Does the Iceberg Profile Discriminate Between Successful and Less Successful Athletes? A Meta-Analysis

JOURNAL OF SPORT & EXERCISE PSYCHOLOGY, 1995, 17, 185-199 O 1995 Human Kinetics Publishers, Inc. Does the Iceberg Profile Discriminate Between Successful and Less Successful Athletes? A Meta-Analysis Allan J. Rowley, Daniel M. Landers, L. Blaine Kyllo, and Jennifer L. Etnier Arizona State University The Profile of Mood States (POMS) is commonly used to measure mental health in athletes. Athletes scoring below norms on scales of tension, depression, confusion, anger, and fatigue, and above norms on vigor, are said to possess a positive profile that graphically depicts an iceberg. However, the predictive power of the iceberg profile has recently been questioned. A metaanalysis was conducted on 33 studies comparing the POMS scores of athletes differing in success to estimate the magnitude of the findings. The overall effect size was calculated to be 0.15. Although this value was significantly different from zero, the amount of variance accounted for was less than 1%. The results suggest that across many different sports and levels of performance, successful athletes possess a mood profile slightly more positive than less successful athletes. However, with such a small and nonrobust effect, the utility of the POMS in predicting athletic success is questionable. Key words: Profile of Mood States, mental health model Morgan's (1985) mental health model states that "positive mental health enhances the likelihood of success in sport, whereas psychopathology is associated with a greater incidence of failure" (p. 79). In examining the mood states of athletes and nonathletes, Morgan found that athletes, especially successful ones, possessed a unique mood profile which he labeled the iceberg profile. This term refers to the graphic picture that raw scores on the Profile of Mood States (POMS) create when they are plotted on a profile sheet (McNair, Lorr, & Droppleman, 1971/1981), with the test norms representing the "water line." If an athlete scores low on "negative" mood scales (tension, depression, anger, fatigue, confusion) and high on the "positive" vigor scale, the plotted curve resembles an The authors are with the Department of Exercise Science and Physical Education at Arizona State University, Box 870404, Tempe, AZ 85287-0404. Request reprints from D.M. Landers.

186 / Rowley, Landers, Kyllo, and Etnier iceberg. Using this terminology, Morgan's hypothesis can be simplified as follows: Successful athletes possess more of an iceberg profile than less successful athletes. Textbook reviews of personality and sport research have given the mental health model considerable attention. Gill (1986) stated that "Morgan offers the most systematic and strongly supported work on the relationship of personality to success in sport" (p. 31), and Cox (1990) stated that "the research evidence is very convincing that the successful athlete is likely to enjoy positive mental health" (p. 38). Others have described the consistency of Morgan's results as "remarkable" (Eysenck, Nias, & Cox, 1982, p. 19). Claiming that successful athletes tend to be mentally healthy is not a provocative concept, and surely, psychopathology and success at nearly anything should be inversely correlated. Yet, when examining the mental health model, a question arises as to whether the relationship between mental health and athletic performance is correlated highly enough to make fine-grain distinctions between athletes possessing similar psychological profiles. After all, athletes do not fall into two categories: psychopaths and normals. Mental health is a continuum, and athletes may fall very close to each other in their overall rating of mental health, or they may be very different. The point is, the smallest divisions between levels of mental health are nearly indistinguishable. Can the mental health model detect fine-grain distinctions among successful and less successful athletes with similar mental health profiles? The empirical evidence presented by Morgan (1985) in support of the mental health model, in fact, came from studies examining fairly homogeneous groups of athletes (elite or advanced athletes from relatively few sports) whose profiles were likely to be similar. The results from these studies, in terms of the ability to predict success from these similar profiles, were very impressive. The general acceptance of the mental health model is based largely on seven of Morgan's early studies that provided convincing research support. In the years following Morgan's early research, however, conflicting evidence has generated criticism and concern as to the usefulness of the mental health model (Landers, 1991; Prapavessis & Grove, 1991). Criticism is leveled at five problem areas in the research: the measures of mental health, the measures of performance, the time span between the measurement of these variables, the interpetation of the findings, and the design methodology. Although other psychological measures have been used to assess mental health, the POMS has been the inventory of choice because of its success in Morgan's early studies. Morgan (1980b) stated that "of all the psychological tests my colleagues and I have experimented with, we have found the Profile of Mood States (POMS) to be the most highly predictive of athletic success" (pp. 93,97). However, situational factors, such as completing the POMS before team selection, can influence the results. Th.e validity of the POMS in such environments is questionable because, wording to Boyle (1987), "the instrument is highly susceptible to distorting influences such as effects of social desirability and other response sets, inadequate self-insight, and even downright dissimulation, given the obvious item transparency" (p. 353). In fact, undergraduate students were able to "fake good" &d produce an iceberg profile when asked to imagine themselves as elite athletes who suspected that POMS data would influence their being selected for a team (Miller & Edgington, 1984).

Iceberg Profile / 187 A second concern with past research is related to the measures of performance that were used. Specifically, successful performance, which the mental health model claims to predict, was never directly assessed in Morgan's research. Instead, success was inferred from placement in a competition or in terms of qualifying for a team (e-g., Morgan & Johnson, 1978; Morgan, O'Connor, Sparling, & Pate, 1987). This definition fails to account for the nonqualifiers and nonplacers who produced career-best performances. As Krane (1992) points out, this use of a between-subjects comparison in terms of performance does not adequately account for how well the athlete actually performs. Thus, the measures of success that have typically been used are not appropriate, nor is it appropriate to rely upon between-subjects comparisons rather than within-subjects analysis. There is a need to incorporate within-subjects analyses rather than betweensubjects analyses because precompetitive mood profiles related to successful and unsuccessful performances may vary between individuals. For instance, one athlete may perform quite well when angry, whereas another athlete with the same mood state may perform poorly. The third criticism of the research in this area is that the time span between the assessment of mental health and performance has been inconsistent (Landers, 1991). Administration of the POMS, a paper-and-pencil measure, is limited to times outside an athlete's precompetition preparatory period and has occurred anywhere from 1 week to 4 years (Morgan & Johnson, 1978) prior to performance. Often, the wording of the instructions for the POMS did not direct athletes to report their moods for the time in which the performance occurred. Additionally, empirical evidence has shown that mood states measured by the POMS are influenced by time of day (Hill & Hill, 1991) and, in swimmers, by level of training intensity (Morgan, Brown, Raglin, O'Connor, & Ellickson, 1987). Therefore, since the mood state likely changes between the time of psychological test administration and the time of performance assessment, the relationship between a "snapshot" of mental health and performance is tenuous at best. Another criticism of research in this area is that early researchers may have misinterpreted their results. In fact, the initial praise for Morgan's research may have resulted from these faulty interpretations. Prapavessis and Grove (1991) indicated that past researchers have wrongly interpreted correlational findings as causal by assuming that a certain mood profile produces good performances. An athlete's outstanding performances and training in the past may cause both a positive mood profile and a successful performance, without the mood state actually causing the performance. Heyman's (1982) reanalysis of a series of articles on the relationship between psychological states and performance support this contention. Heyman (1982) found that athletes exhibiting successful performances had significantly better season records, more experience, and came from superior training programs when compared to unsuccessful athletes. As Heyman concluded "the psychological patterns and cognitions may themselves reflect previous experiences rather than cause or facilitate performance in most of the participants" (p. 299). This evidence supports an overlooked concept in the psychological testing of athletes: the impact of an athlete's physical skills and abilities upon psychological parameters. The final criticism against studies that have investigated the mental health model is that they have been plagued with methodological problems. In spite of Morgan's consistent results, many of the studies claiming to support the iceberg

188 1 Rowley, Landers, Kyllo, and Etnier profile made no statistical comparison to an appropriate control group (Landers, 1991). Rather than compare groups of athletes, researchers have often contrasted an athletic group's scores with POMS norms generated in 1967 by participants in psychological tests and students in psychology classes (e.g., Morgan &Johnson, 1978; Morgan, O'Connor, et al., 1987). As Landers (1991) argued, the normative sample reported in the POMS manual is an inappropriate comparison group for research on athletes. McNair et al. (1971/1981), the authors of the POMS, even cautioned that the normative information "should be considered as very tentative" (p. 19) because the data were obtained from only one university at one point in time. In fact, since Morgan's early work, few studies have found the iceberg profile to be successful at predicting athletic performance. In a review of research, Landers (1991) discovered that only 53% of the comparisons were in the predicted direction. This percentage is not significantly different from the total that would be expected from chance. Unfortunately, a traditional narrative review of the literature is unable to determine the magnitude of an effect, and Landers (1991) could not conclusively estimate the predictive import of the mental health model. However, Glass's (1977) quantitative review method, meta-analysis, integrates the findings of multiple studies that meet present selection criteria in a particular domain in order to arrive at general conclusions on a topic. Meta-analyses quantify the results of all relevant, appropriate, and available studies with a standard metric, the effect size (ES), that can later be subjected to statistical analysis. Combining the ESs of many studies is equivalent to combining all of the participants from all of the studies. This then augments the statistical power of the analysis, which is useful in detecting small, significant trends and which allows for the use of multivariate techniques. Meta-analysis, therefore, can provide what traditional reviews cannot: in this case, a replicable, relatively objective, scaled measure of the mental health model's efficacy in discriminating among athletes. None of the methodological concerns with the past research can be corrected through meta-analysis; however, these concerns can be coded so that the impact that they have on the ESs can be examined. The extent to which mental health, as measured by the POMS, is predictive of athletic performance was tested using meta-analytic techniques. It was hypothesized that, as predicted by Morgan (1985), successful athletes would possess a more positive mood (i.e., iceberg) profile when compared to less successful athletes. Additional a priori hypotheses were proposed with regard to characteristics that could potentially moderate the effects outlined in the original hypothesis. Referring to the mental health model, Morgan (1985) stated that "specific responses will be dependent upon specific stimulus conditions" (p. 71). This suggests that the model's predictive utility may be limited to certain types of activities (Landers, 1991). Therefore, it is hypothesized that the ability of the POMS to predict athletic success will be influenced by the type of activity (endurance or strength) being observed. Additionally, because mood states change over time (Hill & Hill, 1991) and because the mood of the athlete is thought to influence the athlete's performance capability, it is hypothesized that studies that assess mood immediately before competition will be better predictors of performance than those that assess mood well before competition or after competition. As evident in previous meta-analyses in sport psychology (Feltz & Landers,

Iceberg Profile / 189 1983; North, McCullagh, & Tran, 1990), published studies have a larger overall ES than unpublished studies. Thus, it is predicted that effects from published studies will be significantly larger than effects from unpublished studies. Finally, since the POMS is susceptible to distorting influences (Boyle, 1987; Miller & Edgington, 1984), it is hypothesized that the POMS will be less effective as a predictive instrument in environments with strong demand characteristics (i.e., during team selection) because all athletes may respond in a "desirable" manner if they feel that their responses may impact their chances of, for example, making the team. This idea is also supported by Morgan's (1974) caution that studies reporting null findings often do not use lie, guess, or random response scales. Without such safeguards, the truthfulness of the responses may be questioned, and the variability in the responses may be misleadingly small. As such, studies that include any of these distortion scales should produce larger ESs than studies without safeguards. Selection and inclusion of Studies Method All available literature from 197 1 (the copyright date of the POMS) through January 1992 were located using computer databases (PsychLIT, ERIC, Medline, and Dissertation Abstracts International); AAHPERD Convention Abstracts of Research Papers; Completed Research in Health, Physical Education, Recreation, and Dance; Psychological Abstracts; Social Sciences Citation Index; Current Contents; and Microform Publications. Systematic searches of reference lists were also conducted to locate studies not located through the computerized data sets. Glass (1977) has recommended that a priori exclusions of research should be minimized. As such, inclusion criteria were kept broad so that studies included in the analysis had to assess mental health only with the POMS and with at least two groups of athletes (i.e., any individual who is a regular participant in a sport at the time of mood assessment) in the same sport at different levels of success. A total of 94 studies were located using the search strategies described. However, 8 studies could not be included in the analyses because they did not include sufficient data for the calculation of ESs. Additionally, 52 studies were not included because they did not include any comparison of POMS scores between athletes who differed in success. Finally, one study was excluded because the data were presented in another study included in the analysis. Therefore, 33 studies yielding 41 1 ESs were subjected to further analysis. These studies are identified in the reference list by an asterisk. Coding Characteristics Studies were coded for variables that could potentially moderate the effect of mental health on athletic performance. Publication status referred to whether the study was published in a journal or book or was an unpublished thesis, dissertation, or professional presentation. Participant characteristics included the following:

190 / Rowley, Landers, Kyllo, and Etnier 1. The difference between the groups on the level of expertise (see bulleted list), which was coded as same level, one level different, or two levels different 2. The confidence in the expertise level categorization which was simply coded as questionable (unclear) or valid (clear) 3. The sex of participants, which was coded as men only, women only, or men and women combined Activity characteristics included the following: 1. Aerobic (training consists of continuous rhythmic movements using major muscle groups) or strength (any activity not meeting the aerobic definition) 2. The operational definition of performance, which was coded as personal best, ranking, selection for team, starterslnonstarters, placement in a race, finishinglnot finishing a race, winning/losing, or a subjective measure of performance 3. The confidence in this definition of performance, which was coded as questionable (unclear) or valid (clear) Methodological characteristics included the following: 1. Time of mood assessment, which was coded as more than 24 hours prior to performance, less than 24 hours prior to performance, or postperformance 2. Presence of a lie scale, which was coded as included or not 3. Presence of demand characteristics (any situation in which athletes were being evaluated for selection to a team) as present or not 4. Differences in physical training state between groups as different or the same 5. Internal validity, as good (0 threats), moderate (1-2 threats), or poor (3 or more threats) based upon the number of threats to internal validity which existed Effect'sizes were calculated for each scale of the POMS (i.e., tension, depression, anger, vigor, fatigue, and confusion). Expertise categories were based upon the following operational definitions: Elite Ranked in the top 10 in the United States Member of a national team or professional Best marathon: men: under 2.5 hours; women under 2.75 hours Best 10K. women: under 35 minutes Advanced Ranked, but not in top 10 Member of a state team or winner of a state tournament College athletes High School: member of a national team or nationally ranked Black belt in martial arts

Analyses Best marathon: under 4 hours Able to run an ultramarathon (50 miles or more) Iceberg Profile / 191 Intermediate High school athletes Colored belt in martial arts Able to run a marathon, triathlon, or 15 miles per week regularly Novice Ranked as novice Regular participant for less than one year at time of assessment ESs were calculated using Hedges and Olkin's (1985) extension of Glass's (1977) original technique. The pooled standard deviation (weighted to correct for sample size bias) was used in ES calculation because it provides a more precise estimate of the population variance (Hedges, 1981). The formula for this calculation is: ME - MC (NE- 1). SD; + (Nc- 1). SD; ES = ---, where SDp = SDP NE-Nc-2, and where ME = mean of the experimental group; Mc = mean of the control group; SDp = pooled standard deviation; NE = number of participants in the experimental group; Nc = number of participants in the comparison group; SDE = standard deviation of the experimental group; and SDc = standard deviation of the comparison group. To ensure that ESs in support of the iceberg profile were positive values, the means of the successful group were subtracted from the means of the less successful group for all scales except "vigor," in which the less successful group was subtracted from the more successful group. Because studies with small samples sizes may have a biased effect size (Thomas& French, 1986), each effect size was then multiplied by a correction factor designed to yield an unbiased estimate of effect size (Hedges, 1981). This correction factor is: c = 1 - [3/(4m - 9)] where m = N, + N, - 2. Studies that reported more than one set of six ESs (representing the six POMS scales) can pose a threat to the statistical validity of the results. This situation violates the assumption of independence of data points (Bangert-Drowns, 1986). To control for this problem, overall ES was calculated twice. First, all ESs were combined to determine overall ES. Then for the remainder of the analyses only one ES (reflecting the two most extreme success groups) per study was included. The decision to use this particular set of ESs was made to give the mental health model the greatest likelihood of showing an effect. That is, since Morgan (1980a) contends that the relationship between psychological state and performance is linear, greater differences between performance outcome should be reflected in greater differences in POMS scores. Additionally, while effects could be calculated for total mood disturbance scores, these scores represent a sum of the six aspects of the mood scale and thus would not be independent scores. Therefore, these scores were only used in the overall average of effects and were not included in any subsequent analyses.

192 / Rowley, Landers, Kyllo, and Etnier Results The overall average ES (n = 41 1) was calculated to be 0.15 (SD = 0.89), which indicates that successful athletes possess a mood profile approximately one sixth of standard deviation healthier than less successful athletes. Based on the preceding rationale, which prescribed the elimination of the total mood disturbance score and the use of only one set of ESs for the most extreme groups, the overall ES was 0.19 (SD = 1.09, n = 198). The average ES for each study and the coding of the moderating variables can be found in Table 1. Next, three outliers (ESs falling outside of three standard deviations from the mean) were identified and omitted. This did not change the mean ES, but the SD was decreased to 0.71. Following the procedures outlined by Thomas and French (1986), the H-statistic was calculated and found to be significant, x2(194) = 817.69, p <.05, with a critical value of 227.21, indicating that ESs were not homogeneous. This lack of homogeneity warranted further examination of moderator variables. One-way analyses of variance (ANOVAs) were conducted on all of the coded variables to test for significant differences in ESs among the levels of each variable. ANOVAs were used because the distribution of the effects was reasonably close to normal to warrant the use of parametric techniques rather than necessitating the use of nonparametric techniques (Wolf, 1986). The only moderating variable that was found to significantly impact the ESs was the moderating variable that reflected the confidence the coder had in the value assigned to the performance variable, F(1, 193) = 5.652, p <.05. The studies in which the categorization was questionable (i.e., in which performance scores were unclear) had larger ESs (M = 0.51, SD = 0.78, n = 29) than did studies in which the categorization was not questionable (i.e., in which performance scores were quite clear) (M = 0.13, SD = 0.69, n = 166). Discussion All available research included in the meta-analysis indicated that successful athletes, in general, report POMS scores one sixth of a standard deviation higher than less successful athletes in the same sport (ES = 0.15). Despite biasing analyses to allow the mental health model to demonstrate an effect (by using the largest possible set of ESs from each study in the analysis), these results show that POMS scores for successful athletes are less than one fifth (ES = 0.19) of a standard deviation higher than POMS scores for unsuccessful athletes. The magnitude of this effect, according to Cohen's (1988) classification, is very small. Of additional concern is the fact that larger effects were found in those studies in which the definition of success was of questionable clarity (M = 0.51) as compared to studies in which the definition of success was clear (M = 0.13). This does not encourage the use of the mental health model as an instrument to predict performance, or as a dimension for team selection. The lack of strong support for the POMS in predicting athletic success reflects poorly on the mental health model proposed by Morgan (1985). Given that prediction rates of the model have been reported to range from 70% to 80% (Morgan, 1985) and 70% to 90% (Raglin, Morgan, & Luchsinger, 1990), the

Iceberg Profile / 193 ESs generated by the meta-analytic procedures should have been larger. Further, when these ESs are converted to a Pearson's?, they account for less than 1% of the variance explained for performance outcome. That is, of all the factors that determine an athlete's performance, precompetitive mood states as measured by the POMS accounted for less than 1%, and the remaining 99% could be explained by other components (e.g., physiological characteristics, biomechanical status, practice, diet, or other psychological variables). Eight studies were excluded from the analysis due to insufficient data for the calculation of ESs. Four of these studies (DeMers, 1983; Morgan, O'Connor, Ellickson, & Bradley, 1988; Nagle, Morgan, Hellickson, Serfass, & Alexander, 1975; Silva, Shultz, Haslam, Martin, & Murray, 1988) reported that the POMS discriminated between successful and less successful athletes, with the successful athletes possessing a more pronounced iceberg profile. The other four studies found that mood profiles either did not differ between athletes of different levels (Berger & Owen, 1986; Craighead, Privette, Vallianos, & Byrkit, 1986; Durtschi & Weiss, 1986) or that less successful athletes had a more pronounced iceberg profile than did more successful athletes (Riddick, 1984). It appears that, had these studies been included in the analysis, they would not have significantly affected the overall ES. The results of the present analysis mirror Landers's (1991) traditional review of the mental health model. Landers (1991) identified 14 studies that compared the POMS scores of two groups of athletes representing two levels of performance outcome, and a total of 148 comparisons between successful and less successful athletes were made for the six scales of the POMS. Of these comparisons, only 18% were statistically significant, with 16% of the significant results in the direction predicted by the mental health model. Since the sample sizes of the studies were so small, Landers (1991) conducted a second analysis in which statistical significance was ignored, and the 148 comparisons were divided into those that supported and those that did not support the mental health model. As already mentioned, when examined in this manner 53% of the results supported the mental health model. The present meta-analysis yielded a total of 41 1 comparisons, with 54% in the predicted direction. The striking similarity of results adds credence to Landers's (1991) conclusion that the findings "fail to clearly support the predictions of the mental health model" (p. 197). Based on the results of this analysis, future research would benefit from using larger samples. To ensure that adequate statistical power exists (i.e., 0.80), approximately 100 participants are required for 95% confidence in the results (Kraemer & Thiemann, 1987). However, there appears to be little promise in using the POMS and the mental health model in attempts to predict performance in athletes. Although a series of studies with strong research designs may be helpful in demonstrating a larger effect, it is likely to be moderate at best, and the resources may be better spent on more promising endeavors. However, Morgan and his colleagues have recently shifted their focus to a more "dynamic" mental health model, which proposes a within-subjects approach to predicting individual performance by means of consistent monitoring of the athlete's psychological states across different training conditions (Morgan, O'Connor, et al., 1987). The dynamic mental health model accounts for an athlete's individual differences by tracking his or her unique level of success and mood profiles. As such, each athlete will generate his or her own graphic

Table 1 Effect Size, Standard Deviation, and Moderating Variables by Study Study Clarity Time of Lie 2 Publ. No. Overall Overall Diff. in Conf. Operational of mood scale Demand Diff. in Internal & status subjects ES SD expert. in diff. Sex Task def. of perf. measure assessment included? char.? training? validity Bell & Howe (1986) Berer & Owen (1983) Boyce (1987) Pub Pub Unpub 160 27 66 Same One Two Quest Quest M Aerobic Placing in race F Aerobic Ranking M Aerobic Personal best After perf. After pref. 1 24 before Cavanaugh (1982) Unpub 201 Same Mx Strength Starters1 > 24 before non Two Quest M Strength Ranking Quest After perf. Daiss, LeUnes, & Pub 72 Nation (1986) Daus, Wilson, & Pub 41 Freeman (1986) Dryer & Crouch (1987) Pub 40 Frazier (1986) Unpub 68 Yes Yes No No NR Same One Quest M Strength Subjective Quest Two Same Aerobic Ranking Aerobic Personal 1 24 before After perf. Frazier (1988) Pub 113 One best Aerobic Personal best After perf. Frazier & Nagy (1989) Pub 16 Same Aerobic Widlose Quest > 24 before Gondola & Tuckman Pub 396 One Quest Aerobic Personal Quest 5 24 before (1983) Gutmann et al. (1984) Pub 7 Same best Aerobic Selection > 2A before No Yes NR Hagberg et al. (1979) Pub 9 Same Aerobic Selection NR Yes No Same Good Harris (1985) Unpub 131 Same Strength Widlose 124 before No No NR Hassmen & Pub 61 Same Aerobic Finish time > 24 before NR No Same Good Blomstrand (1991) Mod 'D s -' : _h 3. 2

Lindstrom (1990) Unpub 46-0.06 0.49 Same M Aerobic Finished 5 24 before No No Same non Mahoney (1989) Pub 50 0.15 0.15 One Mx Strength Ranking NR No No NR McGowan & Miller Pub 107-0.24 0.31 Combo Quest Mx Strength Subjective > 24 before NO NO NR (1989) McGowan et al. (1990) Pub 55-0.30 0.37 Two Quest Mx Strength Ranking 124 before No No NR Miller & Edgington Pub 20-0.12 0.46 Same F Strength Selection > 24 before No Yes NR (1984) Morgan (1985) Pub 16 0.79 0.30 Same M Strength Selection 24 before NR Yes NR Morgan & Pollock Pub 57-0.15 0.15 One NR Aerobic Ranking After perf. NR No Dierent (1977) Morgan & Johnson Pub 27-0.34 0.22 Same M Aerobic Selection > 24 before Yes Yes NR (1978) Morgn et al. (1987) Pub 27 0.24 0.36 One F Aerobic Personal After perf. No No Different best Poole et al. (1986) Pub 92 0.60 0.11 One F Strength Ranking After perf. No No NR Prapavessis & Grove Pub 12 0.00 0.13 Same Mx Strength Personal 2 24 before No No NR (1991) best Raglin, Morgan, & Pub 22 0.52 0.52 Same F Aerobic Selection > 24 before No Yes Same Luchsinger (1990) Ramadan (1984) Unpub 17-0.04 0.41 Same M Strength Starters1 NR No No NR non Robinson & Howe Pub 17 1.53 0.44 Same M Strength Personal After perf. No No NR (1987) best Silva et al. (1981) Pub 15 0.27 0.52 Same Quest M Strength Selection > 24 before No Yes Different Tharion, Strowman, & Pub 56-0.08 0.26 Same M Aerobic Finished After perf. No No NR Rauch (1988) non Toner (I981) Unpub 81-0.49 0.12 Same F Strength Subjective > 24 before No No NR Wilson, Morley, & Pub 20 2.53 1.92 Two Quest M Aerobic Ranking Quest NR NO No NR Bird (1980) Nore. Pub = published; Unpub = unpublished; Quest = questionable; Diff = difference; Expert. = expertise; Conf = confidence; Def = Definition; Perf. = performance; Char. = characteristics; Mod = moderate; M = male; F = female; Mx = mixed. Mod Mod Mod Good Mod Mod f Mod 3 5?r

196 / Rowley, Landers, Kyllo, and Etnier profile of POMS scores that is consistent with success, which may or may not depict the classic iceberg profile. The dynamic mental health model is likely a more appropriate and effective application of the POMS as it relates to athletic performance. References' Bangert-Drowns, R.L. (1986). Review of developments in meta-analytic method. Psychological Bulletin, 99, 388-399. *Bell, G.J., & Howe, B.L. (1988). Mood state profiles and motivations of triathletes. Journal of Sport Behavior, 11, 66-77. *Berger, B.G., & Owen D.R. (1983). Mood alteration with swimming: Swimmers really do "feel better." Psychosomatic Medicine, 45, 425-433. Berger, B.G., & Owen, D.R. (1986). Mood alteration with swimming: A reexamination. In L. Vander Velden & J.H. Humphrey (Eds.), Psychology and sociology of sport: Current selected research (pp. 97-1 13). New York: AMS Press. *Boyce, L.V. (1987). Psychology fitness, personality, and cognitive strategies of marathon runners as related to success and gender [Abstract]. Completed Research in Health, Physical Education, Recreation, & Dance, 29, 108. Boyle, G.P. (1987). A cross-validation of the factor structui.e of the profile of mood states: Were the factors correctly identified in the first inptance? Psychological Reports, 60, 343-354. *Cavanaugh, S.R. (1982). The mood states of selected collegiate athletes during the season of varsity competition (Doctoral dissertation, Brigham Young Unverisity). Dissertation Abstracts International, 43, 1874A. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (3rd ed.). New York: Academic Press. Cox, R.H. (1990). Sport psychology: Concepts and applications (2nd ed.). Dubuque, IA: Brown. Craighead, D.J., Privette, G., Vallianos, F., & Byrkit, D. (1986). Personality characteristics basketball players, starters, and non-starters. International Journal of Sport Psychology, 17, 110-1 19. *Dais, S., LeUnes, A., & Nation, J. (1986). Mood and locus of control of a sample of college and professional football players. Perceptual and Motor Skills, 63, 733-734. *Daus, A.T., Wilson, J., & Freeman, W.M. (1986). Psychological testing as an auxiliary means of selecting successful college and professional football players. Journal of Sports Medicine, 26, 274-278. DeMers, G.E. (1983, May-June). Emotional states of high-caliber divers. Swimming Technique, pp. 33-35. Durtschi, S.K., & Weiss, M.R. (1986). Psychological Characteristics of elite and nonelite marathon runners. In D.M. Landers (Ed.), Sport and elite performers: The 1984 Olympic scientific congress proceedings (pp. 73-80). Champaign, IL: Human Kinetics. *Dyer, J.B., & Crouch, J.G. (1987). Effects of running on moods: A time series study. Perceptual and Motor Skills, 64, 783-789. 'References marked 'with an asterisk were included in the meta-analysis.

Iceberg Profile / 197 Eysenck, H.J., Nias, D.K.B., & Cox, D.N. (1982). Sport and personality. Advances in Behavior Research and Therapy, 4, 1-56. Feltz, D.L., & Landers, D.M. (1983). The effects of mental practice on motor skill learning and performance: A meta-analysis. Journal of Sport Psychology, 5, 25-57. 'Frazier, S.E. (1986). Psychological characteristics of male and female marathon runners of various performance levels (Doctoral dissertation, Indiana University). Dissertation Abstracts International, 47, 2076. *Frazier, S.E. (1988). Mood state profiles of chronic exercisers with differing abilities. International Journal of Sport Psychology, 19, 65-71. "Frazier, S.E., & Nagy, S. (1989). Mood state changes of women as a function of regular aerobic exercise. Perceptual and Motor Skills, 68, 283-287. Gill, D.L. (1986). Psychological dynamics of sport. Champaign, IL: Human Kinetics. Glass, G.V. (1 977). Integrating findings: The meta-analysis of research. Review of Research in Education, 5, 351-379. 'Gondola, J.C., & Tuckman, B.W. (1983). Extent of training and mood enhancement in women runners. Perceptual and Motor Skills, 57, 333-334. 'Gutmann, M.C., Pollock, M.L., Foster, C., & Schmidt, D. (1984). Training stress in Olympic speed skaters: A psychological perspective. The Physician and Sportsmedicine, 12(12), 45-57. *Hagberg, J.M., Mullin, J.P., Bahrke, M., Limburg, J. (1979). Physiological profiles and selected psychological characteristics of national class American cyclists. Journal of Sports Medicine and Physical Fitness, 19, 341-346. *Harris, D. (1985). The relationship between player mood state and performance outcome of a woman's intercollegiate softball team (Master's thesis, Stephen F. Austin State University). Master's Abstracts International, 23, 503. "Hassmen, P., & Blomstrand, E. (1991). Mood change and marathon running: A pilot study using a Swedish version of the POMS test. Scandinavian Journal of Psychology, 32, 225-232. Hedges, L.V. (1981). Distribution theory for Glass' estimator of effect size and related estimators. Journal of Educational Statistics, 6, 107-128. Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta analysis. New York: Academic Press. Heyman, S.R. (1982). Comparisons of successful and unsuccessful competitors: A reconsideration of methodological questions and data. Journal of Sport Psychology, 4, 259-300. Hill, C.M., & Hill, D.W. (1991). Influence of time of day on responses to the Profile of Mood States. Perceptual and Motor Skills, 72, 434-439. Kraemer, H.C., & Thiemann, S. (1987). How many subjects? Statistical power analysis in research. Newbury Park, CA: Sage. Krane, V. (1992). Conceptual and methodological considerations in sport anxiety research: From the inverted-u hypothesis to catastrophe theory. Quest, 44, 72-87. Landers, D.M. (1991). Optimizing individual performance. In D. Druckman & R.A. Bjork (Eds.), In the mind's eye: Enhancing human performance (pp. 193-246). Washington, DC: National Academy Press. *Lindstrom, D.V. (1990). Personality characteristics of ultramarathoners: Finishers vs. nonfinishers (Master's thesis, San Jose State University). Master's Abstracts International, 28, 644. *Mahoney, M.J. (1989). Psychological predictors of elite and non-elite performance in Olympic weightlifting. International Journal of Sport Psychology, 20, 1-12.

198 / Rowley, Landers, Kyllo, and Etnier McNair, D.M., Lon; M., & Droppleman, L.F. (1981). Profile of mood states manual. San Diego: Educational and Industrial Testing Service. (Original work published 1971) *McGowan, R.W., & Miller, M.J. (1989). Difference in mood states between successful and less successful karate participants. Perceptual and Motor Skills, 68, 505-506. *McGowan, R.W., Miller, M.J., & Henschen, K.P. (1990). Differences in mood states between belt ranks in karate tournament competitors. Perceptual and Motor Skills, 71, 147-150. *Miller, B.P., & Edgington, G.P. (1984). Psychological mood state distortion in a sporting context. Journal of Sport Behavior, 7, 91-94. Morgan, W.P. (1974). Selected psychological considerations in sport. Research Quarterly, 45, 374-390. Morgan, W.P. (1980a). The trait psychology controversy. Research Quarterly, 51, 50-76. Morgan, W.P. (1980b, July). Test of champions: The iceberg profile. Psychology Today, pp. 92-99. *Morgan, W.P. (1985). Selected psychological factors limiting performance: A mental health model. In D.H. Clarke & H.M. Eckert (Eds.), Limits of human performance (pp. 70-80). Champaign, IL: Human Kinetics. Morgan, W.P., Brown, D.R., Raglin, J.S., O'Connor, P.J., & Ellickson, K.A. (1987). Psychological monitoring of overtraining and staleness. British Journal of Sports Medicine, 21, 107-114. *Morgan, W.P., & Johnson, R.W. (1978). Personality characteristics of successful and unsuccessful oarsmen. International Journal of Sport Psychology, 9, 119-133. Morgan, W.P., O'Connor, P.J., Ellickson, K.A., & Bradley, P.W. (1988). Personality structure, mood states, and performance in the elite male distance runners. International Journal of Sport Psychology, 19, 247-263. *Morgan, W.P., O'Connor, P.J., Sparling, P.B., &Pate, R.R. (1987). Psychological characteristics of the elite female distance runner. International Journal of Sports Medicine, ~(SUPP~.), 124-131. *Morgan, W.P., & Pollock, M.L. (1977). Psychologic Characterization of the elite distance runner. Annals of the New York Academy of Sciences, 301, 383-403. Nagle, F.J., Morgan, W.P., Hellickson, R.O., Serfass, R.C., & Alexander J.F. (1975). Spotting success traits in Olympic contenders. The Physician and Sportsmedicine, 3(12), 31-34. North, T.C., McCullagh, P., & Tran, Z.V. (1990). Effect of exercise on depression. Exercise and Sport Science Reviews, 18, 379-415. *Poole, C., Henschen, K.P., Shultz, B., Gordon, R., & Hill, J. (1986). A longitudinal investigation of the psychological profiles of elite collegiate female athletes according to performance level. In L. Unestahl (Ed.), Contemporary sport psychology: Proceedingsfrom the VI World Congress in Sport Psychology (pp. 65-72). Orebro, Sweden: Veje. *Prapavessis, H., & Grove, J.R. (1991). Precompetitive emotions and shooting performance: The mental health and zone of optimal functioning models. The Sport Psychologist, 5, 223-234. *Raglin, J.S., Morgan, W.P., & Luchsinger, A.E. (1990). Mood and self-motivation in successful and unsuccessful female rowers. Medicine and Science in Sports and Exercise, 22, 849-853. *Ramadan, J.M. (1984). Selected physiological, psychological, and anthropometric characteristics of Kuwaiti World Cup soccer team (Doctoral dissertation, Louisiana State University). Dissertation Abstracts International, 46, 924A.

Iceberg Profile 1 199 Riddick, C.C. (1984). Comparative psychological profiles of three groups of female collegians: Competitive swimmers, recreational swimmers, and inactive swimmers. Journal of Sport Behavior, 7, 160-174. *Robinson, D.W., & Howe, B.L. (1987). Causal attribution and mood state relationships of soccer players in a sport achievement setting. Journal of Sport Behavior, 10, 137-146. Silva, J.M., Shultz, B.B., Haslam, R.W., MartinT.P., & Murray, D.F. (1988). Discriminating characteristics of contestants at the United States Olympic wrestling trials. International Journal of Sport Psychology, 16, 79-102. *Silva, J.M., Shultz, B.B., Haslam, R.W., & Murray, D. (1981). A psychophysiological assessment of elite wrestlers. Research Quarterly, 52, 348-358. *Tharion, W.J., Strowman, S.R., & Rauch, T.M. (1988). Profile and changes in moods of ultrarnarathoners. Journal of Sport & Exercise Psychology, 10, 229-235. Thomas, J.R., & French, K.E. (1986). The use of meta-analysis in exercise and sport: A tutorial. Research Quarterly for Exercise and Sport, 57, 196-204. *Toner, M.K. (1 98 1). The relationship of selectedphysical fitness, skill, and mood variables to success in female high school basketball candidates (Doctoral dissertation, Boston University). Dissertation Abstracts International, 42, 3909A. *Wilson, V.E., Morley, N.C., & Bird, E.I. (1980). Mood profiles of marathon runners, joggers, and non-exercisers. Perceptual and Motor Skills, 50, 117-1 18. Wolf, F.M. (1986). Meta-analysis: Quantitative Methods for Research Synthesis. Beverly Hills, CA: Sage. Author Note This article is based on a master's thesis completed by the first author at Arizona State University under the guidance of the second author. Manuscript submitted: June 17, 1994 Revision received: November 12, 1994