knew their roommates well before the beginning of the semester were dropped from all subsequent analyses, resulting in a sample of 156 eligible participants. To reduce the possible confounding relations between culture and the selfconstrual, we usually eliminate noncitizens from analyses. In this study, all of the 156 eligible participants indicated that they were U.S. citizens and English was their first language. Of these 156, 148 (95%) gave us permission to contact their roommates (half of those who refused permission were women). Comparisons of the participants who gave permission to contact their roommates and those who did not revealed no differences in any of the measures described below. After completing these materials, participants were thanked and debriefed. In the remainder of the article, the initial volunteers who signed up for the study and completed the longer questionnaire will be referred to as the “participants” and their roommates will be referred to as the “roommates.” To make it more likely that the roommates would complete and return the questionnaire, they were sent a much shorter questionnaire than that given to the participants. They were asked to complete the questionnaire alone, not to discuss it with their roommate, and to return it within 10 days. In exchange for completing the questionnaire, their names were entered into one of three drawings for $50. The roommates also were asked to write down contact information in case they won the lottery and this sheet was separated from responses immediately on receipt by the researchers. Responses were returned in prepaid envelopes directly to the researchers. A total of 95 roommates (81 women, 14 men) of the 148 eligible roommates returned completed questionnaires for a return rate of 64%. To determine if there was a difference in the relationships of the roommate responders and nonresponders,

516

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

we compared the responses of the participants whose roommates returned the questionnaire to those whose roommates did not on the measures described below (only for the participants who gave permission to contact their roommates). These analyses revealed no significant differences between these two groups. Thus, roommates in relationships described by the participants as close were no more or less likely to respond to the questionnaire than were other roommates. Materials The packet of materials completed by the participants in the first phase of the study included all the measures described below. The shorter instrument sent to the roommates included the measure of the self-construal, the items assessing values and beliefs, and the background items. Both the participants and roommates completed additional unrelated measures that were not included in these analyses. Relational-Interdependent Self-Construal (RISC). The RISC measure (Cross et al., 2000) was used to measure differences in the self. Sample items are as follows: “My close relationships are an important reflection of who I am” and “Overall my close relationships have very little to do with how I feel about myself” (reverse-scored). This scale correlates moderately with other measures of communal orientation, interdependence, and empathy but does not correlate with measures of independence. Previous research generally reveals positive associations with measures of social support and self-disclosure but little or no association with measures of self-esteem or well-being (Cross et al., 2000). Coefficient alphas usually range from .85 to .90; test-retest reliability over a 2month period was .76 (Cross et al., 2000). Responses to the 11-item scale were made on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Cronbach’s α was .88. Relationship closeness. Depth of closeness to the roommate was assessed using items from the Quality of Relationships Inventory (QRI) (Pierce, Sarason, & Sarason, 1991). The QRI contains three subscales, only one of which was used in the current study. Sample items from the depth of closeness subscale include “How significant is this relationship in your life?” and “How much would you miss this person if the two of you could not see or talk with each other for a month?” Responses to the eight items were made using a 5-point Likert-type scale ranging from 1 (not at all) to 5 (quite a lot). Cronbach’s α was .93. Values and beliefs items. The participants were given a list of 15 items and were asked to rate how important each item was to their roommate. This list included items such as “making all As in classes” and “keeping in

touch with his or her family.” Responses were made on a 5-point Likert-type scale ranging from 1 (not at all important) to 5 (extremely important). The roommates were given the same lists (e.g., reframed as “keeping in touch with your family”) and were asked to indicate how important each value or belief statement was to them using the same 5-point Likert-type scale used by the participants. Well-being. The Satisfaction With Life Scale (SWLS) (Diener, Emmons, Larsen, & Griffin, 1985) was used to assess well-being. This five-item measure is widely used and has good psychometric properties. It correlates strongly with other measures of well-being. A sample item is, “In most ways my life is close to my ideal.” Responses were made on a 5-point scale (1 = strongly disagree, 5 = strongly agree); the coefficient α was .84. Background information. The questionnaire packets included items requesting information about gender, age, ethnicity, citizenship status, how long they had lived with their roommate, and how well they knew their roommate before living together. Sex Differences As in other studies, there was a significant sex difference in the participants’ RISC scale scores, Mmen = 5.03, SD = .86, Mwomen = 5.68, SD = .73, t(155) = –4.62, p < .001. There was also a marginal gender difference in the participants’ reports of their depth of closeness to their roommates, Mmen = 2.50, SD = .90, Mwomen = 2.83, SD = .98, t(155) = –1.80, p < .08. Among the roommate pairs, there was also a gender difference in the ideographic agreement scores, Mmen = –.15, SD = .34, Mwomen = .03, SD = .30, t(94) = 2.24, p < .03. There were no interactions with gender in any of these analyses; thus, this variable will not be mentioned further. RESULTS

The Self-Construal and Predictions of Roommates’ Values and Beliefs To assess agreement between the roommates’ selfreports and the participants’ predictions, we used an approach described by Bernieri et al. (1994). They recommend using a type of profile analysis in which correlation coefficients are computed for each target and rater down the set of items. This correlation coefficient reflects similarities in item-to-item profile shapes and is not affected by the problems of elevation and differential elevation that afflict some other accuracy scores. This metric is affected, however, by what Bernieri and his colleagues term “implicit profile accuracy,” which is the extent to which the target has a fairly typical personality profile and the extent to which the judge is aware of this typical profile. For example, if most college students

Cross, Morris / RELATIONAL SELF-CONSTRUAL describe themselves as valuing hard work and close family relationships, and the judge is aware of this prototype, then this knowledge would enhance the judge’s rating of any particular college student. To account for this, the judges’ ratings are correlated with the means of the items generated by the sample as a whole to create an implicit profile accuracy score. The variance attributable to implicit profile accuracy is then removed from the raw accuracy scores to create a measure of ideographic accuracy. This value represents “the degree to which peer judges accurately predicted how their roommates reported their own constellation of trait relationships as they deviated from the most typical (aggregate) values of these traits” (Bernieri et al., 1994, p. 373). Thus, in these data, we first created raw accuracy scores by computing intraclass correlations between the participants’ predictions and the roommates’ selfratings. Next, we created the implicit profile accuracy measure by correlating the participants’ ratings with the mean of each item generated by the sample as a whole (separately for men and women). These correlations were transformed using the r-to-z transformation, then the raw accuracy correlations were regressed on the implicit profile accuracy scores to create the ideographic agreement residual scores. As expected, there was a significant relation between the participants’ RISC scale scores and this ideographic agreement residual, r(95) = .21, p < .05. We hypothesized that the participants’ ratings of the depth of closeness of the relationship moderated the association between their RISC scale scores and the ideographic agreement scores. The ideographic agreement residual scores were regressed on the RISC scale, the depth measure, and the RISC Scale × Depth interaction term. The correlations among the measures used in these analyses are presented in Table 1. The predictor variables were centered prior to the analyses. The results of these regressions are presented in Table 2.2 There was a significant main effect for the RISC scale, indicating that when closeness was controlled, the participants’ self-construal scores continued to predict the agreement score. There was also a significant interaction between the RISC scale and the depth of closeness measure. Examinations of the simple slopes revealed that in relationships that were low in closeness (1 SD below the mean), there was a strong relation between the RISC scale and agreement, β = .44, p < .01 (see Figure 1). For relationships that were high in closeness (1 SD above the mean), there was no relation between the RISC scale and ideographic agreement, β = –.03, ns. Thus, consistent with our predictions, the relational self-construal was a better predictor of the ideographic agreement measure in more distant relationships than in close relationships.

TABLE 1:

517

Descriptive Statistics and Correlations Among Variables Used in Study 1

Participant Variable 1. RISC scale 2. Depth of closeness 3. Ideographic agreement 4. Life satisfaction M SD

1

2

— .14 .21 .06 5.52 .81

— .04 .05 2.75 .97

3

— .12 .00 .24

4

— 3.55 .83

NOTE: N = 156, except for the ideographic agreement score, for which N = 95. RISC = Relational-Interdependent Self-Construal.

TABLE 2:

Predicting Ideographic Agreement for Values and Beliefs, With Participant Self-Construal Scores and Closeness, Study 1

Variable Intercept RISC scale Depth of closeness RISC × Depth

B

SE

.00 .09 .02 –.11

.03 .04 .04 .05

R

2

.09* .20* .06 –.22*

NOTE: N = 94 pairs. RISC = Relational-Interdependent Self-Construal. *p ≤ .05.

The Relational Self-Construal, Closeness of the Relationship, and Well-Being We expected that the self-construal scores would interact with the depth of closeness measure to predict participants’ life satisfaction. Life satisfaction scores were regressed on the participants’ RISC scale scores, depth, and the RISC Scale × Depth interaction term. The predictor variables were centered prior to the analyses. As shown in Table 3, there was a significant interaction between the participants’ RISC scale scores and their closeness scores in predicting life satisfaction (see Figure 2). Analyses of the simple slopes showed that among the participants with high RISC scale scores (1 SD above the mean), depth was positively related to life satisfaction, β = .25, p = .02. Among low relationals (1 SD below the mean), depth scores were negatively related to life satisfaction, β = –.24, p = .06. DISCUSSION

We predicted that when beginning to live with a new roommate, individuals with a highly relational selfconstrual will seek ways to develop and enhance the relationship. One way to do this is to pay attention to those things that are important to the other person. As expected, individuals who scored high on the RISC scale were more likely than low scorers to accurately predict their roommates’ self-reported values and beliefs. Furthermore, the relational self-construal was not strongly

518

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN 3.9

0.2

.25*

3.7

.44** Life Satisfaction

Agreement - Values and Beliefs

3.8

0.15

0.1

0.05

-.03 0

Low RISC

High RISC

-0.05

3.6 3.5

-.24

3.4

Low RISC High RISC

3.3 3.2 3.1

Low Participant Depth -0.1

3

High Participant Depth

Low depth

High depth

Figure 2

The interaction between participants’ RISC scale scores and depth of closeness predicting their life satisfaction, Study 1. NOTE: RISC = Relational-Interdependent Self-Construal.

-0.15

Participant RISC Scale Scores Figure 1

The interaction of the participants’ RISC scale scores and their depth of closeness to the roommate predicting ideographic agreement for the roommate’s values and beliefs, Study 1. NOTE: RISC = Relational-Interdependent Self-Construal.

TABLE 3:

Predicting Participants’ Life Satisfaction With Participant Self-Construal Scores and Depth of Closeness, Study 1

Variable Intercept RISC scale Depth of closeness RISC × Depth

b

SE

3.52 .09 .01 .26

.07 .08 .07 .09

R

2

.06* .09 .01 .23**

NOTE: N = 156. RISC = Relational-Interdependent Self-Construal. *p ≤ .05. **p < .01.

related to agreement in very close relationships but was a strong predictor of agreement in more distant relationships. In these distant relationships, participants with a highly relational self-construal may be seeking out and remembering information about their roommate to work toward more closeness or to avoid conflict. An alternative explanation for this finding is that the highly relational participants were more similar to their roommates than were the lows; as a result, they would appear to be accurate when in fact they were simply projecting their own views onto their roommate. Although most friends (and presumably roommates) select each other based on perceived similarity (Byrne, 1971), these roommates were largely strangers before living together. Therefore, they were less likely than self-selected roommates to share common values and beliefs, making the similarity explanation for these findings unlikely. Do the high relationals simply assume that their roommates are more similar to them than do low relationals, as some findings suggest (Cross, Gore, & Morris, 2002)? If so,

that would suggest that the high relationals would be less accurate than the low relationals, if in fact there is little similarity between them and their roommates. These results showed that the relational self-construal measure moderated the association between closeness scores and well-being. As expected, perception of the depth of their relationship was positively associated with life satisfaction among participants with a highly relational self-construal. We did not, however, expect the negative association between closeness and well-being found for the low relationals. This result contradicts most research findings on close relationships and wellbeing. The cross-sectional design of this study unfortunately leaves the question of the direction of causality unclear. Thus, Study 2 examines the replicability of these findings and whether they persisted over the course of 1 month. Study 2 also used an alternative measure of relational cognition. STUDY 2

Thinking about a relationship can be colored by one’s hopes or wishes for the relationship. In Study 2, we assume that individuals with a highly relational selfconstrual will tend to have positive illusions of their roommate’s feelings about the relationship because they desire a close, or at least harmonious, relationship with their roommate. Such optimistic perspectives may promote relationship-enhancing behaviors and attributions and enhance commitment to the relationship (Murray et al., 1996). Thus, this study assesses participants’ estimates of their roommate’s depth of closeness to the relationship and examines the degree to which the relational self-construal is associated with optimistic estimates.

Cross, Morris / RELATIONAL SELF-CONSTRUAL Although research generally indicates that a close relationship should enhance well-being for all participants, the results of Study 1 suggest that closeness may instead reduce well-being for participants with a low relational self-construal. Thus, in this study, we again examine the association between the relational self-construal, closeness of the relationship, and well-being using prospective data. METHOD

Participants and Procedure Students in undergraduate introductory psychology courses participated in this study for extra course credit. The participants were asked to sign up for the study only if they were living with a same-sex roommate for the first time, although it was acceptable if they knew each other before living together. Initially, we hypothesized that the closeness of the relationship before the students began to live together would moderate some of the processes under investigation in this study. Only 13% of the participants indicated that they were close friends before living together, and the moderation hypotheses were not supported by the data. Therefore, this measure was not included in any of these analyses. The first phase of the study was conducted in groups of 2 to 15 people. Participants provided informed consent, completed a questionnaire packet, and if they chose, gave permission to the researchers to contact their roommates. Participants were then thanked and reminded of their second session 1 month later. A total of 281 introductory psychology students completed the first phase of the experiment. Of this group, 227 were women, 53 were men, and 1 participant did not indicate his or her sex. Twenty-one people were excluded for living with their roommates for more than 4 weeks prior to the beginning of the fall semester and an additional 10 were excluded for living with an oppositesex roommate. To reduce the possible confounding relations between culture and the self-construal, 6 noncitizens were excluded from the analyses. This brought the Time 1 total number of participants to 244 (200 women, 44 men). Of this group, 91% gave permission to contact their roommates (the rate was 91% for women and 89% for men) and there were no significant differences between those who did and did not give permission on any Time 1 variables. A shorter version of the questionnaire was mailed to the roommates. Directions for completing the questionnaire, the incentives provided, and the administration of the questionnaire were the same as in Study 1. Roommates were assured that their responses were completely confidential and would not be shared with anyone. Following receipt of the questionnaires, a debriefing state-

519

ment was sent back to the roommates by mail. A total of 142 (64%) roommates of the final group of participants returned completed questionnaires. Of this group, 123 were women and 19 were men. Analysis of the responses of the participants whose roommates returned the questionnaire with those who did not (only for those who gave permission to contact their roommates) revealed only one significant difference: Participants whose roommates responded to the questionnaire scored higher on the RISC scale (M = 5.53, SD = .67) than did those whose roommates did not respond (M = 5.31, SD = .84), t(219) = –2.09, p < .04. There were no differences between these groups in the variables assessing the quality of the relationships with their roommates. The initial participants returned after 1 month to complete the life satisfaction measure. They were thoroughly debriefed about the purpose of the study. A total of 209 people (171 women, 38 men) from the first phase returned 1 month later, for a return rate of 86%. There were no significant differences on any of the Time 1 measures between those who completed both waves of data collection and those who did not. Materials The participants completed the following measures at Time 1: the relational-interdependent self-construal scale, depth of closeness, life satisfaction, and background information. The participants also rated how they thought their roommate would complete the depth of closeness measure. Roommates completed the relational-interdependent self-construal scale, the depth of closeness measure, and the background items. All of the respondents completed additional measures for another study that will not be discussed further. The measures that were described for Study 1 used the identical response scales in Study 2; thus, they are not described further here. Coefficient alphas for all of the measures are presented in Table 4 with the means and standard deviations. Background information. The questionnaire packets included items requesting information about gender, age, ethnicity, citizenship status, how long they had lived with their roommate, and how well they knew their roommate before living together. Gender Differences Women scored higher than men on the RISC scale, M women = 5.57, SD = .68, M men = 5.01, SD = .74, t(240) = –4.82, p < .001. Women also rated themselves higher than did men on the other variables in this study (participants’ ratings of their depth in the relationship, their perceptions of their roommate’s depth, Time 1 and Time 2 life satisfaction), all ts > –2.3, ps < .02 (two-tailed). Significant interactions with gender found in the

520

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

TABLE 4:

Descriptive Statistics and Correlations Among Variables Used in Study 2 1

Participant variable 1. RISC scale — 2. Depth of closeness .17 3. Estimation of roommates’ depth .23 4. T1 life satisfaction .20 5. T2 life satisfaction .21 Roommate variable 6. Depth of closeness .08 M 5.46 SD .67 α .85

2

3

4

5

6

— .02 –.05

— .79

—

.64 3.05 1.02 .93

.66 3.08 .95 .94

.14 3.70 .79 .86

.13 3.79 .81 .88

Predicting Participants’ Estimates of Their Roommate’s Depth, Study 2

Variable Step 1 Participant depth Roommate’s depth Step 2 RISC scale scores Step 3 RISC × Participant Depth

— .18 .10 .09

TABLE 5:

— 3.27 1.05 .94

NOTE: RISC = Relational-Interdependent Self-Construal, T1 = Time 1, T2 = Time 2.

remainder of the analyses are described in Notes in the Results section. RESULTS AND DISCUSSION

Predicting Roommate Closeness To test the hypothesis that high relationals would view their roommate’s closeness in the relationship more optimistically than the low relationals, the approach suggested by Kenny and Acitelli (2001) for analyzing dyadic data was used. In data of this sort, the participants’ predictions of the roommate’s depth can be broken down into two components: an accuracy effect (the extent to which the participant’s estimate actually reflects the roommate’s feelings of closeness) and a bias effect (the extent to which the participant projects his or her own feelings of closeness onto the roommate). Kenny and Acitelli recommend the use of multiple regression in such cases, in which the predictor variables are the judges’ ratings of themselves and the targets’ self-ratings and the outcome variable is the judges’ estimates of the target’s rating. The path from the participants’ ratings of their own depth to their prediction of their roommate’s depth represents the bias or projection effect and the path from the roommate’s self-rating of depth to the participant’s estimate represents the accuracy effect. In Step 1 of hierarchical regression analyses, the participants’ estimates of their roommate’s depth were regressed on the participants’ and roommates’ ratings of their own depth. In the next step, the participants’ RISC scale scores were entered into the equation. The predictor variables were centered in these analyses; correlations and descriptive statistics are provided in Table 4. As shown in the first two lines of Table 5, there were significant accuracy and bias effects in these data. Step 2 revealed that participants’ RISC scale scores were posi-

R

2

B

SE

.68 .17

.05 .05

.72*** .77*** .18***

.13

.06

.09*

.01*

–.15

.05

–.11*

.01*

NOTE: N = 142. RISC = Relational-Interdependent Self-Construal. *p ≤ .05. ***p < .001.

tively related to their estimates of their roommate’s depth of closeness, controlling for the accuracy and bias effects. This supports our hypothesis that highly relational participants are more optimistic in their evaluations of their roommate’s depth than are low relationals.3 In the final step, we added the RISC × Participant Depth interaction term to determine whether the moderation effects found in Study 1 replicated with this measure of relationship cognition. This interaction term explained a significant proportion of the variance in the dependent variable (see Table 5). Examination of the simple slopes revealed that in relationships perceived as low in depth by the participants (1 SD below the mean), high RISC participants were more optimistic than low RISC participants, β = .17, p < .001 (see Figure 3). Although the participants who perceived their relationships as high in depth (1 SD above the mean) reported higher estimates than the low depth participants, their RISC scale scores were unrelated to their estimates, β = –.06, ns. Thus, controlling for bias and accuracy effects, when participants described their relationships as relatively distant, the high relationals were more optimistic than the low relationals. Self-Construals, Relationship Closeness, and Participants’ Changes in Life Satisfaction To examine change in life satisfaction, the Time 2 measure of well-being was regressed onto the Time 1 measure in Step 1, then the participants’ RISC scale scores and depth of closeness ratings were added to the equation in Step 2. In Step 3, the RISC Scale × Depth term was included in the analysis. As shown in Table 6, there were no significant main effects of the RISC scale or depth but there was a significant interaction between these two predictors (see Figure 4). Analysis of the simple slopes revealed that among participants with highly relational self-construals, there was not a significant association between their depth scores and change in wellbeing, β = .03, p > .50. Among participants with low rela-

Cross, Morris / RELATIONAL SELF-CONSTRUAL

P's Estimate of R's Depth

TABLE 6:

-.06

4

3.5

Low Participant Depth

3

High Participant Depth

2.5

.17***

2

1.5

Low RISC

High RISC

Participant RISC Scale Scores

Figure 3

The interaction of the participants’ RISC scale scores and their depth of closeness to the roommate predicting the participants’ estimates of their roommate’s depth of closeness, controlling for the roommate’s self-reported depth. NOTE: RISC = Relational-Interdependent Self-Construal.

521

Predicting Participants’ Change in Life Satisfaction Across 1 Month as a Function of the Participants’ RISC Scale Scores and the Depth of the Relationship, Study 2 B

SE

.81

.04

.08 –.04

.05 .04

.10

.04

Variable Step 1 T1 life satisfaction Step 2 RISC Depth Step 3 RISC × Depth

R

2

.63*** .79*** .006 .07 –.05 .01* .10*

NOTE: N = 209. RISC = Relational-Interdependent Self-Construal, T1 = Time 1. *p ≤ .05. ***p < .001.

0.1

GENERAL DISCUSSION

How people think about close relationships and the consequences of close relationships for well-being depends on the definition of the self. As hypothesized, participants with highly relational self-construals were better able to predict their roommate’s values and beliefs than were the low relational participants (Study 1). The highly relational participants were also more optimistic in their predictions of their roommate’s depth of commitment to the relationship than were lows (Study 2). These findings are consistent with earlier research showing that high relationals have more close friends, engage in more self-disclosure, and are perceived as more responsive by interaction partners than are low relationals (Cross et al., 2000). The hypothesized interaction between the relational self-construal, depth, and well-being was not supported; in fact, a surprising pattern for low relationals was revealed. We discuss these findings in more detail below. The Relational Self-Construal and Relationship Cognition These studies examined two different types of relationship cognition: predictions of one’s roommates values and beliefs and estimates of one’s roommate’s depth of closeness. Participants with highly relational selfconstruals were better able to predict their roommate’s

.03

Change in Life Satisfaction

tional self-construals, there was a negative association between their depth scores and change in life satisfaction, β = –.16, p < .02. Contrary to the findings of Study 1, closeness to a roommate was not associated with wellbeing for persons with a highly relational self-construal. Surprisingly, but consistent with the findings of Study 1, closeness to a roommate was related to declines in wellbeing for the participants with a low relational selfconstrual.4

0.05 0 -0.05

Low Depth

High Depth -.16*

-0.1 -0.15 -0.2

Low Participant RISC

-0.25

High Participant RISC

Figure 4

The interaction of participant RISC scale scores and depth of commitment predicting change in life satisfaction over 1 month (Study 2). NOTE: RISC = Relational-Interdependent Self-Construal.

responses in Study 1 and were more optimistic about their roommate’s closeness in Study 2. Moreover, the association between the self-construal and these measures of relationship cognition were stronger in lowdepth relationships than in high-depth relationships. Why would individuals with a highly relational selfconstrual spend time and energy attending to and being optimistic about a relationship that is not close? We contend that because high relationals desire a good relationship with their roommate, they seek out information about the roommate, elicit self-disclosures by the roommate, and remember disclosed information in an attempt to work toward a closer relationship. Their hope for a better relationship also may underlie their estimates of the roommate’s closeness and the resulting optimism may justify further efforts to improve the relationship.

522

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

So how can the participants with highly relational selfconstruals be both accurate (in Study 1) and biased (as expressed by optimism in Study 2)? This seeming contradiction is partly a function of the two quite different tasks and analysis techniques used. The measure of relationship cognition in Study 1—predicting the roommate’s values and beliefs—is a relatively objective task likely to be based on explicit disclosures by the roommate. The task of Study 2—predicting the roommate’s ratings of his or her closeness in the relationship—is a more subjective task that may be influenced by explicit disclosures as well as very indirect, implicit, or nonverbal communication. Estimating another’s closeness is also influenced by the perceiver’s own motivations and hopes for the relationship. As others have shown (Murray, Holmes, Bellavia, Griffin, & Dolderman, 2002; Simpson, Ickes, & Blackstone, 1995), biased, inaccurate evaluations of a relationship partner may sometimes serve relationship goals better than accurate evaluations. Accurate knowledge of a roommate’s values and beliefs can direct specific strategies that contribute to closeness (i.e., selecting activities and conversation topics the roommate enjoys), whereas optimistically biased views of their depth of closeness provide motivation to persist with the relationship. The Relational Self-Construal, Depth, and Well-Being Finally, we hypothesized that for highly relational persons, the success of roommate relationships would represent an important goal; consequently, the closeness of that relationship would be associated with well-being. The cross-sectional findings of Study 1 supported this hypothesis but the prospective data of Study 2 revealed no association between depth and well-being for the high relationals. To investigate the possibility that this finding might be artifactual, we examined the variance in the life satisfaction residual score (i.e., regressing Time 2 life satisfaction onto Time 1 life satisfaction) but found no evidence of significant differences between high versus low relationals. Given that persons with highly relational self-construals tend to have a greater number of close relationships than lows (Cross et al., 2000), perhaps they are protected from the negative effects of a disappointing roommate relationship by other satisfying relationships. Both studies revealed surprising results for the participants with a low relational self-construal. In contrast to most other research that finds a positive association between relationship closeness and well-being, these studies revealed a negative association for participants with a low relational self-construal. Unfortunately, we cannot untangle the person from the relationship in these studies. The low relational participants may be especially independent and perceive a close relationship

as infringing on their autonomy. Perhaps they tend to be socially isolated or lack interpersonal skills and have more difficulty dealing with the inevitable problems that arise with roommates. Researchers interested in close relationships have rarely examined involuntary relationships, such as roommate relationships. It may be the involuntary and relatively inescapable nature of these relationships that make closeness (and associated expectations by the roommate) distressing for persons with low relational self-construals. Additional research focused on the attributes of low relationals and situating the roommate experience into the larger context of other close relationships may clarify these unexpected findings. In the meantime, these results point out that the meaning and value of closeness in a relationship depends on the nature of the self-construal. Additional analyses uncovered some unexpected interactions with participant gender in Study 2 (see Notes 3 and 4). Among men, RISC scale scores were negatively related to estimates of the roommate’s depth of closeness, contrasting with the positive association found among women. Similarly, for men, RISC scale scores were negatively related to change in life satisfaction, whereas for women, it was positively related to change. The sample of men in these analyses was very small, and interactions with gender occurred in only one of the two studies, so we are reluctant to speculate about the meaning of these results. Perhaps highly relational men find themselves rooming with others who do not share this concern for relatedness, and culturally prescribed gender roles and expectations for relationships make communicating their concerns, working through differences, and achieving desired levels of closeness especially difficult for these men, leading to reduced well-being. CONCLUSIONS

Why doesn’t everyone feel the same happiness expressed in Shakespeare’s play when thinking about close friends? These studies suggest that the answer lies in the self-construal, at least in part. More important, these studies indicate that cognitive processes that promote the development of relationship closeness are associated with individual differences in the relational selfconstrual. Further consideration of variation in the selfstructure and its consequences will help researchers better understand the diverse ways that individuals participate in important relationships. NOTES 1. Although accuracy research was sharply criticized by Cronbach (1955), there has been a resurgence in interest in the past two decades. Researchers currently are taking into account Cronbach’s criticisms of the then-existing work, especially his comments about the components

Cross, Morris / RELATIONAL SELF-CONSTRUAL of accuracy that must be considered (i.e., elevation, stereotype accuracy, differential elevation, and differential accuracy; for reviews, see Funder, 1995; Kenny & Albright, 1987). New approaches to the measurement of accuracy now explicitly measure and account for these components (e.g., Bernieri, Zuckerman, Koestner, & Rosenthal, 1994). Because our research addresses dyadic relationships, we focus on agreement between partners (i.e., congruence between one person’s self-ratings and his or her roommate’s predictions of those ratings) rather than on the more strictly defined construct of accuracy (see Funder & Colvin, 1988, for a description of the differences between the Cronbachian notion of accuracy and the issue of agreement). 2. One male outlier was dropped from all analyses using the ideographic agreement scores. The results were not significantly different with the inclusion of this case. 3. To test for interactions with participant gender, the variables for gender and the interactions with the main effect terms were entered into the equation. This analysis revealed a significant RelationalInterdependent Self-Construal (RISC) × Gender interaction, β = .93, p < .02. Separate analyses for men and women revealed a positive association between the RISC scale and the prediction variable for women, β = .11, p < .02, n = 122, and a negative association for men, β = –.19, p > .10, n = 19. These findings are necessarily tentative due to the small sample of men. 4. Additional analysis examined the possibility that participant gender interacted with these predictors. These analyses revealed a significant RISC × Gender interaction, β = 1.28, p = .001. When these results were examined separately by gender, the analyses revealed that for men (n = 38), there was a negative main effect of the RISC scale, β = –.28, p = .01, but for women (n = 170) there was a positive effect of the RISC scale, β = .14, p < .01. There were no other significant interactions with gender.

REFERENCES Ambady, N., & Rosenthal, R. (1992). Thin slices of expressive behavior as predictors of interpersonal consequences: A meta-analysis. Psychological Bulletin, 111, 256-274. Andersen, S. M., & Ross, L. (1984). Self-knowledge and social inference: I. The impact of cognitive/affective and behavioral data. Journal of Personality and Social Psychology, 46, 280-293. Baumeister, R. F. (1998). The self. In G. Lindzey, S. Fiske, & D. Gilbert (Eds.), Handbook of social psychology (pp. 680-740). Reading, MA: Addison-Wesley. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529. Berg, J. H. (1984). Development of friendship between roommates. Journal of Personality and Social Psychology, 46, 346-356. Bernieri, F. J., Zuckerman, M., Koestner, R., & Rosenthal, R. (1994). Measuring person perception accuracy: Another look at selfother agreement. Personality and Social Psychology Bulletin, 20, 367378. Brewer, M. B., & Gardner, W. (1996). Who is this “we”? Levels of collective identity and self representations. Journal of Personality and Social Psychology, 71, 83-93. Byrne, D. (1971). The attraction paradigm. New York: Academic Press. Collins, N. L., & Miller, L. C. (1994). Self-disclosure and liking: A meta-analytic review. Psychological Bulletin, 116, 457-475. Cronbach, L. J. (1955). Processes affecting scores on “understanding of others” and “assumed similarity.” Psychological Bulletin, 52, 177193. Cross, S. E., Bacon, P. L., & Morris, M. L. (2000). The relationalinterdependent self-construal and relationships. Journal of Personality and Social Psychology, 78, 791-808. Cross, S. E., Gore, J. S., & Morris, M. L. (2002). The relational selfconstrual, self-consistency, and well-being. Manuscript submitted for publication. Cross, S. E., & Madson, L. (1997). Models of the self: Self-construals and gender. Psychological Bulletin, 122, 5-37.

523

Cross, S. E., Morris, M. L., & Gore, J. (2002). Thinking about oneself and others: The interdependent self-construal and social cognition. Journal of Personality and Social Psychology, 82, 399-418. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71-75. Diener, E., Suh, E., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125, 276-302. Funder, D. C. (1995). On the accuracy of personality judgment: A realistic approach. Psychological Review, 102, 652-670. Funder, D. C., & Colvin, C. R. (1988). Friends and strangers: Acquaintanceships, agreement, and the accuracy of personality judgment. Journal of Personality and Social Psychology, 55, 149-158. Harvey, J. H., & Omarzu, J. (1997). Minding the close relationship. Personality and Social Psychology Review, 1, 224-240. Hays, R. B. (1985). A longitudinal study of friendship development. Journal of Personality and Social Psychology, 48, 909-924. Joiner, T. E., Jr., Vohs, K. D., & Schmidt, N. B. (2000). Social appraisal as correlate, antecedent, and consequence of mental and physical outcomes. Journal of Social and Clinical Psychology, 19, 336-351. Kelley, H. H. (1979). Personal relationships: Their structures and processes. Hillsdale, NJ: Lawrence Erlbaum. Kenny, D. A., & Acitelli, L. K. (2001). Accuracy and bias in the perception of the partner in a close relationship. Journal of Personality and Social Psychology, 80, 439-448. Kenny, D. A., & Albright, L. (1987). Accuracy in interpersonal perception: A social relations analysis. Psychological Bulletin, 102, 390-402. Markus, H., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224-253. Markus, H., & Wurf, E. (1987). The dynamic self-concept: A social psychological perspective. Annual Review of Psychology, 38, 299-337. Murray, S. L., Holmes, J. G., Bellavia, G., Griffin, D. W., & Dolderman, D. (2002). Kindred spirits? The benefits of egocentrism in close relationships. Journal of Personality and Social Psychology, 82, 563-581. Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996). The self-fulfilling nature of positive illusions in romantic relationships: Love is not blind, but prescient. Journal of Personality and Social Psychology, 71, 1155-1180. Myers, D. G., & Diener, E. (1995). Who is happy? Psychological Science, 6, 10-19. Pierce, G. R., Sarason, I. G., & Sarason, B. R. (1991). General and relationship-based perceptions of social support: Are two constructs better than one? Journal of Personality and Social Psychology, 61, 1028-1039. Reis, H. T., & Patrick, B. C. (1996). Attachment and intimacy: Component processes. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 523-563). New York: Guilford. Reis, H. T., & Shaver, P. (1988). Intimacy as an interpersonal process. In S. W. Duck (Ed.), Handbook of personal relationships (pp. 367389). Chichester, UK: Wiley. Sheldon, K. M., Ryan, R. M., Rawsthorne, L. J., & Ilardi, B. (1997). Trait self and true self: Cross-role variation in the Big-Five personality traits and its relations with psychological authenticity and subjective well-being. Journal of Personality and Social Psychology, 73, 1380-1393. Simpson, J. A., Ickes, W., & Blackstone, T. (1995). When the head protects the heart: Empathic accuracy in dating relationships. Journal of Personality and Social Psychology, 69, 629-641. Snydersmith, M. A., & Cacioppo, J. T. (1992). Parsing complex social factors to determine component effects: I. Autonomic activity and reactivity as a function of human association. Journal of Social and Clinical Psychology, 11, 263-278. Surrey, J. (Ed.). (1991). The “self in-relation”: A theory of women’s development. New York: Guilford. Waldo, M. (1984). Roommate communication as related to students’ personal and social adjustment. Journal of College Student Personnel, pp. 39-44. Received June 15, 2001 Revision accepted September 16, 2002