Measuring Friendship Quality in Late Adolescents and Young Adults :

Two studies established the psychometric properties of two friendship questionnaires; one taps respondents' feelings for a friend and satisfaction with the friendship, the other, respondents' assessments of the degree to which a friend fulfills six friendship functions (stimulating companionship, help, intimacy, reliable alliance, self-validation, and emotional security). Factor analysis confirmed the subscale structure of each questionnaire. The subscales showed high internal consistency, distinguished best from casual friends, and did not covary with social desirability. They did covary with the duration of being a best friend and with a self-esteem subscale regarding close friends, but not with other self-esteem measures. Women reported higher positive feelings for their friend than did men, and evaluated the friend higher on friendship functions. Finally, positive feelings and satisfaction covaried with each friendship function subscale. The research here concerns the development and validation of two multi-scale friendship questionnaires-one concerning a respondent's feelings for a friend and friendship, the other concerning the respondent's assessment of the degree to which the friend fulfills six friendship functions. The studies grew out of work on friendship quality in children and young adolescents (Aboud & Mendelson, 1996; Mendelson, Aboud, & Lanthier, 1994). The goal here was to design measures, based on a similar model of friendship, suitable for late adolescents and young adults. Friendships, like other relationships, vary in quality. Although it is difficult to judge friendship quality in behavioral terms, length of the relationship and reciprocated versus nonreciprocated nominations are gross criterion measures of friendship quality. Furthermore, individuals can specify types of friendships, distinguishing, for example, between best friends, good friends, casual friends, and acquaintances (i.e., nonfriends); and such distinctions are also gross criterion measures of quality. Gender also provides a criterion for validating friendship measures, because there is ample evidence that gender differences do exist, with women's friendships characterized by better overall quality, closeness, enjoyment, intimacy, and nurturance (e.g., Bell, 1991; Jones, 1991; Sapadin, 1988; Wright & Scanlon, 1991). Thus, any friendship measure should be sensitive enough to differentiate women's and men's friendships. At the most general level, relationships can be assessed as positive or negative. In these terms, friendship scales have been developed to assess attachment to the friend and conflict.2 Attachment refers to the special 1This paper was completed in 1997; we subsequently published a brief version (Mendelson & Aboud, 1999). Part of the research was presented as a poster (Measuring Friendship Quality in Late Adolescents and Young Adults) at the American Psychological Association, Toronto, ON, August, 1996. We heartily thank the following: Jocelyne Andrews, Sophie Beugnot, Lisa Seidel, and Tsafrir Vanounou collected data reported here as part of their undergraduate theses. Rhonda Amsel offered excellent statistical advice. Finally, Barry Corenblum, Richard Koestner, and Debbie Moskowitz constructively criticized a draft of the manuscript. The research was supported by grants from the Social Sciences and Humanities Research Council of Canada and from the Social Sciences Research Grants Subcommittee of the Faculty of Graduate Studies and Research, McGill University. Send correspondence to Morton J. Mendelson ([email protected]). 2Conflict, which might be considered the opposite of attachment, can certainly be an important aspect of any relationship. Indeed, friends often have conflicts, but may nonetheless be able to resolve them equitably and without bad feelings (Hartup, Laursen, Stewart & Eastenson, 1988). Thus, numerous friendship measures have subscales related to conflict or conflict resolution (Bukowski et al., 1994; Furman & Adler, 1982; Furman McGill Friendship Questionnaires 2 feelings that individuals have for a friend. Mutual liking has often been used as a criterion to identify a friend, but a separate subscale may be used to assess liking in greater depth (cf. Bukowski, Hoza, & Boivin, 1994; Sharabany, 1974; Wright, 1991). One of the measures reported here, taps positive feelings for the friend and satisfaction with the friendship, which will be viewed as criterion measures of undifferentiated friendship quality. Although the two measures are conceptually distinct, they should covary highly because they are both assumed to reflect overall friendship quality. However, an important assumption guiding research on friendship is that it is possible to assess specific qualities of friendships. Consideration of the theoretical foundations of various scales suggested that a framework based on friendship functions (Furman & Buhrmester, 1985; Parker & Asher, 1989)--rather than specific behaviors (Bukowski, et al. 1994; Sharabany, Gershoni, & Hofman, 1981) or motives (Wright, 1991)-is preferable for a number of reasons: It provides a broader scope, yields a better ideal for a mature relationship, and makes it possible to develop analogous, if not identical, measures for different developmental stages. Within a functional approach, a friend is seen as a source of certain social, emotional and instrumental resources that a person seeks (Asher & Parker, 1989; Weiss, 1974). In a review of existing measures (Aboud & Mendelson, 1992), we sought to define friendship functions that were theoretically distinct, that distinguished between friends and nonfriends, and that were associated with affection/satisfaction. We identified six relevant functions (stimulating companionship, help, intimacy, reliable alliance, self-validation and, emotional security). It is assumed that individual friends fulfill some, if not all, of these functions, so measures of the different functions should covary. Nonetheless, the following definitions describe what are clearly six conceptually distinguishable functions of friendship: Stimulating Companionship refers to doing things together that arouse enjoyment, amusement, and excitement. This quality seems to be an important expectation of friends at all ages. Some measures have focused mainly on doing things together (Buhrmester, 1990; Bukowski et al., 1994; Parker & Asher, 1989; Sharabany, 1974), but it seems important to stress the fun and excitement in common activities (Jones, 1991; Wright, 1991). Help refers to providing guidance, assistance, information, advice, and other forms of tangible aid necessary to meet needs or goals. Thus, it need not be reciprocal (Jones, 1991). Help has been assessed in specific subscales (Bukowski et al., 1994; Parker & Asher, 1989; Wright, 1991) and it has also been combined with support (Bukowski et al., 1994; Sharabany, 1974). However, the instrumental aspect of support tapped by help is distinguishable from other aspects of support tapped by Emotional Security and Self-Validation. Intimacy refers to sensitivity to the other's needs and states, providing an accepting context in which personal thoughts and feelings can be openly and honestly expressed, and openly and honestly disclosing personal information about oneself. A number of researchers have Intimacy subscales (e.g., Buhrmester & Furman, 1987), although Mannarino (1976) and Buhrmester (1990) assess it as a composite along with companionship. Sharabany's (1974) Sensitivity and Knowing subscale stresses the importance of knowing without any explicit disclosure. Wright (1991) does not include such a subscale except as it pertains to selfaffirmation or the expression of true feelings. Reliable Alliance refers to being able to count on the continuing availability and loyalty of the friend. This was an important dimension underlying Selman's (1980) distinction between a fair-weather friend who would end the relationship if conflict or strains arose and a more durable friend. It is assessed in a specific Conflict and Betrayal subscale (Parker & Asher, 1989) and in a Trust and Loyalty subscale (Sharabany, 1974). Bukowski and colleagues' (1994) Reliable Alliance subscale concerns self-disclosure, which is referred to here & Buhrmester, 1985; Parker and Asher, 1989; Wright, 1991). However, these constructs are not theoretically analogous to the six friendship functions considered here. Therefore, we are currently developing separate instruments to tap negative feelings for a friend and the incidence of conflict and conflict resolution in a friendship. McGill Friendship Questionnaires 3 as Intimacy; but they combine it with a Transcending Problems subscale that is closer to the definition of Reliable Alliance. Self-Validation refers to perceiving the other as reassuring, agreeing, encouraging, listening, and otherwise helping to maintain one's self-image as a competent and worthwhile person. This is often achieved through social comparison and consensual validation of one's attributes and beliefs. Similar items have been referred to as Attachment (Sharabany, 1974), as Ego Support and Self-Affirmation (Wright, 1991), and as Reflected Appraisal (Bukowski et al., 1994), although that was combined with an Affective Bond subscale as part of Closeness. Emotional Security refers to the comfort and confidence provided by the friend in novel or threatening situations. Although the emotional support provided by a friend is considered to be important, only Wright (1991) includes items in a Security subscale to assess perception of the friend as safe and unthreatening because he or she does not betray one's trust or draw attention to one's weaknesses. Besides assessing the respondent's feelings for the friend and satisfaction with the friendship, it seems important to choose between assessing the functions that the friend is perceived to fulfill and the functions that the respondent reportedly fulfills. Most, but not all, of the items in the reviewed subscales focus on the respondent's perceptions of the friend, which is a better orientation for a number of reasons: It incorporates the functional approach to friendship quality. It avoids the attributions and inferences that are necessary for respondents to rate functions that they fulfill for a friend. Finally, it allows for empirical studies that test the validity of the qualities in enhancing rater's feelings for the friend and satisfaction with the friendship. The goal of this research was to develop one questionnaire to assess feelings for a friend and satisfaction with a friendship and another to assess the degree to which a friend fulfills the six friendship functions. Study 1 involved an initial affection questionnaire and a long version of the functions questionnaire; Study 2 involved a revised affection questionnaire and a short version of the functions questionnaire. In each study, the subscale structure of the questionnaires was examined through factor analyses, and the internal consistency of the subscales was tested through Cronbach's alpha coefficients. Finally, the validity of the measures was examined in several ways: in Study 1, through the degree to which responses covaried with social desirability scores and through the difference in scores for best and casual friends; in Study 2, through the degree to which the friendship measures covaried with the length of the friendship and with self-esteem measures; and, in both studies, through gender differences and through the degree to which affection and satisfaction covaried with the friend's fulfilling the friendship functions. Study 1. McGill Friendship Questionnaires: Subscale Structure and Validity The aim of Study 1 was to develop and validate two questionnaires: one tapping affection for a friend and satisfaction with the friendship and one tapping six friendship functions. Method Participants The participants in Study 1 included 253 undergraduates, of whom 246 designated their gender (149 women and 97 men); the under-representation of men in the sample presumably reflects males' inhibitions about volunteering for research on same-sex friendship (Lewis, Winstead, & Derlega, 1989). The participants were, on average, 22.4 years old (SD = 2.3 years; Range: 18 to 29 years), with 82% from 20 to 25 years. The participants majored in a wide range of departments in Arts, Science, and Engineering, although 28% of them were in Psychology. They were recruited primarily from large classes and were encouraged to participate by an offer of a ticket to two $50 raffles. Measures McGill Friendship Questionnaire-Respondent's Affection (MFQ-RA) (Appendix A). A 16-item questionnaire was designed to tap affection for a friend and satisfaction with the friendship. Some items were McGill Friendship Questionnaires 4 adopted from earlier scales (Asher & Parker, 1989; Bukowski et al., 1994; Furman & Buhrmester, 1985; Wright, 1991) or adapted to sample the domains of interest; new items were written to yield eight items per subscale; and some items were changed after piloting. The items are positive statements about feelings for a specific friend or the friendship, and the respondent indicates degree of agreement on a 9-point scale (-4 to 4), on which four points are labelled (-3 = very much disagree, -1 = somewhat disagree, 1 = somewhat agree, and 3 = very much agree). McGill Friendship Questionnaire-Friend's Functions (MFQ-FF) (Appendix B). A questionnaire was designed to assess the degree to which a friend fulfills the six friendship functions. Again, items from other subscales (Asher & Parker, 1989; Bukowski et al., 1994; Furman & Buhrmester, 1985; Wright, 1991) were adopted or adapted; new items were written to yield eight items per subscale, but with the intention of reducing the subscales to five items; and some items were revised after piloting. Each item is a positive statement about a specific friend fulfilling a friendship function. The respondent indicates "how often [the] friend is or does what the item says" on a 9-point scale (0 8), on which five of the points are labelled (0 = never, 2 = rarely, 4 = once in a while, 6 = fairly often, and 8 = always). Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960). The Marlowe-Crowne Social Desirability Scale is a 33-item questionnaire that taps the extent to which respondents tend to describe themselves in favorable, socially desirable terms to achieve the approval of others. The scale has high internal consistency (Cronbach's alpha = .88), high 1-month test-retest reliability (r = .88), and good validity (Robinson & Shaver, 1973). Procedures Approximately 450 sets of questionnaires were distributed to large psychology classes or by approaching students individually. The purpose of the study was explained, and students were asked to take the questionnaires with them, fill them out, and return them before a designated date. At the beginning of the questionnaire, respondents were asked to write the initials of a best same-sex friend, a good same-sex friend, a casual same-sex friend, and a same-sex acquaintance, which ensured that respondents distinguished among their friends. They were then instructed to complete the questionnaires with reference to their best same-sex friend (on about half the sets) or with reference to their casual same-sex friend (on the other half); that is, respondents were randomly assigned to the bestor casual-friend condition. All sets of questionnaires contained the MFQ-FF and the MFQ-RA in that order, either before or after two questionnaires not relevant here. About 30% of the distributed questionnaire sets ended with the MarloweCrowne Social Desirability Scale. The type of friend and questionnaire order were counterbalanced across distributed sets. About 56% of the sets were returned. Of those, 137 concerned a best friend, and 116, a casual friend; 130 had the MFQ-FF and MFQ-RA first and second, 123 had them third and fourth; and 87 sets included the Marlowe-Crowne Social Desirability Scale. Results Subscale Structure, Descriptive Statistics, and Internal Consistency McGill Friendship Questionnaire-Respondent's Affection. The 16 items of the MFQ-RA were factor analyzed to confirm its subscales structure. Oblique rotation was used here and in subsequent factor analyses because we assumed that the subscales, although conceptually distinct, would be empirically related. The resulting three-factor solution (eigenvalues = 68.5, 7.0, 6.6) accounted for 82.1% of the variance (68.5%, 7.0%, 6.6%); the intercorrelations among the factors ranged from .34 to .60 (Mr = .46). The analysis did not support the two a priori subscales. Rather, five affection items (2, 3, 8, 9, 14) and five satisfaction items (7, 10, 12, 13, 16) loaded higher on Factor 1 than on the other factors (all reported loadings _ .40). Nine of these items were combined into a Positive Feelings subscale. The tenth item (10) was close in wording to the satisfaction items (1, 4, 5) that loaded higher on Factor 2 than on other factors, so it was combined with them into a Satisfaction McGill Friendship Questionnaires 5 scale.3 Three affection items (6, 11, 15) that concerned the respondent's opinion of the friend loaded exclusively on Factor 3; they were not analyzed further because they seemed to tap an evaluative aspect of person perception unrelated to our goals. As in all the measures reported here, the score for each subscale was the mean of its items (not weighted by factor loadings). Table 1a shows the descriptive statistics, Cronbach's alphas, and mean inter-item correlations for the two retained MFQ-RA subscales. The obtained ranges were large, but the distributions were negatively skewed, with means in the top sixth of a scale from -4 to 4. The internal consistency for the subscales was very good. Although the subscales covaried, r(251) = .83, p < .01, they were analyzed independently because they tapped conceptually distinct constructs and their correlation coefficient was lower than their Cronbach's alpha coefficients. McGill Friendship Questionnaire-Friend's Functions. Factor analyses with oblique rotation were used to verify the subscale structure of the MFQ-FF and to reduce the number of items per subscale from eight to five. A six-factor solution was applied to the 48 items, and we retained the five items on each subscale that loaded highest on the factor for the subscale and lowest on the other factors.4 The 30 retained items were subjected to another six-factor solution (eigenvalues = 18.0, 1.5, 1.3, 1.0, .9, .8), which accounted for 78.8% of the overall variance (60.2%, 5.1%, 4.4%, 3.4%, 3.0%, 2.7%); the absolute values of the intercorrelations among the factors ranged from .29 to .61 (M|r| = .45). With only one exception, the five items from each subscale loaded (_ .40) on a single factor (Factor 1: Reliable Alliance; 2: Stimulating Companionship: 3: Help; 4: Emotional Security; 5: Self-Validation; and 6: Intimacy). One item from the Intimacy subscale (15) also loaded on Factor 4 with the Emotional Security items. The factor analysis provided excellent support for the subscale structure of the reduced 30-item MFQ-FF, and all subsequent analyses were conducted with the 5-item subscales. Table 1b shows descriptive statistics, Cronbach's alphas, and inter-item correlations for the MFQ-FF subscales. The obtained ranges spanned virtually the entire possible range, although the distributions tended to be negatively skewed, with means between 5.9 and 6.6 out of a possible 8. The internal consistency was very good Table 1 Descriptive Statistics of McGill Friendship Questionnaires in Study 1 Scale Theoretical Range Observed Range Mean SD Alpha Interitem r (a) MFQ-Respondent's Affection Positive Feelings -4 -4 -1.7 4 2.9 1.3 .96 .74 Satisfaction -4 -4 -3.5 -4 2.5 1.6 .92 .74 (b) MFQ-Friend's Functions Stimulating Companionship 0 8 .4 -8 6.3 1.5 .91 .67 Help 0 8 .0 -8 5.9 1.6 .89 .63 3 The decision to combine the four items on the Satisfaction scale was supported by the fact that they were the only items that specifically mentioned friendship, by the finding that the internal consistency of the Satisfaction scale was, if anything, improved with Item 10 on it (alpha = .92 vs. .91), and by the factor analysis of the revision to the MFQ-RA in Study 2. 4 Two other methods were used to eliminate three items from each subscale: reiteratively computing factor analyses and reiteratively computing Cronbach's alpha coefficients and dropping the worst item on each iteration. These methods yielded results virtually identical to those reported in the text. McGill Friendship Questionnaires 6 Intimacy 0 8 .0 -8 6.1 1.9 .94 .75 Reliable Alliance 0 8 .8 -8 6.6 1.6 .95 .79 Self-Validation 0 8 .4 -8 6.0 1.6 .91 .68 Emotional Security 0 8 .4 -8 6.1 1.7 .92 .71 for the six subscales (alphas: .89 to .95; inter-item rs: .63 to .79). Although the subscales were intercorrelated [Mr(251) = .74; Range: .68 (Intimacy and Stimulating Companionship) to .83 (Intimacy and Emotional Security)], they were analyzed independently because they tapped conceptually distinct functions, because the subscale structure was supported by the results of the factor analysis, and because their intercorrelations were lower than their Cronbach's alpha coefficients. Validity McGill Friendship Questionnaire-Respondent's Affection. Social desirability scores did not predict Positive Feelings, r(82) = -.07, p > .05, or Satisfaction, r(82) = -.01, p > .05. That is, respondents' reports of their feelings for their friend or their satisfaction with the friendship were apparently not influenced by a tendency to respond in a socially desirable way, which supports the discriminant validity of the two subscales. The two subscales were subjected to a Type of Friend x Gender MANOVA. Type of friend was significant multivariately, F(2,241) = 43.39, p < .01, and univariately for both subscales (Table 2a). Participants responding with reference to a best friend rated their positive feelings for the friend and their satisfaction with the friendship higher than did participants responding with reference to a casual friend. Gender was also significant multivariately, F(2,241) = 3.70, p < .05; women reported slightly higher positive feelings for the friend than did men (3.0 vs. 2.7), F(1,241) = 6.73, p < .05, but they did not differ from men in satisfaction with the friendship (2.5 vs. 2.4), F(1,241) = 2.22, p > .05. Table 2 Means (and Standard Deviations) of Friendship Measures for Casual and Best Friends