EVALUATIONS OF THE PSYCHOMETRIC PROPERTIES OF THE RECOVERY-STRESS QUESTIONNAIRE FOR ATHLETES AMONG A SAMPLE OF YOUNG FRENCH TABLE TENNIS PLAYERS 1, 2

Psychological Reports: Measures & Statistics 2014, 114, 2, 326-340. Psychological Reports 2014 EVALUATIONS OF THE PSYCHOMETRIC PROPERTIES OF THE RECOVERY-STRESS QUESTIONNAIRE FOR ATHLETES AMONG A SAMPLE OF YOUNG FRENCH TABLE TENNIS PLAYERS 1, 2 GUILLAUME MARTINENT Center for Research and Innovation in Sport (EA 647) University of Lyon, University of Claude Bernard Lyon 1 JEAN-CLAUDE DECRET French Federation of Table Tennis (FFTT), Paris EDITH FILAIRE Laboratory of Complexity Innovation and Motor and Sport Activities (EA 4532) University of Paris Sud University of Orléans SANDRINE ISOARD-GAUTHEUR Sport and Social Environment Laboratory (EA 3742) University of Joseph Fourier, Grenoble 1 CLAUDE FERRAND Psychology of the Various Stages of Life (EA 2114) University of François Rabelais, Tours Summary. This study used confirmatory factor analyses (CFAs) among a sample of young French table tennis players to test: (a) original 19-factor structure, (b) 14-factor structure recently suggested in literature, and (c) hierarchical factor structure of the Recovery-Stress Questionnaire for Athletes (RESTQ Sport). 148 table tennis players completed the RESTQ Sport and other self-report questionnaires between one to five occasions with a delay of 1 mo. between each completion. Results of CFAs showed: (a) evidence for relative superiority of the original model in comparison to an alternative model recently proposed in literature, (b) a good fit of the data for the 67-item 17-factor model of the RESTQ Sport, and (c) an acceptable fit of the data for the hierarchical model of the RESTQ Sport. Correlations between RESTQ Sport subscales and burnout and motivation subscales also provided evidence for criterionrelated validity of the RESTQ Sport. This study provided support for reliability and validity of the RESTQ Sport. Physical and mental recovery in sport has recently received increasing attention in research and practice ( Kallus & Kellmann, 2000 ; Hanin, 2002 ; Kellmann, 2010 ). Researchers suggested that enhanced recovery allows athletes to train more and improve their overall fitness ( Kellmann, 2010 ). Failure to properly recover from stress of training can cause a state of overtraining and burnout, leading to decrease in sport performance ( Kellmann & Kallus, 2001 ). Achievement of optimal performance and avoidance of overtraining can only 1 Address correspondence to Guillaume Martinent, Centre de Recherche et d'innovation sur le Sport, Université Claude Bernard Lyon 1, 27 29 Boulevard du 11 novembre, 69622 Villeurbanne, France or e-mail ( guillaume.martinent@univ-lyon1.fr ). 2 This research was supported by the French Federation of Table Tennis. DOI 10.2466/03.14.PR0.114k18w2 ISSN 0033-2941

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 327 be achievable if athletes are able to optimally balance training stress with adequate recovery ( Kellmann, Altenberg, Lormes, & Steinacker, 2001 ; Hanin, 2002 ). However, common physiological monitoring (e.g., blood analysis) may take days for feedback. Recent research suggested using psychometric selfreport to continuously monitor athletes' subjective experience of stress and recovery ( Kellmann, 2010 ). Nevertheless, because the process of recovery is an active re-establishment of individual athletes' psychological and physical abilities and not only the elimination of stress, the frequently used Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1992 ) may be inadequate for examining and monitoring recovery ( Kellmann, et al., 2001 ). As a result, Kellmann and Kallus (2001 ) developed the Recovery-Stress Questionnaire for Athletes (RESTQ Sport) based on a biopsychological perspective. This questionnaire comprises 19 four-item subscales and is designed to assess the recovery-stress state of athletes (i.e., the extent to which athletes are physically or mentally stressed and their current capabilities to use individual strategies for recovery). In this way, introduction of the RESTQ Sport (Kellmann & Kallus, 2001 ) could be considered as one of the most significant developments in research on recovery in sport because it allows researchers and practitioners to assess simultaneously athletes' stress and recovery ( Davis, Orzeckb, & Keelan, 2007 ). A hierarchical structure of the RESTQ Sport has also been proposed, leading researchers to use both individual subscales and second-order factors of (general and sport-specific) stress and recovery ( Kellmann, 2011 ). Since introduction of the RESTQ Sport, it has been widely used in various sports and countries ( Kallus & Kellmann, 2000 ; Kellmann, et al., 2001 ; Jürimäe, Mäestu, Purge, Jürimäe, & Sööt, 2002 ; Kellmann & Günther, 2000 ; Filaire, Rouveix, & Duclos, 2009 ). This literature demonstrated: (a) effectiveness of the RESTQ Sport in monitoring individuals during training camps or over an entire season ( Kellmann & Günther, 2000 ), (b) effectiveness of the RESTQ Sport in developing concrete recommendations for intervention ( Kallus & Kellmann, 2000 ), (c) corresponding changes with training volume existing between physiological variables and RESTQ Sport scores (Filaire, et al., 2009 ), and (d) close relationship between recovery-stress state and sport performance ( Jürimäe, et al., 2002 ). However, limitations of the psychometric properties of this scale have been indicated ( Davis, et al., 2007 ). Criticisms have been based on shortcomings in original validation studies conducted by Kellmann and Kallus (2001 ). To assess factor structure of general and sport-specific scales, the authors conducted principal components analyses on each of the 12 general or seven sport-specific subscales, rather than the items. An a priori method of identifying the number of subscales was employed instead of an analysis empirically driven by items comprising subscales. Additionally, it is unknown what psychometric properties the RESTQ Sport might have when analysis is conducted on all RESTQ Sport items, because previous research

328 G. MARTINENT, ET AL. has separately analysed 12 general subscales and seven sport-specific subscales. Davis, et al. (2007) used an empirically derived method on RESTQ Sport items to identify critical components of the RESTQ Sport and to confirm or disconfirm its original factor structure. Results failed to confirm the original model and suggested a factor-structure of eight general and six sport-specific subscales. Davis, et al. (2007) also questioned exclusive use of exploratory factor analysis because there was sufficient theoretical basis for specifying models to be tested before analysis. They suggested that further research could test both original and alternative models through confirmatory factor analyses (CFAs). Acquisition of knowledge is dependent on reliable and valid instruments. Development of a valid and reliable inventory for measuring stress and recovery is an important step to: (a) monitor athletes continuously during training and/or competition and (b) provide an easy assessment of early indicators of overtraining and burnout in athletes ( Kellmann, 2010 ). Although introduction of the RESTQ Sport has led to a growing development of research assessing the recovery-stress state of athletes, further evaluation of this questionnaire in independently drawn samples is required ( Davis, et al., 2007 ). The present study differed from previous RESTQ Sport factor analytic research by performing confirmatory factor analyses simultaneously on general and sportspecific stress and recovery items to test: (a) the original 19-factor structure, (b) the 14-factor structure suggested by Davis, et al. (2007), and (c) a hierarchical structure in which second-order latent variables of general and sport-specific stress and recovery were added. Criterion-related validity evidence was also provided through correlations of RESTQ Sport subscales with motivation and burnout subscales. Stage 1: Factorial Validity and Reliability Evidences of RESTQ Sport M ETHOD Participants A total of 46 female and 102 male table tennis players in intensive training centers ( M hr. of training per week = 15.3, SD = 5.9) ranging in age from 13 to 19 yr. ( M = 14.2, SD = 2.1) voluntarily participated in this study. On average, they had been competing for 6.5 yr. ( SD = 2.3), and they participated in regional ( n = 29), national ( n = 81), or international sport events ( n = 38). This sample represented nearly 90% of all young French table tennis players involved in intensive training centers. These centers receive the best young athletes in France in order to help them reach the highest level of sport performance and achieve good academic results. Measure The RESTQ Sport ( Kellmann & Kallus, 2001 ) consists of 76 items that assess seven general stress subscales (general stress, emotional stress, social

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 329 stress, conflicts/pressure, fatigue, lack of energy, and physical complaints), five general recovery subscales (success, social recovery, physical recovery, general well-being, and sleep quality), three sport-specific stress subscales (disturbed breaks, emotional exhaustion, and injury), and four sport-specific recovery subscales (being in shape, personal accomplishment, self-efficacy, and self-regulation), with four items for each subscale. RESTQ Sport items are written in a simple and easy to understand language. The RESTQ Sport was first translated into French and sent to two bilingual translators who then translated it back into English. Differences were resolved so that the original meaning of each item was considered to be present in the final French version. Subsequently, comprehensibility, acceptability, relevance, and completeness of all items were discussed with eight young table tennis players not involved in this study. No changes were considered necessary. The response scale had participants rate frequency of each item during the past 3 days/nights on a 7-point, Likert-type scale with anchors 0: Never and 6: Always. Procedure Coaches of each athlete were contacted to obtain permission to approach their athletes and ask them to participate in the study. Athletes' participation was voluntary, their anonymity was ensured, and parental consent was required for athletes under 18 years of age. Participants completed the RESTQ Sport between one to five occasions with a delay of 1 mo. between each completion, resulting in 614 cases. Data Analysis The RESTQ Sport is designed to assess athletes' perception of their state of recovery and stress in the prior 3 days. Kellmann and Kallus (2001) showed large fluctuations of RESTQ Sport subscales if the delay between each completion was of 1 mo., as is the case in the present study. Measurement occasions of the current study could thus be considered as independent measurements. This is why the raw data of five measurement occasions ( N = 614) were combined for data analyses. Factorial validity of the RESTQ Sport was examined with a series of CFAs to better identify sources of poor overall model fit. First, tenability of original general and sport-specific RESTQ Sport subscales was tested independently. Second, these models were respectively compared to their alternative models suggested by Davis, et al. (2007). Third, the original general model was compared to an alternative 39-item, 10-factor general model in which problematic items and subscales were discarded. Fourth, original general and sport-specific models were respectively compared to hierarchical models in which second-order latent variables of stress and recovery were added. Fifth, the overall measurement model of the RESTQ Sport including simultaneously general and sport-specific items was tested.

330 G. MARTINENT, ET AL. Models were tested using maximum likelihood estimation on covariance matrices. To achieve a comprehensive evaluation of fit, multiple fit indices were chosen ( Hu & Bentler, 1999 ): chi square (χ 2 ), Bentler Bonett non-normed fit index (NNFI), comparative fit index (CFI), standardised root mean square residual (SRMR), root mean squared error of approximation (RMSEA), and confidence interval of RMSEA (90% CI ). Hu and Bentler (1999 ) recommended as good fit: CFI and NNFI 0.95, SRMR 0.08, and RMSEA 0.06. Akaike information criterion (AIC) and expected crossvalidation index (ECVI) were used for comparison of several models. R ESULTS Kaiser-Meyer-Olkin measure of sampling adequacy (0.95) and Bartlett's Test of Sphericity (approx. χ 2 = 26,368.98, p <.001) provided a minimum 2850 standard that should be passed before CFAs are conducted. No significant multivariate outliers ( p <.001) were identified. Both skewness (< 1.7) and kurtosis (< 2.4) values of all RESTQ Sport items indicated data were approximately normally distributed ( Hair, Black, Babin, & Anderson, 2010 ). CFAs of RESTQ Sports general subscale. Error terms from two similar physical complaints Items 6 and 41 and from two similar sleep quality Items 35 and 45 were correlated in all CFAs. Goodness-of-fit indices of original RESTQ Sport general subscales (48-item, 12-factor model) were slightly below cut off criterion values ( Table 1 ). Loadings of social recovery Items 5 (.29) and 22 (.12), success Item 16 (.37), and conflict Item 31 (.31) were below the cut off criterion of.40. This model was compared to an alternative 45-item, eight-factor model suggested by Davis, et al. (2007). Chi-squared difference tests and AIC and ECVI values provided evidence for relative superiority of original model. Goodness-of-fit indices of alternative model were below cut off criterion values ( Table 1 ). Loadings of seven items (Items 2, 5, 22, 31, 38, 40, and 48) were below.40. Then, an alternative 39-item, 10-factor general model was tested, in which two four-item subscales of success and social recovery and conflict Item 31 were discarded from the original general model. This decision was based on rationale that: (a) correlations between success subscale and other general RESTQ Sport subscales were largely inconsistent with those of the original validation study ( Kellmann & Kallus, 2001 ) 3, (b) reliability of suc- 3 Kellmann and Kallus (2001 ) showed that the subscale of success was: (a) negatively correlated or non-significantly correlated with all seven general stress subscales of the RESTQ Sport and (b) strongly (r >.40) positively correlated with all the other four general recovery subscales of the RESTQ Sport. In contrast, results of the present study showed that the subscale of success was: (a) significantly positively correlated with all seven general stress subscales (M ø =.31, range =.21.46), (b) weakly significantly positively correlated with three general recovery subscales (ø =.16 for general well-being,.27 for somatic relaxation, and.34 for disturbed breaks), and (c) non-significantly correlated with the subscale of sleep quality (ø =.08, p >.05). These results seem to be an indication of a poor criterion-related validity of the subscale of success in the present study.

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 331 cess and social recovery subscales was unacceptable in the present study, 4 and (c) loadings of social recovery, success, and conflict items were below.40. Results suggested that the alternative model provided a good fit to the data and a better fit than original model ( Table 1 ). All λ were significant and higher than.40. This revised model was compared to hierarchical model in which second-order latent variables of general stress and recovery were added. Results suggested that the 39-item, 10-factor general model provided better fit to the data than hierarchical model. Nevertheless, goodnessof-fit indices of the hierarchical model reached cut off criterion values and were only slightly below those of the non-hierarchical model ( Table 1 ). CFAs of RESTQ Sport sport-specific subscales. Error terms from two similar self-efficacy Items 51 and 58, from two similar self-regulation Items 55 and 66, and from two similar burnout/personal accomplishment Items 59 and 76 were correlated in CFAs. Goodness-of-fit indices of the original 28- item, seven-factor sport-specific model of RESTQ Sport reached cut off criterion values ( Table 1 ). All λ were statistically significant and above.40. This model was compared to an alternative 28-item, six-factor sport-specific model suggested by Davis, et al. (2007). Chi-squared difference tests and AIC and ECVI values provided evidence for relative superiority of the original sportspecific model ( Table 1 ). Goodness-of-fit indices of Davis, et al. s (2007) sportspecific model were largely below cut off criterion values ( Table 1 ). The original sport-specific model was also compared to a hierarchical sport-specific model in which second-order latent variables of sport-specific stress and recovery were added. Results suggested that the original sport-specific model provided better fit to the data than the hierarchical sport-specific model even if goodness-of-fit indices of the hierarchical model were only slightly below those of the non-hierarchical model ( Table 1 ). CFA of RESTQ Sport overall measurement model. The 67-item, 17-factor model of the RESTQ Sport provided a good fit to the data ( Table 1 ). All λ were significant ( t > 1.96). Standardised factor loadings and error variances are shown in Fig. 1. Inter-scale correlations and correlations between latent constructs are presented in Table 2. Pearson's correlations were consistent with those of the English original validation study. Results showed: (a) significant positive correlations between stress subscales (.46 < r s <.76), (b) statistically significant positive correlations between recovery subscales (.16 < r <.70) except for correlation between sleep quality and personal accomplishment ( r =.06, p >.05), (c) significant negative correlations s between 4 Cronbach s α coefficients of success and social recovery subscales were respectively.56 and.51, indicating that reliability of these two subscales was unacceptably low. Reliability of these two subscales was also unacceptable when problematic items were discarded from analyses (Success α =.59 when Item 16 was discarded; social recovery α =.56 when Items 5 and 22 were discarded).

332 G. MARTINENT, ET AL. TABLE 1 F IT INDICES FOR GENERAL SCALES, SPORT-SPECIFIC SCALES, AND OVERALL MEASUREMENT MODELS OF RESTQ SPORT χ 2 df CFI NNFI SRMR RMSEA 90% CI AIC ECVI Δ χ 2 Δ df General scales a 48-item, 12-factor general model 3653.58 1012 0.96 0.96 0.089 0.065 0.063 0.068 3981.58 6.50 b 45-item, 8-factor general model 4481.09 915 0.96 0.96 0.082 0.080 0.077 0.082 4721.09 7.70 827.51 97 c 39-item, 10-factor general model 2333.83 655 0.97 0.97 0.060 0.065 0.062 0.068 2583.83 4.22 1319.75 357 d Hierarchical general model 2776.01 689 0.97 0.97 0.070 0.070 0.068 0.073 2958.01 4.83 442.18 34 Sport-specific scales a 28-item, 7-factor sport-specific model 1308.20 326 0.96 0.95 0.067 0.070 0.066 0.074 1468.20 2.40 b 28-item, 6-factor sport-specific model 2371.25 332 0.93 0.91 0.093 0.100 0.096 0.100 2519.25 4.11 1063.05 6 d Hierarchical sport-specific model 1527.92 339 0.94 0.94 0.100 0.076 0.072 0.080 1661.92 2.71 219.72 13 Overall measurement models 67-item, 17-factor model 5814.91 2003 0.97 0.97 0.066 0.056 0.054 0.057 6364.91 10.38 d Hierarchical model 8116.65 2116 0.96 0.96 0.095 0.068 0.066 0.070 8440.65 13.77 2301.74 113 Note. a Original measurement model ( Kellmann & Kallus, 2001 ); b c Models suggested by Davis, et al. (2007 ); 48-item, 12-factor general model in which two four-item subscales of success and social recovery and conflict Item 31 were discarded; d Second-order latent variables of general stress, general recovery, sport-specific stress, and/or sport-specific recovery were added. p <.001.

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 333 TABLE 2 D ESCRIPTIVE STATISTICS AND CORRELATION MATRIXES OF 67-ITEM, 17-FACTOR RESTQ SPORT SUBSCALES, SMS SUBSCALES, AND ABQ SUBSCALES Scale and Subscale M SD Pearson s Correlation 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 General stress subscales of RESTQ Sport 1. General stress 1.10 1.00. 85.91.85.88.64.81.85.61.69.67.70.77.62.44.10*.47.20 2. Emotional stress 1.45 0.98.76. 79.91.90.70.90.86.63.70.76.64.76.65.48.01*.45.12 3. Social stress 1.48 1.06.73.76. 83.86.55.76.73.45.52.62.55.62.53.29.02*.36.13 4. Conflicts/pressure 1.90 1.22.71.69.70. 75.65.84.80.50.54.67.62.70.62.35.15.37.01* 5. Fatigue 2.09 1.16.53.58.49.53. 81.67.93.55.34.65.67.85.80.49.13.17.05* 6. Lack of energy 1.78 0.91.61.65.58.58.51. 65.79.55.52.67.72.80.67.45.06*.45.09* 7. Physical complaints 1.67 1.05.69.67.60.60.72.55. 76.73.55.75.73.91.89.61.10*.29.05* General recovery subscales of RESTQ Sport 8. Physical recovery 3.28 0.91.44.44.33.33.34 0.34.44. 65.78.74.41.52.59.95.39.67.48 9. General well-being 4.14 0.88.58.57.45.41.28 0.39.43.62. 80.67.43.49.35.63.33.63.39 10. Sleep quality 4.09 1.16.58.62.55.56.56 0.51.62.49.51. 82.59.59.57.67.23.53.27 Sport-specific stress subscales of RESTQ Sport 11. Disturbed breaks 1.36 1.08.58.52.47.50.56.53.58.26.35.52. 82.77.66.33.09*.13.05* 12. Emotional exhaustion 1.70 1.06.63.60.54.53.63.58.66.34.42.50.59. 71.84.46.11*.24.00* 13. Injury 2.11 1.25.53.53.46.49.66.50.69 37.29.50.55.63. 80.50.17.18.04* (continued on next page) Note. Correlations between latent constructs for higher data matrix; Pearson s correlations for lower data matrix; IM = intrinsic motivation. Values on the diagonal in italics are Cronbach s α. * p >.05.

334 G. MARTINENT, ET AL. TABLE 2 (CONT D) D ESCRIPTIVE STATISTICS AND CORRELATION MATRIXES OF 67-ITEM, 17-FACTOR RESTQ SPORT SUBSCALES, SMS SUBSCALES, AND ABQ SUBSCALES Scale and Subscale M SD Pearson s Correlation 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Sport-specific recovery subscales of RESTQ Sport 14. Being in shape 3.45 1.06.38.40.26.27.42.34.49.70.54.53.30.37.41. 84 15. Personal accomplishment 2.83 1.10.01*.06*.10.18.12.11.11.25.20.06*.11.08.13.30. 68.47.68.53.79.85 16. Self-efficacy 3.44 1.11.37.35.28.27.14.29.21.52.50.37.10.24.15.53.47. 82.85 17. Self-regulation 3.21 1.17.15.08.08.03*.05*.05*.03*.37.30.16.06*.04*.05*.42.54.64. 80 Motivation subscales (SMS) 18. IM to known and toward accomplisment 3.62 0.80.23.17.06*.12.13.17.21.28.32.23.19.19.17.36.28.29.30 19. IM to experience stimulation 4.13 0.86.17.06*.00*.00*.01*.06*.09.21.24.14.12.15.06*.30.28.27.35 20. Identified regulation 2.66 0.88.02*.10.14.08.08.11.04*.11.09.01*.05*.05*.08.08*.24.04*.11 21. Introjected regulation 2.61 0.82.18.25.29.26.18.18.18.12.00*.11.16.14.19.07*.21.01*.10 22. External regulation 1.48 0.64.38.37.31.31.22.35.29.08*.27.26.37.42.22.15.09.14.10 23. Amotivation 1.43 0.64.45.43.34.32.22.38.28.16.39.29.33.49.23.25.11.38.30 Burnout subscales (ABQ) 24. Reduced accomplishment 2.43 0.73.47.52.32.43.36.47.38.22.45.40.34.49.37.34.10.52.24 25. Sport devaluation 1.78 0.83.34.41.34.29.22.33.24.10.29.26.28.41.22.16.06*.33.25 26. Emotional and physical exhaustion 2.82 0.90.25.29.15.27.58.28.47.03*.08.26.34.49.49.17.20.13.25 Note. Correlations between latent constructs for higher data matrix; Pearson's correlations for lower data matrix; IM = intrinsic motivation. Values on the diagonal in italics are Cronbach's α. * p >.05.

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 335.37 Item 21.79.39 Item 23.78.66.56 Item 29.77.41 Item 44.61 Item 4.62.50 Item 7.71.69.53 Item 27.77.40 Item 36.32 Item 20.82.36 Item 25.80.67.55 Item 37.71.50 Item 47.39 Item 11.78.64.59 Item 17.72.48 Item 43.64 Item 1.45 Item 15.48 Item 24.30 Item 34.77 Item 3.56 Item 10.70 Item 30.67 Item 39.60.74.72.84.48.66.55.57 General stress Emotional stress Social stress Conflicts / pressure Fatigue Lack of energy.47 Item 6.70 Item 14.57 Item 19.50 Item 41.60 Item 8.84 Item 12.38 Item 28.78 Item 38.65 Item 9.47 Item 33.43 Item 42.40 Item 46.48 Item 18.38 Item 26.44 Item 35.66 Item 45.46 Item 50.44 Item 57.40 Item 65.52 Item 71.48 Item 55.65 Item 61.73.55.61.71.63.40.79.47.59.73.75.77.72.79.75.58.73.75.77.69 Physical complaints Physical recovery General well-being Sleep quality Disturbed breaks.53 Item 53.51 Item 62.80 Item 67.65 Item 75.43 Item 49.57 Item 56.42 Item 63.55 Item 72.63 Item 52.30 Item 60.36 Item 68.38 Item 74.54 Item 54.75 Item 59.68 Item 69.83 Item 76.46 Item 51.54 Item 58.58 Item 64.46 Item 70.69.70.45.59.75.66.76.67.61.84.80.79.68.50.57.41.73.68.65.73 Selfefficacy.72 Selfregulation.67 Item 66.55.59.75 Item 73.44 Emotional exhaustion Fitness / injury Being in shape Personal accomplishment F IG. 1. Measurement model of the 17-factor, 67-item model of the RESTQ Sport. Circles represent latent constructs, and squares represent manifest variables. All parameters are standardized and significant at p <.05. Residual variances are shown in small circles. general recovery subscales and stress subscales (.26 < r s <.62), (d) significant negative correlations between sport-specific recovery subscales of being in shape and self-efficacy and stress subscales (.10 < r s <.49), and (e) non-significant or weak positive significant correlations between sportspecific recovery subscales of personal accomplishment and self-regulation and stress subscales (.01 < r s <.18). Correlations between latent constructs were stronger and consistent with Pearson's correlations. Confidence intervals ( ± 2 SD ) support discriminant validity, insofar as intervals did not include 1.0 ( Anderson & Gerbin, 1988 ). The 67-item, 17-factor model was compared to a hierarchical model in which four second-order latent variables of general and sport-specific stress and recovery were added. The 67- item, 17-factor model provided better fit to the data than the hierarchical model. Nevertheless, goodness-of-fit indices of the hierarchical model approach those of the first-order model ( Table 1 ). Reliability. The reliability of the RESTQ Sport was assessed by examining the Cronbach's αs of the RESTQ Sport subscales. Alpha coefficients indicated that reliability of each of the 17 subscales was acceptable, with Cronbach's αs ranging from.65 to.85 ( Table 2 ). Cronbach's α for lack of energy (α =.65), somatic relaxation (α =.65), and personal accomplishment (α =.68) subscales met criteria for an adequate internal consistency (α >.60) because these subscales had only four items ( Hair, et al., 2010 ).

336 G. MARTINENT, ET AL. Stage 2: Criterion-related Validity Evidence of RESTQ Sport M ETHOD Participants, Procedure, and Data Analysis Participants also completed the Sport Motivation Scale (SMS) and Athlete Burnout Questionnaire (ABQ) on each measurement occasion. Criterion-related convergent validity evidence was provided through correlations with motivation and burnout. Correlations are interpreted using Cohen's (1988) criteria (i.e., small.30; medium =.30 to.50; large.50). Measures The French version of the SMS ( Brière, Vallerand, Blais, & Pelletier, 1995 ) was used to measure motivation. SMS is derived from tenets of the Self-Determination Theory (SDT; Deci & Ryan, 1985 ) and comprises six subscales that assess different types of motivation: intrinsic motivation to know and move toward accomplishment, intrinsic motivation to experience stimulation, identified regulation, introjected regulation, and external regulation and amotivation. Previous research lent credence to validity and reliability of the SMS ( Brière, et al., 1995 ; Li & Harmer 1996 ). Each item was rated on a 5-point Likert-type scale with anchors 1: Does not correspond at all and 5: Corresponds exactly. Alpha coefficients indicated that reliabilities were acceptable with Cronbach's αs of.76,.88,.87,.83,.79, and.83, respectively. The French version of the ABQ ( Isoard-Gautheur, Oger, Guillet, & Martin-Krumm, 2010 ) was used to measure athlete burnout. The French ABQ consists of 12 items that assess reduced accomplishment, sport devaluation, and emotional and physical exhaustion, with four items measuring each dimension. Isoard-Gautheur, et al. (2010) used CFA to assess psychometric properties of the ABQ. They concluded there was a good fit for the hypothesized model on a sample of 895 French youth involved in competitive sport or physical education at school (age range = 11 19 yr.; NNFI = 0.95, CFI = 0.96, GFI = 0.95, RMSEA = 0.07; Isoard-Gautheur, et al., 2010 ). Each item was rated on a 5-point Likert-type scale with anchors 1: Almost never and 5: Most of the time. Alpha coefficients indicated that reliabilities were acceptable with Cronbach's αs of.70,.82, and.90. R ESULTS Motivation Intrinsic motivation to know and move toward accomplishment correlated negatively with stress subscales (.12 < r s <.23) and positively with recovery subscales (.23 < r s <.36). Intrinsic motivation to experience stimulation showed non-significant or negative significant correlations

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 337 with stress subscales (.00 < r s <.17) and positive significant correlations with recovery subscales (.14 < r s <.35). Identified regulation showed nonsignificant to weak positive correlations with stress and recovery subscales (.01 < r s <.24). Introjected regulation showed positive significant correlations with stress subscales (.14 < r s <.29) and non-significant or low significant correlations with recovery subscales (.11 < r s <.21). External regulation and amotivation correlated positively with stress subscales (.22 < r s <.49) and negatively with recovery subscales (.10 < r s <.38; see Table 2 for more details). Athlete burnout. Reduced accomplishment correlated positively with stress subscales (.32 < r s <.49) and negatively with recovery subscales (.10 < r s <.52). Sport devaluation correlated positively with stress subscales (.22 < r s <.41) and negatively with recovery subscales (.06 < r s <.33). Emotional and physical exhaustion correlated (a) positively with stress subscales (.15 < r <.58), self-efficacy ( r =.13), personal accomplishment ( r =.17), and s self-regulation ( r =.25), and (b) negatively with general well-being ( r =.08), being in shape ( r =.17), and sleep quality ( r =.26; see Table 2 for more details). D ISCUSSION Construct validity of the RESTQ Sport was supported by several arguments. CFAs revealed acceptable fits between participants responses and the 39-item, 10-factor model of general subscales; 28-item, seven-factor model of sport-specific subscales; and the 67-item, 17-factor model of the RESTQ Sport overall measurement model. These results are consistent with the original factor structure of the RESTQ Sport and do not support an alternative model recently suggested by Davis, et al. (2007 ). An improved model fit of the RESTQ Sport was obtained after deleting success and social relaxation items. Previous research already indicated potential psychometric weakness for these subscales. Kellmann and Kallus (2001 ) reported α coefficients of less than.60 for the success subscale on two samples of the validation study, whereas a factor analysis conducted by Davis, et al. (2007 ) highlighted low loadings (<.36) for three items originally designed to measure success and social recovery (Items 16, 22, and 49). Inter-scale correlations were consistent with those of the original validation study ( Kellmann & Kallus, 2001 ), suggesting that 17 dimensions of the RESTQ Sport are tapping unique, yet correlated, dimensions of athletes' recovery and stress. Based on exploratory factor analyses of the RESTQ Sport, previous research suggested using second-order factors of general and sport-specific stress and recovery in addition to individual RESTQ Sport subscales ( Kellmann & Kallus, 2001 ; Kellmann 2010 ). In the present study, the hierarchical model produced fit values that were only marginally lower than that of the

338 G. MARTINENT, ET AL. first-order model. Marsh (1987 ) remarked that fit of a second-order model cannot be better than fit of an equivalent first-order structure. Therefore, he suggested that if fit of a higher model approaches that of a first-order model, a hierarchical structure should be preferred because it is more parsimonious. As a result, it is suggested that the hierarchical model of the RESTQ Sport should be adopted by researchers interested in a general measure of recovery-stress state of athletes (e.g., for structural equation modeling). For those examining relationships between specific recovery and stress dimensions and other concepts or outcomes, the 17-factor model of the RESTQ Sport would likely be most applicable since it provides a more in-depth assessment. Another validity evidence of the RESTQ Sport is that its subscales related to athlete burnout and motivation in accordance with theoretical expectations ( Deci & Ryan, 1985 ; Kallus & Kellmann, 2000 ). Consistent with self-determination theory ( Deci & Ryan, 1985 ), stress subscales correlated positively with the less self-determined types of motivation (external regulation and amotivation) and negatively with the more self-determined types of motivation (intrinsic motivation), whereas recovery subscales correlated negatively with external regulation and amotivation and positively with intrinsic motivation. As expected, identified and introjected regulations also showed non-significant or weak significant correlations with RESTQ Sport subscales. Burnout is conceptualized as the result of chronic exposure to stress and insufficient recovery leading to detrimental consequences such as drastic performance decrements or dropout of sport ( Kallus & Kellmann, 2000 ). Consistent with this conceptualization, results showed that stress subscales correlated positively with all three dimensions of sport burnout, whereas recovery subscales correlated negatively with reduced accomplishment and sport devaluation. Nevertheless, a somewhat surprising result concerned correlations between emotional and physical exhaustion and three RESTQ Sport sportspecific recovery subscales. In contrast with previous research ( Kellmann & Kallus, 2001 ; Isoard-Gautheur, et al., 2010 ), emotional and physical exhaustion correlated positively with personal accomplishment, self-efficacy, and self-regulation. These results could be explained by the characteristics of the sample. Participants were characterized by low scores of stress, reduced accomplishment, and sport devaluation. As a result, when participants reported relatively high scores on emotional and physical exhaustion, this could simply signify that they had trained hard, leading to high scores for personal accomplishment, self-efficacy, and self-regulation. Further research should clarify whether the unexpected positive relationships observed in the present study between emotional and physical exhaustion and personal accomplishment, self-efficacy, and self-regulation are robust results that could be generalized to other samples (e.g., athletes reporting high scores for stress and burnout).

PSYCHOMETRIC EVALUATIONS OF RESTQ SPORT 339 Given the specificity of the sample (young table tennis players in intensive training centers), caution should be exercised before generalizing the findings. We chose this sample because these athletes seemed vulnerable to stress and burnout as they must cope with lots of stressors in their everyday lives (e.g., practice and training, academics, competition, and relationships; Kellmann & Kallus, 2001 ). More validation studies are required to evaluate whether results are invariant across type of sport, characteristic of athletes, or age. It seems also important to note that the sample size is rather small. Notwithstanding these criticisms, this study enhanced previous stress and recovery researches by showing that the RESTQ Sport is a reliable and valid tool for estimating the recovery-stress state of athletes. To our knowledge, the present study is the first to empirically test psychometric properties of the RESTQ Sport through CFAs. Results of CFAs strongly supported the 67-item, 17-factor model and provided evidence for factorial validity and reliability of the RESTQ Sport. Correlations between RESTQ Sport, ABQ, and SMS subscales provided evidence for criterion-related validity of the RESTQ Sport. REFERENCES A NDERSON, J. C., & GERBIN, D. W. (1988) Structural equation modeling in practice: a review and recommended two-step approach. Psychological Bulletin, 103, 411-423. B RIÈRE, N. M., VALLERAND, R. J., BLAIS, M. R., & PELLETIER, L. G. (1995) On the development and validation of the French form of the Sport Motivation Scale. International Journal of Sport Psychology, 26, 465-489. C OHEN, J. (1988) Statistical power analysis for the behavioral sciences. (2nd ed.) Hillsdale, NJ : Erlbaum. D AVIS, H., O RZECKB, T., & K EELAN, P. (2007) Psychometric item evaluations of the Recovery-Stress Questionnaire for Athletes. Psychology of Sport and Exercise, 8, 917-938. D ECI, E. L., & RYAN, R. M. (1985) Intrinsic motivation and self-determination in human behavior. New York : Plenum. F ILAIRE, E., R OUVEIX, M., & DUCLOS, M. (2009) Training and 24-hr urinary catecholamine excretion. International Journal of Sports Medicine, 30, 33-39. H AIR, J. F., B LACK, W. C., B ABIN, B. J., & ANDERSON, R. E. (2010) Multivariate data analysis. (7th ed.) Englewood Cliffs, NJ : Prentice-Hall. H ANIN, Y. L. (2002) Individually optimal recovery in sports: an application of the IZOF model. In M. Kellmann (Ed.), Enhancing recovery: preventing underperformance in athletes. Champaign, IL : Human Kinetics. Pp. 199-217. H U, L., & BENTLER, P. M. (1999) Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. I SOARD-GAUTHEUR, S., O GER, M., GUILLET, E., & MARTIN-KRUMM, C. (2010) Validation of a French version of the Athlete Burnout Questionnaire (ABQ) in competitive sport and physical education context. European Journal of Psychological Assessment, 26, 203-211.

340 G. MARTINENT, ET AL. J ÜRIMÄE, J., M ÄESTU, J., P URGE, P., J ÜRIMÄE, T., & S ÖÖT, T. (2002 ) Relations among heavy training stress, mood state, and performance for male junior rowers. Perceptual & Motor Skills, 95, 520-526. K ALLUS, K. W., & K ELLMANN, M. (2000 ) Burnout in athletes and coaches. In Y. L. Hanin (Ed.), Emotions in sport. Champaign, IL : Human Kinetics. Pp. 209-230. K ELLMANN, M. (2010 ) Preventing overtraining in athletes in high-intensity sports and stress/recovery monitoring. Scandinavian Journal of Medicine & Science in Sports, 20, 95-102. K ELLMANN, M., A LTENBERG, D., LORMES, W., & S TEINACKER, J. M. (2001 ) Assessing stress and recovery during preparation for the world championships in rowing. The Sport Psychologist, 15, 151-167. K ELLMANN, M., & GÜNTHER, K. D. (2000 ) Changes in stress and recovery in elite rowers during preparation for the Olympic games. Medicine & Science in Sports & Exercise, 32, 676-683. K ELLMANN, M., & KALLUS, K. W. ( 2001 ) The Recovery-Stress Questionnaire for Athletes: user manual. Champaign, IL : Human Kinetics. LI, F., & HARMER, P. H. (1996 ) Testing the simplex assumption underlying the sport motivation scale: a structural equation modeling analysis. Research Quarterly for Exercise and Sport, 67, 396-405. M ARSH, H. W. (1987 ) The hierarchical structure of self-concept and the application of hierarchical confirmatory factor analysis. Journal of Educational Measurement, 24, 17-39. M CNAIR, D., LORR, M., & DROPPLEMAN, L. F. (1992 ) Profile of Mood States manual. San Diego, CA : Educational and Industrial Testing Service. Accepted January 30, 2014.

Copyright of Psychological Reports is the property of Ammons Scientific, Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.