Merrill-Palmer Quarterly

The Development of Achievement Motivation in Peer Contexts

The current study contributes to a still small, but growing, literature indicating that interactions with friends play an important role in the development of children's school-related attitudes and outcomes (for reviews, see Altermatt & Kenney-Benson, 2006; Berndt, 1999; Wigfield et al., 2006). Children who select friends who do well in school, who are actively involved in classroom activities, and who hold positive achievement-related beliefs tend to adopt similar characteristics over time (Altermatt & Pomerantz, 2003; Berndt & Keefe, 1995; Ide, Parkerson, Haertel, & Walberg, 1981; Kindermann, 1993; Ryan, 2001). Moreover, children who are involved in relationships characterized by high levels of positive friendship qualities (e.g., intimacy and social support) experience stronger classroom engagement and academic performance (Berndt & Keefe, 1995; Cauce, 1986; Kurdek & Sinclair, 1988) than children who are involved in lower quality relationships.

The processes by which children's achievement attitudes and outcomes are shaped by interactions with friends are less well understood. Although researchers have identified a number of potentially important mechanisms (e.g., reinforcement, social comparison, and help giving), rarely have these processes been directly observed in achievement settings or empirically linked to changes in children's achievement-related behaviors and beliefs over time (for reviews, see Berndt, 1999; Ryan, 2000; Wentzel, 1999). Berndt and his colleagues (Berndt, Laychak, & Park, 1990) have conducted one of the few studies to examine the mechanisms of friends' influence in the achievement domain directly, using quasi-experimental observational methods. In that study, eighth-grade students were presented with hypothetical dilemmas in which they had to make a decision that reflected either a high level of achievement motivation (e.g., studying for an examination) or a low level of achievement motivation (e.g., going to a rock concert instead). Children made their decisions both before and after talking with a friend. Consistent with the notion that reinforcement is an important mechanism of influence, adolescents were most likely to shift toward high-motivation alternatives when friends supported these alternatives in their conversations.

The present study extends the work of Berndt and his colleagues by directly observing children's conversations with friends immediately after children experience an achievement-related failure to examine the nature of [End Page 456] these interactions and their relation to changes in children's responses to failure from postfailure to postdiscussion. Directly observing children's interactions with friends in the context of achievement-related failure is important in at least three respects. First, children's responses to failure have been broadly implicated in a variety of models of achievement motivation (for a review, see Wigfield et al., 2006) and are thought to play a key role in predicting children's long-term educational and occupational choices (Dweck, 1986). Second, although positive and negative feedback from adults has been clearly linked to the development and maintenance of learned helpless responses to failure, the degree to which other features of children's social interactions-particularly their social interactions with equal-status peers-are related to learned helpless responses to failure are not well understood (Nolen-Hoeksema, Wolfson, Mumme, & Guskin, 1995; Wigfield et al., 2006). Finally, peer influence has most often been assessed indirectly, using survey methodologies (Berndt et al., 1990). By examining children's interactions with friends in the context of achievement-related failure, we are able to directly observe a number of mechanisms of peer influence (e.g., help giving, social comparison) that have been identified in several recent reviews of the literature as potentially important contributors to the development of children's achievement-related behaviors and beliefs (for reviews, see Berndt, 1999; Ryan, 2000; Wentzel, 1999).

Gender and Relative Performance as Moderating Variables

Evidence for gender difference in learned helpless responses is mixed, with some studies reporting gender differences (e.g., Dweck & Bush, 1976; Dweck, Goetz, & Strauss, 1980; Dweck & Repucci, 1973; Nicholls, 1975) and others reporting no gender differences (e.g., Kamins & Dweck, 1999; Ziegert et al., 2001). When studies do find gender differences, the usual pattern is that girls are more likely than boys to exhibit maladaptive learned helpless responses to failure (Ziegert et al., 2001; Wigfield et al., 2006). These gender differences have been linked in part to students' interactions with important adult figures in their lives. For example, Dweck, Davidson, Nelson, and Enna (1978) found that boys and girls receive different types of feedback in the classroom setting and that the types of feedback received by girls (e.g., criticism focused on intellectual inadequacies) predicts the negative attribution style (i.e., blaming failure on ability) characteristic of learned helplessness.

The current study extends this work by examining the degree to which boys' and girls' conversations with friends differ and the degree to which these conversations might contribute to gender differences in learned helplessness. Prior research suggests that girls' interactions with friends are more [End Page 457] likely than boys' interactions to be characterized by high levels of help, validation, and caring (e.g., Berndt, Hawkins, & Hoyle, 1986; Berndt & Keefe, 1995; Brendgen, Markiewicz, Doyle, & Bukowski, 2001; Parker & Asher, 1993; for a review, see Rose and Rudolph, 2006). Much of this research is based on children's self-reports, however, and it remains unclear to what degree these gender differences will be apparent in observations of actual conversations between friends, especially in the context of an achievement-related failure. Assuming that gender differences do emerge, it still remains unclear what the consequences might be for children's responses to failure. For example, some research indicates that the social support-seeking characteristic of girls' interactions during times of stress may heighten rather than reduce anxiety (Altermatt, 2007; Costanza, Derlega, & Winstead, 1988; Harlow & Cantor, 1994; Rose, 2002; Rose, Carlson, and Waller, 2007).

The current study also examines the ways in which relative performance may influence the nature and consequences of children's conversations with friends following failure. There is clear evidence that children use social comparison information to evaluate their competencies and to guide their behavior (Ruble, Boggiano, Feldman, & Loebl, 1980). As a result, we expect children and friends to communicate differently when both children experience failure than when one child experiences failure and the other succeeds. For example, we expect children to seek more help when friends succeed than when friends fail, as children are more likely to expect help seeking to be effective when a friend has successfully completed the task (see Ryan, Pintrich, & Midgley, 2001). It remains unclear whether the consequences of help seeking (or other features of children's conversations) will vary according to friends' performance. On the one hand, one might expect help seeking to lead to better outcomes (i.e., fewer learned helpless responses) when friends succeed than when friends fail, as successful friends can indeed provide more effective help and model positive achievement beliefs (e.g., high expectations for future success). On the other hand, one might expect help seeking to predict fewer learned helpless responses when a friend also experiences challenge, as friends can work together to find a solution without the negative self-evaluative consequences that upward social comparison might bring. Indeed, prior research indicates that a key reason for the avoidance of help seeking is the fear that others will view one as incompetent (Ryan et al., 2001).

Overview

The overall purpose of present study is, then, to examine the role that fourth- through sixth-grade children's interactions with friends play in predicting maladaptive learned helpless responses to failure in an achievement [End Page 458] setting. The specific goals of this study are threefold. First, the research provides novel information regarding the nature of children's conversations with friends as children cope with a specific achievement-related challenge. Second, the research examines both focal child and friend statements as potentially important predictors of changes in children's responses to failure from postfailure to postdiscussion. Third, the research examines whether the nature of friends' discourse or its relation to changes in children's responses to failure varies by the gender of the dyad or by relative performance of its members.

In coding children's conversations, we focused our attention on two general categories of statements. First, we coded the frequency with which children and their friends made off-task statements, that is, statements that were not task relevant (e.g., "What are you wearing to school tomorrow?"). We expected that when off-task talk was prevalent in friends' conversations, children would report increasingly more learned helpless responses to failure. This finding would be consistent with evidence from the coping literature that suggests that problem-focused, engaged coping strategies (including help seeking) generally predict better adjustment than emotion-focused, disengaged coping strategies (including cognitive avoidance) (Compas et al., 2001). It would also be consistent with Hokoda and Fincham's (1995) finding that mothers who continued to engage their children in a difficult task by, for example, reassuring them of their high ability and making teaching statements were less likely to raise children identified as learned helpless. Second, we coded a range of on-task statements. These included statements in which children were simply exchanging task-relevant information (i.e., "My blocks were green and white") and those statements that were either directly or indirectly evaluative (e.g., negative task statements such as "I only got one done" or help-seeking statements). These on-task evaluative statements were coded into categories similar to those used in prior research examining children's discourse in the classroom setting (see Altermatt et al., 2002; Frey & Ruble, 1985, 1987). We again expected statements that reflected problem-focused or engaged coping (e.g., help-seeking statements or help-giving statements) to predict positive responses to failure. In contrast, statements that reflected emotion-focused or disengaged coping-for example, discounting statements (e.g., "I don't care") and negative task statements (e.g., "I hate these kinds of puzzles")-were expected to predict negative responses to failure (see Compas et al., 2001).

Importantly, the present study examines statements made by children and their friends. Although many prior studies have examined the behaviors of only one interaction partner, some recent research examining adult-child interactions (e.g., Hokoda and Fincham, 1995) and peer interactions (e.g., Altermatt, Pomerantz, Ruble, Frey, & Greulich, 2002; Sage and Kindermann, [End Page 459] 1999) suggest that both types of statements can be important predictors of changes in children's achievement-related beliefs. Moreover, these studies suggest that examining statements in the natural sequence in which they occur (e.g., to determine what happens after a child asks for help from a peer) can be helpful in explaining why and in what contexts particular statement types are predictive of changes in achievement-related beliefs over time. For example, Altermatt et al. (2002) found that elementary school students who frequently sought help from peers in the classroom setting reported positive changes in their competence perceptions over the academic year; however, this effect held only for girls. Sequential analyses provided some insights into this finding, revealing that girls were significantly more likely than boys to receive the help they requested. In contrast, boys were significantly more likely than girls to have peers respond by evaluating themselves negatively (e.g., "I'm really bad at math"). Context is expected to be similarly important in the present study, and sequential analyses are, again, used to provide at least initial insight into when and why particular types of statements might be associated with changes in children's learned helpless responses from post-failure to postdiscussion. For instance, although negative performance statements (e.g., "I didn't get any of them right!") are expected to be associated with maladaptive learned helpless responses to failure in the present study because they are emotion focused rather than problem focused, we expect that this is more likely to be the case in dyads where friends respond by evaluating their own performance positively (e.g., "Really? I got them all!"). This finding would be consistent with research indicating that elementary school children use social-comparative information to judge their capabilities (e.g., Butler, 1989; Ruble, Boggiano, Feldman, & Loebl, 1980; Ruble, Feldman, & Boggiano, 1976).

Methods

Measures

Discourse

Children's discourse was transcribed and later coded by three research assistants. The unit of analysis was a thought unit defined as a sentence, phrase, or fragment that conveys a single, complete idea (Bakeman and Gottman, 1997). Each statement was coded according to the person who made the statement (i.e., focal child or friend) and the content of the statement made. A broad off-task category was used to code all statements that were non-task relevant (e.g., "Did you go to the basketball game?"). A broad on-task, nonevaluative category was used to code all statements in which participants were exchanging information that was task relevant but nonevaluative (e.g., "The puzzle looked like a star"). The remainder of children's statements represented on-task, evaluative discourse and were coded into 19 statement types similar to those used by Altermatt et al. (2002) and Frey and Ruble (1985, 1987) in earlier studies of children's classroom discourse. Among these statement types were discounting statements (e.g., "I don't care"), help-seeking statements (e.g., "How did you do it?"), negative performance statements (e.g., "I didn't get any of them"), and negative task statements (e.g., "I hate these kinds of puzzles").

Checks on observers' reliability in categorizing children's discourse were conducted by having all three coders code 20% of the transcripts. Reliability coefficients were estimated following Cohen (1960). Averaging across statement types and observers, percentage agreement in coding statements was 94% (Cohen's kappa = .90). Mean kappas between each pair of observers ranged from .86 to .96. All statement types are listed, with examples and individual kappas, in Table 1. [End Page 462]

Click for larger view
View full resolution

Table 1.

Statement Types and Examples

[End Page 463]

Responses to failure

Focal children's responses to failure were assessed at two time points: (1) postfailure (i.e., immediately after the focal child failed to solve the puzzles but before he or she discussed the puzzles with a friend) and (2) postdiscussion (i.e., immediately after the focal child had the opportunity to discuss the failure with a friend). Children were asked to complete 11 Likert-scale items (range = 1 to 5) that were similar to those used in other research designed to assess children's learned helpless responses to failure (e.g., Kamins & Dweck, 1999; Ziegert et al., 2001). Specifically, children were asked to report on their self-perceptions of competence (e.g., "I am good at solving puzzles"; 3 items), ability attributions for failure (e.g., "I did poorly on the puzzles because I'm not smart"; 2 items), expectations for future success (e.g., "I am confident that I will do well on the next set of puzzles"; 3 items), and persistence (e.g., "I want to do more puzzles"; 3 items). All items were scored so that higher numbers reflected more maladaptive learned helpless responses to failure (i.e., negative self-evaluations, attributions for failure to ability, low expectations for future success, and a desire not to persist). Because these items were highly correlated, participants' responses were averaged to form a single scale ranging from 1 (maladaptive) to 5 (adaptive). These items yielded reliable scales at both postfailure (α = .86) and postdiscussion (α = .86).

Results

Four broad sets of analyses were carried out. First, a set of preliminary analyses was conducted to examine the efficacy of the success/failure manipulation and to investigate potential gender and/or relative performance (i.e., friend failure vs. friend success) differences in children's responses to failure. Second, descriptive statistics were calculated to examine the content of focal children's discourse with friends, and analyses were conducted to test the degree to which children's discourse varied by gender or relative performance. Third, a series of regression analyses was conducted to examine relations between children's discourse with friends and changes in their responses to failure from postfailure to postdiscussion. Fourth, sequential analyses were conducted to examine the context in which focal child and friend statements were made (e.g., What typically happens after children evaluate the task negatively? How often do friends mirror the negative evaluation? How often do they provide help?). By understanding the context in which these statements are made, we hoped to gain at least initial insight into why and in what contexts particular statements predict maladaptive responses to failure. One might expect, for example, that negative task statements (e.g., "I hate these kinds of puzzles") [End Page 464] would be more likely to predict learned helpless responses to failure if they are frequently repeated by friends (e.g., "Yeah. They're dumb.") than if friends frequently offer assistance.

Children's Discourse

The content of children's discourse was examined by calculating two sets of statistics: (a) the raw frequency with which each type of statement was made and (b) means and standard deviations representing the percentage of children's statements that, on average, fell into each category. Percentages for off-task statements were calculated by dividing the number of off-task statements each child made by the total number of statements made by that child. Percentages for each of the on-task statement types were calculated by dividing the total number of statements made by the child in each category [End Page 465] (e.g., help seeking) by the total number of on-task statements made by that child. Raw frequencies, means, and standard deviations are presented, separately for focal children and friends, in Table 2. To control for individual differences in the total number of statements made and received, we based all analyses on percentages rather than raw frequencies. For these analyses, percentages were arcsine transformed to better approximate the normal distribution. For ease of interpretation, the nontransformed percentages are reported in Table 2 and in the text.

Click for larger view
View full resolution

Figure 1.

Focal Children's Learned Helplessness by Gender and Time.

To assess whether children's discourse varied by the gender of the participants or the performance of friends, we computed a series of gender (male, female) × relative performance (friend success, friend failure) ANOVAs for focal child and friend statements. To avoid making conclusions about statement types that occurred very infrequently, we included in the analyses only statements with a mean frequency greater than 1%. To further control for Type 1 errors, we used the Benjamini-Hochberg (1995) correction. This correction is similar to the Bonferroni correction but has been favored by statisticians (e.g., Williams, Jones, & Tukey, 1999) because its power remains stable as the number of comparisons becomes large. Only comparisons that remained significant after the Benjamini-Hochberg correction was employed are presented here.

Off-task statements. The percentage of children's statements that were off-task did not differ by gender or by relative performance for either focal children or friends (all ps > .05).

On-task statements. Analyses for focal child statements revealed a gender main effect for performance checks. Specifically, male focal children [End Page 466] (M = 5.68, SD = 7.92) were more likely than female focal children (M = 2.89, SD = 3.81) to ask friends about their performance (F[1,112] = 6.88, p = .01). Relative performance main effects emerged for help-seeking, negative performance, and negative task statements (Fs[1,112] > 6.80, ps < .01). Specifically, focal children were more likely to seek help from their friends (M_success = 2.85, SD = 4.82; M_failure = 0.41, SD = 1.03), state that they had performed poorly (M_success = 6.59, SD = 5.85; M_failure = 4.44, SD = 3.27), and evaluate the task negatively (M_success = 2.97, SD = 4.41; M_failure = 1.19, SD = 2.65) when their friends succeeded than when they failed.

For friend statements, no gender main effects emerged. Relative performance main effects emerged for discounting, help-giving, positive performance, negative performance, negative self-evaluative, task easy, task difficult, and task positive statements (Fs[1,112] > 7.04, ps < .01). These analyses showed that friends were more likely to offer help to the focal child (M_success = 6.16, SD = 10.11; M_failure = 1.21, SD = 3.39), state that they had performed well (M_success = 4.74, SD = 3.95; M_failure = .17, SD = 0.75), comment that the task was easy (M_success = 2.95, SD = 4.52; M_failure = .81, SD = 1.98), and evaluate the task positively (M_success = 1.75, SD = 3.00; M_failure = .44, SD = 1.09) following success rather than failure. In contrast, friends were more likely to discount their performance (M_success = .52, SD = 1.38; M_failure = 1.96, SD = 3.00), state that they had performed poorly (M_success = 0.28, SD = 1.14; M_failure = 4.90, SD = 4.29), evaluate themselves negatively (M_success = .34, SD = 1.14; M_failure = 2.51, SD = 3.42), and state that the task was difficult (M_success = .16, SD = 2.39; M_failure = 5.89, SD = 4.89) when they failed than when they succeeded. In addition, one gender by relative performance interaction emerged for performance checks (F[1,112] = 5.84, p < .05). Follow-up analyses revealed that male friends (M= 7.13, SD = 4.71) were more likely than female friends (M= 3.35, SD = 5.15) to ask focal children about their performance (e.g., "How many did you get?") in the success condition (t[56] = -2.54, p = .01). No gender differences emerged in the failure condition (t[56] < 1, ns).

Predicting Responses to Failure

To examine relations between children's discourse and changes in focal children's responses to failure from postfailure to postdiscussion, we conducted separate hierarchical regression analyses for each statement type. To minimize the chance of Type 1 errors, we included only statements with a mean frequency greater than 1%. At Step 1 in these analyses, we entered focal children's postfailure responses to failure. Adjusting for children's responses immediately after failure (but prior to discussions with friends) [End Page 467]

Click for larger view
View full resolution

Table 2.

Descriptive Statistics: Focal Child and Friend Statements

[Begin Page 470]

when predicting their postdiscussion responses is important to ensure that significant findings indicate that children's discourse predicts changes in children's responses rather than that discourse and responses to failure are related concurrently. Statement type, gender, and condition were entered at Step 2. Two-way interaction terms (statement type × gender, statement type × condition, gender × condition) were entered at Step 3. A three-way interaction term (statement type × gender × condition) was entered at Step 4. Significant interactions were broken down using procedures developed by Aiken and West (1991). All continuous independent variables were centered.

Off-task statements. A significant statement type main effect emerged for off-task statements made by friends (β = .13, t[114] = 2.29, p < .05). Focal children whose friends made off-task statements more frequently reported more learned helpless responses to failure than focal children whose friends made off-task statements less frequently.

On-task focal child statements. Significant statement type main effects emerged for discounting statements (β = .19, t[111] = 2.29, p < .05) and for task negative statements (β = .20, t[111] = 2.39, p < .05). Specifically, focal children reported more learned helpless reactions to failure when they more frequently discounted their failure (e.g., "It doesn't matter") and when they more frequently evaluated the task negatively (e.g., "I hate these kinds of puzzles"). A significant statement main effect also emerged for help-seeking statements (β = -.81, t[111] = -2.90, p < .01) such that focal children who sought help relatively frequently from peers reported less of a learned helpless response to failure than focal children who sought help relatively infrequently. This effect was qualified, however, by a significant statement type × relative performance interaction (β = .67, t[111] = 2.57, p = .01). Follow-up analyses indicated that help seeking was associated with significantly lower learned helpless responses in the failure condition (β = -.23, t[55] = -2.98, p < .01) but not in the success condition (β = -.11, t[55] = -1.54, ns). Finally, a significant statement type by gender interaction emerged for negative performance statements (β = .24, t[111] = 3.37, p = .001). Analyses conducted separately by gender revealed that males who made relatively frequent negative performance statements (e.g., "I didn't get any of them") reported more learned helpless responses to failure (β = .27, t[37] = 2.92, ps < .01). In contrast, negative performance statements were related to significantly lower learned helpless responses to failure for females (β = -.17, t[73] = -2.68, p < .01).

On-task friend statements. For friend statements, significant statement type main effects emerged for help-giving statements (β = -.44, t[111] = -2.09, p < .05) and task negative statements (β = .14, t[111] = 2.16, p < .05). Regardless of gender or condition, focal children reported more adaptive [End Page 470] reactions to failure when their friends engaged in relatively high levels of help giving (e.g., "Let me show you how to do it") but fewer adaptive reactions to failure when their friends evaluated the task negatively relatively frequently (e.g., "I hated these puzzles").

Sequential Analyses

To further investigate the findings from our regression analyses, we used sequential data analysis techniques to examine children's on-task evaluative discourse in the natural sequence in which it occurred. To reiterate, we expected that by examining the types of statements that followed from focal child and friend statements, we might better understand why certain features of children's discourse (e.g., discounting) were associated with more maladaptive learned helpless responses to failure from postfailure to postdiscussion.

Sequential analysis begins with a simple frequency count of the number of times an antecedent event i (e.g., the focal child asks for help) occurs in sequence with a consequent event j (e.g., the friend provides help) at some lag, or interval. In the current study, lags of +1 were used. That is, we determined the frequency with which each antecedent event (i.e., the event at Lag 0) was immediately followed by each consequent event (i.e., the event at Lag +1).

Methods designed to test for the independence of antecedent and consequent events have been developed by Sackett (1979) and modified by Allison and Liker (1982) and by Bakeman and Quera (1995). We used Bakeman and Quera's method in the present study because, consistent with the procedures used in our study, Bakeman and Quera's method allows consecutive events to be assigned the same code (e.g., one negative self-evaluative statement could be followed by another) and allows for codes to be pooled across multiple children. As noted by Bakeman and Gottman (1997), pooling is often necessary when-as in the present study-multiple children are observed and relatively few instances of codable behavior are observed for each individual. The purpose of testing for independence is to determine whether one type of event follows another type of event more (or less) often than would be expected by chance. In the language of sequential analysis, the conditional probability (i.e., the probability that event j occurs given that event i has just occurred) is compared with the expected probability that such a sequence will occur based on the probability that each event occurs alone (i.e., based on the unconditional probabilities of the two events). The z statistic is compared to the standard normal distribution. If the null hypothesis of independence is rejected, then it can be said that i and [End Page 471] j are dependent; that is, event j follows event i more (or less) often than expected by chance.

Gender and relative performance differences in the level of dependency between statement types were also examined. In order to adjust for gender and relative performance differences in total verbal output, z-scores were transformed to phi coefficients, r_φ (see Bakeman & Gottman, 1997). Differences between phi coefficients were then examined using standard procedures for testing the difference between two independent correlation coefficients (see Glass & Hopkins, 1984; Hays, 1994). These procedures yield a z statistic (referred to here as z_diff) that can be compared to the standard normal distribution.

Analysis overview. In consideration of our goal to employ sequential analyses to gain insight into why children's evaluative discourse predicts their responses to failure from postfailure to postdiscussion and to minimize the possibility of Type 1 errors (see Bakeman & Gottman, 1997), we limited our sequential analyses to those statement types that emerged as significant predictors of children's failure responses in our regression analyses. For the statement type main effects that emerged for discounting and negative task statements made by the focal child and for help-giving and negative task statements made by friends, we were interested in determining which statements followed more (or less) often than expected by chance. For the statement type × relative performance interaction that emerged for focal children's help-seeking statements, we were interested in whether the prevalence of particular sequences differed for children in the friend-success and friend-failure conditions. For the statement type × gender interaction that emerged for focal children's negative performance statements, we were interested in whether the prevalence of particular sequences differed for males and females.

Focal child statements. Earlier regression analyses on focal child statements revealed two unqualified statement type main effects, one statement type by relative performance interaction, and one statement type by gender interaction. Sequential analyses examining each of these effects will be discussed in turn.

First, focal children who frequently discounted their failure (e.g., "I don't really care") reported more negative learned helpless responses to failure than did focal children who did so relatively infrequently. As shown in Table 3, sequential analyses examining these discounting statements revealed that focal children continued to discount their performance and that friends mirrored these discounting statements (e.g., "Yeah. It doesn't matter.") more often than expected by chance (zs > 11.95, ps < .001). Focal children followed up with a negative performance statement (e.g., "I didn't get any of [End Page 472] them"), and friends responded by providing help to the focal child (e.g., "You just need to . . .") less often than expected by chance (zs > -2.20, ps < .05).

Click for larger view
View full resolution

Table 3.

Results of Sequential Analyses Examining Statement Type Effects for Statements Made by Focal Children

Second, focal children who frequently evaluated the task negatively ("Those puzzles were stupid") responded more negatively to failure than did focal children who did so relatively infrequently. As shown in Table 3, sequential analyses revealed that focal children repeated negative task statements and friends mirrored these negative task statements more often than expected by chance (zs > 8.15, ps < .001). At the same time, focal children followed up by making negative performance statements (e.g., "I didn't get any"), and friends followed up by giving help to the focal child (e.g., "What you need to do is . . .") less often than expected by chance (zs > -2.84, ps < .01).

Third, focal children who sought help from friends relatively frequently responded more positively to failure than focal children who avoided help seeking, but only when friends also failed. As shown in Table 4, sequential analyses revealed that focal children were more likely to continue to seek [End Page 473]

Click for larger view
View full resolution

Table 4.

Results of Sequential Analyses Examining Statement Type × Relative Performance Interaction for Help-Seeking Statements Made by Focal Children

[End Page 474]

help from friends (e.g., "How did you do it?") when friends succeeded than when friends failed (z > 4.28, p < .001). Somewhat surprisingly, however, successful friends were less likely to respond to requests for help with assistance than were friends who also failed (z = -2.86, p < .01).

Finally, regression analyses revealed that negative performance statements (e.g., "I got zero right") predicted negative responses to failure for boys but positive responses to failure for girls. As shown in Table 5, sequential analyses indicated that boys' friends were significantly more likely than girls' friends to respond to these statements by asking about the focal child's performance (e.g., "You didn't get any of them?!") (z_diff = -2.29, p < .05) and marginally more likely to respond by evaluating their own performance positively (e.g., "I got them all") (z_diff = -1.74, p < .10). Girls' friends were, in contrast, more likely to mirror their friends' negative performance statements (e.g., "I didn't get any either") (z_diff = 3.77, p < .001).

Friend statements. Earlier regression analyses on friend statements revealed two statement type main effects. Sequential analyses examining these effects are discussed in turn.

First, focal children who received help relatively frequently from friends reported fewer maladaptive learned helpless responses to failure than did focal children who received help infrequently. As shown in Table 6, sequential analyses revealed that friends continued to give help to the focal child significantly more often than expected by chance and that focal children continued to seek help from friends significantly more often than expected (zs > 7.70, ps < .001). In contrast, focal children responded to help giving from friends by making negative performance statements (e.g., "I didn't get any of them") and task difficult statements (e.g., "They were so hard") less often than expected by chance (zs > -4.20, ps < .001).

Second, focal children whose friends relatively frequently evaluated the task negatively (e.g., "I hate doing puzzles") responded more negatively to failure. As shown in Table 6, sequential analyses showed that both focal children and friends continued to evaluate the task negatively more often than expected by chance (zs > 9.95, ps < .01). In contrast, focal children followed negative task statements with negative performance statements (e.g., "I only got one"), and friends followed negative statements by giving help to the focal children significantly less often than expected by chance (zs > -2.60, ps < .01).

Discussion

Our first goal was to better understand the nature of children's conversations with friends following failure. Not surprisingly, the content of friends' [End Page 475]

Click for larger view
View full resolution

Table 5.

Results of Sequential Analyses Examining Statement Type × Gender Interaction for Negative Performance Statements Made by Focal Children

[End Page 476]

Click for larger view
View full resolution

Table 6.

Results of Sequential Analyses Examining Statement Type Effects for Statements Made by Friends

discourse differed depending on their performance. For example, friends were more likely to state that they had performed well and help the focal child when they succeeded and were more likely to state that they had performed poorly and suggest the task was difficult when they failed. More interesting is that focal children's discourse varied according to their friends' performance. When friends succeeded, focal children were more likely to communicate that they had performed poorly, that they didn't like doing the task, and that they needed help. These findings support prior research indicating that children in elementary school use social comparison information to evaluate themselves and to guide their behaviors (Ruble et al., 1980) and are willing to express their weaknesses to friends in the service of receiving help (Parker & Asher, 1993).

Although boys' and girls' conversations were generally similar, some gender differences did emerge. First, male focal children asked about their friends' performance more than female focal children did. One reason for [End Page 477] this finding might be that male friends were less forthcoming about their performance and thus needed to be asked several times before they revealed how they had performed. Follow-up sequential analyses support this interpretation. Specifically, male focal children followed up a performance check with another performance check 15% of the time, a rate significantly higher than expected by chance (z = 2.93, p < .05). In contrast, female focal children followed up a performance check with another performance check only 4% of the time, a rate that did not differ from chance (z = .62, ns). This finding is consistent with prior research indicating that girls self-disclose more than boys (Lansford & Parker, 1999; McNelles & Connolly, 1999; Parker & Asher, 1993; Rose, 2002) and that boys may be more interested in maintaining their privacy (Rose & Asher, 2004). A significant gender difference also emerged in the frequency with which friends asked about the performance of the focal children. Here, male friends were more likely than female friends to inquire about the focal child's performance, but only in the success condition. One reason for this might be that in asking the focal child about their performance, male friends were hoping to be given the opportunity to talk about their own good performance. This finding supports the notion that boys are more likely than girls to want to present themselves in a positive light (Rose & Asher, 2004), while girls may be more likely to worry about hurting others (see Rose & Rudolph, 2006) and may be more sensitive to potential distress in others (see Rose, 2002).

Our next set of analyses focused on understanding how particular features of children's conversations might predict changes in responses to failure from postfailure to postdiscussion. We began by examining the consequences of engaging in high versus low levels of off-task talk (e.g., talking about a birthday party rather than the experimental task). As expected, focal children whose friends engaged in high levels of off-task talk reported more maladaptive learned helpless responses to failure than focal children whose friends engaged in low levels of off-task talk. This finding is consistent with evidence that engaged problem-focused coping strategies (e.g., help seeking) are associated with better adjustment than more disengaged emotion-focused coping strategies (e.g., cognitive avoidance) (Compas et al., 2001). Increasingly negative learned helpless responses to failure were also predicted when focal children or their friends made negative task statements ("Those puzzles were stupid") or when focal children discounted the failure by suggesting it was unimportant ("It doesn't matter"). One reason for these findings, as suggested by our sequential analyses, is that these types of negative/dismissive statements tended to lead to conversations that remained emotion-focused rather than problem-focused. Indeed, sequential analyses revealed that when focal children or friends evaluated the task [End Page 478] negatively or focal children discounted their failures, the most frequent consequence was a mirroring of these statements either by the focal child or the friend (e.g., "I hated those puzzles." "Uh-huh. Me too."). Moreover, in all three cases these statements led friends to give help to the focal child significantly less often than expected by chance.

This lack of help giving is important given that focal children benefited from receiving help from friends. That is, focal children who received relatively high levels of help from friends reported fewer learned helpless responses to failure than those who received less help. The results of help seeking were somewhat more complex. Consistent with the literature that children benefit from adaptive help seeking (i.e., seeking help when they need it), focal children who frequently asked for help reported fewer learned helpless responses to failure than focal children who asked for help relatively infrequently. Importantly, however, this effect was qualified by a significant relative performance interaction, indicating that the benefits of help seeking held only when friends had also failed. Follow-up sequential analyses provide some insight into this phenomenon, showing that children were less likely to receive help immediately after asking for it when friends succeeded than when friends failed. Children in the friend-success condition frequently had to ask for help again. It appears that focal children eventually received the help they needed (as indicated by the finding that, overall, children received more help in the success condition than in the failure condition), but help didn't come as quickly when friends succeeded as when friends failed. The reason may be that friends realized the potentially negative self-evaluative consequences of needing to ask for help (see Ryan, Pintrich, & Midgley, 2001) and held off providing it to help their friends save face. This approach may, however, have led to negative outcomes, leading children to feel less competent than if their friend had offered immediate assistance.

Only one gender difference emerged in predicting changes in children's responses to failure from the features of their conversations with friends. Although girls and boys did not differ in the degree to which they commented on their poor performance (e.g., "I didn't get any"), the consequences of making negative performance statements differed for boys and girls. For boys, these statements predicted fewer adaptive responses to failure. For girls, they predicted more adaptive responses.

Our sequential analyses again provide some insight into these findings. Boys' friends were significantly more likely than girls' friends to respond to negative performance statements (e.g., "I didn't get any") by checking on the focal child's performance, in effect asking them to repeat the negative performance statements (e.g., "How many did you miss?). Male friends [End Page 479] were also marginally more likely to respond by making positive performance statements (e.g., "Really? I got them all."). These types of responses likely highlighted the focal child's poor performance in ways that made them less than anxious to tackle additional puzzles. In contrast, girls were significantly more likely than boys to have friends respond by mirroring the negative performance statements. Thus, the social comparison information that girls received was far less damaging than that received by boys.

Still, it is not entirely clear why making negative performance statements-and having those statements mirrored by friends-would lead to more positive responses to failure. One way to make sense of these findings is in light of the corumination phenomenon recently identified by Rose (2002). Corumination is characterized by repeated discussion of the same problem and mutual encouragement of discussing the problem, as in the following conversation between two girls in the present study.

Focal child: How did you do?

Friend: Um . . . I only got one right out of all eight puzzles I did.

Focal Child: I didn't even get any done.

Friend: I was pretty close sometimes, but I didn't get it.

Focal child: Yeah, I think that my problem was, um, was sometimes I couldn't find the blocks that I wanted.

Friend: Yeah, sometimes I couldn't find . . . I had the same problem. Sometimes I couldn't find the blocks that I needed.

Focal child: Yeah. I couldn't get any of them.

Research by Rose and her colleagues shows that girls are more likely to coruminate than are boys and that corumination has its trade-offs (Rose, 2002; Rose et al., 2007). Specifically, Rose finds that corumination may lead girls to feel better about their friendships (because corumination is socially supportive) but also to worry more (because it leads girls to ruminate on their failures). In the case of the present study, the trade-offs may also involve immediate versus long-term outcomes. These types of coruminative interactions may boost girls' immediate confidence insofar as they feel socially supported by a friend. However, these types of interactions may not boost girls' confidence for long as girls come to face the reality that they've performed poorly and may have to face similar challenges alone in the future. In the end, then, good feelings about the friendship may stand the test of time, but negative feelings about the self may remain and may contribute to girls' greater tendency to engage in learned helpless behavior.

Future research will certainly be needed to replicate and further explore the findings reported here. In this research, several limitations of [End Page 480] the present study should be addressed. First, although the present study extends prior work by examining relations between children's discourse with friends and changes in children's responses to failure, the degree to which particular conversational features predict other outcomes-including children's behaviors and children's feelings about their friends-was not examined. It is likely that the consequences of children's statements may differ from outcome to outcome. For example, while off-task talk was associated with fewer adaptive responses to failure in the current study, this same conversational feature may lead children to evaluate their friendships more positively. Second, future research will be important in examining children's interactions with peers following failure in the classroom or in other achievement settings. Given that children interact with a variety of peers in the classroom setting-only some of whom may be friends-it is very possible that children's conversations will differ (e.g., children may be less willing to ask for help from nonfriends because of self-presentation concerns) and that the consequences of particular conversational features might differ as well (e.g., children who ask for help, particularly from non-friends, in a classroom setting may be judged harshly, leading to fewer positive achievement-related beliefs). Third, while the current study is important in demonstrating that changes in children's learned helpless responses to failure over time can be predicted by the features of their conversations with friends, potentially important third variables (e.g., personality or friendship characteristics) cannot be ruled out. Future research, perhaps directly manipulating feedback from friends, will be important in making causal links between the features of children's discourse and changes in their achievement-related beliefs.

Still, the present study contributes to the literature in several ways. First, it represents one of only a handful of studies to examine the processes by which social interactions between friends contribute to children's achievement-related behaviors and beliefs (see also Altermatt et al., 2002; Berndt et al., 1990; Sage & Kindermann, 1999). Second, this study answers calls emanating from achievement motivation, coping, and friendship literatures to supplement self-reports of children's coping behaviors and friendship experiences with observations of actual social interactions between close friends (Berndt & Hanna, 1995; Compas et al., 2001; Denton & Zarbatany, 1996; Rubin et al., 2006). Given evidence that unsuccessful coping with daily stressors can lead to negative and long-term interpersonal, psychological, and achievement-related outcomes (Folkman, 1984; Repetti, 1996; Rutter, 1979; Windle & Windle, 1996), the current study has important practical implications for improving the social and emotional well-being of children and for promoting school success, [End Page 481] suggesting in particular that friends can help children to cope effectively with failure by providing assistance and social support. Teachers may play an important role in this process by attending to the classroom climates they create. Prior research has shown, for example, that students are more likely to ask for help when teachers emphasize that learning from mistakes is an important part of the educational process and students feel that they are a part of a caring, supportive, and friendly social community (see Ryan, Gheen, & Midgley, 1998; Wigfield et al., 2006).