Athletic Identity in Marathon Runners: Functional Focus or Dysfunctional Commitment?

Horton, Robert S. ; Mack, Diane E.

Previous research on athletic identity (Brewer Van Raalte, & Linder, 1993; AI) suggests that strong Al may force an athlete to neglect other aspects of life in order to fulfill the athlete role. This project assessed the effect of Al on life priorities and athletic experiences. Two hundred thirty-six runners completed a questionnaire assessing demographic information, AI, life priorities, commitment to sport, sport performance, and psychological, physical, and social consequences of marathon training. Bivariate and extreme groups analyses investigated the relationship between scores on the Athletic Identity Measurement Scale (Brewer, Van Raalte, & Linder, 1993) and each of the variables mentioned above. There was no evidence that runners with high AI were neglecting other aspects of life in order to fulfill the role of an athlete. Relative to low AI, high AI was associated with better athletic performance, more commitment to running, expanded social network, and relatively more frequent experience of both pos itive and negative effects of marathon training. The relevance of age of athlete in the assessment of AI is discussed.

The prevailing view of the self-concept is that it is a multidimensional structure that includes all of a person's thoughts and feelings about the self within various aspects of life (Carver, Reynolds, & Scheier, 1994; Higgins, 1987; Linville, 1985; 1987; Showers, 1992). The multidimensional nature of the self-concept allows people to activate different dimensions of the self at different times, and behavior and information processing will tend to vary depending upon the dimension of the self that is active in a given situation (Markus & Nurius, 1987). For instance, when the athlete role is activated during intense training, loud laughter may be interpreted as disruptive. On the other hand, when the social dimension is activated during a night out with friends, the same person would interpret this laughter as pleasing.

Given that people can move among many dimensions of the self, researchers have become interested in the relative salience of some common dimensions of the self-concept (Markus, 1977; Markus & Zajonc, 1985). According to Stryker (1978), identity salience can be conceptualized as the probability that a given identity will be activated in a given situation. For example, the extent to which athletics is an important part of a person's self-concept will determine how likely it is that thoughts and behaviors associated with the athlete role (e.g., motivation, competition, social relationships, performance) will be expressed in, or used to interpret, a given situation.

Recent research has confirmed the importance of the athlete role in the way people define themselves. Brewer, Van Raalte, and Linder (1993) have shown that athletic identity (i.e., the extent to which a person identifies with the athlete role) is a unique and important dimension of the self-concept that can be regarded as both a cognitive structure (a schema) and a social role. As a cognitive structure, AI provides a framework for interpreting information, determines how an athlete copes with career-threatening situations, and inspires behavior consistent with the athlete role. As a social role, AI may be determined by the perceptions of those close to the athlete. Often, an individual whose friends, family members, or coaches emphasize the athletic dimension of the individual will internalize the perceptions of these important people and will define the self as others define him or her: as an athlete (Mead, 1934). In addition, people strong in AI may surround themselves with other athletes who encourage a s elf-definition centered on athletics.

The strength of AI in a person's self-concept varies with past and current athletic experience and relative success or failure within the athletic domain. However, AI is an important part of the self-concept in athletes and non-athletes alike (Brewer et al., 1993; Cornelius, 1995; Murphy, Petitpas, & Brewer, 1996; Perna, Zaichowsky, & Bocknek, 1996). Though individuals currently participating in organized sports manifest higher levels of AI than non-participants (Brewer et al., 1993), AI is an important dimension in the self-concepts of most individuals. Further, having a self-concept that includes a dimension focused on physical performance and appearance has been found to be related to physical fitness (Marsh, 1993), athletic performance (Porat, Lufi, & Tenenbaum, 1989), personality traits such as extroversion and masculinity (Colley, Roberts, & Chipps, 1985), and global self-esteem (Marsh, Perry, Horsely, & Roche, 1995). AI has been empirically linked to health and fitness benefits (Brewer et al., 1993), increased participation in physical activity and exercise (Anderson & Cychosz, 1990; Fox & Corbin, 1986; Kendzierski, 1988), and increased social relationships and confidence (Petitpas, 1978). Further, Stryker and Serpe (1994) argue that increased salience of a dimension of the self will increase commitment to thoughts and behaviors consistent with that dimension. Thus, strong AT should increase commitment to athletic participation.

While there are potential benefits to a strong Al, detrimental effects of Al have also been demonstrated. The problems linked to strong AI arise when there is a commitment to the role of the athlete at the expense of other aspects of life. Overcommitment to the athlete role may lead to two types of problems. First, overcommitment to the athlete role may lead to dysfunctional practices within the athlete role: over training, anxiety when not training, or in extreme cases, the use of performance enhancing drugs (Coen & Ogles, 1993; Hughes & Coakley, 1991). Second, overcommitment to the athlete role can restrict the development of a multidimensional self-concept. Linville's (1987) research suggests that high self-complexity (i.e., maintaining a self-concept that includes numerous, independent dimensions) protects the self-concept in the event of failure in any one dimension. Eldridge (1983) argues that the importance placed on the athlete role may conflict with other roles and activities. Consistent with this r easoning, the detriments linked to strong Al include depression and lack of effective adjustment upon athletic career termination (Baillie & Danish, 1992; Blinde & Greendorfer, 1985; Brewer, 1993; Petitpas, 1978), poor physical and emotional health, social isolation (Brewer et al., 1993; Hughes & Coakley, 1991), and career immaturity (Murphy et al., 1996). Having identified the potential detriments of Al, it is important to note that researchers have confirmed that Al does not necessarily lead to dysfunctional commitment (Brown & Hartley, 1998, Cornelius, 1995).

The current investigation builds on previous research in a number of important ways. The primary objective of the project was to assess the relative importance of different life roles (hereafter referred to as "relative role importance") in the self-concepts of a unique population of athletes: adult marathon runners. Many of the studies that have investigated Al have sampled college athletes (e.g. Murphy et. al., 1996). It is possible that the negative consequences associated with Al are more a result of the age of participants sampled than they are of the athletic dimension under investigation. The relative role importance for these adult runners was assessed in order to provide a direct measure of the extent to which Al dominates the self-concept to the exclusion of other life roles. Previous research has not assessed the relative importance of AI as compared to other life roles. Marathon runners should assign the athlete role high relative importance within the self-concept. Previous research suggests tha t marathon runners who are high in athletic identity should assign high relative importance to the athlete role and low relative importance to non-athlete roles, such as family, friendship, and romantic partner roles (Baillie & Danish, 1992; Blinde & Greendorfer, 1985; Brewer, 1993).

A second objective of the project was to assess the relationship between AI and a number of different social (social network, feelings of social isolation), behavioral (e.g., sleep disturbance, illness), and psychological (e.g., body image, self-confidence, anxiety level) consequences of running. Previous research on AI has assessed isolated effects of commitment to the athlete role. This project investigates a diverse collection of social, physical, and psychological variables in order to assess more completely the effects of AI. The importance one assigns to a task and one's experiences during that task tend to be positively related (Lewicki, 1984; Steele, 1988). Thus, strong AI should be associated with greater experience of positive social, physical, and psychological consequences of marathon training (i.e., expanded social network, greater commitment to running) than should weak AI. Further, strong AI should be associated with fewer experiences of negative training consequences (i.e., sleep disturbances , appetite loss, etc.).

Among the psychological variables we assessed, commitment to sport is one that has received a great deal of attention (Carpenter, Scanlan, Simons, & Lobel, 1993; Scanlan et al., 1993). Previous research implies that salience of the AI can lead to increased commitment to athletics (Stryker & Serpe, 1994), yet this relation has yet to be directly tested. Commitment to running and different psychological components of commitment to running (enjoyment, involvement, social constraints, and investment) were assessed. Following from the work of Stryker and Serpe (1994), strong AI should lead to greater commitment to sport and greater experience of each of the components of commitment than does weak AI.

A final objective of this project was to assess the effects of AI on sport performance. Danish (1983) asserts that the training and competition involved in successful athletic performance may necessitate a unitary conception of the self. This argument implies that AI is related to an athlete's performance, yet this relationship has not been empirically tested. This investigation included just such an empirical test with the expectation that strong AI would be associated with better athletic performance than would weak AI.

Method

Participants

Two hundred thirty-six marathon runners (n = 176 males and n = 60 females) participated in the current study. Participants ranged in age from 19 to 72 years (M = 40.81; SD = 10.23) and reported consistently running for an average of 11.2 years (SD = 7.93). Participants had completed anywhere from 1 to 127 marathons (M = 10.2, SD = 16.43; median = 4.0) with personal best times ranging from 2 hours 29 minutes to 5 hours 47 minutes (M = 3:43; SD 37 minutes).

Materials

Demographic Questionnaire. Each participant completed a demographic questionnaire which included items assessing gender, age, running and athletic history, and social network. Two items were included to investigate the social network of the participants. Participants indicated what proportion of their good friends were runners and whether they had ever had a relationship end as a direct or indirect result of running.

Athletic Identity Measurement Scale (AIMS). The AIMS is a ten-item scale designed to assess the degree to which a participant identifies him or herself as an athlete. The items on the AIMS were modified slightly in order to assess identification with the role of "runner" as opposed to the general role of "athlete." For example, item #1 was changed from "I consider myself an athlete" to "I consider myself a runner." Participants indicated on a seven-point Likert scale the degree to which they agreed with each item. The scale had endpoints "strongly disagree" (corresponding to a rating of 1) and "strongly agree" (corresponding to a rating of 7). High scores on the AIMS indicate stronger identification with the athlete role than do low scores. The internal consistency of the modified scale ([alpha] = .86) was comparable to the internal consistency ([alpha] = .93) obtained by Brewer and colleagues (1993) for the unmodified version of the AIMS.

Life roles inventory The life roles inventory was adapted from the work of Stryker and Serpe (1994) who assessed the psychological centrality of six different life roles (academic, athletic, family, personal involvement (friendship), dating, and extracurricular) for college students. This inventory evaluated psychological centrality of each role through the use of pairwise comparisons. Each role was compared to every other role individually. Participants indicated which role in each pair was more important to how they viewed themselves. Each time a role was rated more important than another role, the preferred role received a "1" and the non-preferred role received a "0." The scores were then summed for each role and a "I" was added to the final sum. Thus, the range of possible scores for each role was 1-6. A score of six indicated that the role was rated as more important than was each of the other five roles. A score of one indicated that the role was rated as less important than was each of the other role s. Stryker and Serpe (1994) found satisfactory internal consistency for the assessment of individual life roles (Cronbach's alphas range from .81 to .89).

The labels for three of the roles were altered slightly co clarify the roles or to apply them, to the population under consideration. "Personal involvement" was changed to "Friendship," "Dating" became "Romantic partner," and "Academic" was modified to "Academic/Occupational."

Training effects assessment. In order to assess various positive and negative effects of marathon training, participants were given a list of 17 possible effects of marathon training. These items were compiled from various studies of elite athletes, marathon runners, and the effects of marathon training (Coen & Ogles, 1993; Hackney, Pearman, & Nowacki, 1990; Ziegler, 1991). Participants were asked to indicate which effects they experienced and the frequency of those effects. Frequency of each effect was rated on a 7-point Likert-type scale anchored at the extremes by 1 (not at all) to 7 (all the time).

The item "improved sense of identity" was included as a measure of the effect of marathon training on the self-concept. The item "increased social network" was included to further investigate the social impact of marathon training. The remaining fifteen items were subjected to an exploratory factor analysis using principal components analysis and direct oblimin rotation. This analysis resulted in a five-factor structure that accounted for 61.8% of the variance. All items loaded onto their respective factors above .49, which is classified as acceptable for item retention (Comrey & Lee, 1992). The solution also yielded a simple structure with each question loading onto only one factor (Table 1). Two of the factors included only two items and thus, were not interpretable (Clark & Watson, 1995; Comrey & Lee, 1992). The three factors that remained were interpreted as (a) a social consequences factor, (b) a positive consequences factor, and (c) a general negative consequences factor. These three factors accounted for 46.7% of the variance.

Cronbach's alphas on these three factors were modest but satisfactory given the few items that comprised each factor (social = .70, positive .77, and general negative = .56). Composite scores for these three factors were computed by averaging across the items that loaded onto each factor. These composite scores were used in subsequent bivariate and extreme groups analyses.

The Sport Commitment Scale (SCS). The SCS (Scanlan, Simons, Carpenter, Schmidt, & Keeler, 1993) assesses an athlete's desire and resolve to continue participating in a sport. This model effectively predicts youth athlete's commitment to sport (Scanlan, et al., 1993) and is based on the Investment Model developed and espoused by Rusbult (1980) to explain commitment in relationships. The items used to assess each component of the model show satisfactory internal consistency and discriminant validity (cf. Scanlan et al, 1993).

The SCS used in the current research included items assessing commitment to running (n = 4), enjoyment of running (n = 4; a positive affective response to sport participation), personal investment in running (n =2; resources put into an activity that can not be recovered upon termination of participation in the sport), social constraints to continue running (n = 3; social expectations and norms that create feelings of obligation to continue participation in a sport), and involvement opportunities in running (n = 3; valued opportunities that are present only through continued involvement). Each item was rated on a 1 (not at all) to 7 (very) Likerttype scale.

Reliability analyses on the items assessing each construct revealed satisfactory internal consistency in all cases. Internal consistency ranged from moderate for the involvement opportunities subscale (alpha = .62) to high for the enjoyment subscale (alpha = .96). Total scores for commitment, enjoyment, investment, social constraints, and involvement opportunities were computed by averaging the items assessing each construct.

Procedure

A total of five hundred and forty questionnaires were released to potential participants through two mediums. Questionnaires were placed in the race packets of five hundred runners entered in a marathon. Forty questionnaires were placed at a local running store to be distributed to members of a track club who had recently (within the past year) completed a marathon. Each questionnaire included a stamped return envelope so questionnaires could be returned. Two hundred and thirty-six questionnaires were returned, giving a response rate of 44%.

Results

Athletic Identity

The mean score on the AIMS within the entire 236 runner sample was 40.92 (SD = 9.27). Scores on the ALMS were divided into extreme groups with cut-points at the 33rd (corresponding to an AI score of 37) and 67th (corresponding to an AI score of 44) percentiles. Participants with an AIMS score below the 33rd percentile were classified as "Low AI" participants (M = 30.97, SD = 4.77, n = 79). Participants with an AIMS score above the 67th percentile were classified as "high AI" participants (M = 51.09, SD = 5.28, n = 79). The AIMS scores of these extreme groups were significantly different than one another, F(1,156) = 630.76, p [less than].001. These groups were then compared on variables related to importance of life roles, training effects, commitment to running, and running performance

Life Roles

An analysis of variance (ANOVA) was conducted with level of AL (high/low) as the independent variable and relative importance of each role (e.g., athlete, family, friendship, romantic partner, academic/occupational, and extracurricular) as the dependent variable (See Table 2). A Bonferroni adjustment was calculated to account for multiple analyses. Consequently, a significance level of .0083 for each analysis was employed. The analysis of variance on the athlete role revealed that high AI participants rated the role as significantly more important, relative to other roles, than did low AI participants (M = 3.49 and 2.54, respectively; F(1,149) = 28.19, p [less than] .001). High AI and low AI participants did not differ in their ratings of the relative importance of the other five roles: family, romantic partner, academic/occupational, friendship, and extracurricular (p[greater than].05).

Social Network

In order to assess the effects of marathon running on an individual's social network, participants responded to two items. One item assessed the degree to which their social network had expanded due to marathon training and the other assessed the percentage of their good friends who were runners. High AI participants reported an expanded social network as a result of running more frequently than did low AI participants, F( 1, 156) = 16.70, p [less than].001. In addition, results of a bivariate analysis across the full sample demonstrated a positive correlation between AI and the proportion of good friends identified as runners (r = .45, p [less than].001). Finally, an extreme groups ANOVA revealed a significant difference between the percentage of good friends identified by high (M = .50, SD = .31) and low AI (M = .19, SD = .21) participants as runners, F( 1, 154) = 49.64, p [less than].001.

Training Effects Assessment

The training effects assessment was broken down into three components via factor analysis (described previously). Bivariate analyses and extreme groups analyses were conducted on each of the three factors. A Bonferroni adjustment accounted for the multiple (three) analyses of each type and resulted in an alpha of .017.

Social consequences. A bivariate analysis between the composite social score and AIMS score revealed that AI was associated with social consequences of training (r = .32, p [less than] .001). Further, high AI runners scored higher on this factor (M = 2.92) than did low AI runners (M = 2.04), t(156) = -4.38, p [less than] .001. Remembering that the social consequences factor included negative social consequences of running (i.e., isolation from non-running friends; decreased time with family/friends), these data indicate that runners with strong AI were experiencing negative social consequences of training more frequently than were runners with weak AI.

Positive consequences. This factor included items assessing body image, self-confidence, overall self-image, energy level, and anxiety level. The composite score for these items was significantly associated with AI (r = .31, p [less than] .001). Extreme groups analysis was consistent with this result as high AI runners experienced positive effects of training more often (M = 5.35) than did runners with weak AI (M = 4.72), t(156) = -3.33, p [less than] .002.

General negative consequences. This factor included items assessing financial difficulties, increased susceptibility to illness, and decreased occupational performance. The composite score for these items was significantly associated with AI(r = .l6, p [less than] .05). In addition, high AI participants (M = 1.83) experienced these negative consequences more frequently than did low AI participants (M = 1.50), t(156) = -2.52, p [less than] .017.

Commitment to Running

In addition to the training effects described above, we assessed different components of psychological commitment to running using the Sport Commitment Scale (Scanlan, et. al., 1993). Bivariate and extreme groups analyses were conducted to examine the relationship between AI and each component. Using the entire sample, significant correlations were found between AI and the five subscales of the Sport Commitment Model. Results of these correlational analyses indicated that commitment to running (r = .47, p [less than] .001), enjoyment of running (r= .39, p [less than].001), investment in running (r= .47, p [less than].001), involvement opportunities in running (r = .46, p [less than] .001), and perceived social constraints to continue running (r .35, p [less than] .001) increase with AI.

Extreme groups analyses were conducted on the five subscales with a Bonferroni adjustment to account for multiple analyses. Consequently, a significance level of .01 for each of the five analyses was employed. These analyses revealed significant differences between levels of AI on all five variables. Consistent with predictions, high AI participants expressed greater commitment to running, F( 1, 155) = 37.72, p [less than]001, greater enjoyment of running, F( 1, 155) = 21.44,p [less than].001, greater investment in running, F( 1, 155) = 32.87p [less than].001, greater involvement in running opportunities, F(l, 155) = 51.00,p [less than].001, and greater perceived social constraints to continue running, F( F(1, 155) = 24.02, p [less than].001, than did participants with low AI (see Table 4).

Athletic identity and Athletic Performance

The relationship between AI and participants' reported personal best time was investigated with the expectation that high AI participants would report faster personal best times than would low AI participants (See Table 5). This prediction was tested with two analyses. First, using the entire sample, a bivariate correlation between participants' personal best time in a marathon and their score on the AIMS was calculated. This correlation proved significant (r = -.17, p [less than] .05), indicating that stronger AI was associated with faster personal best times. Extreme groups analysis found no significant difference between high AI and low AI runners with respect to personal best time. This was a surprising finding, so post-hoc investigation was warranted.

An ANOVA was conducted with performance serving as the independent variable and each participant's AIMS score serving as the dependent variable. Each participant's personal best time in a marathon was coded in 30-minute increments on a 1-6 point scale with higher numbers representing slower times (see Table 5). Results revealed a significant difference between levels of personal best time on AI, F(6, 217) = 2.48, p [less than].03. Least Significant Difference analyses revealed that group 1 runners had significantly higher AI than did group 3, 4, and 5 runners. Also, group 2 runners had significantly higher AI than did group 4 runners.

Taken together, the bivariate and ANOVA results suggest that the faster the personal best time in a marathon, the stronger the AI will be. Interestingly, the groups that manifested the strongest AI were those with the fastest and the slowest personal best times (Mean AI = 44.29, 43.88, respectively). This result is discussed further below.

Discussion

The purpose of the current study was to investigate how AI affects (a) the relative importance for marathon runners of different life roles and (b) social, psychological, and behavioral consequences of marathon training. Previous research (e.g., Murphy et al., 1996) has suggested that strong AI is detrimental to athletes' development in other aspects of life. It has been speculated that AI dominates the self-concept and leads to neglect of other life roles. In the current study, participants with high AI rated the athlete role as relatively more important than did those with low AI. However, there were no significant differences between the two groups on the relative importance ratings of other life roles (e.g., family, romantic partner, etc.). Consequently, there was no evidence that runners with strong AI were neglecting other aspects of life in order to fulfill the needs of the athlete role. AI is independent of other aspects of the self and is more salient for some runners than it is for others. This does not mean that runners who have salient AI must decrease the importance or salience of other life roles. Strong AI does not preclude the development of a multidimensional self-concept.

In addition to challenging the idea that AI dominates the self-concept to the exclusion of other life roles, the current investigation observed a number of positive consequences of AI. AI was associated with greater experience of positive psychological consequences of training such as enhanced body image, increased self-confidence, and decreased anxiety. In addition, high levels of AI were associated with greater enjoyment of running and greater overall commitment to running than were lower levels of AI. On the other hand, AI was also associated with certain negative consequences of training. However, overall, the data stand in stark contrast to previous AI research in the suggestion that strong AI can benefit the psychological and physical experience of athletic training.

The effect of AI on runners' social lives revealed ambiguous data and deserves some attention here. AI was directly related to expanded social network and proportion of friends who were runners. However, AI was also associated with negative social consequences such as increased social isolation, decreased time with family and/or non-running friends, and decreased social activity. It appears that runners with strong AI form new relationships with runners and thus, expand their overall social network. However, the social network that these runners had prior to running or outside of running suffers. Running takes the place of social activity and limits the time one can spend with friends not involved in running.

The effects of AI on training and relative role importance are very different from those observed in other studies. One way to account for these differences is to look at the age of participants. The mean age of participants in the current study was almost 41 years old. Much of the previous research on AI has focused on college athletes (e.g., Murphy et al., 1996). One might predict that college athletes would manifest stronger AI than would older athletes and thus, would experience more negative effects of AI than would older athletes. The current investigation was unable to directly compare the AI of college-age athletes with that of adult athletes, for only eight participants of our 236 participants were of college age. However, comparing across studies, the magnitude of AI in the "high AI" runners in this study was comparable to the levels of AI manifested by intercollegiate athletes sampled in other studies (see Brewer, Van Raalte, & Linder, 1993, study 2). Further, the range of scores in the current st udy was comparable to the range observed in a sample of college athletes of different abilities (see Brewer, Van Raalte, & Linder, 1993, study 1). So, if the age of athlete does not predict differences in the magnitude of AI, how is it that age affects Al and its consequences?

The structure of the self-concept can affect behavior and thought independently of the content of the self-concept (Showers, 1992; Linville, 1985; DeSteno & Salovey, 1997). College and adult athletes may exhibit the same levels of Al; however, the way AI is organized in relation to other dimensions of the self may be very different in the two groups. Specifically, college and adult athletes may differ in the complexity of the self-concept: the number and independence of important dimensions of the self-concept (Linville, 1985; 1987).

In practical terms, college athletes may have fewer life roles that they regard as important to the self. Few college athletes concern themselves with concerns such as establishing a career or raising a family Lacking obligations to these, and other, life roles, the athletic dimension will comprise a greater percentage of the overall self-concept. Adult athletes may necessarily have more diverse self-concepts than do college athletes as a result of the numerous non-athletic obligations that accompany adult life. In the self-concepts of adult athletes, Al must compete with many other life dimensions to exert influence on behavior and thought.

With reference to the independence of self-concept dimensions, college athletes may regard the athletic dimension of the self as the basis for other dimensions of the self. For instance, some college athletes may regard athletics as a potential career, or a college athlete may believe that his or her friends and/or romantic partner are dependent upon his or her athletic prowess. If an athlete feels that his or her athletic dimension is the key to other dimensions of the self, this dimension will be highly important and may come to dominate the self-concept in a dysfunctional manner. In adult, non-professional athletes, like those in this study, the athletic dimension is less likely to be used as a basis for other dimensions of the self-concept. In these older athletes, self-concept dimensions may be more independent of one another out of necessity and experience. The current research used a nomothetic assessment of relative role importance and thus, was unable to directly assess self-complexity. Using an idi ographic approach to assess the self-concept would allow investigation of the diversity and independence of self-concept dimensions (see Linville, 1987).

In addition to important findings regarding relative role importance and athletic training, the current investigation makes a unique contribution by investigating the relation between AI and athletic performance. AI was correlated with personal best time in a marathon such that participants with faster personal best times manifested higher levels of AI. This correlation was weak; however, the result is consistent with Danish (1983) who hypothesized that strong AI provides the motivation and discipline necessary for intense training and success in high-level athletics. Until now, this intuitive relationship lacked empirical support.

The tenuous nature of the AI-performance correlation is best understood by looking at the mean AIMS scores for runners of different abilities. The analysis of variance using performance group as the independent variable revealed that the runners with the strongest AI were those that had either the fastest or the slowest personal best times. This result has theoretical basis. It is well established that traits and abilities that are seen as descriptive are regarded as more important to possess than are those that are non-descriptive (Pelham & Swann, 1989). Success in a self-concept dimension will lead to increased importance of that dimension, whereas failure in a self-concept dimension will lead to deflated importance of that dimension (Lewicki, 1984; Steele, 1988). Indeed, it is not surprising that runners who are successful (i.e., have fast personal best times in a marathon) exhibit strong AI. Given this reasoning, would one expect runners who run marathons "slowly" to decrease the relative importance of t he athletic role?

Anyone who has run a marathon knows that running a "slow" marathon is a far cry from "failure," so the expectation of deflated importance of the athlete role has no merit. This does not explain, however, why runners who do not run fast times would exhibit particularly strong AI. Speaking to this issue, research on cognitive dissonance suggests that the evaluation of a task is directly related to the amount of effort one must expend to accomplish the task (e.g., Aronson & Mills, 1959). Expending a great amount of effort will increase the attractiveness of a goal and can change the self-concept if salient external justification for effort is not available. "Slow" marathon runners win no medals. They break no records. Justification for the long hours of training and the fatigue must come from within. These runners develop strong AI to justify the effort that they expend in pursuit of the marathon finish line.

It is true that AI is associated with a number of psychological, behavioral, social, training, and performance consequences; however, it is important to acknowledge that the relationships observed in this study may be bi-directional. Just as AI may be a cause of the consequences observed, it may be a result of these factors as well. The methodology employed in the current investigation does not establish causal direction within the observed relationships; however, future research may address the issue of causation by using experimental methods to examine the effects of AI.

AI has been regarded as an important, yet potentially damaging, aspect of an athlete's self-concept. This study challenges that notion. AI certainly does influence multiple aspects of an athlete's life. However, athletic identity does not, necessarily, lead to stunted development in other life roles. Identification with the role of an athlete is similar to other dimensions of the self in that, identification with this role to the exclusion of other roles will be detrimental to an individual's life functioning. We are not suggesting that AI never results in negative consequences. However, it is clear from this project that strong AI can benefit athletic performance and lead to positive psychological and physical experience of athletic training. In addition, AI can diversify the self-image, thus providing a buffer against depression (Linville, 1987) and boosting general self-esteem (Marsh, et a]., 1995).

One question we have yet to address is the extent to which our findings generalize beyond our sample of 236 runners. Forty-four percent of runners who received questionnaires returned it. This return rate is similar to other questionnaire studies (Neumon, 1997), yet it does raise the question of self-selection. The runners who completed the questionnaire may be qualitatively different than those who did not return the questionnaire. However, the sample included a diverse group of runners who varied highly in marathon experience, marathon performance, motivation, life experience, and AI. This diversity suggests that this sample of runners may be representative of a larger population of runners.

A different question of external validity is the extent to which these data may be generalized from marathon runners to athletes in different sports. Marathon training may be a particularly unique activity in that, unlike sports in which training prepares an athlete for a season, marathon training is often focused on one event. However, the challenge of marathon training lies not in the event, but in maintaining mental and physical fortitude during months of accumulating training fatigue. In this way, marathon training is similar to training in all sports. Thus, there is no reason to think that AI inspired by marathon training will be qualitatively different than that inspired by different sports. The objective of the project was not to differentiate marathon training from other sports. On the contrary, the objective was to assess potential benefits and detriments of AI for training and performance in all sports and to refute the findings that Al is an exclusionary dimension of the self-concept.

Robert S. Horton, Department of Psychology, University of North Carolina at Chapel Hill. Chapel Hill, North Carolina, U.S.A.: Diane E. Mack, Department of Physical Education, Brock University, St. Catherines, Ontario, Canada. The authors would like to acknowledge the contributions of Michelle Ritter-Taylor to the conceptual and operational development of this project.

Notes

(1.) The data set was analyzed with and without 46 runners who may he classified as exercise addicts. The different analyses revealed exactly similar results on all variables.

Though the questionnaire did not include a measure of addiction to running, marathon number and marathon frequency offered rough indicators of potential addiction. Runners who had run more than 8 marathons AND had run one or more marathons per year since they began running were classified as exercise addicts.

References

Anderson, D. F., & Cychosz, C. M. (1990). Prediction of participation in exercise among employees of a medical center from their perceived susceptibility benefits, barriers, social influences, cues to action, physical self-efficacy, and exercise identity. Paper presented at the annual meeting of the North American Society for the Psychology of Sport and Physical Activity, Houston, TX, USA.

Aronson, E., & Mills, J. (1959). The effect of severity of initiation on liking for a group. Journal of Abnormal and Social Psychology, 59, 177-181.

Baillie, P. H. E., & Danish, S.J. (1992). Understanding career transition in athletes. The Sport Psychologist, 6, 77-98.

Blinde, E. M., & Greendorfer, S. L. (1985). A reconceptualization of the process of leaving the role of competitive athlete. International Review for the Sociology of Sport, 20, 87-93.

Brewer, B. W. (1993). Self-identity and specific vulnerability to depressed mood. Journal of Personality, 61, 343-364.

Brewer, B. W., Van Raalte, J. L., & Linder, D. E. (1993). Athletic Identity: Hercules' muscles or Achilles heel? International Journal of Sport Psychology 24, 237-254.

Brown, C. & Hartley, D. L. (1998). Athletic identity and career maturity of male college student athletes. International Journal of Sport Psychology, 29, 17-26.

Carpenter, P. J., Scanlan, T. K., Simons, J. P., Lobel, M. (1993). A test of the sport commitment model using structural equation modeling. Journal of Sport and Exercise Psychology, 15, 119-133.

Carver, C. S., Reynolds, S. L., Scheier, M. F. (1994). The possible selves of optimists and pessimists. Journal of Research in Personality 28, 133-141.

Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in objective scale development. Psychological Assessment, 7, 309-319.

Coen, S. P., & Ogles, B. M. (1993). Psychological characteristics of the obligatory runner: A critical examination of the anorexia analogue hypothesis. Journal of Sport and Exercise Psychology, 15, 338-354.

Colley, A., Roberts, N., & Chipps, A. (1985). Sex-role identity, personality and participation in team and individual sports by males and females. International Journal of Sport Psychology, 16, 103-112.

Comrey, A. L. & Lee, H. B. (1992). A first course in factor analysis. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Cornelius, A. (1995). The relationship between athletic identity, peer and faculty socialization, and college student development. Journal of College Student Development, 36, 560-573.

Danish, S. J. (1983). Musing about personal competence: The contributions of sport, health, and fitness. American Journal of Community Psychology 11, 221-240.

DeSteno, D. A. & Salovey, P. (1997). The effects of mood on the structure of the self-concept. Cognition & Emotion, 11, 351-372.

Eldridge, W. D. (1983). The importance of psychotherapy for athletic related orthopedic injuries among adults. International Journal of Sport Psychology, 14, 203-211.

Fox, K. R., & Corbin , C. B. (1986). An extension to a model of physical involvement: A preliminary investigation into the role of perceived importance of physical abilities. In J. Watkins, T. Reilly, & L. Burwitz (Eds.), Sports science (pp. 223-228). London: Spon.

Hackney, A. C., Pearman, S. N., & Nowacki, J. M. (1990). Physiological profiles of over-trained and stale athletes: A review. Journal of Applied Sport Psychology, 2, 21-33.

Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340.

Hughes, R., & Coakley, J. (1991). Positive Deviance Among Athletes: The implications of overconformity to the sport ethic. Sociology of Sport Journal, 8, 307-325.

Kendzierski, D. (1988). Self-schemata and exercise. Basic and Applied Social Psychology, 9, 45-59.

Lewicki, P. (1983). Self-image bias in person perception. Journal of Personality and Social Psychology, 45, 384-393.

Lewicki, P. (1984). Self-schema and social information processing. Journal of Personality and Social Psychology, 47, 1177-1190.

Linville, P. W. (1985). Self-complexity and affective extremity: Don't put all of your eggs in one cognitive basket. Social Cognition, 3, 94-120.

Linville, P. W. (1987). Self-complexity: A cognitive buffer against stress-related illness and depression. Journal of Personality and Social Psychology, 52, 663-676.

Marsh, H. W. (1993). Physical fitness self-concept: Relations of physical fitness to field and technical indicators for boy and girls aged 9-25. Journal of Sport and Exercise Psychology, 15, 184-206.

Marsh, H. W., Perry, C., Horsely, C., & Roche, L. (1995). Multidimensional self-concepts of elite athletes: How do they differ from the general population? Journal of Sport and Exercise Psychology, 17, 70-83.

Markus, H., (1977). Self-schemata and processing information about the self. Journal of Personality and Social Psychology, 35, 63-78.

Markus, H., & Nurius, P. (1986). Possible selves. American Psychologist, 41, 954-969.

Markus, H., & Zajonc, R. B. (1985). The Cognitive Perspective in Social Psychology. In The Handbook of Social Psychology, Vol. 1, Lindsey, G. & Aronson, E. (Eds.), pp. 137-230, New York: Random House.

Mead, G. H. (1934). Mind, Self, and Society. Chicago: University of Chicago Press.

Murphy, G. M., Petitpas, A. J., & Brewer, B. W. (1996). Identity foreclosure, athletic identity, and career maturity in intercollegiate athletes. Sports Psychologist, 10, 239-246.

Neumon, W. L. (1997). Social research methods: Qualitative and quantitative approaches (3rd ed.) Allyn and Bacon: Needham Heights, MA.

Pelham, B. W. & Swann, W. B., Jr. (1989). From self-conceptions to self-worth: On the sources and structure of global self-esteem. Journal of Personality and Social Psychology, 57, 672-680.

Perna, F.M., Zaichkowsky, L., Bocknek, G. (1996). The association of mentoring with psychosocial development among male athletes at termination of college career. Journal of Applied Sport Psychology, 8, 76-88.

Petitpas, A. J. (1978). Identity foreclosure: A unique challenge. Personnel and Guidance Journal, 56, 558-561.

Porat, Y., Lufi, D., Tenebaum, G. (1989). Psychological components contribute to select young female gymnasts. International Journal of Sport Psychology, 20, 279-286.

Rusbult, C.E. (1980). Commitment and satisfaction in romantic associations: A test of the investment model. Journal of Experimental Social Psychology, 16, 172-186.

Scanlan, T. K., Simons, J. P., Carpenter, P. J., Schmidt, G. W., Keeler, B. (1993). The Sport Commitment Model: Measurement Development for the Youth-Sport Domain. Journal of Sport and Exercise Psychology, 15, 16-38.

Showers, C. (1992). Compartmentalization of positive and negative self-knowledge: Keeping bad apples out of the bunch. Journal of Personality and Social Psychology, 62, 1036-1049.

Steele, (1988). The psychology of self-affirmation: Sustaining the integrity of the self. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol.21, pp. 261-302). New York: Academic Press.

Stryker, S. (1978). Status inconsistency and role conflict. Annual Review of Sociology, 4, 57-90.

Stryker, S., & Serpe, R. T. (1994). Identity salience and psychological centrality: Equivalent, overlapping, or complementary concepts? Social Psychology Quarterly, 57, 16-35.

Ziegler, S. G. (1991). Perceived benefits of marathon running in males and females. Sex Roles, 25, 119-127.

 Pattern Matrix for Principal Components
 Analysis with Direct Oblimin Rotation
 on Eleven Physical and Psychological
 Effects of Marathon Training
 Factor
Effect 1 2 3 [*] 4 5 [*]
Decreased social activity .79 .02 .10 -.01 -.06
Decreased time with family/friends .73 .07 .08 .03 .03
Isolation from non-running friends .75 .03 -.12 .10 -.01
Enhanced self-image .03 .88 .02 -.04 -.06
Increased self-confidence .18 .80 .08 -.12 .10
Enhanced body image .12 .74 -.04 .08 -.03
Increased energy level -.09 .57 -.30 -.03 .36
Decreased anxiety level -.18 .62 .13 .08 -.07
Chronic fatigue .15 .09 .82 .06 -.09
Chronic soreness in knees/ankles -.02 -.03 .68 .01 .26
Increased susceptibility to illness -.14 .08 .26 .79 -.06
Financial difficulties .09 -.02 -.29 .76 .13
Decreased job performance .19 -.05 .05 .58 -.01
Appetite loss -.09 .05 .05 .04 .83
Sleep disturbance .33 -.12 .27 .05 .50
Eigenvalue 3.3 2.6 1.2 1.1 1.0
Percent of variance explained 21.97 17.37 8.25 7.33 6.84
(*.)non-interpretable factor
 Mean Relative Importance Ratings of Six
 Life Roles for Runners High or Low in AI
 Level of AI
 Full Sample High (n = 79) Low (n = 78)
Life role Mean Rank Mean Rank Mean Rank
Academic/Occ. 3.33 3 3.36 4 3.65 3
Athlete [*] 2.94 5 3.49 3 2.54 5
Extracurricular 2.00 6 1.82 6 2.15 6
Family 5.00 1 4.88 1 5.15 1
Friendship 3.24 4 3.30 5 3.06 4
Romantic Partner 4.48 2 4.12 2 4.57 2
(*.)p[less than].001
 Mean Levels of Social, Positive, and Negative
 Consequences of Marathon Running in Runners
 High and Low Levels of AI
 Level of AI
 High Low
Social consequences [*] 2.92(1.40) 2.04(1.09)
Positive consequences [*] 5.35 (.98) 4.72(1.37)
General negative consequences [*] 1.83 (.93) 1.50 (.72)
(*.)p [less than] .017 (Bonferroni-Adjusted Significance Level)
 Mean Scores of Commitment to, Enjoyment
 of, Investment in, Perceived Social Constraints
 to and Involvement in Running for
 Participants High and Low in AI
 Level of AI
Variable High Low
Commitment [*] 6.58 (.48) 5.92 (.82)
Enjoyment [*] 6.78 (.40) 6.31 (.81)
Investment [*] 5.94 (.92) 5.03 (1.05)
Social Constraints [*] .14 (1.17) 1.42 (.57)
Involvement [*] 6.12 (.89) 4.87 (1.27)
(*.)p[less than].001
 Mean Levels of AI in Runners Grouped by
 Personal Best Time in a Marathon
Personal best time (interval) Coding N Mean AI
[less than]3:00 1 28 44.29(8.77)
3:00-3:29.59 2 61 42.36(8.77)
3:30-3:59.59 3 73 39.64(9.84)
4:00-4:29.59 4 38 37.82(9.05)
4:30-4:59.59 5 16 38.06(6.36)
[greater than]=5:00 6 8 43.88(7.81)
Totals 224 [*] 40.69
(*.)12 participants did not report personal best times.