lable at ScienceDirect

Psychology of Sport and Exercise 15 (2014) 227e237

Contents lists avai

Psychology of Sport and Exercise

journal homepage: www.elsevier .com/locate/psychsport

Evaluation of the psychometric properties of a modified Positive andNegative Affect Schedule including a direction scale (PANAS-D) amongFrench athletes

Michel Nicolas a,1,2, Guillaume Martinent b,*,1, Mickaël Campo c

aUniversité de Bourgogne, Laboratory of Socio Psychology and Management of Sport, SPMS (EA 4180), Faculty of Sport Sciences, Dijon, FrancebUniversité de Lyon, Université Claude Bernard Lyon 1, Centre de Recherche et d’Innovation sur le Sport, 27e29 Boulevard du 11 Novembre, 69622Villeurbanne, FrancecUniversité de Rouen, CETAPS, Faculty of Sport Sciences, Bd Siegfried, 76 821 Mont Saint Aignan Cedex, France

a r t i c l e i n f o

Article history:Received 15 July 2013Received in revised form7 January 2014Accepted 10 January 2014Available online 2 February 2014

Keywords:Affective statesConfirmatory factor analysisDirectionIncremental validityIntensityPANAS

* Corresponding author. Tel.: þ33 4 72 43 28 38; faE-mail addresses: michel.nicolas@u-bourgogne.f

martinent@univ-lyon1.fr (G. Martinent), mickael.camp1 Note: The first two authors played equal roles in

and should both be considered as first authors.2 Tel.: þ33 3 80 39 90 11.

1469-0292/$ e see front matter � 2014 Elsevier Ltd.http://dx.doi.org/10.1016/j.psychsport.2014.01.005

a b s t r a c t

Objectives: The goal of these studies was to provide validity and reliability evidence of a modified Pos-itive and Negative Affect Schedule (PANAS) including a direction scale (PANAS-D). Study 1 tested thevalidity and reliability of the PANAS-D to measure both intensity and direction of affects. Study 2examined the relationships between direction of affects and selected variables (i.e., coping, attainment ofachievement goals and sport satisfaction) by controlling for intensity of affects.Method: A total of 306 and 296 athletes (studies 1 and 2) completed the PANAS-D and other self-reportquestionnaires. Data were analysed with reliability, confirmatory factor analyses (study 1) and correla-tional analyses (studies 1 and 2).Design: Cross-sectional with self-reported questionnaires.Results: In study 1, the 4-factor structure of the PANAS-D (intensity and direction of positive affect andnegative affect) fitted the data adequately. Multiple-group CFAs showed that PANAS-D was partiallyinvariant across the two measurement occasions (before and after competition). The patterns of re-lationships between PANAS-D, attainment of achievement goals and coping provided evidence for thecriterion-related validity of the PANAS-D. In study 2, direction of positive affect and negative affect wereassociated with selected outcomes (i.e., coping, attainment of sport achievement goals, and/or sportsatisfaction) after intensity of these affective states were held constant.Conclusions: This study provided support for the reliability and validity of the PANAS-D (study 1) and theincremental validity of the direction of affective states (study 2), supporting the distinction betweenathletes’ intensity and direction of affective states.

� 2014 Elsevier Ltd. All rights reserved.

Two theoretical conceptions have guided research on thestructure of affective states. From the categorical perspective, theyare organized in distinct categories such as anger, anxiety, orhappiness (Lazarus, 2000). From the dimensional perspective, af-fective states are categorised into higher-order dimensions (e.g.,positive and negative affects) on the basis of the relationshipsamong discrete emotions (Watson, Clark, & Tellegen, 1988). Aconsensus has emerged that both conceptions have relative

x: þ33 4 72 43 28 46.r (M. Nicolas), guillaume.o@univ-rouen.fr (M. Campo).the preparation of this article

All rights reserved.

advantages and limitations (Lazarus, 2000). Categorical conceptu-alization offers the advantage for more refined discrimination ofpsychological meanings whereas dimensional conceptualizationoffers the advantage for a parsimonious representation of theglobal affective space (Lazarus, 2000).

Researchers in sport psychology have traditionally focused onthe intensity of affective states with a predominant focus on pre-competitive anxiety based on the rationale that this affectivestate is thought to affect athletic performance (Mellalieu, Hanton, &Fletcher, 2006). This literature has consistently shown that athletescan experience a wide range of affects e at various intensities e

likely to facilitate or impair sport performance (e.g., Hanin, 2007;Martinent, Campo, & Ferrand, 2012). In recent years, growingempirical attention has been allocated to the direction componentof affective states in an effort to further disentangle the anxiety

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237228

performance relationship; i.e., perceived facilitating or debilitatingeffects of athletes’ affective states on their performance (e.g., Hanin,2007; Martinent & Ferrand, 2009). Despite their respective pleasantand unpleasant valence, positive and negative affective states canbe perceived by athletes either as facilitating or debilitating fortheir sport performance (Hanin, 2007). A positive affective state e

experienced at a particular intensity level e could thus be inter-preted as facilitating for performance for a certain athlete at aparticular point in time and as debilitative for the same athlete atother points in time (Lazarus, 2000; Martinent et al., 2012;Martinent, Nicolas, Gaudreau, & Campo, 2013).

Originally, the addition of a direction scale (Jones & Swain, 1992)to the Competitive State Anxiety Inventory-2 (CSAI-2, Martens,Burton, Vealey, Bump, & Smith, 1990) has resulted in a plethora ofpublications investigating the direction of athletes’ pre-competitivestate anxiety (see for a reviewMellalieu et al., 2006). Especially, thisliterature has consistently shown that anxiety direction is associ-ated with key athletic outcomes (e.g., sport performance) aftercontrolling for anxiety intensity (Mellalieu et al., 2006). Specifically,the more elite or better performers in competition interpreted theintensity of their anxiety symptoms as more facilitative comparedto less elite or poorer performers, despite no differences in anxietyintensity levels (Jones & Swain, 1992; Mellalieu et al., 2006).However, this reliance on anxiety is problematic because athletes’affective experiences cannot be accurately described by the pres-ence or lack of anxiety (Martinent & Ferrand, 2009). Examiningwhether the direction of affective states other than anxiety wasassociated with theoretically selected variables after intensity ofthese affective states were held constant would provide strongevidence for the usefulness and relevance of the concept of direc-tion of affective states in sport settings. Specifically, highlightingsignificant relationships between direction of affective states andimportant competitive outcomes such as attainment of achieve-ment goals, utilization of coping strategies or sport satisfactioncontrolling for intensity of affective states would clearly demon-strate that direction of affective states do not overlap with intensityof affective states in their relationships with relevant theoreticallycompetitive outcome variables.

Although several scholars have outlined the promises ofconsidering the direction of affective states in addition to theirintensity (Hanin, 2007; Martinent et al., 2012; Martinent & Ferrand,2009), few studies have explored simultaneously the intensity anddirection of positive and negative affective states (Martinent et al.,2013; Mellalieu, Hanton, & Jones, 2003; Robazza & Bortoli, 2007).Mellalieu et al. (2003) investigated how facilitators and debilitatorsof cognitive and perceived physiological symptoms associated withcompetitive anxiety differed in their experience of precompetitiveaffective states (nature, intensity and direction) in preparation forand with regard to actual performance. Findings showed thatanxious facilitators differed significantly from debilitators in regardto the nature of these symptoms as well as their direction withrespect to preparation for and actual performance. Performers whointerpreted anxiety symptoms as facilitative labelled significantlymore positive affective states and experienced these affects asmorefacilitating (large effect sizes) than did individuals who interpretedtheir anxiety symptoms as debilitative (Mellalieu et al., 2003).Robazza and Bortoli (2007) extended the notion of directionalperceptions beyond anxiety to anger by assessing rugby players’perception of the facilitative or debilitative effects of trait angersymptoms. Their findings revealed a general tendency of rugbyplayers to experience amoderate frequency of anger symptoms andto interpret their symptoms as facilitative rather than debilitative(Robazza & Bortoli, 2007). Martinent et al. (2013) explored affectiveprofiles of athletes e based on affect intensity and direction e

before and during the competition. Four similar profiles of athletes

were identified for the two measurement occasions: high positiveaffect facilitators, facilitators, low affect debilitators, and highnegative affect debilitators. However, the analyses they computeddid not shed light on the incremental validity of the direction ofaffective states.

A significant limitation in this area was that these studies havemainly investigated direction of affective states in using self-reportmeasures not psychometrically validated (e.g., Martinent et al.,2013; Mellalieu et al., 2003; Robazza & Bortoli, 2007). Most of thetime, researchers added the direction scale by Jones and Swain(1992) to an existing published self-report without testingwhether the modified self-report (including both intensity anddirection scales) was a psychometrically sound questionnaire (seefor an exception on the CSAI-2 Revised Martinent, Ferrand, Guillet,& Gautheur, 2010). Another limitation of previous studies was thatstatistical analyses performed did not shed light on the incrementalvalidity of the intensity and direction of affective states. Thisapproach prevents examining the independent contribution of theintensity and direction dimensions of affects in their relationshipswith other variables.

Based on the dimensional approach of affective states, Watsonet al. (1988) developed the Positive and Negative Affect Schedule(PANAS) comprising two 10-item adjective checklist subscales.During scale development, the PANAS items were empiricallyderived from a larger list of 27 adjectives within nine mood cate-gories (attentive, excited, proud, strong, distressed, angry, fearful,guilty, and nervous), which were originally proposed by Zevon andTellegen (1982). Positive Affect (PA) reflects the extent to which aperson feels enthusiastic, active, and alert (Watson et al., 1988),with low PA being a state of sadness and lethargy. Negative Affect(NA) is conceptualized as a general dimension of subjective distressand unpleasurable engagement that subsumes a variety of aversivemood states including anger, nervousness, or fear, with low NAbeing a state of calmness and serenity (Watson et al., 1988). ThePANAS is one of the most frequently used instruments to assessaffect intensity (level) in social (e.g., Koestner, Lekes, Powers, &Chicoine, 2002), health (e.g., Brown & Ryan, 2003), educational(e.g., Sideridis, 2005) as well as sport and exercise psychology (e.g.,Crocker, 1997; Gaudreau & Blondin, 2002). Of particular importancein the context of the present study, the PANAS in this present formassesses only the intensity of positive and negative affective statesand not their direction (i.e., perceiving an affective state as bene-ficial or harmful for sport performance).

As a result, the main purpose of this current research was toprovide validity and reliability evidence of a modified PANASincluding a direction scale (PANAS-D). Specifically, through twostudies, we tested the validity and reliability of the PANAS-D formeasuring both the intensity and direction of affective states (study1) and we examined the relationships between direction of affec-tive states and theoretically selected variables by controlling forintensity of affective states (study 2).

Study 1

Acknowledging the frequent use of the PANAS in sport and ex-ercise literature, Crocker (1997) tested its factorial structure with aconfirmatory factor analysis (CFA) on a sample of youth sport par-ticipants. Results supported the hypothesized two-factor model.However, the large residual of several items suggested that thetwo-factor model did not fully account for the conceptual speci-ficity of adjectives such as irritated, distressed, and upset. Thisresult might suggest the existence of a latent construct not includedin the hypothesized model; the negative affect scale of the PANAS,designed as a general dimension of distress, might comprisedistinct subscales of discrete, yet interrelated, categories of

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237 229

negative affect (Gaudreau, Sanchez, & Blondin, 2006). To furtherdevelop this line or research, Mehrabian (1997) investigated thefactorial structure of the PANAS using an exploratory factor anal-ysis. Their results provided evidence for the tenability of both atwo- and a three-factor model. In the latter model, the negativeaffect itemswere divided into two conceptually meaningful factors:Afraid (scared, nervous, afraid, guilty, ashamed, and jittery) andUpset (distressed, irritated, hostile, and upset).

Based on these initial results, the structure of the PANAS hasrecently received more empirical attention. To date, to the best ofour knowledge, 12 studies have employed confirmatory factoranalysis (CFA) to provide validity evidence based on the internalstructure of the PANAS (Crawford & Henry, 2004; Crocker, 1997;Gaudreau et al., 2006; Joiner, Sandin, Chorot, Lostao, & Marquina,1997; Killgore, 2000; Lonigan, Hooe, David, & Kistner, 1999;Melvin & Molloy, 2000; Merz & Roesch, 2011; Merz et al., 2013;Molloy, Pallant, & Kantas, 2001; Terracciano, McCrae, & Costa,2003; Tuccitto, Giacobbi, & Leite, 2010). These CFA researcheshave provided equivocal validity evidence based on the fit indicesof the original two-factor structure of the PANAS (see for a detailedreview on this issue Tuccitto et al., 2010). In complementary ways,using average variance explained values (i.e., the extent to whichaffect dimensions account for item variance), previous CFAs havefound that the latent PA and NA factors explain around only 30.00%of the variance in their items (see for a detailed review on this topicTuccitto et al., 2010). Given the widespread use of the PANAS, theinability of researchers to provide strong validity evidence for theoriginal factor structure of the PANAS suggests that the validity ofthe PANAS score interpretations requires further examination.

One possible explanation for these equivocal findings was thatWatson et al. (1988) did not incorporate the content categoriesfrom Zevon and Tellegen (1982) into the factor structure of thePANAS. As an illustration of this issue, Crocker (1997) has suggestedthat misspecification in the two-factormeasurementmodel may bereduced by allowing for 13 correlated uniquenesses amongredundant items derived from within the same single contentcategory of the nine categories originally proposed by Zevon andTellegen (1982). For example, hostile and irritable were bothderived from the “angry” category and thus have overlappingcontent, creating a flawed measurement model. While controver-sial in CFA research, such a practice is justified in the PANAS contextbecause the demonstration of estimated error correlations suggeststhat theoretically meaningful latent factors (affect content cate-gories) may have been erroneously omitted from the CFA model(Merz et al., 2013; Merz & Roesch, 2011; Tuccitto et al., 2010).Previous research showed that model fit is enhanced when ac-counting for these redundancies (Crawford & Henry, 2004; Merzet al., 2013; Merz & Roesch, 2011; Tuccitto et al., 2010).

Another possible explanation for the aforementioned equivocalfindings based on the fit indices of the original two-factor structureof the PANAS was the existence of a three-factor structure of thePANAS (Gaudreau et al., 2006; Killgore, 2000; Mehrabian, 1997;Merz et al., 2013). The three-factor structure of the PANAS hasreceived some support in studies that used CFA and allowed forcorrelated uniquenesses among the redundant items from Zevonand Tellegen’s (1982) check-list (Gaudreau et al., 2006). However,other studies have challenged Mehrabian’s (1997) model. Forexample, Crawford and Henry (2004) found that the two-factormodel of the PANAS allowing for correlated uniquenesses fit bet-ter than the three-factor model (notably, correlated uniquenesseswere not included in the three-factor model in this study). It hasalso been suggested that both the two-and three-factor structuresare plausible, but that the three-factor model provides superior fit(Killgore, 2000; Merz et al., 2013). Conversely, another studyshowed that the two-factor model provided superior fit than the

three-factor model if the correlated uniqueness terms of redundantitems were permitted to correlate (Tuccitto et al., 2010). Thus, thebest fitting structure of the PANAS remains a relatively unansweredquestion, particularly when accounting for overlapping item-levelunique variance.

To explore the criterion related validity of the PANAS-D, wecollected data on attainment of sport achievement goals andcoping. In the cognitive motivational relational theory of Lazarus(CMRT, 1991, 2000), affective states arise from the transaction be-tween athlete and his environment through appraisals (Lazarus,2000). Based on the premise that primary appraisals refer towhether a competitive situation is personally relevant andcongruent to the athlete’s goals (Lazarus, 2000), attainment of sportachievement goals can be considered as both an antecedent and aconsequence of affective states experienced by athletes in compe-tition. Lazarus (1991) also postulated that coping mediates therelationship between appraisals and the subsequent affective statesand thus can be considered as essential in the affect elicitation. Inthis perspective, coping is defined as the constantly changingbehavioural and cognitive mechanisms used to manage theongoing internal and external demands of a specific stressful situ-ation (Lazarus & Folkman, 1984). Although athletes are usingvarious coping strategies, hierarchical models of coping have beenproposed to regroup coping strategies into a meaningful andparsimonious set of coping dimensions: (a) Task-oriented copingrepresents strategies aimed at dealing directly with the stressfulsituation and the resulting thoughts and affects (e.g., effortexpenditure, thought control), (b) Disengagement-oriented copinginvolves the strategies through which a personwithdraws from theprocess of actively striving toward the realization of desirableoutcomes (e.g., behavioural disengagement, denial), and (c)distraction-oriented coping represents strategies used to momen-tarily focus the attention on external and/or internal stimuli un-related to the stressful situation such as distancing or mentaldistraction (Gaudreau & Blondin, 2002).

Previous research confirmed theoretical predictions of theCMRT by showing that: (a) PA intensity correlated significantlywith the attainment of sport achievement goals (e.g., Amiot,Gaudreau, & Blanchard, 2004), (b) PA intensity was negativelycorrelated with disengagement-oriented coping and positivelycorrelated with task-oriented coping (Gaudreau & Blondin, 2002),(c) NA intensity was positively correlated with distraction- anddisengagement-oriented coping (Gaudreau & Blondin, 2002). Toour knowledge, no study investigated the relationship betweencoping and direction of affective states. Nevertheless, based onprevious direction anxiety research, we could also hypothesize thata facilitative interpretation of PA and NA correlated negatively withdisengagement-oriented coping and/or positively with task-oriented coping (e.g., Mellalieu et al., 2006).

Because acquisition of knowledge depends upon valid in-struments, the development of an inventory for measuring bothintensity and direction of athletes’ affective states is an importantstep. Although the PANAS has been used to assess level (intensity)of affects experienced by athletes before and/or during thecompetition (e.g., Crocker, 1997; Gaudreau & Blondin, 2002), theintroduction of a direction scale (PANAS-D) in addition to the in-tensity scale commonly used in the sport literature require to testpsychometric properties of the PANAS-D. As a result, study 1 aimedat evaluating the validity and reliability of the PANAS-D. We hy-pothesized that the model fit of the two-factor structure wouldimprove for both intensity scale, direction scale and overall mea-surement model (including simultaneously intensity and directionscales) by allowing for the 13 correlated uniquenesses based onZevon and Tellegen’s (1982) checklist. We also examined whetherMehrabian’s (1997) three-factor structure of the PANAS would

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237230

enhance the measurement model; no hypothesis was specified asthis was considered exploratory. Furthermore, this study sought toassess criterion related validity of the PANAS-D with a sample ofathletes who reported concurrently their pre-competitive affectiveexperience and retrospectively their affective states experiencedduring competitive. Criterion-related validity looks at the rela-tionship between test scores and outcomes. In the present study,we examined the relationships between PANAS-D scores andtheoretically selected outcomes (i.e., attainment of sport achieve-ment goals and coping) based on the CMRT of Lazarus (2000). Wehypothesize that the PANAS-D will show evidence of criterion-related validity, as indicating by the correlations between thePANAS-D subscales and the attainment of sport achievement goalsand coping subscales.

Stage 1: factorial validity and reliability evidence of thePANAS-D

Method

ParticipantsThree hundred and six French athletes (112 females and 194

males) ranging in age from 15 to 39 years (M ¼ 22.23 years,SD ¼ 4.90) volunteered to participate in the study. On average theyhad been competing in their sport for 10.84 years (SD ¼ 5.73) andthey trained 7.52 h per week (SD ¼ 4.49). They participated indepartmental (N ¼ 35), regional (N ¼ 122), national (N ¼ 125) orinternational sport events (N ¼ 24).

MeasureWe used the French version of the PANAS (Gaudreau et al.,

2006), which includes the two original 10-item adjective check-list subscales (Watson et al., 1988) of PA and NA. Using the“moment” instruction (i.e., “right now, that is, at the presentmoment”, Watson et al., 1988, p. 1070), participants were asked torate the intensity of each symptom on a scale of 1 (not at all or veryslightly) to 5 (extremely). In addition, participants rated on thedirection scale from �3 to þ3 the degree to which the intensity ofeach symptom experienced was either debilitative or facilitative tosubsequent performance (Jones & Swain, 1992).

ProcedureThe research was conducted in accordance with international

ethical guidelines that are consistent with APA norms. The coachesof each team/athlete were contacted to obtain permission toapproach their athletes and ask them to participate in the study.Participation was voluntary and parental consent was required forathletes under 18 years of age. The participants completed thePANAS-D on two occasions: (a) within 2 h before the competitiveevent ea time frame consistently used in the literature and regar-ded as acceptable since it did not interfere with the athletes’preparation routines (Mellalieu et al., 2006) e and (b) within 2 hafter the competition using a single question “Howwould you havereported experiencing these affective states during the competi-tion?” (Gaudreau et al., 2006).

Data analysisTo assess the factorial validity of PANAS-D scores, multiple a

priori models were specified and tested in Lisrel 8.71 (Jöreskog &Sörbom, 2004) using maximum likelihood estimation. Multiple fitindices were chosen to achieve a comprehensive evaluation of fit:the chi-square (c2), the comparative fit index (CFI), the stand-ardised root mean square residual (SRMR), the root mean squarederror of approximation (RMSEA), and the confidence interval ofRMSEA (90% CI). For the CFI, values above .90 are traditionally

considered reasonable model fit, whereas newer recommendationssuggest values close to .95 (Hu & Bentler, 1999). For the RMSEA andthe SRMR, values below .08 are traditionally considered reasonablemodel fit whereas newer recommendations suggest values close to.05 (Hu & Bentler, 1999). The Akaike information criterion (AIC) andthe expected cross-validation index (ECVI) were used for compar-ison with alternative models (MacCallum & Austin, 2000).

The pre-competitive completion of the PANAS-D was used as acalibration sample in which several models were tested andcompared. The post-competition completion of the PANAS-D(validation sample) was used to determine whether the bestfitting PANAS-D models identified in the calibration sample couldbe replicated with a sample of athletes who completed the PANAS-D at different stage of competition. First, a series of original two-factor (PA, NA) models were specified and tested for intensityscale, direction scale and overall measurement model (includingsimultaneously intensity and direction scales). Model 1awas a two-factor model without correlated uniquenesses. Model 1b was atwo-factor model with correlated uniquenesses from Zevon andTellegen (1982). Next, a series of Mehrabian’s (1997) three-factor(PA, Ufraid, and Upset) models were specified and tested for in-tensity scale, direction scale and overall measurement model. Thisthree-factor structure of the PANAS-D was tested without corre-lated uniquenesses representing content categories (model 2a), andwith the Zevon and Tellegen (1982) correlated uniquenesses(model 2b). Finally, the best-fitting model identified in the cali-bration sample was also tested for invariance across the pre- andpost-competitive completions. Based on a statistical methodologyproposed by Gregorich (2006), we successively tested configuralinvariance (no equality constraints), metric invariance (equal itemloadings), strong invariance (equal item loading and item interceptsconcurrently) and strict invariance (equal item loadings, item in-tercepts and item error variances concurrently). The differencebetween two multi-sample models (e.g., metric and strong) wasjudged based on the difference of CFI value. A value equal to or lessthan .010 indicates no difference between models and thus tena-bility of equality constraints (Cheung & Rensvold, 2002).

Results

Calibration on the pre-competitive completion of the PANAS-DTable 1 presents fit indices for the CFA models for the original

PANAS-D. The 2-factor models without correlated residuals(Models 1a) of the intensity and direction scales did not fit well; the4-factor model without correlated residuals had an implausible fitaccording to the newer recommendations for the RMSEA, SRMRand CFI. The 2-factor models with correlated residuals (Models 1b)did not fit well for the intensity and direction scales. However, the4-factor model with correlated residuals fit reasonably well ac-cording to the CFI, RMSEA and SRMR. Comparisons of the nestedmodels are also available in Table 1. The 2-factor models withcorrelated residuals (Models 1b) fit significantly better than themore restrictive 2-factor models without correlated residuals(Models 1a) for the intensity scale, direction scale and overallmeasurement model (see Dc2, ECVI and AIC values in Table 1).

The 3-factor models without correlated residuals (Models 2a) ofthe intensity and direction scales did not fit well; the 6-factormodel without correlated residuals had a plausible fit accordingto the RMSEA but CFI and SRMR were slightly below the newerrecommendations (Hu & Bentler, 1999). The 3-factor models withcorrelated residuals (Models 2b) did not fit well for the intensityand direction scales. However, the 6-factor model with correlatedresiduals fit reasonablywell according to the CFI, RMSEA and SRMR.Comparisons of the nested models for the 3-factor structure of thePANAS-D showed that the 3-factor models with correlated

Table 1Fit indices for the calibration and cross-validation samples of the PANAS-D (study 1).

Model c2 p df CFI SRMR RMSEA 90% CI AIC ECVI Dc2 Ddf

Calibration sample1a 2-Factor intensity modela 539.03 <.001 169 .896 .074 .085 .077e.093 621.03 2.041a 2-Factor direction modela 542.48 <.001 169 .899 .071 .085 .077e.093 624.48 2.051a 4-Factor modela 1527.07 <.001 734 .915 .063 .060 .055e.064 1699.06 5.571b 2-Factor intensity modelb 388.20 <.001 156 .925 .068 .070 .061e.079 496.20 1.63 150.83* 131b 2-Factor direction modelb 426.67 <.001 156 .920 .067 .075 .070e084 534.67 1.75 115.81* 131b 4-Factor modelb 1263.79 <.001 708 .933 .060 .051 .046e.055 1487.79 4.88 263.28* 262a 3-Factor intensity modelc 501.54 <.001 167 .903 .073 .081 .073e.089 587.54 1.932a 3-Factor direction modelc 541.16 <.001 167 .898 .071 .086 .078e.094 627.16 2.062a 6-Factor modelc 1478.13 <.001 725 .918 .062 .058 .054e.063 1668.13 5.472b 3-Factor intensity modeld 361.05 <.001 154 .933 .069 .066 .057e.075 473.05 1.55 140.49* 132b 3-Factor direction modeld 426.69 <.001 154 .920 .067 .076 .068e.085 538.69 1.77 114.47* 132b 6-Factor modeld 1229.01 <.001 699 .936 .059 .050 .045e.054 1471.01 4.82 249.12* 26Cross-validation sample1b 4-Factor modelb 1297.62 <.001 708 .924 .057 .052 .048e.057 1521.62 4.992b 6-Factor modeld 1268.45 <.001 699 .926 .056 .052 .047e.056 1510.45 4.95

Multiple groups models c2 p df CFI SRMR RMSEA 90% CI AIC ECVI DCFI

1b Configural invariance 2586.24 <.001 1426 .927 .062 .052 .048e.055 3014.24 4.941b Metric invariance 2789.72 <.001 1462 .920 .070 .055 .051e.058 3145.72 5.16 .0071b Strong invariance 3519.80 <.001 1498 .886 .072 .066 .064e.069 3963.80 6.50 .0341b Partial strong invariancee 2944.61 <.001 1494 .910 .070 .056 .053e.059 3396.61 5.57 .0101b Partial strict invariance 3165.42 <.001 1530 .901 .074 .059 .056e.062 3545.42 5.81 .009

*p < .001.a Original 2-factor model (Watson et al., 1988).b 2-factor model with correlated uniquenesses from Zevon and Tellegen (1982).c 3-factor model (Mehrabian, 1997).d 3-factor model with correlated uniquenesses.e Item intercepts for intensity items 4 and 11 and direction items 11 and 15 were not constrained to equality across samples.

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237 231

residuals (Models 2b) fit significantly better than the morerestrictive 3-factor models without correlated residuals (Models2a) for the intensity scale, direction scale and overall measurementmodel (see Table 1). In sum, both the 4- and 6-factor models withcorrelated residuals of the PANAS-D provided reasonably good fit tothe data.

Table 2Standardized factor loadings for the best-fitting models of the PANAS-D (study 1).

4-factor model 1b 6-factor model 2b

PAI PAD PAI PAD

Item 1. Intéressé(e) [interested]c. .54a/.68b .56/.56 .54/.68 .56/.56Item 3. Excité(e) [excited]d. .57/.63 .38/.48 .57/.63 .37/.49Item 5. Fort(e) [strong]f. .59/.58 .53/.54 .59/.58 .53/.55Item 9. Enthousiaste [enthusiastic]d. .67/.72 .61/.63 .67/.72 .61/.62Item 10. Fier(e) [proud]e. .62/.62 .44/.51 .62/.62 .43/.51Item 12. Alerte [alert]c. .56/.48 .37/.23 .56/.48 .37/.23Item 14. Inspiré(e) [inspired]d. .65/.63 .50/.62 .65/.63 .50/.61Item 16. Déterminé(e) [determined]e. .70/.65 .63/.53 .70/.65 .64/.53Item 17. Attentif(ve) [attentive]c. .71/.61 .49/.49 .71/.61 .49/.48Item 19. Actif(ve) [active]f. .66/.67 .56/.51 .66/.67 .56/.51

NAI NAD UI UDItem 2. Angoissé(e) [distressed]g. .80/.73 .38/.54 .53/.70 .38/.53Item 4. Fâché(e) [upset]g. .16/.56 .67/.51 .06/.47 .67/.51Item 8. Hostile [hostile]h. .29/.45 .61/.68 .25/.45 .61/.68Item 11. Irrité(e) [irritable]h. .28/.64 .66/.79 .21/.61 .67/.80

AI ADItem 6. Coupable [guilty]i. .06/.38 .70/.65 .09/.37 .70/.65Item 7. Effrayé(e) [scared]k. .47/.59 .65/.80 .48/.57 .65/.80Item 13. Honteux(se) [ashamed]i. .19/.47 .66/.64 .19/.47 .66/.64Item 15. Nerveux(se) [nervous]j. .87/.76 .34/.48 .83/.73 .34/.48Item 18. Agité(e) [jittery]j. .43/.53 .58/.59 .48/.51 .58/.59Item 20. Craintif(ve) [afraid]k. .69/.52 .67/.76 .67/.48 .67/.76

Note. a Calibration sample; b Cross-validation sample. cek denotes the itemuniqueness terms permitted to correlate as suggested by Zevon and Tellegen (1982).PA ¼ Positive affect; NA ¼ Negative Affect; A ¼ Afraid; U ¼ Upset; I ¼ Intensity;D ¼ Direction.

Cross-validation on the post-competitive completion of the PANAS-D

The 4-factor and 6-factor models with correlated residuals alsoprovided reasonably good fit to the data of the post-competitivecompletion (Table 1). The standardized factor loadings of thesetwo models (Models 1b and 2b) for both the calibration and cross-validation samples are presented in Table 2. For the 4-factormodels with correlated residuals (Model 1b), standardized factorloadings were relatively large for PA (ls ranged from .37 to .72excepted for item 12 in the cross-validation sample). For NA, allstandardized factor loadings were also relatively large in the cross-validation sample (ls ranged from .38 to .80). In the calibrationsample, whereas the majority of standardized factor loadings werein an acceptable range (ls ranged from .38 to .87), a fewstandardizedfactor loadings were below .35 (i.e., intensity items 4, 6, 8, 11 and 13and direction item 15). As shown in Table 3, PA intensity was un-correlated with NA intensity (ø¼ .11 and�.01, p> .05). PA intensitywaspositivelycorrelatedwith PAdirection (ø¼ .79 and .66, p< .001)and NA direction (øcalibration sample¼ .11, p> .05; øvalidation sample¼ .17,p < .05). NA intensity was uncorrelated with PA direction (ø ¼ .04and �.08, p > .05) and negatively correlated with NA direction(øcalibration sample ¼ �.13, p < .05; øvalidation sample ¼ �.01, p > .05). PAdirection was positively correlated with NA direction (ø ¼ .28 and.24, p < .001).

For the 6-factor model with correlated residuals (Model 2b),standardized factor loadings were relatively large for PA (ls rangedfrom .37 to .72 excepted for item 12 in the cross-validation sample).For NA, all standardized factor loadings were relatively large in the

cross-validation sample (ls ranged from .37 to .80). In the calibra-tion sample, whereas the majority of standardized factor loadingswere in an acceptable range (ls ranged from .38 to .83), a fewstandardized factor loadings were below .35 (i.e., intensity items 4,6, 8, 11 and 13 and direction item 15). However, the confidenceintervals of two correlations e upset and afraid intensity (ø ¼ .99and .99) as well as upset and afraid direction (ø¼ .99 and .98) e did

Table 3Means, standard deviations, Cronbach’s alphas, average variance extracted, composite reliability values, correlations between latent constructs of the PANAS-D, and corre-lations of the PANAS-D subscales with the A-SAGS and CICS subscales (study 1).

Pre-competitive affective states Competitive affective states M SD Cronbach’s r

PAI NAI PAD NAD PAI NAI PAD NAD Alphas AVE

Pre-competitive affective statesPositive affect intensity 3.39 .66 .86 .43 .87Negative affect intensity .11 1.78 .51 .76 .26 .71Positive affect direction .79* .04 1.48 .59 .76 .28 .78Negative affect direction .11 �.13* .28* .06 .82 .84 .40 .85Competitive affective statesPositive affect intensity .43* .11* .37* .10 2.91 .75 .86 .42 .87Negative affect intensity .00 .37* �.03 �.05 �.01 1.81 .63 .83 .35 .83Positive affect direction .15* .09 .37* .20* .66* �.08 .98 .67 .77 .29 .78Negative affect direction .03 �.09 .12* .52* .17* �.01 .24* �.02 .81 .87 .46 .88Pre-competitive copingTask-oriented coping .42* .31* .40* .12* .32* .12* .20* �.01 2.68 .57 .85Distraction-oriented coping .13* .40* .17* .00 .18* .25* .11 .00 2.08 .65 .74Disengagement-oriented coping �.13* .37* �.08 .02 �.02 .23* �.01 �.02 1.60 .55 .77Competitive copingTask-oriented coping .45* .34* .39* .14* .42* .15* .29* .08 2.58 .50 .85Distraction-oriented coping .05 .41* .07 �.11* .10 .37* .10 �.08 1.78 .68 .80Disengagement-oriented coping �.08 .28* �.09 .06 �.21* .45* �.20* �.07 1.89 .71 .82Attainment of sport achievement goals .45* .03 .26* .08 .21* .06 �.03 �.01 5.61 .85 .95

Note. PA ¼ Positive affect; NA ¼ Negative affect; I ¼ Intensity; D ¼ Direction; AVE ¼ Average Variance Extracted. *p < .05.

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237232

not support discriminant validity, insofar as the intervals include1.0 (Anderson & Gerbin, 1988). These results clearly suggested thatit seems impossible to distinguish empirically upset and afraid. As aresult, these two factors should be merged together in one uniquefactor (NA), as it is the case in the 4-factor structure of the PANAS-D.

Invariance across the calibration and cross-validation samplesAs shown in Table 1, the configural multiple-sample CFA model

(4-factor structure of the PANAS-D) fitted the data adequately,showing that the number of latent constructs can be assumed to beidentical across pre-competitive and competitive samples. The nextmodel constrained corresponding factor loadings to be equal acrossgroups. The metric multiple-sample CFA model fitted the dataadequately. Additionally, the difference of CFI value between theconfigural and metric models was less than .010 (see Table 1),providing evidence of metric invariance across samples. The thirdmodel tested strong factorial invariance by additionally imposingequality constraints on corresponding item intercepts. This modelwas rejected based on the fit indices and D CFI (see Table 1). Lisreloutput suggested that the cross-group equality constraint on theintercept for intensity items 4 and 11 and direction items 11 and 15contributed most strongly to the lack of fit. The fourth model freelyestimated this parameter in both groups and provided evidence ofpartial strong factorial invariance. The fifth model tested partialstrict factorial invariance by imposing additional cross-groupequality constraints on all corresponding item residual variancesexcept those for intensity items 4 and 11 and for direction items 11and 15. This model was accepted (D CFI¼ .009), providing evidenceof partial strict factorial invariance.

ReliabilityThe alpha coefficients indicated that the reliability of each of the

four PANAS-D subscales was acceptable, with Cronbach’s alphacoefficients ranging from .76 to .87 (see Table 3). To further assessthe internal reliability of the PANAS-D, we also provided AverageVariance Extracted (AVE) values and composite reliability values(r). AVE values, computed as the sum of squared standardizedfactor loadings divided by the sum of squared standardized factorloadings plus the sum of error variances, described the variancecaptured by measurement errors as opposed to the varianceattributable to the latent factors. A value of .50 or greater indicates a

good reliability as the variance of the construct is greater than theerror variance (Fornell & Larcker,1981). Composite reliability values(i.e., r ¼ (sum of standardized loadings)2/{(sum of standardizedloadings)2 þ (sum of error variances)}) measure the overall reli-ability of a collection of heterogeneous but similar items (Raykov,2001). A value of .70 or greater indicates a good reliability(Raykov, 2001). AVE values ranged from .26 to .46 (see Table 3),suggesting a relatively poor reliability for the four PANAS-D sub-scales. In contrast, the r values indicated that the reliability of eachof the four PANAS-D subscales was acceptable, with r ranging from.71 to .88 (see Table 3).

Stage 2: criterion-related validity evidence of the PANAS-D

Method

ProcedureThe participants (N ¼ 306) also completed the Coping Inventory

for Competitive Sport (CICS, Gaudreau & Blondin, 2002) within 2 hboth before and after the competitive event and the Attainment ofSport Achievement Goal Scale (A-SAGS, Amiot et al., 2004) within2 h after the competitive event.

MeasuresThe CICS (Gaudreau & Blondin, 2002) is a French questionnaire

containing 39 items. Consistent with previous research (e.g.,Gaudreau, El Ali, & Marivain, 2005), the ten subscales were orga-nized in the 3 s-order dimensions of task-oriented (mental imagery,thought control, effort expenditure, seeking support, logical anal-ysis, and relaxation), distraction-oriented (mental distraction anddistancing), and disengagement-oriented coping (venting of un-pleasant emotions and disengagement/resignation). Previousstudies provided support for the reliability and validity of the CICS(e.g., Gaudreau et al., 2005). Each itemwas rated on a 5-point Likertscale ranging from 1 (does not correspond at all) to 5 (correspondsvery strongly). The alpha coefficients indicated that the reliabilitywas acceptable with Cronbach’s alpha coefficients varying from .74to .85 (Table 3).

The A-SAGS (Amiot et al., 2004) is a French questionnaire con-taining 12 items measuring three theoretically driven criteria usedby athletes to evaluate their level of mastery, self-referenced, and

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237 233

normative subjective sport achievement. Previous research lentcredence to the validity and reliability of the A-SAGS (Amiot et al.,2004; Gaudreau & Antl, 2008; Martinent et al., 2013). Results ofconfirmatory factor analyses supported both the hypothesizedthree-factor structure of goal attainment at a lower level and ageneral index of goal attainment at a higher level (Amiot et al.,2004; Gaudreau & Antl, 2008). Consistent with previous research(e.g., Gaudreau & Antl, 2008; Nicolas, Gaudreau, & Franche, 2011), ageneral index of goal attainment averaging the three subscales wasused in this study. Each item was rated on a 7-point Likert scaleranging from 1 (not at all) to 7 (totally). The alpha coefficientindicated that the reliability was goodwith a Cronbach alpha of .95.

Results

Pre-competitive and competitive copingResults showed that: (a) pre-competitive and competitive task-

oriented coping correlated significantly with pre-competitive andcompetitive PA intensity, NA intensity, PA direction and pre-competitive NA direction; (b) pre-competitive and competitivedisengagement- and distraction-oriented coping correlated signif-icantly with pre-competitive and competitive NA intensity; and (c)pre-competitive and/or competitive disengagement-orientedcoping showed significant negative correlations with pre-competitive PA intensity and competitive PA intensity and direc-tion (Table 3).

Attainment of achievement goalsAttainment of sport achievement goals correlated significantly

with pre-competitive and competitive PA intensity as well as pre-competitive PA direction (Table 3).

Discussion

Study 1 aimed at evaluating the validity and reliability of thePANAS-D among a sample of athletes who completed this checklistat a different stage of a performance related situation (pre- andpost-competitive completions). For these two distinct measurement oc-casions, CFAs revealed acceptablefits for the4-factor structure of thePANAS-D (intensity and direction of PA and NA) with correlateduniquenesses from Zevon and Tellegen (1982). Specifically, modelcomparison tests revealed that the models permitting theoreticallymeaningful error correlations demonstrated a statistically signifi-cant improved fit over those that did not. Therefore, our resultsprovided further evidence supporting a PANAS factor structure thataccounts for Zevon and Tellegen’s (1982) mood content categories(Crawford & Henry, 2004; Merz et al., 2013; Tuccitto et al., 2010).Recent research also suggested the existence of a 3-factor structureof the PANAS in which the negative affect items were divided intothe two factors of afraid and upset (e.g., Gaudreau et al., 2006;Killgore, 2000; Merz et al., 2013). In contrast with this literature,results of the present study clearly disconfirmed the 6-factorstructure of the PANAS-D (intensity and direction of PA, afraid andupset) based on the rationale that afraid and upset factors were notdistinguishable empirically (for both intensity and direction scales).

The literature on the CFA of PANAS has suggested that PA and NAscores exhibit acceptable scale reliabilities, as indicated by thealpha coefficients and r values, but typically explain around only30.00% of the variance in their items, as indicated by AVE values(e.g., Crawford & Henry, 2004; Joiner et al., 1997; Lonigan et al.,1999). In our study, AVE values for the 4-factor model with corre-lated residuals of the PANAS-D ranged from .42 to .43 for PA in-tensity, from .26 to .35 for NA intensity, from .28 to .29 for PAdirection and from .40 to .46 for NA direction. Therefore, in com-parison with previous findings, it appears that our Cronbach’s

alpha, r and AVE results have provided comparable support for thevalidity of PANAS-D score interpretations.

Confirming previous research on anxiety symptoms (e.g.,Martinent et al., 2010), the inter-scale correlations of the PANAS-Demphasized the importance of measuring the intensity and direc-tion dimensions of affective response separately. The proportions ofshared variances between intensity and direction subscales of thePANAS-D support the view that these dimensions should beconsidered as, and thereforemeasured as, separate dimensions thatindependently contribute to the individual’s affective experience. Aseries of multiple-sample CFAs also tested the invariance ofparameter estimates across the pre- and post-competitive com-pletions of the PANAS-D. Overall, our results indicated that thePANAS-D was partially invariant with only a few differences inspecific parameter estimates across the samples. Of particularimportance, all the 40 factor loadings were not significantlydifferent across the pre- and post-competitive completions of thePANAS-D. As a result, items are measuring the same attribute acrosssamples of athletes differing with respect to the stages of a sportcompetition in which they completed the PANAS-D. This findingsupports the use of the PANAS-D for the assessment of affectivestates both before and after a sport competition.

Another support for the validity of the PANAS-D is that its sub-scales related to coping and attainment of sport achievement goalsin accord with the theoretical expectations and previous findings(e.g., Amiot et al., 2004; Gaudreau & Blondin, 2002; Mellalieu et al.,2006): (a) PA intensity and direction correlated positively with theattainment of sport achievement goals, (b) pre-competitive andcompetitive PA intensity correlatednegativelywithpre-competitiveand/or competitive disengagement-oriented coping and positivelywith pre-competitive and competitive task-oriented coping, (c) pre-competitive and competitiveNA intensity correlated positivelywithpre-competitive and competitive distraction- and disengagement-oriented coping, and (d) pre-competitive and/or competitive PAdirection correlated negatively with competitive disengagement-oriented coping and positively with pre-competitive and/orcompetitive task- and distraction-oriented coping.

In sum, the results of this study advance previous research byshowing that the PANAS-D is a reliable and valid tool for estimateathlete’s pre-competitive and competitive intensity and directionof affective states. Because the validity and reliability of self-reportand theory testing are inextricably linked, more validation studiesare required to evaluate whether results are invariant across thetype of sport, age or level.

Study 2

Providing reliability, structural as well as criterion-related val-idity evidence of the PANAS-D was not sufficient to assert unam-biguously the significance of a distinction between athletes’intensity and direction of affective sates. Demonstrating that di-rection of affective states is related to theoretically selected vari-ables controlling for athletes’ intensity of affective states (and vice-versa) would provide a prominent argument in favour of thedistinction between the intensity and direction components ofathletes’ affective states. As a result, in study 2 we moved beyondthe zero-order correlations between affective states and othervariables and examined incremental validity of the two dimensionsof athletes’ affective states. Specifically, we examined the rela-tionship between direction of affective states experienced by ath-letes during competition and three theoretically selected variablesfrom a CMRT perspective (i.e., coping strategies used duringcompetition, sport satisfaction and attainment of sport achieve-ment goals after competition) controlling for the intensity of af-fective states. We also examined whether intensity of affective

Table 4Zero-order and partial correlations between PANAS-D scores and coping, attainment of sport achievement goals and sport satisfaction scores (Study 2).

M SD Cronbach’salphas

Zero-order correlations NAD Partial correlations NAD

PAI PAD NAI PAI PAD NAI

Task-oriented coping 2.66 .56 .85 .42* .26* .06 �.02 .34* .02 .05 �.01Distraction-oriented coping 1.77 .67 .79 .03 �.16* .27* �.20* .15* �.22* .24* -.15*Disengagement-oriented coping 2.04 .76 .80 �.05 �.22* .50* �.16* .11 �.24* .49* �.08Attainment of sport achievement goals 3.95 1.89 .92 .34* .32* �.09 .15* .20* .15* �.07 .13*Sport satisfaction 4.49 1.14 .80 .21* .18* .00 .06 .13* .07 .01 .06M 3.18 1.02 2.02 �.05SD .68 .89 .68 .88Cronbach’s Alphas .80 .83 .81 .83

Note. PA ¼ Positive affect; NA ¼ Negative affect; I ¼ Intensity; D ¼ Direction. *p < .05.

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237234

states was associated with these variables controlling for the di-rection of affective states.

Coping, attainment of sport achievement goals and sport satis-faction were selected for two reasons: (a) based on the CMRT byLazarus (1991, 2000), they were expected to be related to athletes’affective states; and (b) previous research showed significant re-lationshipsbetween intensityand/ordirectionof affective states andeach of these variables (e.g., Amiot et al., 2004; Diener, Emmons,Larsen, & Griffin, 1985; Gaudreau & Blondin, 2002). As previouslydiscussed, from a CMRT perspective, attainment of sport achieve-ment goals can be considered as a consequence of affective statesexperienced by athletes during competition. In addition, copingmediates the relationship between appraisals and the subsequentaffective states. Thus, both constructs can be considered as potentialantecedents of affective states (Lazarus,1991, 2000).Moreover, sportsatisfaction could be conceptualized as a consequence of athletes’affective states from a CMRT approach (Lazarus, 1991).

Previous research as well as results from study 1 lent credenceto these hypotheses by showing that intensity and/or direction ofaffective states were consistently associated with attainment ofsport achievement goals, coping and satisfaction (e.g., Amiot et al.,2004; Diener et al., 1985; Gaudreau & Blondin, 2002). Specifically,task-oriented-coping correlated positively with PA intensity anddirection whereas distraction- and disengagement oriented copingcorrelated positively with NA intensity (e.g., study 1; Gaudreau &Blondin, 2002). Attainment of sport achievement goal correlatedpositively with PA intensity and direction whereas satisfactioncorrelated positively with PA intensity (e.g., study 1; Amiot et al.,2004; Diener et al., 1985). As a consequence, based on theoreticalexpectations (Lazarus, 1991, 2000) and previous findings (e.g.,study 1; Amiot et al., 2004; Diener et al., 1985; Gaudreau & Blondin,2002), we hypothesized that direction of PA and NA will be asso-ciated with attainment of sport achievement goals, coping and/orsport satisfaction after intensity of these affects are held constant.

Method

Participants

Two hundred and ninety six French athletes (98 females and 198males) ranging in age from 14 to 45 years (M ¼ 21.61 years,SD ¼ 6.32) volunteered to participate in the study. On average theyhave been competing in their sport for 9.25 years (SD ¼ 4.06) andthey train 6.45 h per week (SD¼ 4.58). They participated in regional(N ¼ 160), national (N ¼ 119) or international sport events (N ¼ 17).

Measures

Based on CFAs results of study 1, the 4-factor structure of thePANAS-D designed to measure intensity and direction of PA (10

items) and NA (10 items) was used in study 2. Using the “moment”instruction (i.e., “right now, that is, at the presentmoment”, Watsonet al. (1988, p. 1070), the participants were asked to rate the in-tensity of each affective state on a scale of 1 (not at all or veryslightly) to 5 (extremely) as well as the direction of each affect on ascale from �3 (very debilitative) to þ3 (very facilitative). The alphacoefficients indicated that the reliability was acceptable withCronbach’s alpha coefficients varying from .80 to .83 (Table 4).

The CICS (Gaudreau & Blondin, 2002), described in detail inStudy 1, was used to measure coping strategies used by athletesduring competition. The alpha coefficients indicated that the reli-ability was acceptable with Cronbach’s alpha coefficients varyingfrom .79 to .85 (Table 4).

The A-SAGS (Amiot et al., 2004), described in detail in study 1,was used to measure attainment of sport achievement goals afterthe competition. The alpha coefficient indicated that the reliabilitywas good with a Cronbach’s alpha of .92.

An adaptation to the sporting context of the French version(Blais, Vallerand, Pelletier, & Brière, 1989) of the Satisfaction WithLife Scale (SWLS, Diener et al.,1985) was used to assess participants’global satisfaction with their sport lives and circumstances. Theadaptation consisted of minor changes in the wording of someitems to target the sport context. For example, the item “So far Ihave gotten the important things I want in life” was slightlymodified as follow: “So far I have gotten the important things Iwant in my sport”. The participants were asked to rate theiragreement with each of the five items on a scale of 1 (does notcorrespond at all) to 7 (corresponds very strongly). As we slightlymodified the wording of some items, we performed a confirmatoryfactor analysis with a maximum likelihood estimation procedure.Fit indices indicated that the measurement model was good(c2(5) ¼ 10.04, p ¼ .07, GFI ¼ .99, CFI ¼ .99, SRMR ¼ .03). The alphacoefficient indicated that the reliability was acceptable with aCronbach’s alpha of .80.

Procedure

The research was conducted in accordance with internationalethical guidelines that are consistent with APA norms. The coachesof each team/athlete were contacted to obtain permission toapproach their athletes and ask them to participate in the study.Participation was voluntary and parental consent was required forathletes under 18 years of age. The participants completed thePANAS-D, CICS, A-SAGS and adapted SWLS within 2 h after thecompetition. When completing the CICS and the PANAS-D, partic-ipants were instructed to indicate to which extent the items rep-resented the things that they had done or thought during thecompetition they had just completed, or how they had felt duringthis competition (e.g., Gaudreau & Blondin, 2002; Gaudreau et al.,2006). As for the A-SAGS and the adapted SWLS, participants

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237 235

were instructed to indicate to which extent the items representedthe things that they thought right now.

Data analysis

First, zero-order correlations between PANAS-D scores and CICS,A-SAGS and adapted SWLS scores were computed to examine therelationships between affective states and the three theoreticallyselected variables from a CMRT perspective. Second, partial corre-lations were computed to examine the incremental validity of thetwo dimensions of affective states (intensity and direction). Spe-cifically, for each dimension of the PANAS-D (PA, NA), wecomputed: (a) partial correlations between intensity of affectivestates and coping, attainment of sport achievement goals and sportsatisfaction controlling for direction of affective states; and (b)partial correlations between direction of affective states andcoping, attainment of sport achievement goals and sport satisfac-tion controlling for intensity of affective states.

Results

Zero-order correlations

Table 4 presents the zero-order correlations between the PANAS-D scores and the coping, sport satisfaction and attainment of sportachievement goals scores. Results showed distinct patterns of zero-order correlations between intensity and direction components ofaffective states. Whereas PA intensity and direction were positivelycorrelated with task-oriented coping, attainment of sport achieve-ment goals and sport satisfaction, PA direction (and not intensity)was negatively correlated with distraction- and disengagement-oriented coping. Whereas NA intensity and direction were signifi-cantly correlated with distraction- and disengagement-orientedcoping, NA direction (and not intensity) was positively correlatedwith attainment of sport achievement goals. Because our centralquestion pertained to the incremental validity of the two di-mensions of affective states (intensity and direction), we thenfocused on the partial correlations between affective states and thethree theoretically selected variables from a CMRT perspective.

Partial correlations

Table 4 presents the partial correlations between the PANAS-Dscores and the coping, sport satisfaction and attainment of sportachievement goals scores. After controlling for PA intensity, PA di-rection correlated negatively with distraction- and disengagement-oriented coping. After controlling for PA direction, PA intensitycorrelated positively with attainment of sport achievement goals,task- and distraction-oriented coping. After controlling for NA in-tensity, NAdirection correlatednegativelywithdistraction-orientedcoping and positively with attainment of sport achievement goals.After controlling for NA direction, NA intensity correlated positivelywith distraction- and disengagement-oriented coping.

Discussion

Replicating results of study 1, we demonstrated the importanceof measuring the intensity and direction dimensions of affectivestates separately. This statement was based on the rationale thatdistinct patterns of zero-order correlations between the intensityand direction components of each dimension of affective statesmeasured by the PANAS-D (PA, NA) have been identified. Never-theless, these results seem insufficient to assert unambiguously therelevance of a distinction between athletes’ intensity and directionof affective sates. It is why the main purpose of study 2 was to

examine the incremental validity of the two dimensions of athletes’affective states e intensity and direction e in their associationswith three theoretically selected variables from a CMRT perspective(Lazarus, 1991) e coping, attainment of sport achievement goalsand sport satisfaction.

Consistent with our hypotheses, our results indicated that bothintensity and direction of affective states were incrementally valid.Specifically, the five significant zero-order correlations identifiedbetween intensity of affective states and coping, attainment of sportachievement goals and sport satisfaction were always statisticallysignificant after controlling for the direction of affective states.Additionally, the non-significant zero-order correlation between PAintensity and distraction-oriented coping became significant aftercontrolling for PAdirection. Similarly, over the eight significant zero-order correlations identified between direction of affective statesand coping, attainment of sport achievement goals and sport satis-faction, five correlations were always statistically significant aftercontrolling for the intensity of affective states. The significant zero-order correlations between PA direction and task-oriented coping,PA direction and sport satisfaction as well as NA direction anddisengagement-oriented coping became non-significant after con-trolling for PA or NA intensity. As a whole, these results demon-strated the incremental validity of the intensity and direction of PAand NA and provided further evidence of the relevance of adistinction between athletes’ intensity and direction of affects.

General discussion

The present investigation yielded three primary findings: (a)The PANAS-D is a valid and reliable self-report questionnaire tomeasure both intensity and direction components of athletes’ af-fective states both before and during competition (study 1); (b)after statistically controlling for intensity of affective states, direc-tion of affective states was associated with coping and attainmentof sport achievement goals (study 2); and (c) after statisticallycontrolling for direction of affective states, intensity of affectivestates was associated with coping, attainment of sport achievementgoals and sport satisfaction (study 2). These results strongly sup-port the relevance of a distinction between the intensity and di-rection components of positive and negative affective states.

The relationships observed between athletes’ intensity of af-fective states and the three selected variables from a CMRTapproach are in line with theoretical expectations (Lazarus, 1991,2000) and previous research (e.g., Amiot et al., 2004; Dieneret al., 1985; Gaudreau & Blondin, 2002). PA intensity correlatedpositively with task-oriented coping and attainment of sportachievement goals whereas NA intensity correlated positively withdistraction- and disengagement-oriented coping. To the best of ourknowledge, previous study did not examine relationships betweenathletes’ direction of affective states (other than anxiety) andcoping, attainment of sport achievement goals and sport satisfac-tion. Our results revealed that PA direction correlated positivelywith task-oriented coping, attainment of sport achievement goalsand sport satisfaction as well as negatively with disengagement-oriented coping. NA direction correlated negatively with distrac-tion- and disengagement-oriented coping and positively withattainment of sport achievement goals even if this relationship wassignificant in study 2 but not in study 1.

Limitations and future directions

Because all variable used in the two studies were measuredusing a single source of data (self-report questionnaires), commonmethod bias could have distorted the findings. Future researchshould complement self-reported datawith informant-ratings (e.g.,

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237236

coach), physiological indicators of affective states (e.g., autonomicnervous system indicators) or objective indicators of performance(e.g., race time). Another limitation refers to the time frame used toassess affective states. In study 1, it can be argued that, 2 h before acompetition, not all athletes may have entered a stage duringwhich they start preparing themselves for competition. In study 1and 2, it can also be argued that a retrospective measurement ofaffective states experienced by athletes during competition isinsufficient because affective states are susceptible to change upduring the different phases of a sport competition. Future researchshould try to monitor affective states at multiple points during thecompetition across naturally segmented performance episodessuch as periods, innings, or rounds (Martinent et al., 2012).

Another limitation of the studies presented here was that sta-tistical analyses performed did not shed light on the interactionsbetween the intensity and direction components of the athleteaffective states. For example, Martinent et al. (2013) proposed analternative methodology (cluster analysis) that may provide re-searchers and practitioners with a useful way to examine complexnaturally-occurring combinations of athletes’ intensity and direc-tion of affective states. Their results indicated that multivariateprofiles of affect should be seen as useful heuristics to explicateconsequential self-regulatory and achievement correlates of affec-tive states both before and during competitions (Martinent et al.,2013).

Conclusion

Although research on direction of affective states (other thananxiety) is in its incipient stages, these studies suggest that in-tensity and direction of affective states should be conceptualized as,and therefore measured as, separate dimensions that indepen-dently contribute to the athlete’s affective experience related tocompetition. These studies also suggest that measuring simulta-neously intensity and direction of affective states with a self-reportquestionnaire such as the PANAS-Dmakes it possible to explore theindependent contributions of the intensity and direction compo-nents of affective states to psychological functioning and adjust-ment of athletes with regard to sport performance.

Acknowledgements

Wewant to thank Patrick Gaudreau for his help in planning anddesigning this research project.

References

Amiot, C. E., Gaudreau, P., & Blanchard, C. M. (2004). Self-determination, coping, andgoal attainment in sport. Journal of Sport & Exercise Psychology, 26(3), 396e411.

Anderson, J. C., & Gerbin, D. W. (1988). Structural equation modelling in practice: areview and recommended two-step approach. Psychological Bulletin, 103, 411e423. http://dx.doi.org/10.1037/0033-2909.103.3.411.

Blais, M. R., Vallerand, R. J., Pelletier, L. G., & Brière, N. M. (1989). The satisfactionscale: CanadianeFrench validation of the satisfaction with life scale. CanadianJournal of Behavioural Science, 21, 210e223. http://dx.doi.org/10.1037/h0079854.

Brown, K. W., & Ryan, M. R. (2003). The benefits of being present: mindfulness andits role in psychological well-being. Journal of Personality and Social Psychology,84, 822e884. http://dx.doi.org/10.1037/0022-3514.84.4.822.

Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes fortesting measurement invariance. Structural Equation Modeling, 9, 233e255.http://dx.doi.org/10.1207/S15328007SEM0902_5.

Crawford, J. R., & Henry, J. D. (2004). The Positive and Negative Affect Schedule(PANAS): construct validity, measurement properties and normative data in alarge non-clinical sample. British Journal of Clinical Psychology, 43, 245e265.http://dx.doi.org/10.1348/0144665031752934.

Crocker, P. R. (1997). A confirmatory factor analysis of the Positive and NegativeAffect Schedule (PANAS) with a youth sport sample. Journal of Sport & ExercisePsychology, 19(1), 91e97.

Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with lifescale. Journal of Personality Assessment, 49, 71e75. http://dx.doi.org/10.1207/s15327752jpa4901_13.

Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with un-observable variables and measurement error. Journal of Marketing Research,18(1), 39e50.

Gaudreau, P., & Antl, S. (2008). Athletes’ broad dimensions of dispositionalperfectionism: examining changes in life satisfaction and the mediating role ofsport-related motivation and coping. Journal of Sport & Exercise Psychology,30(3), 356e382.

Gaudreau, P., & Blondin, J.-P. (2002). Development of a questionnaire for theassessment of coping strategies employed by athletes in competitive sportsettings. Psychology of Sport and Exercise, 3, 1e34. http://dx.doi.org/10.1016/S1469-0292(01)00017-6.

Gaudreau, P., El Ali, M., & Marivain, T. (2005). Factor structure of the coping in-ventory for competitive sport with a sample of participants at the 2001 NewYork marathon. Psychology of Sport and Exercise, 6, 271e288. http://dx.doi.org/10.1016/j.psychsport.2004.01.002.

Gaudreau, P., Sanchez, X., & Blondin, J.-P. (2006). Positive and negative affectivestates in a performance-related setting: testing the factorial structure of thePANAS across two samples of FrencheCanadian participants. European Journalof Psychological Assessment, 22, 240e249. http://dx.doi.org/10.1027/1015-5759.22.4.240.

Gregorich, S. E. (2006). Do self-report instruments allow meaningful comparisonsacross diverse population groups? Testing measurement invariance using theconfirmatory factor analysis framework. Medical Care, 44, S78eS94. http://dx.doi.org/10.1097/01.mlr.0000245454.12228.8f.

Hanin, Y. L. (2007). Emotions in sport: current issues and perspectives. InG. Tenenbaum, & R. C. Eklund (Eds.), Handbook of sport psychology (3rd ed.) (pp.31e57). New York: Wiley & Sons.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structureanalysis: conventional criteria versus new alternatives. Structural EquationModeling, 6, 1e55. http://dx.doi.org/10.1080/10705519909540118.

Joiner, T. E., Sandin, B., Chorot, P., Lostao, L., & Marquina, G. (1997). Development andfactor analytic validation of the SPANAS among women in Spain: (More) cross-cultural convergence in the structure of mood. Journal of Personality Assessment,68, 600e615. http://dx.doi.org/10.1207/s15327752jpa6803_8.

Jones, G., & Swain, A. B. J. (1992). Intensity and direction dimensions of competitivestate anxiety and relationships with competitiveness. Perceptual and MotorSkills, 74(2), 467e472.

Jöreskog, K., & Sörbom, D. (2004). LISREL 8.71 for Windows (Computer Software).Lincolnwood, IL: Scientific Software International, Inc.

Killgore, W. D. S. (2000). Evidence for a third factor on the Positive and NegativeAffect Schedule in a college student sample. Perceptual and Motor Skills, 90,147e152. http://dx.doi.org/10.2466/pms.2000.90.1.147.

Koestner, R., Lekes, N., Powers, T. A., & Chicoine, E. (2002). Attaining personal goals:self-concordance plus implementation intentions equals success. Journal ofPersonality and Social Psychology, 83, 231e244. http://dx.doi.org/10.1037/0022-3514.83.1.231.

Lazarus, R. S. (1991). Emotion and adaptation. Oxford, UK: Oxford University Press.Lazarus, R. S. (2000). How emotions influence performance in competitive sports.

The Sport Psychologist, 14(3), 229e252.Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal and coping. New York: Springer.Lonigan, C. J., Hooe, E. S., David, C. F., & Kistner, J. A. (1999). Positive and negative

affectivity in children: confirmatory factor analysis of a two-factor model andits relation to symptoms of anxiety and depression. Journal of Consulting andClinical Psychology, 67, 374e386. http://dx.doi.org/10.1037/0022-006X.67.3.374.

MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equationmodelling in psychological research. Annual Review of Psychology, 51, 201e226.http://dx.doi.org/10.1146/annurev.psych.51.1.201.

Martens, R., Burton, D., Vealey, R. S., Bump, L., & Smith, D. E. (1990). Developmentand validation of the Competitive State Anxiety Inventory-2 (CSAI-2). InR. Martens, R. S. Vealey, & D. Burton (Eds.), Competitive anxiety in sport (pp. 117e213). Champaign, IL: Human Kinetics.

Martinent, G., Campo, M., & Ferrand, C. (2012). A descriptive study of emotionalprocess during competition: nature, frequency, direction, duration and co-occurrence of discrete emotions. Psychology of Sport and Exercise, 13, 142e151.http://dx.doi.org/10.1016/j.psychsport.2011.10.006.

Martinent, G., & Ferrand, C. (2009). A naturalistic qualitative study of the directionalinterpretation process of discrete emotions during high-stakes table tennismatches. Journal of Sport & Exercise Psychology, 31(3), 318e336.

Martinent, G., Ferrand, C., Guillet, E., & Gautheur, S. (2010). Validity of the Frenchversion of the Competitive State Anxiety Inventory-2 Revised (CSAI-2R)including frequency and direction scales. Psychology of Sport and Exercise, 11,51e57. http://dx.doi.org/10.1016/j.psychsport.2009.05.001.

Martinent, G., Nicolas, M., Gaudreau, P., & Campo, M. (2013). A cluster analysis ofaffective states before and during competition. Journal of Sport & Exercise Psy-chology, 35(6), 600e611.

Mehrabian, A. (1997). Comparison of the PAD and PANAS as models for describingemotions and for differentiating anxiety from depression. Journal of Psychopa-thology and Behavioral Assessment, 19, 331e357. http://dx.doi.org/10.1007/BF02229025.

Mellalieu, S. D., Hanton, S., & Fletcher, D. (2006). A competitive anxiety review:recent directions in sport psychology research. In S. Hanton, & S. D. Mellalieu(Eds.), Literature reviews in sport psychology (pp. 1e45). New York: Nova Science.

M. Nicolas et al. / Psychology of Sport and Exercise 15 (2014) 227e237 237

Mellalieu, S. D., Hanton, S., & Jones, G. (2003). Emotional labeling and competitiveanxiety in preparation and competition. The Sport Psychologist, 17(2), 157e174.

Melvin, G. A., & Molloy, G. N. (2000). Some psychometric properties of the Positiveand Negative Affect Schedule among Australian youth. Psychological Reports, 86,1209e1212. http://dx.doi.org/10.2466/pr0.2000.86.3c.1209.

Merz, E. L.,Malcarne,V. L., Roesch, S. C., Ko, C.M., Emerson,M., Roma,V.G., et al. (2013).Psychometric properties of Positive and Negative Affect Schedule (PANAS) orig-inal and short forms in an African American community sample. Journal of Af-fective Disorders, 151, 942e949. http://dx.doi.org/10.1016/j.jad.2013.08.011.

Merz, E. L., & Roesch, S. C. (2011). Modeling trait and state variation using multilevelfactor analysis with PANAS daily diary data. Journal of Research in Personality, 45,2e9. http://dx.doi.org/10.1016/j.jrp.2010.11.003.

Molloy, G. N., Pallant, J. F., & Kantas, A. (2001). A psychometric comparison of thePositive and Negative Affect Schedule across sex and age. Psychological Reports,88, 861e862. http://dx.doi.org/10.2466/pr0.2001.88.3.861.

Nicolas, M., Gaudreau, P., & Franche, V. (2011). Perception of coaching behaviors,coping, and achievement in a sport competition. Journal of Sport and ExercisePsychology, 33(4), 460e468.

Raykov, T. (2001). Estimationof congeneric scale reliabilityusing covariance structureanalysis with nonlinear constraints. British Journal of Mathematical and StatisticalPsychology, 54, 315e323. http://dx.doi.org/10.1348/000711001159582.

Robazza, C., & Bortoli, L. (2007). Perceived impact of anger and anxiety on sportingperformance in rugby players. Psychology of Sport and Exercise, 8, 875e896.http://dx.doi.org/10.1016/j.psychsport.2006.07.005.

Sideridis, G. D. (2005). Goal orientation, academic achievement, and depression:evidence in favour of a revised goal theory framework. Journal of EducationalPsychology, 97, 366e375. http://dx.doi.org/10.1037/0022-0663.97.3.366.

Terracciano, A., McCrae, R., & Costa, P. T. (2003). Factorial and construct validity ofthe Italian Positive and Negative Affect Schedule (PANAS). European Journal ofPsychological Assessment, 19, 131e141. http://dx.doi.org/10.1027//1015-5759.19.2.131.

Tuccitto, D. E., Giacobbi, P. R., & Leite, W. L. (2010). The internal structure of positiveand negative affect: a confirmatory factor analysis of the PANAS. Educationaland Psychological Measurement, 70, 125e141. http://dx.doi.org/10.1177/0013164409344522.

Watson, D., Clark, L. E., & Tellegen, A. (1988). Development and validation of briefmeasures of positive and negative affect: the PANAS Scales. Journal of Person-ality and Social Psychology, 54, 1063e1070. http://dx.doi.org/10.1037/0022-3514.54.6.1063.

Zevon, M. A., & Tellegen, A. (1982). The structure of mood change: an idiographic/nomethetic analysis. Journal of Personality and Social Psychology, 43, 111e122.http://dx.doi.org/10.1037/0022-3514.43.1.111.