Children’s Conceptions of Friendship: A Multimethod Study of Developmental Changes

Abstract

Although previous research has examined the development of children’s conceptions of friendship, two major limitations currently exist. First, previous investigators have relied principally on a single methodological approach-the open-ended interview. Second, little is known about children’s expectations of other peer relationships or how friendship expectations are distinguished from them. These two issues were addressed in the present study by administering second-, fourth-, and sixth-grade children three measures: (a) an open-ended interview, (b) a story-recognition task, and (c) a questionnaire about the importance of different characteristics to friendships and acquaintances. In general, friendship expectations based on dispositional characteristics increased in saliency with age, whereas characteristics referring to overt behaviors remained constant. Several substantive differences were found among the three measures. Children recognized the importance of different friendship expectations before spontaneously reporting them in the interview. Moreover, the recognition tasks yielded a more complete set of friendship expectations than the interview did. Finally, children of all ages differentiated between expectations for friends and acquaintances, but the degree of differentiation increased with age.

Children’s perceptions of others and of friends in particular have been found to undergo marked developmental changes. For example, young elementary school-aged children describe others primarily in terms of overt concrete behaviors and physical attributes, whereas older children describe others’ dispositions—that is, underlying personality characteristics that require an inference from the observed behaviors (Livesley & Bromley, 1973). Similarly, young children view friends as playmates whom they like and spend time around; later, they begin to describe the admirable qualities of their friends; finally, children emphasize the importance of acceptance, loyalty, intimacy, and common interests to friendship (Bigelow, 1977; Bigelow & LaGaipa, 1975).

Although the existing research provides a solid foundation for understanding the development of children’s friendship conceptions, two major limitations currently exist. First, previous research has relied almost exclusively on a single methodological approach—the open-ended interview. Certainly, interviews permit children to respond naturally and, thus, may be invaluable for generating hypotheses about the content of children’s conceptions. At the same time, children may only describe selected aspects of their conceptions in an interview, focusing on those features that are most salient, easily remembered, or commonly discussed by friends.

Several alternative procedures seem promising. In other areas of social-cognitive research, investigators have used picture or story stimuli to elicit more extensive and complete responses from school-age children (Bierman, in press). Similarly, quantitative ratings of the significance of different characteristics may also prove valuable. Although such structured approaches may provide more information about children’s conceptions than interviews alone, they require that response options be specified in advance and constrain answers to the options presented. The results of previous interview studies, however, do seem to provide a solid basis for selecting relevant characteristics and, hence, make the use of structured techniques advantageous.

A second major limitation of existing research has been the absence of information about children’s conceptions of other peer relationships. LaGaipa (1979) found that adolescents and adults differentiate among various types of positive relationships, whereas Hayes (1978) found that preschool children describe liked and disliked peers differently. As yet, however, no one has determined whether elementary school children differentiate among various types of positive relationships such as friendships and acquaintanceships. Our understanding of children’s friendship conceptions could be enhanced by identifying the specific areas of differentiation and the developmental changes in them.

The purpose of the current study was to address these two limitations in the research. Three different methods were used to examine friendship conceptions. First, an open-ended interview was conducted to measure spontaneous productions of friendship expectations. Next, the children were administered a structured story-recognition task in which they heard a series of vignettes about friends and then were asked why the children were friends. Finally, the children were asked to rate the importance of each of a series of specific characteristics for friendships and for acquaintanceships.

On the basis of the previous research, we hypothesized that young children’s conceptions would be comprised of a limited number of characteristics that refer to overt behaviors, such as sharing or playing together. It was expected that older children would have acquired additional friendship expectations. It was anticipated that these later ones would show some continuity with earlier characteristics, reflecting similar dimensions of support, association, and so on, but they would be dispositional in nature, reflecting the inference of intentions, motives, and traits appropriate to a friend’s role. We expected these changes would be observed on all measures, but the children would identify characteristics on the recognition task prior to reporting them spontaneously during an interview. Moreover, it was anticipated that even after acquiring the dispositional features, the older children would continue to identify the behavioral features on the recognition task but might not mention them as frequently in the open-ended interview. This prediction was based on the expectation that the dispositional characteristics would become central components of their conceptions, but the children would continue to recognize that the behavioral characteristics were elements of a friendship. Finally, it was hypothesized that on the questionnaire the children would differentiate between acquaintances and friendships, and the degree of such differentiation would increase with age.

Method

Subjects

The subjects were 84 second-, fourth-, and sixth-grade children. Each sex was equally represented in all grades. Seventy-five subjects were interviewed at a local elementary school, and the other nine were recruited from the same area and interviewed at a nearby university. The sample principally consisted of middle- to upper-middle-class Caucasian children.

Procedure

The experiment was comprised of three tasks—an interview, a story-recognition task, and a written questionnaire. First, each child was seen in an individual session in which they were administered the interview and story-recognition task. Two female graduate students served as the interviewers. Following these tasks, the questionnaire was administered in the classroom at the school or individually to the children who participated at the university setting.

List of Friendship Expectations

The construction of the measures and coding schemes required the advance specification of friendship expectations We reviewed previous interview studies (Austin & Thompson, 1948; Bigelow, 1977; Bigelow & LaGaipa, 1975; Livesley & Bromley, 1973) and extracted the terms most commonly used by children to describe friends or liked persons. To reduce this list to a manageable set, we grouped these characteristics into conceptually similar sets. Based on previous research, we distinguished between characteristics referring to concrete behaviors and those referring to underlying dispositions that encompass and transcend specific behaviors. For example, “help” and “sharing secrets” were categorized as behavioral characteristics, whereas “being considerate of the other’s feelings” and “loyal” were classified as dispositional characteristics. Additionally, the behavioral and dispositional categories were each subdivided into five conceptual domains: (a) prosocial support, (b) intimacy, (c) association, (d) affection, and (e) similarity. These domains were selected because they appeared to be important elements of children’s friendship (see Furman, 1982; Hartup, 1978) and because research has shown that older elementary school children do differentiate among these domains in describing their relationships (Furman, 1984). With this scheme it was possible to include all of the most commonly reported characteristics and identify behavioral and dispositional referents for each domain.

The combination of the behavioral/dispositional and domain differentiations yielded a set of 10 main categories (and an 11th miscellaneous category). These were as follows (a) behavioral support (sharing/helping), (b) dispositional support (dependable for help/considerate of other’s feelings), (c) behavioral intimacy (sharing secrets/talking), (d) dispositional intimacy (trusting/understanding), (e) behavioral association (propinquity/companionship), (f) dispositional association (continued association—i.e, faithful/loyal), (g) behavioral similarity (shared activities/mutual play), (h) dispositional similarity (common interests and beliefs), (i) behavioral affection (overt liking), and (j) dispositional affection (acceptance/admiration). Although overt liking is not a behavior per se, it was included at the lower level because it appeared to be based on more overt characteristics than were feelings such as acceptance, admiration, or respect. A miscellaneous category was also included to incorporate any other characteristics. These categories were used for coding the interview responses and for constructing and scoring the two structured measures.

Dependent Measures

Interview

The interview was comprised of five open-ended questions designed to probe friendship conceptions. Sample questions are “What is a friend?” and “What are the things people should do to be good friends?” In order to foster extensive responses by children of all ages, we included standardized probes after each question (e.g., “Tell me more”). Two naive research assistants served as coders. Interrater agreement was calculated on each of the II categories. Kappas ranged from .74 to .95 (M = .88).

Story-recognition task

Although the previous task was intended to measure children’s spontaneous expressions of friendship expectations, the story-recognition task was designed to measure the recognition of these expectations For this task the children heard 10 short vignettes describing interactions between two friends. After a vignette was read, they were asked why they thought the characters were friends.

Each of the five domains anticipated to be central to children’s friendship conceptions was presented in two stories. Every story contained descriptions and examples of both a behavioral characteristic and a corresponding dispositional characteristic. Order of presentation of the two characteristics was counterbalanced across stories. For example, to represent the domain of support, “helping each other” was selected as a behavioral characteristic, and “being able to depend and rely on each other” served as a dispositional one. The story went as follows:

Mary and Linda are friends. They depend and rely on each other. One time Mary was sick and missed school. She was depending on Linda to bring her homework. Linda wanted to play outside but she knew Mary needed the homework, so she brought Mary’s homework to her instead of playing. Mary and Linda help each other. For example, Linda helps Mary do her homework. What makes Linda and Mary friends?

Questionnaire

The questionnaire was comprised of 43 items in which the children used a 9-point Likert scale for rating the importance of specific characteristics to a relationship (e.g., “How important is it for friends to help each other?”). Parallel questions were asked about the importance of a characteristic to friends and to acquaintances (“kids who are just classmates”). The specific friendship expectations that were assessed were identical to those used in the story-recognition task. Items were presented in a random order except that parallel questions for friends and acquaintances were not presented sequentially. Additionally, five filler items were evenly distributed throughout the questionnaire to detect a general positivity bias. These items consisted of characteristics that were not expected to differentiate between friends and acquaintances (e g., “How important is it for friends to sing the same songs?”). Each item was read aloud to the children.

Results

Saliency of Friendship Expectations

First, the open-ended interview data were examined to determine whether the 10 characteristics were central facets of children’s friendship conceptions. Each of the five behavioral characteristics was mentioned by the majority of children (M = 67%). The five dispositional ones were mentioned less frequently (M = 19%), although examination of the data indicates that most of these proportions increased with grade level (see Grade Differences).

These 10 expectations also provided a relatively complete picture of the children’s friendship expectations. In fact, 79% of all responses could be classified into these categories. The remaining characteristics were generally idiosyncratic ones. Only a category for diffuse positive comments (e.g., “nice”) would have incorporated many of the remaining descriptors (14%). Thus, the present recognition and rating tasks do appear to incorporate almost all the categories of friendship expectations commonly reported by children.

As expected, almost all characteristics were more commonly mentioned on the recognition task than on the interview. The only one reported equally frequently was behavioral association, with almost all children mentioning it on both tasks. It is interesting that the proportional differences between the two tasks were generally larger on the dispositional characteristics (M = 71%) than on the behavioral characteristics (M = 24%).

Grade Differences

Grade and sex differences were examined in a multivariate analysis of variance (MANOVA) of all dependent variables. The MANOVA revealed a significant grade effect (p < .001), but neither the sex effect nor the sex by grade interaction was significant. Next, univariate analyses were conducted to determine the basis of the significant grade effect. Follow-up analyses of significant univariate effects were examined with Student Newman-Keul’s tests.

A summary of results is presented in the first three columns of Table 1. The five behavioral characteristics were commonly reported by children at all levels; hence, few consistent developmental effects were observed. The only characteristics showing any evidence of increases with age were behavioral intimacy, interview F(2, 78) = 5.77, p < .01, questionnaire F(2, 74) = 3.16, p < .05, and behavioral association, interview F(2, 78) = 3.62, p < .05. Analyses of the interview measures of behavioral similarity and affection revealed developmental decreases in scores, F(2, 78) = 3.14, p < .05, F(2, 78) = 3.98, p < .05, respectively. Post-hoc comparisons indicating the specific nature of these age differences are shown in Table 1.

Table 1.

Grade Effects in Friendship Conceptions

Expectations	Open-ended interview	Story recognition	Questionnaire (friends)	Differentiated gradesa	Size of differentiationb
Behavioral
Support	ns	ns	ns	4 + 6	ns
Intimacy	2 < 4 + 6	ns	4 < 6	All	2 + 4 < 6
Association	2 < 6	ns	ns	All	ns
Affection	4 > 6	ns	ns	All	2 + 4 < 6
Similarity	2 + 4 > 6	ns	ns	All	2 + 4 < 6
Dispositional
Support	2 + 4 < 6	2 < 4 + 6	4 < 6	All	2 < 6
Intimacy	2 + 4 < 6	2 < 4 + 6	2 + 4 < 6	All	2 < 6
Association	2 + 4 < 6	ns	ns	4 + 6	2 < 4 < 6
Affection	4 < 6	2 < 4 + 6	ns	4 + 6	2 < 6
Similarity	ns	ns	ns	4 + 6	2 + 4 < 6
Other
Miscellaneous/Filler	ns	ns	2 > 6	None	ns

Note Numbers represent grade levels. Numbers that are separated by < or > are significantly different at p < .05.

^aThe differentiated grade column lists the grades in which the friends and acquaintance ratings on the questionnaire differ significantly (ps < .05).

^bThe size of differentiation column reports whether the size of this differentiation on the questionnaire differed significantly across the three grade levels (ps < .05).

In contrast, when the dispositional characteristics were examined, many more significant developmental increases were found. In particular, significant grade effects were found on the following characteristics: (a) dispositional support, interview F(2, 78) = 10.88, p < .01, story-recognition F(2, 78) = 6.78, p < .01, questionnaire F(2, 74) = 4.33, p < .05; (b) dispositional intimacy, interview F(2, 78) = 17.94, p < .001, story-recognition F(2, 78) = 3.14, p < .05, questionnaire F(2, 74) = 5.12, p < .01); (c) dispositional association, interview F(2, 78) = 12.05, p < .001; and (d) dispositional affection, interview F(2, 78) = 4.87; p < .05, story-recognition F(2, 74) = 4.48, p < .05. No significant differences were found on the characteristic of dispositional similarity, although a near-significant effect was observed on the story-recognition measure, F(2, 78) = 2.58, p < . 10.

No significant effects were observed on the miscellaneous categories of the interview and recognition tasks (ps > .10). Analyses of the combined ratings of the questionnaire’s filler items did reveal a significant grade effect, F(2, 74) = 4.50, p < .01. Analyses were also conducted to rule out the possibility that the developmental changes on the interview simply reflected increases in verbal fluency. No significant differences were found in the total number of words said during the interview, but a significant grade effect was found in the total number of characteristics reported, F(2, 78) = 18.08, p < .001. Scores increased with grade level with each grade significantly different from the other two (ps < .05).

Differentiation of Relationships

To determine whether the children differentiated between friends and acquaintances, scores on the rating scale were subjected to three-way analyses of variance (ANOVAS) in which grade, sex, and relationship status (friend/acquaintance) were factors. Highly significant effects for relationship status were found for all 10 expectations, all Fs(1, 74) > 16, ps < .001. All were reported to be more important in friendships. In contrast, the children did not differentiate between the two types of relationships on the filler items.

The significant main effect of relationship status was usually modified by an interaction between relationship and grade. In seven instances this interaction reached statistical significance (p < .05), and in two other cases it approached significance (p < .10). The behavioral-association characteristic and the summed filler items were the only ones for which the interaction was nonsignificant. Follow-up analyses revealed that fourth- and sixth-grade children significantly differentiated between their friendship and acquaintance ratings on all 10 characteristics but not on the filler items (see Table 1, column 4). In contrast, on almost half of the characteristics (4 of 10) the second graders did not make a significant differentiation in their ratings. The size of the differentiation also increased developmentally (see Table 1, column 5). For each of the five dispositional characteristics, the difference between the friend and acquaintance ratings was greater for the sixth graders than for the fourth graders. Similar effects were found on three of the five behavioral characteristics.

Discussion

The obtained pattern of results both supports and expands upon prior research. Consistent with previous investigations, expectations based on friendly behaviors appear to be well established by the second grade and continue to be important characteristics of friendship for fourth and sixth graders. In particular, on the interview task a high percentage of children in all age groups referred to behavioral characteristics such as helping and common activities. A similar pattern was found on the two new measures employed. Specifically, on the story-recognition task the vast majority of children identified these features as characteristic of friendships, and on the questionnaire they rated them as more important for friendships than acquaintances.

In contrast, the acquisition of dispositional features continued during the elementary school years. Dispositional support and intimacy snowed increases on all three measures and dispositional affection showed increases on two measures (interview and recognition). Developmental changes were also found on the interview measure of dispositional association but not on the other two. A high percentage of all children (95%) recognized this feature as characteristic of friendship, but the proportion of children spontaneously reporting it increased from 7% to 54% between the second and sixth grade. Dispositional similarity was rarely mentioned by children of any age during the interview (M = 6%).

In general, the similarities in findings across the three measures suggest that the developmental trends found in previous research are not limited to interview measures. Moreover, no developmental changes were found in the total numbers of words, even though marked changes were found in the total number of expectations. This pattern of results may be a function of the present interviewing technique of prompting children to provide full responses. It appeared that these prompts increased the amount of talk by younger children, but much of this talk was a simple elaboration of their comments.

Although many consistent patterns were observed across the three measures, the differences among them are of equal interest. First, the proportion of children reporting a feature was almost always greater on the recognition task than on the open-ended interview. This pattern is quite consistent with previous research on semantic development (see Clark & Clark, 1977). Children often recognize that a concept is applicable before spontaneously producing the concept in their descriptions. Differences between the proportions on the recognition and the interview task measures were also much greater on the dispositional features than on the behavioral features. Apparently all ages were able to both recognize and spontaneously report behavioral features, but many of the younger children were only able to recognize the dispositional ones.

As children grew older, reports of behavioral similarity and affection decreased on the interview task but not on the recognition task or questionnaire. This pattern of findings is consistent with Bigelow’s (1977) hypothesis that friendship expectations are acquired in a cumulative manner. That is, even after children have acquired the more developmentally advanced features, the ones that were acquired at an earlier point are not discarded but instead continue to be facets of their friendship conceptions. When asked if they are part of their conceptions, as on a recognition task, they indicate that they are. However, they may be less likely to report such characteristics spontaneously because other more advanced ones have taken precedence.

The increases on behavioral intimacy and association on the interview task are more difficult to interpret. Perhaps like the dispositional ones, these two are still developing during the elementary school years. All children may recognize that these characteristics are important, but some of the younger children may not be able to spontaneously report them.

The present findings also support LaGaipa’s (1979) hypothesis that children begin to differentiate among relationships prior to adolescence. Even the second graders differentiated between friends and acquaintances on some characteristics. At all ages, however, the differentiation is selective; the children did not believe the filler items differentiated the two relationships. Moreover, the developmental process of differentiation continues during the elementary school years. Second graders displayed significant differences in their ratings on 6 of the 10 features, whereas for the fourth and sixth grade children this differentiation was apparent on all 10 features. It is interesting that the absence of differentiations by the second graders usually occurred on dispositional characteristics that may not have been part of their friendship conceptions. Additionally, for the majority of the characteristics, the difference between friend and acquaintance ratings increased developmentally. Such changes in differentiation were particularly likely to occur on the dispositional features, suggesting that the acquisition of friendship expectations and the differentiation among relationships may be related. Finally, subsequent investigations may want to examine the functions served by acquaintanceships, because children appeared to have positive expectations for them as well as for friendships.

Although the present study focused on friendship conceptions, they provide indirect support for various theories concerning the nature of children’s friendships. For example, the frequent references to behavioral association underscores the importance of common activities in young children’s friendships (Furman, 1982). Sullivan’s (1953) theory about the development of chumships during preadolescence is supported by the indications that dispositional indices of intimacy (trust and understanding) and affection (acceptance) emerge at this time. It is interesting that, although behavioral similarity was commonly reported, dispositional similarity (similar values or attitudes) was not a very salient characteristic of these children’s friendships. Because similar values and attitudes have been found to be important characteristics of adolescent friendships (Duck, 1975), the present findings suggest that the kind of similarity that facilitates friendships may change developmentally.

Several issues warrant further examination. Although this study was the first to use multiple measures, it should be recognized that all of the measures relied on verbal skills. For example, although the questionnaire did not require expressive skills, the children’s responses may have been influenced by their level of receptive vocabulary. As noted previously, the findings are not congruent with a verbal fluency explanation, but further research should incorporate alternative visual or nonverbal techniques.

The current findings also suggest that research is needed on children of a younger age. The five behavioral features were all well established by the second grade, indicating that the initial formation of friendship conceptions must occur at an earlier age. Other research does, in fact, suggest that some of these characteristics are acquired during the preschool years (Furman & Bierman, in press; Hayes, 1978).

Although several issues require further research, the current findings underscore the importance of using multiple methodological procedures. Although consistent findings were often found with the three procedures used here, the differences that were observed proved to be of important theoretical significance. Additionally, the examination of children’s conceptions of acquaintanceship both provided valuable information about the development of children’s conceptions of friendship and was of interest itself. The present approach to the study of children’s social-role concepts can expand and enrich our knowledge of the social-cognitive framework with which children view and understand their social environment.