A field study is reported in which friendships were sampled at 1 month, 4 months and 12 months duration. Friendship variables were measured at each time-point using the Acquaintance Description Form (ADF), and the degree of value similarity was also assessed. Scores on 8 ADF variables were regressed on common similarity (that was easy to find in others) and uncommon similarity (that was not usually shared with others). Common similarity correlated significantly with liking and voluntarily interdependence in 1-month old friendships, but not in the 4-months or 12-months old friendships. In contrast, uncommon similarity correlated significantly with self-referent support variables (ego-support, stimulation and self-affirmation) in 4-months old friendships, but not in the 1-month and 12-months old friendships. A similar pattern of results was obtained in a follow-up of the 1-month old friendships after 4 months. The results suggest that common similarity is associated early on in friendship formation with general, social attraction responses whereas uncommon similarity subsequently contributes to the perception of self-referent support within the context of a personalized relationship. The results are considered in terms of predictive filtering of acquaintances, social identity/self-categorization processes, and self-referent motivation.
Major theories of friendship hold that a primary function is to provide support for the self. The latter has been conceptualized variously as support for personality (Duck, 1977b), for personal identity (Duck and Lea, 1983), for self-concept (Bailey, Finney and Helm, 1975), or as ego support and self-referent affirmation (Wright, 1984). A second feature that has been frequently remarked on is the increasingly personalistic focus of developing friendships, in which friends interact to reveal their “real” versus “role” selves (Kurth, 1970); where a private culture forms, and unique pair norms for dyadic communication and interaction emerge (Hornstein, 1985; Rands and Levinger, 1979; Knapp, 1978; McCall, 1970); and where partners develop a communal relationship in which friends respond to one another’s needs as if they were their own (Clark and Mills, 1979).
Wright (1984) proposed that friendships are formed and maintained because they facilitate the fulfilment or expression of self-referent motivation. That is, friendship development is one aspect of behaviour which is motivated by concern for the well-being and worth of the entity which is identified as self. A valued friendship is one which provides the individual with a variety of self-referent rewards: ego-support, stimulation and self-affirmation. According to this view, the process of friendship formation involves increasing amounts of voluntary interdependence and is marked by a shift from perceiving one’s acquaintance in terms of role occupancy or membership of particular classes and categories, to seeing and reacting to the other as an individual.
Similarly, Duck (1977a) has observed that while attitudinal agreement leads to positive regard between strangers (Byrne, 1971), “deeper” and more idiosyncratic forms of similarity are attractive in the context of a personal relationship. This distinction reflects a general shift that occurs in the course of relationship development such that individuals respond to each other first as stimulus objects, role occupants or stereotypes, and only later (once their personalities have emerged through acquaintanceship) do they perceive each other as stimulus persons, and interact as such.
The process has also been explored within a framework provided by social identity theory (Hogg, 1987, 1992). This draws on the distinction between social identity and personal identity that has become central to theorizing about the self (Turner, 1982; Tajfel and Turner, 1986). Under this approach, social identity derives from our membership of social categories (e.g., nationality, sex, occupation, and also more short-lived group memberships), whereas personal identity denotes specific attributes of an individual and descriptions that emerge from close and enduring personal relationships. Associated with social-personal identity continuum is a behavioural continuum ranging from group behaviour to idiosyncratic interpersonal behaviour. It is argued that in forming personal relationships there is a move from social attraction, which is generated by self-categorization processes and is grounded in group membership, to personal attraction which is interindividual attraction based upon idiosyncratic preferences and firmly rooted in close personal relationships.
In sum, several theoretical perspectives suggest that the major motivation for friendship, as distinct from social relations, is the provision of self-referent support; that comparisons between self and other, or self-categorizations, are instrumental in friendship formation, and that this process leads to a growing particularization of attraction, from social attraction to personal attraction.
Furthermore, it can be argued that value similarity may provide one basis for relevant self-categorizations to be made and for the perception of self-referent support in friendship. Values, by definition, have important implications for self-conception (McCall, 1977; Rokeach, 1973). Someone who accepts our values is perceived as accepting us, while rejection of our values implies that we are rejected (Precker, 1952). Indeed, studies of friendship and courtship suggest that value similarity is a ubiquitous feature of close personal relationships (see Coombs, 1966 and Lea, 1983, for reviews). However, there is little if any empirical evidence that directly supports theoretical analyses regarding the functions of similarity in friendship, partly because previous studies have employed a single dependent measure (such as sociometric choice, or a global attraction rating such as Byrne’s, 1971 IJS) that provides no clue as to why value similarity was attractive. One study examined which interpersonal needs might be satisfied by value similarity in friendship and found some support for its putative predictive and supportive functions Lea, 1983). Individuals who had a high need for intraception (to analyze and predict the behaviour of self and others) and a high need for deference (to get suggestions from others, to find out what others think, and to conform), and a low need for autonomy were more likely to have friends who were similar in values.
One aim of this paper therefore is to investigate the functions of value similarity in friendship, particularly in regard to support for the self, through the use of the multi-dimensional Acquaintance Description Form (ADF: Wright, 1985). The ADF measures four friendship rewards: utilitarian support, ego-support, stimulation and self-affirmation and additional scales measure other aspects of the relationship: maintenance difficulty, voluntary interdependence, general favourability, and personalized interest or concern.
A second aim is to relate value similarity to the increasingly personalistic focus of developing friendships. A previous study involving the ADF suggested that the discovery of value similarity at different points in relationship development might serve different functions. Acquaintances and friends who were value similar interacted more with each other, but friends additionally perceived a value similar partner to be more stimulating whereas acquaintances did not (i.Wright and Crawford, 1971;). However the precise significance of value similarity in the context of relationship development requires systematic study.
Taken as a whole, longitudinal studies of value similarity and relationship development suggest a moderate effect between pre-existing value similarity and subsequent attraction in the first few months of acquaintanceship (Pierce and Schwartz, 1974; Duck and Craig, 1978; Hill and Stull, 1981). The strength of the similarity–attraction effect increases with longer acquaintanceship and is mediated by greater and/or more accurate perceptions of value similarity (Newcomb, 1961; ;Curry and Kenny, 1974). The relationship between value similarity, trust (or lack of interpersonal risk) and self-disclosure is also suggested by several studies (Lundstedt, 1966; Nelson-Jones and Strong, 1976; Hlasny and McCarrey, 1980).
In considering the interplay between these processes and properties of relationships in the context of friendship development, Lea and Duck (1982) argued that values that are important for the individual’s self-concept, but are relatively unlikely to be shared with others are revealed later than values that are commonly held in the population at large. This is because commonly held values are less likely to be rejected by others and consequently their revelation carries less risk to the individual’s self-esteem. As a result, similarities and differences on common values tend to become relevant for relationship development earlier than any similarities on values that are uncommon in the population. However, it was also argued that similarity in uncommon values is particularly attractive precisely because it is more difficult to find. Similarity in these values has particular relevance for the individual’s identity because these values contribute to the person’s self-concept and sense of individuality (Duck and Lea, 1983). The discovery of similarity with another when it is difficult to find provides support for the self without compromising that person’s sense of uniqueness (cf. Snyder and Fromkin, 1980).
There is already some empirical support for the sequential relevance of common and uncommon value similarity for friendship development. In one experiment subjects were led to believe that they were about to interact with a stranger and were given a choice of discussion topics. They reported that they would feel more comfortable and would be more willing to talk about topics relating to those values which subsequent analysis showed that they shared with a large proportion of the population (common values) than they would talking about their uncommon values. Individuals who were high in socio-evaluative anxiety were particularly likely to seek out common forms of value similarity at an early stage in acquaintanceship (Lea, 1983). In another study in which student friendships were sampled after different lengths of acquaintanceship, common similarity was found to be most attractive within the first month whereas uncommon similarity was not attractive until after 3 months. Uncommon similarity was also lower in unreciprocated friendship pairs that were at a more tentative stage of development than mutual friendships.
From the foregoing analysis, it was hypothesized in general terms that common similarities would correlate with non-specific attraction responses early on in acquaintanceship (i.e., at a friendly relations or social attraction phase) that would promote further interaction and relationship development, whereas uncommon similarites that were revealed later when a relationship had become more personalized would be perceived to have more specific self-referent functions. Direct evidence for this perspective was sought from the following study of friendship in which friendship pairs were sampled after one, four and twelve months of acquaintance and the degree of common and uncommon value similarity was assessed as well as friends’ perceptions of their partner and of the relationship. In a subsequent follow-up, 1-month friendships were again assessed on the similarity and ADF variables after 4 months.
Hypotheses for the study were generated from consideration of the expected relationships at different time-points between the two types of value similarity, and the various friendship variables measured by the ADF. For 1 month friendships significant positive correlations were expected between common value similarity and the degree to which partners liked each other (GF) and were voluntarily interdependent in their activities (VID). Common similarity was not expected to correlate with any of the self-referent rewards, such as ego-support, stimulation and self affirmation at this stage, and it was also predicted that uncommon value similarity would be independent from all friendship (ADF) variables at this stage (since these values would not have been fully revealed). In contrast, for 4-month friendships it was hypothesized that there would be significant positive relationships between uncommon value similarity and the degree to which subjects perceived their friends to be providing them with ego-support, stimulation and self affirmation (ESV, SV, SAV). Also, by four months no significant correlations between common value similarity and ADF variables were expected because common similarities would no longer be salient in the context of the discovery of more rewarding similarities. By 12-months uncommon value similarity was also predicted to be no longer relevant for friendship development in accordance with previous research concerning sequential influences on relationship development (Duck and Craig, 1978; Lea and Duck, 1982). Elsewhere it hs been reported that similarities that are yet more specific and idiosyncratic become relevant for friendship by this stage, but these other variables were not measured in this study (e.g., Duck and Spencer, 1972; Lea, 1979).
One hundred and twenty-eight male and female British undergraduate students resident at two adjacent campus colleges were recruited to take part in a study of friendship. Subjects were invited to report to the laboratory with a friend of their choice where they initially completed a Friendship Choice Questionnaire (FCQ). The FCQ is a six-item measure designed to minimize the inclusion of mere acquaintances and heterosexual partners in samples of mutual, platonic friendships (Lea, 1983, 1989). The identification of suitable friendship pairs was made from an inspection of subjects’ responses on the FCQ. In particular, friendships had to have endured for either 1, 4 or 12 months and friends had to agree about the length of their friendship. One hundred and six subjects (82 per cent of volunteers) met the criteria. One subject who subsequently failed to complete the ADF correctly was dropped. The final sample was split by self-reported friendship duration as follows: 1 month n=40, 4 months n=37, and 12 months n=28.
Pairs were separately presented with the Allport-Vernon Study of Values (AVSV: Allport, Vernon and Lindzey, 1951; Richardson, 1965), which measures personal values on six dimensions, and the revised (1979) long form of the Acquaintance Description Form (ADF: Wright, 1985), which subjects completed in respect of their partner. Participation took around 90 minutes for which subjects were paid.
The ADF is an 80-item self-report measure of 8 friendship variables. Four scales measure the degree to which a subject regards a friend as providing various rewards: utility value (UV: willing to use time and resources to help the subject meet needs or reach personal goals), ego-support value (ESV: encouraging, reassuring and, in general, behaving in ways that help the subject maintain an impression of him or herself as a competent, worthwhile person), stimulation value (SV: interesting, stimulating, and capable of fostering an expansion or elaboration of the subject’s knowledge, perspectives, or repertoire of favoured activities), and self-affirmation value (SAV: behaving in ways that facilitate the recognition and expression of the subject’s more important and highly valued self attributes). In addition there are scales that assess the relationship through the subject’s own behaviour: maintenance difficulty (DTM: the degree to which it necessary to spend time and effort clarifying actions or comments, correcting or avoiding misunderstandings or, in general, exercising patience and restraint to keep his or her relationship with a friend from deteriorating or dissolving.), liking or general favourability (GF: responding to a friend in a globally positive way), voluntary interdependence (VID: committing free or otherwise uncommitted time to interaction with a friend apart from pressures or constraints external to the relationship itself.) and person-qua-person perception (PQP: responding to a friend with a personalized interest or concern, i.e., as unique, genuine, and irreplaceable in the relationship.). The reliability and validity of the ADF have been extensively researched and uniformly high scores are obtained across the various scales (Wright, 1985). Factor analysis revealed an oblique four-factor structure which accounted for 79 per cent of the common variance in the scale scores (Lea, 1989). These factors measure Utilitarian Support (UV, VID, ESV), Self-Referent Support (SV, SAV, ESV), Maintenance Difficulty (DTM, -GF) and Friendship Strength (VID, PQP).
Subjects completed the ADF in the standard way in respect of their partner (the “Target Person”, or “TP”) by indicating the degree of applicability or probability of occurrence of a series of statements. For example, one item from the ESV scale was “If I have an argument or disagreement with someone, I can count on TP to stand behind me and give me support when (s)he thinks I am right”. Another item, from the VID scale, was “When I plan for leisure time activities, I make a point to get in touch with TP to see if we can arrange to do things together”. Subjects recorded their responses on anchored 7-point scales. Scale-scores could range from 0 to 60.
The completed AVSVs were scored from the British manual for the test (Richardson, 1965). Based on these data, three procedures previously used in research on value similarity and friendship were employed to measure value importance, value commoness, and value similarity (Lea, 1983; Lea and Duck, 1982). For each subject, the two values on which the highest test scores were recorded were idenitifed as that individual’s most important values. The relative frequency with which these values were held to be important by the relevant student population was assessed by comparing the scores on these values with the mean scores obtained by a reference group of 400 students resident at the university. The mean scores, which were similar to scores for a sample of British students reported in the test manual (Richardson, 1965), provided a comparative index of the importance of the values to the relevant student population in which acquaintanceships and friendships developed. The value which obtained the higher difference score from the reference group mean was designated as the subject’s uncommon value and the other value as his or her common value. Thus, the statistical rationale for the procedure rested on the observation that the more a subject’s score deviates from the average score for a population the fewer people there are in that population with the same or a similar score. On average, individual scores for uncommon values were found to deviate by approximately two standard deviations from the population mean, while common values deviated by just under one standard deviation (Lea, 1983.;).
The validity of the value importance and value uncommonness measures has been assessed in various studies. For example, in a study of values of 117 individuals, those values classed as uncommon in the sample were no more or less important to the individuals concerned than were those values identified as common, indicating that the two measures were independent. In a study of the construct validity of the measures, 47 subjects ranked the AVSV values in order of importance to themselves as individuals and again in order of importance “to students as a whole”. Value importance and value commoness, measured from subjects’ AVSV scores, correlated significantly with perceived value importance and subjects’ estimates of value importance in the student population, respectively (Lea and Duck, 1982; Lea, 1983).
Similarity scores between friends were calculated in accordance with standard practice (e.g., McCarthy, 1981) by comparing absolute difference scores against a base-line of mean absolute difference scores obtained from nominal (no friendship) pairings. The procedure was carried out separately for each subject, once for his or her uncommon value and once for the common value. Each subject’s value score was subtracted from those of every other subject in the sample and the absolute difference scores were classified as belonging to either a friendship pair or a nominal (no friendship) pair. The mean of the difference scores for nominal pairings was then calculated and the subject’s similarity score with his or her friend was obtained by subtracting the difference score (friend) from the mean difference score (nominal). In a post hoc validity study of the similarity measure with a sample of 38 subjects, 72% of friends identified as similar in their important-uncommon values by this procedure reported typically agreeing with their friend after value-relevant conversations, and 78% of dissimilar friends reported disagreement (z = 2.26; p = 0.012).
Results and Discussion
A full correlation matrix of all the variables was calculated for each friendship group (1, 4 and 12 months) and, additionally, standard multiple regression analyses were performed in which each ADF scale was regressed onto the common similarity (COM) and uncommon similarity (UNC) predictor variables. However, correlations between the two similarity predictors for each group were low and non-significant (as expected) and consequently differences between raw correlations and standardized regression coefficients were slight (in no case greater than ±0.03). Thus, the simple correlations provided sufficient estimates of direct effects between similarity and ADF variables and will be presented here.
There were no hypotheses related to the descriptive statistics for the sample, but it was nevertheless interesting to note a consistent lack of variation in the mean scores on each scale for the three friendship groups. The invariance supports previous observations that subjective ratings of friendship behaviours reach a peak within only a few weeks of acquaintance and thereafter remain stable in continuing relationships (Hays, 1984; Wright, pers. comm.). However, maintenance difficulty (DTM) increased by a little over one-half SD with friendship length, which ties in with previous observations that tension and strain increase concomitantly with relationship development (Wright, 1978).
In contrast, there was substantial variation in the correlations between value similarity and friendship ratings dependent on the length of the friendships under study that were broadly in accordance with the predictions. For 1-month friendships, common value similarity correlated significantly with liking (GF: r(38) = 0.34, p < .05) and voluntary interdependence (VID: r(38) = 0.31, p < .01), as predicted, whereas there were no significant correlations with uncommon similarity, as expected. Indeed, only one correlation was greater than ±0.1. Also, neither type of similarity correlated significantly with any of the self-referent support variables (ESV, SV, SAV), as predicted.
A contrasting pattern of correlations that was in-line with predictions was obtained for 4-month friendships. Correlations between common similarity and the ADF scales were all low and nonsignificant, but there were significant correlations between uncommon similarity and the three self-referent support variables: ego-support (ESV: r(35) = 0.35, p < .05), stimulation (SV: r(35) = 0.40, p < .05), and self-affirmation (SAV: r(35) = 0.36, p < .05). An additional correlation was recorded with utility value (UV: r(35) = 0.33, p < .05). For 12 month-old friendships there were no significant correlations between the friendship scales and either type of similarity, as predicted.
The results provide encouraging support for the hypotheses. Common value similarity was associated with liking and with voluntary interdependence in beginning friendships, but was not attractive in longer-standing friendships. In contrast, uncommon similarity was not attractive in beginning friendships, but was associated with various classes of self-referent support in later friendships. Neither type of similarity was salient for 12-month old friendships.
The 1-month old friendships were followed-up after three further months of acquaintanceship, at which point the correlations between value similarity and the friendship variables were re-examined. This phase provided an opportunity to cross-validate the earlier results for the crucial 4-month stage when only uncommon similarity was predicted to be salient by tracking the same relationships as acquaintanceship increased over time. The hypotheses were therefore broadly the same as before. That is, significant correlations were expected between uncommon similarity and the various self-referent rewards at 4 months, but no significant correlations with common similarity were predicted. However, whereas the cross-sectional study sampled only continuing mutual friendships at each time-point, the longitudinal phase followed-up relationships 3 months after the initial friendship choices had been made and should therefore have sampled a wider variation in the quality of relationships. That is, in the intervening period before follow-up it was expected that not all friendships would have developed uniformally, but that a proportion would have either frozen their development or else declined in intimacy. The increased variability in relationship quality at a given stage of acquaintanceship (which should be reflected in increased variability in ADF scores), should provide clearer evidence of the relationship between the similarity and ADF variables in friendship development.
Subjects in the 1 month friendship group were contacted again three months later and invited to return to the laboratory with the same partners. Thirty out of the original sample of 40 friends (i.e., 75%) responded and again completed the AVSV and the ADF. The questionnaires were scored as before to provide measures, after 4 months of acquaintanceship, of uncommon and common similarity between friends relative to nominal pairs, and of the various friendship variables. Means and SD’s were calculated for all the time-2 variables and were recalculated for the time-1 variables in order to allow comparisons to be made. Correlation analysis of the two types of similarity with the ADF variables was carried out separately for the two time-points at which the data had been gathered. In addition, comparisons between the two sets of correlations were carried out separately for each scale by application of the Pearson-Filey formula (Kenny, 1979). However, cross-lag correlational (CLC) analysis was ruled out, a priori, since the fundamental assumption of stationarity (lack of change over time in the strength and direction of the causes of a variable) is incompatible with stage models of development, such as described here. Unstationarity in the data is indicated by unequal synchronous correlations at the two time-points (which is precisely what was hypothesized and observed), but while unequal correlations can indicate a causal relationship between the variables, they violate the stationarity assumption of CLC analysis and invalidate an interpretation of causality from comparison of the cross-lag correlations (Kenny, 1979).
Comparisons of the descriptive statistics showed that differences in scale means and variances between the original and follow-up samples were small and clearly nonsignificant and that the follow-up sample was therefore representative of the original sample of 1-month-old friendships. Comparisons between the data at 1 month and 4 months for the new sample showed no significant differences in the mean ADF scores, but a significant increase in the variability of the scores on GF, as expected (F(29,29) = 2.04, p < .05) and VID (F(29,29) = 2.35, p < .05). The increased variabilities were on the scales that had significant correlations with common similarity at time 1, and supports the contention that at 4 months the sample contained a greater diversity of intimate and non intimate relationships than at time 1.
The results of the correlation analysis suggest how common similarity was associated with the changes in the ways the relationships were perceived by subjects over time. For 1 month friendships there were significant positive correlations between common value similarity and GF (r(28) = 0.40, p < .05), and between common similarity and VID (r(28) = 0.32, p < .05), similar to the pattern of correlations observed in the larger sample, and again suggesting that the subsample was representative of the larger sample from which it was drawn. Correlations between uncommon similarity and the friendship scales were all nonsignificant. As before, only one correlation out of 8 was greater than ±0.1.
At follow-up, after 4 months of friendship, common similarity was no longer positively correlated with liking; indeed common similarity had come to be associated with disliking at this time (GF: r(28) = –0.34, p < .05). The difference between the correlations at time 1 and time 2 was significant (GF: rdiff = 0.70, z = 2.99, p < .01). There was also a significant positive correlation between common similarity and maintenance difficulty after 4 months (DTM: r(28) = 0.34, p < .05), and the difference between the correlations at time 1 and time 2 was significant (DTM: rdiff = 0.46, z = 2.04, p < .05).
These pattern of correlations also provide some confirmation of the relationship between uncommon value similarity and self-referent reward after 4 months of friendship. Significant though moderate positive correlations were observed between uncommon value similarity and two measures of self referent reward, ego-support (ESV: r(28) = 0.35, p < .05) and stimulation (SV:r(28) = 0.33, p < .05), as predicted. For the third self-referent reward scale (SAV) the correlation was smaller and nonsignificant, but in the predicted direction (r(28) = 0.14, n.s.). There were no other significant correlations involving uncommon similarity.
It was proposed that two types of value similarity, common similarity that is easy to find and uncommon similarity that is difficult to find, serve different sequential functions in friendship development and that these functions relate to the transition from social attraction to the formation of a personal relationship and to the perception of self-referent support. Common similarity, it was argued, is instrumental early on in friendship development, by reason of its early discovery, but ceases to be attractive later, once less common but more rewarding similarities are revealed. Several aspects of the results support this general hypothesis. First, common similarity correlated with general favourability and voluntary interdependence after 1 month of friendship, but not after 4 months or 12 months. In the follow-up of 1 month-old friendships, common similarity was no longer attractive three months later. Indeed, there was evidence in those relationships that common similarity had come to be disliked by this time, and that it was associated with increased tension and strain. Presumably this was more the case in relationships where other more rewarding factors (e.g., uncommon similarity) had not emerged and/or the relationship was in decline, although this is a question for further empirical work.
Secondly, uncommon similarity only became attractive subsequent to common similarity when it was significantly associated with perceptions of self-referent support. Whereas after only 1 month of friendship uncommon similarity did not correlate with any of the measured friendship variables, after 4 months it was positively associated with perceptions of ego-support, stimulation and self-affirmation. An almost identical pattern of results was again obtained in the follow-up of 1 month friendships after 4 months.
These results suggest that the different types of value similarity relate systematically to various features of friendship development identified by Wright (1978, 1984). Wright argues that friendship development progresses through a sequence of phases that are characterised by successively increasing levels of personalization of the relationship and voluntary interdependence, driven by the anticipation or perception of self-referent rewards obtaining from the relationship. Thus, friendship begins with initial contact, in which interaction is role-bound or perfunctory, but which provides the opportunity for friendship to occur; it proceeds to friendly relations, in which personalized informal interaction takes place signifying a degree of recognition of personal distinctiveness and anticipation of potential rewards from further interaction. Based on the levels of anticipated and actual rewards, friendship progresses to become more personalized, and involves increasing amounts of voluntary interaction and progressively more intimate self-disclosure. The actual rewardingness of friendship follows from the facilitated expression of behavioural tendencies related to the self and its attributes expressed in terms of ego-support, stimulation, and self-referent affirmation.
However, it should be noted that perceived levels of interpersonal rewards did not increase significantly with friendship length; the results suggest instead that the criteria by which partners judged their friendships to be attractive or rewarding changed with acquaintanceship such that different factors came to be associated at different times with continuance of the relationship. The fact that 12-month-old friendships perceived very similar levels of rewards from their relationships even though these perceptions were no longer associated with value similarity also fits this interpretation.
These observations also support the predictive filtering account of friendship development which holds that different sorts of cognitive criteria become salient at different stages of friendship development (Duck, 1977b). Similarities of different types are argued to be one class of criteria that relate to the anticipation or perception of support for personal identity (Duck and Lea, 1983). Superficial, easily found similarities that are revealed early on in friendship are argued to function as predictive cues to future support that may be obtained from the subsequent discovery of ‘deeper’ similarities. While earlier filters include physical attractiveness, interaction style and superficial similarities (including agreement on specific attitudes), later filters are increasingly more personalized forms of similarity (Duck, 1977b). Results from numerous studies suggest that similarities in idiosyncratic personal constructs become important in well-developed friendships (Duck and Spencer, 1972; Lea, 1979; Neimeyer and Neimeyer, 1981). In addition, the salience of personal construct similarity for relationship development (after approximately 8-12 months in student friendships) coincides with the decline in the importance of value similarity observed here and elsewhere (Duck and Craig, 1978; Lea and Duck, 1982).
It is also interesting to briefly explore the results in relation to the distinction between social and personal attraction developed within social identity theory (Hogg, 1987, 1992; Hogg and Abrams, 1988). This distinction is similar in general terms to the idea (described earlier) that relationship development involves a shift from regarding one another in terms of social stereotypes to perceiving each other as persons (Duck, 1977a, Wright, 1978, 1984). However, under the social identity approach the distinction between social attraction and personal attraction is also made explicit at the level of generative process. Social attraction is based on group processes (e.g., conformity, stereotyping, intergroup differentiation) whereas personal attraction is not. Following this approach, social self-categorizations (at the start of acquaintanceship) generate intragroup attraction, that is liking and interdependence (represented here by the GF and VID scales of the ADF) because they render self and other stereotypically identical (the other is liked because the self is liked) and because they contribute to self-esteem (the positively evaluated characteristics of the ingroup stereotype represented by the other are confered on self). With increasing acquaintanceship however, personal attraction based on idiosyncratic characteristics of the other and of the developing relationship become more salient (represented here by the (personal) self-referent support scales, ESV, SAV and SV).
Strictly speaking, from a self-categorization perspective the distinction is drawn between group-prototypical similarity and idiosyncratic interpersonal similarity rather than on differences in the commonality of similarity; (Hogg, 1992). However differences in the timing of the disclosure of common and uncommon similarity during acquaintanceship mean that common similarities are discovered at a time when social self-categorizations are likely to be more salient whereas uncommon similarities are discovered later when personal idenity is likely to be more salient.
In sum, the present approach to relationship development underlines the importance of locating perceptions of different types of similarity to on-going processes in social attraction and personal relationship development. This approach is further illustrated by contrasting attitudinal agreement with value similarity in the context of relationship development. The important conceptual distinctions between values and attitudes have been overlooked in the attraction literature which often treats them as interchangeable concepts and data. Whereas attitudes are focused upon specific objects and situations, values are overarching, evaluative beliefs that guide actions and judgements across objects and situations. Individuals’ values also show much greater stability over time than their attitudes. Values, it has been suggested, provide enduring standards, or “dominant frames of reference”, for guiding and justifying the actions, attitudes, comparisons and evaluations of self and other and as such have far greater implications for self-conception (Arsenian, 1970; Newcomb, 1950; Rokeach, 1968, 1973).
These distinctions also extend to attitudinal agreement and value similarity in the context of relationship development. In social interaction values are often not explicitly revealed as such, but can be inferred only indirectly from consistencies in attitudes and behaviour observed over time (Altman and Taylor, 1973; Jourard, 1964; Lea, 1983). Although the attractiveness of attitudinal agreement with a stranger relates to individuals’ levels of socio-evaluative anxiety and fear of rejection (Smith and Jeffery, 1970; Goldstein and Rosenfeld, 1969), the correlation between attitudinal agreement and self-esteem, which would provide crucial evidence of the link with self-concept support, has not been found (Griffitt, 1966; Guthwin, 1970; Hendrick and Page, 1970). Furthermore, the attractiveness of attitudinal aggreement observed between strangers and new acquaintances (Byrne, 1971), declines with the discovery of value similarity in developing relationships (Beech, 1967; McCarthy and Duck, 1976, 1979; Duck and Craig, 1978).
All in all, it can be argued that although attitudinal agreement may provide consensual validation of the correctness of holding specific attitudes, and as such leads to positive regard particularly between strangers, there is little evidence to suggest that attitudinal agreement in relationships—in contrast to more fundamental similarities, such as in values—has many implications for individuals’ self concept or identity, or indeed for the relationship itself. One possibility suggested by the present analysis is that the discovery of attitudinal agreement at the stage of initial acquaintance sets up an initial self-other categorization, and thereby contributes to liking as social attraction. Further communication of attitudes permits comparisons to be made in terms of overarching values and any similarities perceived there contribute to increasingly personalized attraction in accordance with the processes of relationship development that have been described. While this is a matter for further research, the study reported here suggests that both the behavioural features of friendship and the perception of self-referent support from the relationship relate systematically to self-disclosure and discovery of similarities in values that differ in their tendency to be shared with others.