STUDY 1: A QUANTITATIVE APPROACH TO UNDERSTANDING MENTAL TOUGHNESS AND STRESS/RECOVERY
There is widespread agreement on the importance of MT in sport (Jones et al. 2002; Krane & Williams, 1988; Loehr, 1982, 1986) with athletes and coaches attributing
successful performances to being mentally tough (Gould et al., 1987, 2002). Researchers are now beginning to develop programs aimed at increasing athletes’ MT (Gucciardi et al. 2009c, 2009d), with the goal being enhanced performance, but the extent of the benefits (or detriments) of increasing MT are not well known.
Athletes have idiosyncratic MT profiles with varied characteristics that may help or hinder performance. For example, athletes who are able to reduce their emotional reaction to adversity, described as stress minimisation (Middleton, 2005), may be able to perform well regardless of the adversity they face. Running training programs to develop this characteristic of MT may help athletes to manage adverse situations without detrimental emotional reactions, and therefore maintain a healthy balance between stress and recovery. Nevertheless, trying to increase other MT characteristics may be detrimental for athletes in terms of the stress/recovery balance. Middleton (2005) described perseverance as,
“persisting in or remaining constant to a purpose, idea, or task in the face of obstacles, discouragement or adversity” (p. 87). This MT characteristic would appear to be important for athletes to help achieve their goals and overcome obstacles, but trying to increase perseverance in athletes who already have high levels of perseverance may teach them to persevere too much (e.g., when injured, ill, or excessively fatigued) leading towards OT and reduced performance. Training to increase many MT characteristics in individuals
without a thorough understanding of all the potential outcomes might lead to problems with athlete wellbeing and performance.
Research investigating the stability of MT is in its infancy, however it appears that some aspects of MT may be stable and some more changeable. Horsburgh et al. (2009) concluded that MT consists of some characteristics that are stable traits and other
characteristics that may be developed. Middleton et al. (2004a) supported the idea that MT
has some stable characteristics and other characteristics that are malleable by separating attributes in their study into actions (e.g., focussing of attention) and personal
characteristics (self-belief). Bull et al. (2005) identified that MT characteristics
encompassed the development of athletes’ enduring personality characteristics indicating the stability of MT. Connaughton, Wadey, et al. (2008) proposed that levels of MT attributes could fluctuate and therefore required maintenance throughout the remainder of athletes’ careers. As seems to be consistent in MT research there is further work to be carried out to enhance our understanding regarding MT stability.
Researchers have primarily focussed on understanding the MT construct and the characteristics associated with MT. An important step to understanding how MT might influence health and wellbeing is to investigate which MT characteristics could be
beneficial to health and performance, and which, if any, could be detrimental, especially at high levels. Further, little research has addressed the impact of MT on stress-recovery imbalance (SRI) and on indicators of impending OT. Given that SRI may arise due to a failure to manage stress or take time for recovery in adverse situations (and most mental toughness definitions contain a capacity to cope in the face of adversity. See Middleton et
al., 2004a), investigating the relationship between MT and SRI may increase knowledge of why some athletes are susceptible to OT.
In this study, I had a number of aims in relation to MT and SRI. I will explain the aims in relation to the order of the Results section to maintain continuity. The first aim was to examine the stability of mental toughness over time. There are disagreements in the MT research with some researchers claiming that mental toughness is stable and trait-like (Clough et al., 2002; Horsburgh et al., 2009; Nicholls et al., 2008) and others describing MT as transient and state-like (Bull et al., 2005). I predicted that some MT attributes to remain stable (e.g., mental self-concept), and other attributes (e.g., task-specific attention) would be less stable during the period of testing. To investigate MT stability, I correlated the MTI attributes measured at two times (test-retest reliability) during a competitive season under potentially different stress conditions (i.e., during a competition week, during a noncompetition week).
The second aim of this study was to investigate if stress and recovery levels vary depending on whether athletes’ are involved in a competitive match or during a training- only week. The RESTQ-Sport has been used to measure athletes’ changes in stress and recovery corresponding to training load (Kellmann & Günther, 2000; Jürimäe, Mäestu, Purge, & Jürimäe, 2004; Jürimäe, Mäestu, Purge, Jürimäe, & Sööt, 2002). I expected scores on the RESTQ-Sport in a week without the stress of a competitive match to reflect lower stress and increased recovery levels compared to scores in a week after a competitive match.
The primary aim of this study was to investigate how MT and SRI may be related. I anticipated that high levels of MT attributes would be associated with low stress scores and
high recovery scores. To investigate this hypothesis I examined the correlations between the Mental Toughness Inventory (MTI; Middleton, 2005) subscales and the subscales of the Recovery-Stress-Questionnaire for Athletes (RESTQ-Sport; Kellmann & Kallus, 2001). To test for the presence of linear relationships (as MT increases does stress decrease and recovery increase?) and curvilinear relationships (as MT increases does stress decrease up to a point then increase again and does recovery increase to a point and then decrease again?), I examined both linear and quadratic relationships.
Method Participants
The participants for this study were 107 male athletes competing in a variety of sports in Victoria, Australia. Participants’ ages ranged from 16 to 34 years (M = 21.5 years, SD = 4.1 years). Participants were recruited from the Australian Football League (AFL, n = 43), Victorian Football League (VFL, n = 33), field hockey (n = 17), and soccer (n = 14). Years involved in sport at their current levels ranged from 1 to 14 years (M = 4.00, SD = 3.32). AFL represents a demanding contact sport that has a 22-week yearly season with a 4- week final series culminating in two teams competing in the Grand Final. The VFL is a reserve league for the AFL following the same competitive structure. The VFL is a semi- professional league and is a statewide competition, but has more limited press and
television coverage than the AFL. The recruited field hockey players were either part of the national Australian team (n = 5), or represented the state (Victorian) team (n = 12). Soccer players represented either the national Australian under 18s team (n = 9) or the state (Victorian) team (n = 5).
Measures
Demographic information form. On an information sheet developed for this study, I collected demographic information on age, sport and level of competition in which each athlete was involved, and years involved in that sport (see Appendix A).
Mental Toughness Inventory. (MTI; Middleton, Marsh, Martin, Richards, & Perry, 2004b, Middleton, 2005). This 36-item questionnaire was designed to measure 12 attributes (3 items for each) of MT identified during the development of the MTI (i.e., self- efficacy, mental self-concept, potential, task familiarity, personal bests, goal commitment, perseverance, task-specific attention, task value, positivity, stress minimisation, positive comparisons). The MTI subscales have high internal consistency across sub-elite and elite athletes. Middleton et al. reported that the MTI is strong conceptually, because it has a strong theoretical background. Participants are asked to self-report on the items, using an 8- point true-false response scale from 1 (completely false) to 8 (completely true). The MTI has a minimum score of 3 and a maximum of 24 for each of the 12 attributes (see Appendix B).
Recovery-Stress Questionnaire for Athletes. (RESTQ-Sport; Kellmann & Kallus, 2001). The RESTQ-Sport is a 52-item self-report measure of general stress and recovery scales along with sport-specific stress and recovery scales. General stress consists of the following eight subscales: general stress, emotional stress, social stress, conflicts/pressure, fatigue, lack of energy, and physical complaints. General recovery consists of five
subscales: success, social recovery, physical recovery, general wellbeing, and sleep quality. Sport-specific stress consists of three subscales: disturbed breaks, burnout/emotional exhaustion, and fitness/injury. Finally, sport-specific recovery consists of four subscales:
fitness/being in shape, burnout/personal accomplishments, self-efficacy, and self- regulation. Subscale scores are added together to give a total score for each of the four major scales. Participants respond on a 7-point Likert scale from 0 (never) to 6 (always). The RESTQ-Sport identifies current stress/recovery states of athletes, which is important in identifying and addressing the potential for OT. Cronbach’s α values for each of the 19 subscales range from .72 to .93 (Kellmann & Kallus, 2001), demonstrating good internal consistency. Test-retest reliability of each subscale was highly stable after 24 hours with decreasing stability after 3 days (Kellmann & Kallus, 2001). Scores for the four major subscales were used in the data analysis (see Appendix C).
Procedure
After the Victoria University Human Research Ethics Committee granted approval for this study, I approached a number of coaches and explained the project to them. I identified the coaches through the Victorian Institute of Sport (VIS) and through contacts within Victoria University. Five coaches agreed that I could approach the athletes with whom they worked at designated training sessions to explain the study. When I met each athlete group, I provided verbal information about the study, and I supplied participation information sheets and informed consent forms (see Appendix D). All athletes who were present at the agreed-upon training sessions gave consent to be involved in this project. Parents or guardians of participants under the age of 18 completed a parental/guardian informed consent form (see Appendix E). During the first data collection period, all athletes had been involved in a competitive match/competition within the previous 2 days. In all, 125 athletes completed the questionnaires at this time. The questionnaires took
because numerous questions were unanswered. I administered the questionnaires again between 4 and 6 weeks after the initial data collection. None of the participants had been involved in a competitive match the week prior to the second time of testing. As before, athletes completed the questionnaires prior to a training session. Of the original
participants, a number of athletes were unavailable during the second data collection period. One athlete left a number of questions blank, so his data did not contribute to this part of the study. Overall, 108 athletes fully completed the questionnaires at both data collection periods. At the end of the second session, I thanked all the athletes for their participation and offered them a debriefing session if needed.
Data Analysis
I used descriptive statistics (Ms, SDs) to examine the variables (demographic information, RESTQ-Sport scores, and MTI scores). I ran test-retest reliability correlations between the two occasions of administering the MTI to determine the stability of MT attributes. To detect differences in RESTQ-Sport scales from a competition week to a week without competition I ran paired t tests for the four general and sport-specific stress and recovery REST-Q scales. To address the question of how MT attributes might be related to general and sport-specific stress and recovery, I ran 4 (stress and recovery scales) x 12 (MTI subscales) correlation matrixes, using linear and curvilinear (quadratic) analysis. I analysed the MTI and RESTQ-Sport results from Time 1 for these correlations.
Results
In the first section, I present descriptive results that outline athlete scores for MT on the MTI during the week with a competitive match and the week without a competitive match. Test-retest reliabilities for MTI attributes are displayed in Table 3.1. I then present
descriptive results of the RESTQ-Sport for the two data collection sessions. Next, I examine differences in stress/recovery as measured by the REST-Q sport between the two data collection occasions. In the main section, I investigate the primary aim of this study and display correlations between MTI attributes and RESTQ-Sport scales using linear and quadratic analyses.
Preliminary analyses were performed on all data with regard to statistical
assumptions (outliers, normality) and there were no outliers or missing data. I conducted tests for normality and the data was within tolerances of normality for correlational analyses.
Patterns of Mental Toughness
The sample showed approximately normal distributions for the 12 subscales, although the study would need to have a much larger sample size to conclude that there are normal distributions in the population. Due to employing multiple statistical tests and experiment-wise error rates, more stringent p levels were used.
Participants demonstrated a range of scores for the MTI, which has a minimum score of 3 and a maximum of 24. All scores were within the top half of the MTI range as displayed in Table 3.1.
Table 3.1
Means, Standard Deviations, Reliability Coefficients for MTI Attributes Measured during a Week with a Competitive Match and the No Competition Week and Cronbach’s alphas
Note. * indicates p < .001
There appears to be little variability between the MTI attributes. The highest score being seen in personal bests (n = 21.38) and the lowest score in the attribute of positivity (n =
MTI attributes With competitive match Without competitive match M SD M SD Test-retest coefficient Cronbach’s alpha Goal commitment 19.41 3.20 20.39 2.90 .61* .84 Task value 19.67 3.27 19.98 3.21 .64* .75 Personal bests 20.44 2.79 21.38 2.45 .64* .80 Perseverance 19.44 2.97 19.82 2.80 .62* .78 Task-specific attention 16.67 3.76 17.69 3.96 .30 .84 Self-efficacy 17.57 3.93 16.91 3.82 .35 .65 Potential 18.73 3.66 18.56 3.51 .80* .84 Mental self-concept 17.84 3.81 17.89 3.72 .84* .91 Positive comparisons 20.06 3.59 21.01 3.25 .31 .60 Positivity 16.63 3.38 16.99 2.99 .51* .78 Stress minimisation 16.66 3.95 17.00 4.01 .48* .79 Task familiarity 17.11 3.39 17.32 3.35 .84* .75
16.63). All MTI mean scores are within the top third of the scale indicating that athletes involved in this study generally felt strong on all aspects of MT as measured by the MTI.
During the week without a competitive match, changes in mean scores for all subscales were less than +/- 1 from the first data collection period. Cronbach’s alphas were used to determine the internal consistency reliability estimates of the MTI. Reliability was within a range of .60 to .91. At this point in the analysis, the scores on the MTI look stable with an acceptable reliability for a 3-item subscale. The definitions of MT have included suggestions of stable personality traits but also the potential for transient mood states. To investigate further I ran test-retest correlations to determine the stability of MT attributes, and results are displayed in Table 3.1. The reliability coefficients display a large range (from .30 to .84) indicating that some attributes of MT may be more stable than other attributes. If one looks at the means and standard deviations only, then the subscales look quite stable, but the test-retest correlations tell a different story. The means may not be changing much, but the low correlations for some subscales indicate that individuals are changing their scores substantially on some subscales (indication of instability). In this case, looking at the means for the two data collection times is deceptive. If a subscale is supposedly measuring a stable underlying variable, then the test-retest correlation should be about .80 or better. To illustrate how little scores on “positive comparisons” are related to each other at times 1 and 2, one should square the r value and get the percentage of shared variance (i.e., .096 or 9.6%). What the results in Table 3.1 suggest is that most of the subscales have poor test-retest reliability, and whatever they are measuring is not trait-like. Only three subscales, mental self-concept (r = .84), task familiarity (r = .84), and potential (r = .80) reach an acceptable level of test-retest reliability coefficients. Data for both
periods were collected under similar but slightly different conditions, but the change of having a competition one week and not the other week, in the middle of a season, should not have such a dramatic effect (at least for some subscales) on subscale reliability. The different testing conditions go somewhat against the basic principle for test-retest analysis, but if the MTI is supposed to measure trait-like attributes, then there is evidence for only a few subscales having adequate stability. Further analysis of the MTI subscales was
conducted only on those scales that reached a cut-off point for test-retest reliability (e.g., ≥ .80). The choice of this cut-off was to be in keeping with traditional psychometric standards and not use the more liberal cut-off of .70
Patterns of Stress and Recovery
The RESTQ-Sport scales are presented in Table 3.2, namely general stress, general recovery, sport-specific stress, and sport-specific recovery. Each scale consists of a
different number of subscales; therefore, each scale has a different maximum. The
minimum score on each item was zero and the maximum was six. General stress consists of seven subscales, so the maximum score is 42; general recovery has five subscales with a maximum score of 30. Sport-specific stress consists of three subscales and has a maximum score of 18, and sport-specific recovery consists of four subscales with a maximum score of 24. There are no published norms for the RESTQ-Sport scales as the questionnaire has been used primarily for monitoring changes in athletes’ stress and recovery states.
Table 3.2
Means and Standard Deviations for RESTQ-Sport Scales during a Competition Week, and the No Competition Week.
Competition week No competition week
M SD M SD
General stress 13.90 4.66 12.34 4.65
General recovery 16.37 3.87 16.26 3.69
Sport-specific stress 6.76 2.34 6.31 2.54 Sport-specific recovery 13.00 3.56 13.52 2.99
The maximum in the general stress scale is 42, so a mean score of 13.9 appears to be low indicating that the athletes did not feel high levels of general stress. Sport-specific stress is similar in that the maximum score is 18, and athletes scored a mean value of 6.76,
suggesting low stress levels. The maximum score for the general recovery scale is 30 and sport-specific recovery is 24, so both recovery scores displayed in table 3.2 were only slightly above the mid-point of the scale. The stress/recovery profiles display athletes who indicate they are feeling low levels of stress and moderate levels of recovery. The
differences in all four scores between competition and no-competition weeks are so small that they are minor and probably meaningless fluctuations in measurement. For example, on the scale with the largest change, general stress, the mean difference is only 1.56 points. What a mean change of this magnitude means for a scale that can range from 0 to 42 is “probably not much.”
Stress and recovery over two periods. Although the general stress scale displayed a small decrease over the two periods, the initial stress scores in both stress scales were low
in terms of the scoring range. The low stress scores do not leave much room for stress to decrease during the week without competition. Similarly, both recovery scores were in the top half of the scoring range, and, so, if participants were already feeling recovered during a competitive week, they may not have felt notably more recovered during the no-match week.
Relationships between Mental Toughness Attributes and RESTQ-Sport Scales Linear correlations. I carried out linear correlation analyses, using Pearson’s product moment correlation coefficient (r), to determine how the three MT attributes that had adequate test-retest reliability (i.e., potential, mental self-concept, task familiarity) were associated with the four RESTQ-Sport scales. Because most of the MTI subscales do not have adequate reliability (and demonstrate instability), any further analysis of their relationships with the REST-Q scales would result in correlations that are also unreliable