British Journal of Health Psychology (2003), 8, 19-36 {© 2003 The Britsh Psychological Society Relationships between personality, an extended theory of planned behaviour model and exercise behaviour Ryan E. Rhodes'* and Kerry S. Courneya” University of Victoria, Canada 2University of Alberta, Canada Objectives. The purpose of this study was to investigate the theory of planned behaviour's (TPB) mediating hypothesis between the five-factor model of personality and exercise behaviour using an extended TPB model including concepts of affective and instrumental attitude, injunctive and descriptive norm, controllability, and self- efficacy. It was hypothesized that extraversion’s activity facet would have a significant direct effect on exercise behaviour while controlling for the TPB, based on the presupposition that activity may represent a disposition that predicts exercise beyond planned behaviour. Design. To test the replicability of these findings, we examined this research question with undergraduate students prospectively and cancer survivors, using a cross-sectional design. Results. Using structural equation modelling, the results indicated that activity had a significant effect (p< .05) on exercise behaviour (study | = .20; study 2 = .31) while controlling for the TPB. Conclusions. This study suggests the importance of extraversion’s activity facet on exercise behaviour, even when controlling for a TPB model with additional social- cognitive concepts and disparate population samples. ‘The physical benefits of exercise have been well documented (Bouchard, Shephard, & Stephens, 1994). Also, research suggests that the benefits of regular exercise extend beyond the primary prevention of chronic diseases, as regular exercise has been demonstrated to aid cancer rehabilitation by improving mental well-being and quality of life (Courneya, Mackey, & Jones, 2000; Courneya, Mackey, & Rhodes, in press). Unfortunately, a majority of the adult population are either sedentary or exercise with a frequency, duration, and intensity too low to derive any of these substantial advantages * Requests for reprints should be addressed to Ryan E. Rhodes, School of Physical Education, University of Victoria, PO Box 3015 STN CSC, V8W 3P1, Canada (e-mail: rhodes @uvca).20 Ryan E. Rhodes and Kerry S, Courneya (Stephens & Caspersen, 1994). Therefore, it is important to understand what factors influence regular exercise in order to implement effective intervention strate One leading theoretical model explaining informational and motivational influenc on behaviour is the theory of planned behaviour (TPB) (Ajzen, 1985, 1988, 1991). Further, empirical reviews of the TPB have supported a relationship for the prediction of many disparate health behaviours including exercise (Armitage & Conner, 2002; Godin & Kok, 1996; Hausenblaus, Carron, & Mack, 1997). Support for the TPB in the exerci domain is therefore robust, ranging from young adults (eg. Ci Bobick, & Schinke, 1999) to diseased populations such as cancer patients (¢.g. Courneya & Friedenreich, 1999). ‘The TPB suggests that the proximal determinant of volitional behaviour is one’s intention to engage in that behaviour. Attitudes and subjective norms are suggested to exert their effects upon behaviour through intentions. Attitudes are the overall evaluations of performing the behaviour by the individual. Subjective norms assess the social pressures on the individual to perform or not to perform a behaviour. ‘The TPB also tries to predict behaviours that are not completely volitional by incorporating perceptions of control over performance of the behaviour as an additional predictor of intention and behaviour (Ajzen, 1991), Perceived behavioural control (PBC) is the individual's perception of the extent to which performance of the behaviour is easy or difficult (Ajzen, 1991), ‘The TPB is hypothesized to mediate the relationship between external concepts and behaviour (Ajzen, 1985). Specifically, all concepts external to the TPB proper are hypothesized to influence a behaviour through attitude, subjective norm, and PBC, ‘This allows for interventions to target TPB concepts with an underlying insight into the antecedents of social cognition. Personality traits, defined as generalized dimensions of individual differences in tendencies to show consistent patterns of thoughts, feelings, and actions (McCrae & Costa, 1990), are considered to be concepts external to the TPB. However, recent research examining the mediating effect of the TPB between personality and exercise behaviour among younger females has demonstrated that the TPB is insufficient to account for personality completely (Courneya, Bobick, & Schinke, 1999; Rhodes, courneya, & Jones, in press). In this research, the five-factor model of personality (FFM) was utilized as a framework of personality. The FFM has emerged as the dominant framework for studying personality (Digman, 1990; McCrae & John, 1992) and its potential utility in the health domain has been duly noted (Marshall, Wortman, Vickers, Kusulas, & Hervig, 1994; Smith & Williams, 1992). The FFM is a version of trait theory that views human nature from the perspective of consistent and enduring individual differences (McCrae & John, 1992). The five personality dimensions of the FFM as highlighted by Costa and McCrae (1992) are neuroticism (N), extraversion (E), openness to experience (0), aggreeableness (A), and conscientiousness (C). A major advantage of the FFM is that it provides a comprehensive yet parsimonious taxonomy of personality traits at the highest hierarchical level of trait description (Digman, 1990; McCrae & John, 1992). Previous exercise behaviour research using the FFM revealed that extraversion had a dierct effect on exercise behaviour when controlling for the TPB (Courneya et al., 1999). Further, Rhodes ef al, (in press) found that E’s facet of activity possessed the significant effect upon exercise behaviour rather than general E or the facets of positive affect and sociability. The activity facet describes people who often keep busy, act vigorously, talk rapidly, and are energetic and forceful (McCrae & Costa, 1990). By this description, active individuals may engage in exercise behaviour beyond their planned intention because they frequently seek situations where the opportunity to be activeTPB, FFM and exercise 21 presents itself. In contrast, in osed to general activity would experi- ence fewer opportunities than their active counterparts to exercise beyond their planned exercise behaviour. ‘Theoretically, this suggests that the TPB is not adequate to fully explain the relationship between personality and exercise behaviour. However, this research included convenience samples of young females and requires replication on more disparate samples for evidence of generalizability. Further, the research used aggregate measures of TPB concepts. Current TPB models are recommended to include two specific sub-components within each TPB concept (Ajzen, 2002a), and it has been argued that research with extended TPB models may eliminate the direct effect of personality on exercise behaviour (Courneya ef al., 1999; Rhodes et al., in press). Ajzen (2002a) suggests that cach TPB concept comprises two specific sub-compo- nents, Specifically, attitude is hypothesized to be composed of affective (¢.g. enjoyable/ unenjoyable) and instrumental (e.g. beneficial/harmful) evaluations toward a behaviour. Empirical support for this conceptualization is based on the work of Rosenberg (1956) identifying distinct concepts of ‘hot’ and ‘cold’ attitude, and has been consistently found in TPB research (Ajzen, 1991; Ajzen & Driver, 1991; Conner & Armitage, 1998). Similar to attitude, factor analysis of PBC has led to distinct components of self- efficacy (e.g. ease/difficulty, confidence) and controllability (¢.g. personal control over behaviour, appraisal of whether the behaviour is completely up to the actor; Ajzen, 2002b). Empirical research has repeatedly demonstrated that these two concepts can be distinguished across a broad range of behaviours (Ajzen, 2002a; Conner & Armitage, 1998). Finally, subjective norm is hypothesized to include both the more traditionally measured injunctive component (¢.g. whether one believes that their social network wants them to perform the behaviour) as well as a descriptive component (whether ‘one’s social network performs a behaviour; Ajzen, 2002a). The recent addition of descriptive norm reflects growing empirical support for its predictive validity upon behavioural intention (Conner & Sparks, 1996; Nucifora, Gallois, & Kashima, 1993; Sheeran & Orbell, 1999; White, Terry, & Hogg, 1994) and theoretical support by extending normative pressure to include social pressure of belonging to a group who may or may not perform the behaviour (White et al., 1994). ‘The purpose of this study was to investigate the mediating hypothesis between personality, exercise, and the TPB using an extended TPB model that includes all sub- components of attitude, subjective norm, and PBC, Because personality is considered stable over time, active individuals were hypothesized to engage in exercise behaviour beyond their planned behaviour because they frequently seek situations where the ‘opportunity to be active presents itself. Therefore, activity was still hypothesized to have a direct effect on exercise behaviour, even using an extended TPB model. To test the replicability of the findings, we examined these research questions with two distinct samples of undergraduate students and cancer survivors. Method Participants Participants for study 1 were 303 (223 female and 80 male) undergraduate students participating in the study for extra credit in their introductory psychology course. The participants attended large group sessions during January and February, completing the22 Ryan E. Rhodes and Kerry S. Courneya self-report measures of the TPB at their own pace. Three hundred of these participants completed a 2-week follow-up measure of exercise behaviour. The mean age of participants was 19.99 (SD = 3.65 years), and the mean year in university for the sample was 1.67 (SD = 1.02). The variance/covariance matrix for the indicators of the relevant concepts was created using list-wise deletion of missing data and resulted in an analysis of 300 participants Participants for study 2 were breast, prostate, colon, and lung cancer survivors, diagnosed between January 1998 and December 1998, who had completed surgery and adjuvant therapy. Eligible participants (N = 802) were randomly selected via computer by the provincial cancer registry. Recruitment began with the physicians of potential participants being contacted by mail and telephone. Those potential participants who received physician approval (NW = 577) were subsequently mailed a selfadministered questionnaire package that included a notice of research study, instructions for participants, two copies of informed consent, a questionnaire booklet, and a self addressed stamped envelope. Participants were asked to complete and return one copy of the informed consent along with the questionnaires at their earliest convenience. The mail protocol was based on the Total Design Method (Dillman, 1983) to maximize response rates. This method consists of: (1) mailing the initial questionnaire packet, (2) mailing a postcard reminder 1 week later, and (3) mailing a second questionnaire packet 3 weeks later to all those who did not respond to the first mailing and postcard reminder. Completed questionnaires were received from 302 (52.3%) of the participants (203 breast cancer, 83 prostate cancer, 12 colon cancer, 4 lung cancer), The descriptive profile of the participants was: mean age of 60.7 (SD = 11.2) years, 69.8% were female, 77.4% were married or common-law, 40.6% had a university education, and 36.6% were currently working full or part-time. Further, 75.4% reported having surgery, 43.1% had chemotherapy, and 83.4% had radiotherapy. The variance/covariance matrix for the indicators of the relevant concepts was created using list-wise deletion of missing data and resulted in an analysis of 272 participants. Instruments Regular exercise was defined in study 1 as activities performed at a vigorous intensity three or more times per week for at least 30 minutes cach time, and in study 2 as activities performed at least at a moderate intensity three or more times per week for at least 20 minutes each time. Participants were asked to use these definitions when answering all exercise-related questions. The different exercise intensities between younger adults and cancer survivors reflect the differences in the American College of Sports Medicine (2000) guidelines for exercise among younger and older adults. Younger adults typically need higher intensities to benefit from exercise than older/ patient populations. Exercise altitude was measured using 7-point bipolar adjective scales, as suggested by Ajzen (2002a). The three items utilized to tap the instrumental attitude concept were: useful-useless, wise-foolish, beneficial-harmful. In contrast, the three items uscd to tap the affective attitude concept were: enjoyable-unenjoyable, interesting- boring, relaxing-stressful. The statement that preceded the adjectives was ‘For me, exercising regularly over the next 2 weeks will be . . .". Scales utilizing these items in previous exercise research have demonstrated acceptable reliability among both undergraduate students (Courneya ef a/., 1999) and cancer survivors (Courneya & Fricdenreich, 1999).TPB, FFM and exercise 23 Subjective norm was measured by 7-point scales that ranged from 1 (strongly disagree) to 7 (strongly agree). For measurement of the énjunctive norm concept, the items were: (1) ‘Most people in my social network want me to exercise regularly in the next 2 weeks’, and (2) ‘Most people in my social network would approve if I exercised regularly in the next 2 weeks’. Scales utilizing these items in previous exercise research have demonstrated acceptable reliability among both undergraduate students (Courneya ef al., 1999) and cancer survivors (Courneya & Friedenreich, 1999). For measurement of the descriptive norm concept, the items were: (1) ‘most of my friends exercise regularly’, (2) ‘most of my family members exercise regularly’, and (3) ‘Most of my co-workers exercise regularly’, These items were created in a format similar to that suggested by Ajzen (2002a), and the salient referents utilized are based on the research studies of Sallis, Grossman, Pinski, Patterson, & Nader (1987) into the important social referents of exercise behaviour, Perceived behavioural control was measured by three items used to tap the controllability concept and three items used to tap the self-efficacy concept. These items were adapted for the exercise domain from previous research by Conner and colleagues (Armitage & Conner, 1999a, 1999b), who identified that the self-efficacy and volitional control scales had acceptable reliabilities and good concept differentiation in principal components analysis. Self-efficacy was measured by the following items: (1) ‘How confident are you that you will be able to exercise regularly in the next 2 weeks?”; on a 7-point scale ranging from 1 (very unconfident) to 7 (very confident), (2) ‘How confident are you over the next 2 weeks that you could overcome obstacles that prevent you from exercising regularly?’; on a 7-point scale from 1 (completely unconfident) to 7 (completely confident), and (3) ‘I believe I have the ability to regularly exercise in the next 2 weeks’; on a 7-point scale from 1 (definitely do not) to 7 (definitely do), Controllability was measured with the following items: (1) ‘Whether or not I exercise regularly in the next 2 weeks is entirely up to me’; on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree), (2) ‘How much personal control do you feel you have over exercising regularly in the next 2 weeks?’; on a 7-point scale ranging from 1 (very little control) to 7 (complete control), and (3) ‘How much do you feel that exercising regularly in the next 2 weeks is beyond your control?’ (reversed scored); on a 7-point scale ranging from 1 (not at all) to 7 (very much). Exercise intention was assessed by three items as follows: (1) ‘In the next 2 weeks, my goal is to exercise’; rated on a 7-point scale from 1 (not at all) to 7 (every day), (2) ‘Over the next 2 weeks, I intend to exercise _ times per week’ rated on an open scale (Courneya, 1994), and (3) ‘I intend to exercise at least every other day over the next 2 weeks’; rated on a 7-point scale from 1 (strongly disagree) to 7 (strongly agree). These items have indicated excellent reliability in previous exercise research among both undergraduate students (Courneya ef al., 1999) and cancer survivors (Courneya & Friedenreich, 1999). Exercise behaviour was measured by the Godin Leisure Time Exercise Question- naire (GLTEQ; Godin, Jobin, & Bouillon, 1986; Godin & Shephard, 1985). The instru- ment contains three open-ended questions covering the frequency of mild (e.g. easy walking), moderate (€.g. fast walking), and strenuous (e.g. jogging) exercise completed during free time. The duration of these intensities was set for at least 30 minutes for study 1 and at least 20 minutes for study 2. An independent evaluation of this measure found it to be easily administered, brief, and reliable, and to possess concurrent validity based on various criteria including objective activity monitors and fitness indexes GJacobs, Ainsworth, Hartman, & Leon, 1993). The GLTEQ demonstrated a I-month24 Ryan E. Rhodes and Kerry S. Courneya test-retest reliability of 62 and concurrent validity coefficients of 52 with an objective activity indicator (CALTRAC accelerometer), .56 with VO3 max (as measured by expired gases), and —.43 with percentage body fat (as measured by hydrostatic weighing). These levels of reliability and validity compared very favourably to nine other self-report measures of exercise that were examined (Jacobs et al., 1993). Mild and moderate exercise were not included as an indicator of exercise behaviour in study 1 because of their incongruence with our definition of regular exercise in the social cognitive measures. However, strenuous and moderate exercise were summed to produce an exercise frequency indicator for study 2 and remain congruent with its respective social cognitive indicators. Personality was measured using the NEO Five Factor Inventory (NEO-FFD) (Costa & McCrae, 1992). This 60-item inventory measures five broad dimensions of personality (N, E, 0, A, ©), answered on a 5-point scale from ‘strongly disagree’ to ‘strongly agree Previous research and development of the NEO-FFI have suggested it as a reliable and valid instrument for the measurement of the FFM (Costa & McCrae, 1992). For the analysis, these items were modelled as facets per domain using Saucier's (1998) di of N (anxiety, depression, and self-reproach), E (positive affect, sociability, and activity), © (aesthetic interests, intellectual interests, and unconventionality), A (pro-social orientation and non-antagonistic orientation), and C (order, dependability, and goal- striving). Saucicr’s (1998) facet structure suggested reproducibility on two population samples and acceptable scale reliability. Model specification and analysis procedure All research questions were investigated using structural equation modelling. Structural equation modelling allows for both statistical significance tests for the size of each theoretical relation in the model and the assessment of overall model fit. Further specific conceptualizations in measurement and structure are investigated simultancously and free of measurement error. Models were estimated with maximum likelihood procedures and assessed using LISREL 8.20 for Windows (Joreskog & Sérbom, 1997). For specification of the latent concepts (i.e. estimations of the concept free of measurement error), the loading for each concept’s first indicator was pre-set to 1.0 in the structural equation model to create a seale. Based upon previous research identify- ing the proportion of error in self reported exercise (Courneya et al., 1997), the exercise behaviour indicator was given a fixed error estimate of 40% for the structural equation models, Although multiple indicators are desirable, @ priori fixed crror estimates on single indicators still allow for the researcher to exert meaning upon the latent concept and place theoretical constraints within the model (Hayduk, 1996). Exogenous con- cepts of attitude, subjective norm, and PBC were freed to correlate. For the analysis, a sequential step-wise procedure was implemented for model parsimony. First, all FEM concepts were modelled with effects upon exercise behaviour. ‘The facet scales provided by Saucier (1998) were utilized as indicators, and the FFM concepts at the structural level were freed to correlate based on research questioning the orthogonality of the FFM (Church & Burke, 1994; Rhodes ef at., in press; Silva ef al., 1994), Next, any FFM concept with a significant effect upon exercise behaviour was modelled using the proposed second and first-order structure of the personality trait (Costa & McCrae, 1995) and the facet structure proposed by Saucier (1998). This procedure divides variance of the observed personality measures into (1) variance resulting from the general latent concept, (2) variance resulting from the unique aspectsTPB, FFM and exercise 25 of the latent facets, and (3) variance resulting from measurement error. The second- order concept's variance was fixed at 1.0 for estimation of the loadings of its respective latent facets. Finally, any personality trait from the fact/general domain analysis with a significant effect upon exercise behaviour was subsequently included in the TPB model and freed upon TPB concepts and exercise behaviour. Assessment of model fit A number of statistics exist to assess the adequacy of structural models (Bollen, 1989). The most useful statistic for testing nested and alternative models is the chi-squared statistic (x). The x? goodness-offit test assesses the adequacy of the theorized model's creation of a covariance matrix and estimated coefficients in comparison with the observed covariance matrix. Models that result in a created covariance matrix that significantly deviates from the observed covariance matrix are judged to be inadequate. For comparison of nested and alternative models, the x’ difference value versus degrees of freedom provides a statistical test for which model fits the observed data better. However, the x? test has been criticized as an insufficient test alone to adequately assess model fit, generally because of sample size and power estimation problems or assumptions (Hu & Bentler, 1995). Therefore, inclusion of absolute and incremental fit indices is recommended (Hu & Bentler, 1999). Absolute fit indices assess how well an 4 priori mode! reproduces the sample data, while incremental fit indices measure the proportionate improvement in fit by comparing a target model with a more restricted baseline model, For the current study, the root mean square error of approximation (RMSEA) was included as an absolute fit index, and the comparative fit index (CF) was included as an index of incremental fit. General rules of thumb for acceptability of model fit using these indexes are >.94 for the CFI and <0.7 for RMSEA (Hu & Bentler, 1999). Results ‘The model investigating the FFM and exercise behaviour suggested a poor fit of the data for both study 1 (x7(76, N = 300) = 242.37, p <.01, CFI = .85, RMSEA = .09) and study 2 (7(76, N = 272) = 296.13, p <.01, CFI=.81, RMSEA = .10), Modification indices and standardized residuals suggested a scattered assortment of multiple changes in the measurement model of the FFM that would significantly improve the model fit but capitalize on chance with the sample data sets. The poor fit of the NEO-FFI has been reported and addressed in previous research utilizing structural equation modelling (Tokar, Fischer, Snell, & Harik-Williams, 1999) and indicates that the data fail to fit the parsimonious structure outlined by the NEO-FFI scales. This poor fit is a current limitation of the FFM model, despite its popularity of use (see MacCallum & Austin, 2000, for a review of this issue). In spite of the relatively poor model fit in both studies, concept loadings were all significant (p <.05) except for the unconventionality indicator for O in study 1 (see ‘Table 1). This item cluster was also the least reliable in the original research developing these NEO-FFI facet traits (Saucier, 1998) and suggests that unconventionality as presently measured is a poor indicator of openness to experience. The structural effects for both studies are presented in Fig. 1. Only E had a significant effect (p <.05) upon exercise behaviour of .28 and .22 for study 1 and study 2, respectively. Further, several26 Ryan E. Rhodes and Kerry S. Courneya Table 1. Factor loadings for the five-factor model of personality and exercise behaviour among undergraduate students (US) and cancer survivors (CS) Mean sD Factor loading Error variance us cs ous) cs us cs) uss csS Neuroticism Anxiety 2i 477075) 08371 7 49 AI* Depression 197 157085080 B3#B4* 32 30 Self-reproach 1.79 1.40 08! 0.82 a Te AI* AT* Extraversion Positive affect 270 «259061068 7H 73 ASK ATH Sociability 245 223 0.67 0.64 .61* Ae 62* 76* Activity 260 214 064 «069 6I* 4* 63 BOF Openness Aesthetic interests «2.35 «2.25099 08510071049 Intellectual interests 276 «= 234074 «069 AI* = STF RYT Unconventionality 2.20 1.81 OSS 0.58 2 26* 99% 93* Agreeableness Non-antagonsitic 2.33 2.67 059 0.53 87 BH 2S* are Prosocial orientation 3.26 326 050 045 -S7* —S6* BTEC Conscientiousness Orderliness 235 2.82 071 0.62 a” 67 50* a Goal striving 294 282 068 «06879 73" 37* Age Dependability 2.93 315 0.58 0.52 4 .78* 46 39 Exercise behaviour 195242 «20028377 7 40 40 *p <.05. large correlations were estimated for the FFM concepts, suggesting significant (p < .05) shared variance, especially between N and E, and E and A. Given that only E had the significant effect upon exercise behaviour for both undergraduates and cancer survivors, the next analysis modelled only E and its facets of positive affect, sociability, and activity as latent concepts. Congruent with our hypothesis, the effect of activity on exercise behaviour in addition to E was freed within the model. This model again suggested a moderately poor fit of the data for both study 1 (x°(61, N = 300) = 158.92, p<.01, CFI=.88, RMSEA —.07) and study 2 02(61, N = 272) = 190.07, p<.01, CFI=.80, RMSEA = .08). Modification indices and standardized residuals again suggested a scattered assortment of multiple changes in the measurement model of E that would significantly improve the model fit but capitalize on chance with the sample data sets. However, no significant changes were suggested for the structural effects upon exercise behaviour, Table 2 details the means and standard deviations for the indicators as well as the factor loadings and error variances for the model. All loadings for the latent concepts were statistically significant (p< 05). Further, the structural model (see Fig. 2) found all loadings on the second order E concept significant (p < .05) for both studies. Finally, both study 1 and study 2 found the effect of activity upon exercise behaviour significant (p <.05) with effects of 50 and .37, respectively, whereas general extraversion was non-significant (p > .10).TPB, FFM and exercise 27 Openness to experience Exercise behaviour Agreeableness Conscientiousness Figure I. Effects of the fivefactor model of personality on exercise behaviour. Note: All effects are standardized; *p <.05. Top: undergraduate students; bottom: cancer survivors.28 Ryan E. Rhodes and Kerry S. Courneya Positive Activiy afet Figure 2. Effect of extraversion’s facets on exercise behaviour. Note: All effects are standardized: *p <.05. Top: undergraduate students; bottom: cancer survivors, Table 2. Factor loadings for extraversion and exercise behaviour among undergraduate students (US) and cancer survivors (CS) Mean sD Factor loading Error variance us cs us cs us cs ous cs Extraversion Positive affect #7 318 298 «079 «(087 36 33 87 7 #12 230-209 094 «099 34* 42 gr BF #37 278 «2.64 080 086 63+ Bf 30" 29% Had 255 260 «096 «099 7I* S7*# 49% 6B Sociability #2 251 205 0951.03.63 63 6I* 60" #I7 321 317077076 = 70*o7*SI* SSH #27 192 1911.02.05, S637 BY #57 216 1.80 1.08 1.09 38* 25* Bg 8g Activity #22 265 «223 «085 «095 «= 4735 7B* BT #32 226 «183096 «0.98 s2F SH IF BBE HAT 274 «(193087 IS a AGI 79H #52 276 258 «= 1.01. 097 7074 SIH AS Exercise behaviour 1952420 -2000-«-283 77 7 40 40 *p<.05.TPB, FFM and exercise 29 The final model included only the activity trait freed upon concepts of the TPB and exercise behaviour. This model suggested a marginal fit of the data for both study 1 (x*(245, N = 300) = 387.72, p<.05, CFI=.95, RMSEA = .04) and study 2 (°(245, N = 272) = 403.03, p <.05, CFI = .94, RMSEA = .05). Further, the modifica- tion indices and standardized residuals suggested minimal changes that would signifi- cantly improve the model fit, supporting that the effects within the model can be interpreted with moderate confidence. ‘Table 3 details the means and standard devia- tions for the indicators as well as the factor loadings and error variances for both study 1 and study 2. All loadings for the latent concepts were statistically significant (p <,05) Further, no correlated structural disturbance terms (i.e. inter-factor correlations) for all TPB concepts exceeded .70, suggesting concept distinctness in both models. Figure 3 details the structural effects of the model. Specifically, activity had a significant effect (p<.05) upon all concepts in the model, including exercise behaviour (study 1 = —.20; study 2 = .31) while controlling for intention, self-efficacy, and con- trollability. For the ‘TPB, affective attitude had a significant effect (p <.05) upon intention for cancer survivors, whereas instrumental attitude had a significant effect (p <.05) upon intention for undergraduate students. Neither study identified significant effects for injunctive and descriptive norm. However, self-efficacy had the largest significant effect (<.05) upon intention for both undergraduate students and cancer survivors, but neither self-efficacy nor controllability were significant upon exercise behaviour while controlling for intention and activity, Finally, intention had the largest significant effect (p < .05) upon exercise behaviour for both studies. cussion ‘This study investigated the mediating hypothesis between personality and exercise, for an extended TPB model including concepts of affective and instrumental attitude, injunctive and descriptive norm, and controllability and self-efficacy. It was hypothe- sized that E's activity facet would have a significant direct effect upon exercise behaviour while controlling for the TPB. First, however, the complete FFM was examined for effects on exercise behaviour, and only E had a significant effect (p < .05) for both undergraduate students and cancer survivor samples, Previous research with exercise behaviour and the FFM has identified N, E, and C as significant correlates (Courneya et al., 1999; Courneya & Helisten, 1998) However, research that has regressed exercise behaviour on the FFM simultaneously has found that only E has a significant effect (Courneya et al., 1999; Rhodes ef ai., in press) ‘This suggests that the significant variance found in N and C is also carried through E. Given that only E had the significant effect upon exercise behaviour, the next analysis modelled only E and its facets of positive affect, sociability, and activity as latent concepts. Both studies found only the effect of activity on exercise behaviour significant as hypothesized, suggesting that it is the specific variance of activity, and not the common variance of E of specific variances of positive affect and sociability, that is responsible for the effect of personality upon exercise behaviour. This is in agreement with the initial hypothesis and replicates the earlier findings of Rhodes et al. (in press) that activity is the key component of extraversion with an influence upon exercise behaviour. One principal limitation of these analyses is the poor model fit from the current conceptualization and measurement of the NEO-FFI. Currently, the failing model of the NEO-FFI suggests that the conceptualization is inadequate and limits the confidence of30 Ryan E. Rhodes and Kerry S. Courneya teva Figure 3. Effect of extraversion’s activity facet on the theory of planned behaviour and exercise behaviour. Note: All effects are standardized; *p <.05. Top; undergraduate students; bottom: cancer survivors. applied researchers who wish to explore the effects of personality on behaviour. Poor fits of the NEO-PI, the longer instrument from which the NEO-FFI was derived, have also been consistently cited (e.g. Church & Burke, 1994; Silva et al., 1994). McCrae, Zonderman, Costa, Bond, and Paunonen (1996) have argued that the FFM lacks a simple structure and should be considered an amalgamation of all the underlying facet traits to various degrees of factor loading. However, this approach results in a fully saturated model (ie. all pathways are free) and an absence of testing power. Theories generate hypotheses that should be testable and observable by the most optimal means. ‘Therefore, a model in which every item is an indicator of every concept is of little theoretical value. Further, without a clear structure to test, the NEO-FFI would have no merit for five distinct scales instead of a single personality scale. Clearly, a five-factor structure is well supported in personality research (Digman, 1990), and the NEO-FFI can be improved to better represent this structure, Future research needs to improve the conceptualization of the model, the measurement of concepts, or perhaps both in order to improve correspondence with observed data. Suggestions of improving model fit in the measurement model may be to use only the ‘purest’ indicators for each concept Gn representativeness and relevance of their respective domain), thereby reducing measurement error and proportionality constraints (Hayduk, 1996), or acknowledge some crossoadings of personality concepts in an a priori measurement model. At the structural level, modelling personality concepts at the lower-order level in correlated clusters, as highlighted by Saucier and Ostendorf (1999), rather than higher-order structures, may also improve model fit. It remains likely, given the current misfit of the NEO-FFI, that all of these strategies are required in some part to improve personality measurement and conceptualization, Despite the poor fit of the previous models, the final model included only the activityTPB, FFM and exercise 31 Table 3. Factor loadings for activity, the theory of planned behaviour and exercise among undergraduate students (US) and cancer survivors (CS) Mean sD Factor loading Error variance us cs us cs us cs ous cs Activity #22 265 «223 «08 «09S 3330 #32 226 «183096 «098 Alt 62 #47 274 «193087 IS 386 BTR #52 276 «258 «1.0097 Affective attitude Enjoyable 529 487 «152 16476 8TH Interesting 499 480 «14315275704 Relaxing 505 508 141 143 «65700736746 Instrumental attitude Useful 623 591) 086 127 kIT Wise 630 618 078 097 66 82 S7 33 Beneficial 641 601 084 1.09708 21 Injunctive norm “Want! 521 53113815675 89H ‘Approve’ 633 5971038125 Cn) 35 Descriptive norm Friends exercise 440394169 17075 84S 30 Family exercise 392 «4.12, 186 199 BSB Co-worker exercise 392 380138163, S784 30 Self-efficacy SEI 542 502175 1914S 58 SE2 519 535 «14715568 77 54 AL SE3 610 600 «131 142788 25 Controllability PBCI 61S 622138 1362 8B PBC2 604 616128 1349S aT PBC3 542 540 180 198 76 64 42 58 Intention INT I 452 484 «136 1709498120 INT2 318373 165195 INT3 494 497-206-214 HC (aT Exercise behaviour 195-242 «200-2837 A trait freed upon concepts of the TPB and exercise behaviour, and fit the data reasonably well. The key finding was that activity had a significant effect (p <.05) upon exercise behaviour while controlling for intention, controllability, and self-efficacy. This suggests robustness of the original findings of Rhodes ef al. (2001), and that even an extended TPB model is insufficient to provide a complete mediation between personality and exercise behaviour. ‘The personality trait of activity appears to account for prediction of exercise behaviour beyond one’s planned behaviour. Active individuals may engage in exercise32 Ryan E. Rhodes and Kerry S. Courneya behaviour beyond their planning because they frequently seek situations where the opportunity to be active presents itself. In contrast, individuals less disposed to general activity would experience fewer opportunitics than their active counterparts to exercise beyond their planned exercise behaviour. This hypothesis is in accordance with the research on extraversion by Eysenck (1981), suggesting that introverts are hypothesized to possess higher base rates of arousal than extraverts, motivating them to withdraw from stimulation. Conversely, extraverts actively seek out stimulation in order to compensate for their low arousal, and one key facet for this outlet may be activity. Alternatively, social cognitions may always precede behaviour, as theoretically surmised (Ajzen, 1991), but lack the temporal stability in measurement to mediate more stable predictors. Temporal stability is a principal limitation outlined in the TPB (Ajzen, 2002a), and exercise intentions have been demonstrated to be unstable even over a 2-day duration (Courneya & McAuley, 1993). In contrast, temporal stability of personality has been well demonstrated (McCrae ef al., 2000). Therefore, personality may not pose a theoretical threat to the TPB but may add predictive value because of a greater temporal stability. Moreover, extending the TPB further, with even more social- cognitive concepts, is unlikely to account for the effects of personality on behaviour because of this temporal stability limitation ‘Though this suggests that the addition of personality to the TPB may have little theoretical value, the practical significance may be paramount. Exercise researchers are generally interested in predicting future behaviour as accurately as possible so as to deliver useful interventions. If social cognitions fluctuate, while personality is a very stable predictor, interventions will be likely to benefit from consideration of one’s personality along with social cognitions. Consequently, a practical implication of this research may be that personality needs to be taken into account when developing exercise interventions, rather than assuming that targeting social cognitions will subsume these differences. For example, individuals with positive social cognitions but lower on activity may require special intervention efforts to increase the ante- cedents of behavioural intention. In comparison, individuals higher on activity should benefit from TPB-based interventions but are also more likely to exercise beyond these efforts. Still, a formal examination of this temporal stability hypothesis is required, utilizing personality and the TPB in a longitudinal analysis. A few avenues of future research may improve the understanding of the linear relationships between personality, social cognition, and exercise behaviour. First, previous rescarch has suggested that personality may not be best suited for the study of exercise as a continuous variable, but rather in the form of stages of exercise preparedness (Rhodes et a/., 2001). This research revealed that precontemplators (those individuals not considering the adoption of regular exercise) were not differ- entiated from any other stage of exercise behaviour based on personality. One possible explanation for this finding may be related to the theorized motivational properties of personality. The five-factor domains are non-health- and non-exercise-specific, and are hypothesized to exert influence through temperamental, experiential, and instrumental effects (Costa & McCrae, 1992). Currently, activity may be the personality trait best suited for a continuous variable of exercise behaviour, as it probably provides the strongest temperamental effect towards actively secking exercise. However, other personality characteristics may exert an instrumental influence on exercise behaviour once an individual has made the decision to exercise. Using C as an example, it may be that high-C individuals are no more likely to decide to exercise than low-C individuals,TPB, FFM and exercise 33 but once they do decide to exercise, their high-C personality makes it more likely that they will follow through on their commitment. Therefore, researchers may wish to investigate the possibility of decision to exercise as a moderator of personality and exercise behaviour. Second, only three facets of each larger personality domain were utilized in the current models, Although these domains were found to be reliable facets of the NEO-FFI in previous research (Saucier, 1998), more facets are likely to encompass these large domains (Costa & McCrae, 1992). For example, E may include facets of assertiveness, and adventurousness, which are not represented in the present facet structure. In either case, activity still seems the most theoretically relevant personality facet for extraversion in the exercise domain. However, more facets of N and C may yield more information about the specific effects of facets in these domains. Third, interactions among the FFM general factors and facets may yield a more complete description of the exercise personality. For example, a high activity, low-N, and high-C individual may be more predictive of regular exercise behaviour than an individual high on activity alone. Though the current study lacks the sample size, future researchers may wish to investigate these possible personality styles for improved clarity of the relationships between personality, social cognition, and exercise behaviour. One interesting finding not hypothesized in the study was the differential predictive utility of affective and instrumental attitude between samples. Affective attitude had a significant influence on intention for cancer survivors, whereas instrumental attitude had a significant influence on intention for undergraduate students. Possible reasons for this difference between the attitude components may be age discrepancies and/or from the cancer experience, since attitude means, variabilities, and inter-concept correlations are comparable. Perhaps the psychological effects from the cancer experience and/or physical impairments as a result of cancer treatment or aging may amplify the importance of the affective component for exercise intention despite instrumental attitude. In contrast, perhaps undergraduate students intend to act more on instru- mental evaluations towards the health benefits of exercise, Certainly, more research is warranted to explore the underlying beliefs of attitude with different populations for a greater insight into the reasons for these disparities. Other avenues for future direction may also prove beneficial. One possibility is an examination of the effect of personality and the TPB with different exercise intensities, Although this research was focused on and corresponded with optimal exercise intensity prescriptions from the American College of Sports Medicine (2000), different personality traits may have effects on frequency of lighter exercise intensities. Differ- ences in gender may also alter the findings. The current research lacked the sample size and subsequent power to adequately test this possibility, but examination is warranted in future research Nevertheless, the findings of this research suggest important considerations for TPB- based intervention, underscoring the need for distinct tailored interventions of person- ality and social cognition within the exercise domain. Limitations of the previous studies beyond the measurement of personality warrant mention. First, our hypothe- sized causal effects in these structural equation models represent one conceptualization of the relationships between the measured indicators. Though the models were found to have generally acceptable fits of the observed variance/covariance matrices, several alternative models may represent the data equally well or better. Second, the cross- sectional design of study 2 limits the accuracy of a modelled concept of exercise behaviour. The TPB concepts were set to predict 2 weeks of prospective behaviour,34 Ryan E. Rhodes and Kerry S. Courneya which may not be optimally represented by the past 2 weeks’ behaviour. However, previous research among similar cancer survivor populations suggests that exercise behaviour is very stable with high correlations between past and present behaviour (Courneya & Friedenreich, 1999). Further, no exercise intervention was delivered in the study, so changes in exercise behaviour over the following 2 weeks were not hypothe- sized, Still, future research is required to examine these multiple TPB concepts as causal influences upon exercise behaviour among cancer survivors, using a prospective design. Finally, none of the measures in this study include lie scales, though the NEO-FFI and ‘TPB do reverse code items in order to attempt to break response sets, This suggests that dishonest responses to the measures by some participants may bias the results. Never- theless, there were very few missing data, and the internal consistencies/factor loadings for the measures were high (indicating consistent and logical responses). References Ajzen, 1. (1985). From intention to actions: A theory of planned behavior. In J. Kuhl & J. Beckmann (Eds), Action-control: From cognition to bebavior (pp. 11-39). Heidelberg: Springer. Ajzen, I. (1991). The theory of planned behavior. In Organizational Bebavior and Human Decision Processes, 50, 179-211 Ajzen, I, (2002a). Construction of a standard questionnaire for the theory of planned behavior. Retrieved from the World Wide Web: http://www-unix.oit.umass.edu/~aizen/ Ajzen, I. (2002b). Perceived behavioral control, self-efficacy, locus of control, and the theory of planned behavior. Journal of Applied Social Psychology, 32, 665-683. Ajzen, L., & Driver, B. L. (1991). Prediction of leisure participation from behavioral, normative, and control beliefs: An application of the theory of planned behavior. Leisure Sciences, 13, 185-204. American College of Sports Medicine (2000). Guidelines for Exercise Testing and Prescription Baltimore, MD: Lippincott, Williams & Wilkins Armitage, C. J., & Conner, M, (1999a). Distinguishing perceptions of control from selfefficacy: Predicting consumption of a low-fat diet using the theory of planned behavior. Journal of Applied Social Psychology, 29, 72-90. Armitage, C. J., & Conner, M. (1999b). The theory of planned behaviour: Assessment of predictive validity and perceived control. British Journal of Social Psychology, 38, 35-54 Armitage, C. J., & Conner, M. (2002). Efficacy of the theory of planned behaviour: A meta-analytic review. British Journal of Social Psychology, 40, 471-499. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Bouchard, C., Shephard, R. J., & Stephens, T. (1994). Physical activity, fitness, and health. Champaign, [L: Human Kinetics. Church, . A., & Burke, P. J. (1994), Exploratory and confirmatory tests of the big five and ‘Tellegen’s three- and four-dimensional models. fournal of Personality and Social Psychology, 66, 93-114 Conner, M., & Armitage, C. J. (1998). Extending the theory of planned behavior: A review and avenues for further research. Journal of Applied Social Psychology, 28, 1429-1464 Conner, M., & Sparks, P. (1996). The theory of planned behavior and health behaviors. In M. Conner & P. Norman (Eds.), Predicting health bebaviour: Research and practice with social cognition models (pp. 121-162). Buckingham, UK: Open University Press, Costa, PT, Jn, & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-PER) and NEO Five Factor Inventory (NEO-FFD) Professional Manual, Odessa, FL: Psychological Assessment Resources. Costa, PT, Jr, & McCrae, R. R. 1995). Domains and facets: Hierarchical personality assessment using the revised NEO Personality Inventory. Journal of Personality Assessment, 64, 21-50.TPB, FFM and exercise 35 Courneya, K. S. (1994). Predicting repeated behavior from intention: The issue of scale correspondence. Journal of Applied Social Psycbology, 24, 580-594. Courneya, K. 8., Bobick, T. M., & Schinke, R. J. (1999). Does the theory of planned behavior mediate the relationship between personality and exercise behavior. Basic and Applied Social Psycbology, 21, 317-324. Courneya, K. S., Estabrooks, P A., & Nigg, C. R. (1997). A simple reinforcement strategy for increasing attendance at a fitness facility. Health Education and Bebavior, 24, 708-715. Courneya, K. S., & Friedenreich, C. M. (1999). Utility of the theory of planned behavior for understanding exercise during breast cancer treatment, Psycho-Oncology, 8, 112-122. Courneya, K. S., & Hellsten, L, A. (1998). Personality correlates of exercise behavior, motives, barriers and preferences: An application of the five-factor model. Personality and Individual Differences, 24, 625-633. Courneya, K. S., Mackey, J., & Jones, L. W. (2000). Coping with cancer: Can exercise help? Physician and Sports Medicine, 28, 49-73, Courneya, K. S., Mackey, J., & Rhodes, R. E. (in press). Exercise and cancers other than breast. In L. M. LeMura & S. P. von Duvillard (Eds.), Clinical exercise physiology: Application and phystological principles. Baltimore, MD: Lippincott, Williams & Wilkins. Courneya, K. S., & McAuley, E. (1993). Can short-range intentions predict physical activity participation? Perceptual and Motor Skills, 77, 115-122 Digman, J. M. (1990). Personality structure: Emergence of the five-factor model. Annual Review of Psychology, 41, 417-440. Dillman, D. A. (1983). Mail and other self-administered questionnaires. In P. H. Rossi, J. D. Wright, & A.B. Anderson (Eds.), Handbook of survey research (pp. 359-378). Toronto: Academic Press. Eysenck, H. J. Ed.) (1981). A model for personality. Berlin: Springer-Verlag. Godin, G., Jobin, J., & Bouillon, J. (1986). Assessment of leisure time exercise behavior by self-report: A concurrent validity study. Canadian Journal of Public Health, 77, 359-361. Godin, G., & Kok, G. (1996). The theory of planned behavior: A review of its applications to health-related behaviors. American Journal of Health Promotion, 11(2), 87-98. Godin, G., & Shephard, R. J. (1985). A simple method to assess exercise behavior in the community. Canadian Journal of Applied Sport Science, 10, 141-146. Hausenblaus, H. A., Carron, A. V., & Mack, D. E. (1997). Application of the theories of reasoned action and planned behavior to exercise behavior: A meta-analysis. Journal of Sport and Exercise Psychology, 19, 36-51. Hayduk, L. A. (1969). LISREL issues, debates, and strategies. Baltimore, MD: Johns Hopkins University Press. Hu, L., & Bentler, P. M. (1995). Evaluating model fit, In R, H, Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp.76-99). Thousand Oaks, CA: Sage. Hu, L., & Bentler, P.M, (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. Jacobs, D. R., Ainsworth, B. E., Hartman, T. J., & Leon, A. S. (1993). A simultaneous evaluation of ten commonly used physical activity questionnaires. Medicine and Science in Sports and Exerctse, 25, 81-91. Joreskog, K., & Sérbom, D. (1997). Lisrael 8.20 for Windows. Scientific Software International, Inc.: Lincolnwood, IL. MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation modeling in psychological research. Annual Review of Psychology, 51, 201-226. McCrae, R. R., & Costa, P. T., Jr. (1990). Personality in adulthood (2nd ed.). New York: Guilford Press, McCrae, R. R., & John, O, P. (1992). An introduction to the five-factor model and its applications. Journal of Personality, 60, 175-215. McCrae, R. R., Costa, P. T., Ostendorf, F, Angleitner, A. Hrebickova, M., Avia, M. D., Sanz, J., & Sanchez-Bernardos, M. L. (2000). Nature over nurture: Temperament, personality, and life span development. Journal of Personality and Social Psychology, 78, 173-186.36 Ryan E. Rhodes and Kerry S. Courneya McCrae, R. R., Zonderman, A. B., Costa, P. T., Bond, M. H., & Paunonen, S. V. (1996). Evaluating replicability of factors in the revised NEO personality inventory: Confirmatory factor analysis versus procrustes rotation. Journal of Personality and Soctal Psychology, 70, 552-566, Marshall, G. N., Wortman, C. B., Vickers, R.R., Kusulas, J. W., & Hervig, L. K, (1994), The fivefactor model of personality as a framework for personality health research. Journal of Personality and Social Psychology, 67, 278-286. Nucifora, J., Gallois, C., & Kashima, Y. (1993). Influences on condom use among undergraduates: ‘Testing the theories of reasoned action and planned behavior. In D. J. Terry, C. Gallois, & M McCamish (Eds.), The theory of reasoned action: Its application to AIDS preventive bebavior (pp.47-64). New York: Pergamon Press. Rhodes, R. E., Courneya, K. S., & Bobick, T: (2001). Personality and exercise participation across the breast cancer experience, Psycho-Oncology, in press. Rhodes, R. E., Courneya, K. S., & Jones, L. W. Gin press). Relationships between extraversion, the theory of planned behaviour, and exercise: The unique role of extraversion's activity facet. Journal of Applied Social Psychology. Rosenberg, M. J. (1956). Cognitive structure and attitudinal affect. Journal of Abnormal and Social Psychology, 53, 367-372 J. F, Grossman, M.S., Pinski, R.B., Patterson, I. L., & Nader, P. R, (1987). The development of scales to measure social support for diet and exercise behaviors. American Journal of Epidemiology, 127, 933-941. Saucier, G. (1998). Replicable item-cluster subcomponents in the NEO Five-Factor Inventory. Journal of Personality Assessment, 70, 263-276. uucier, G., & Ostendorf, E (1999). Hierarchical subcomponents of the big five personality factors’ A crosslanguage replication. Journal of Personality and Social Psychology, 76, 613-627. Sheeran, P, & Orbell, S, (1999). Augmenting the theory of planned behavior: Roles for anticipated regret and descriptive norms. Journal of Applied Social Psychology, 29, 2107-2142. Silva, E, Avia, D., Sanz, J., Martinez-Arias, R., Grana, J. L., & Sanchez-Bernardos, L. (1994). The fivefactor model—I. Contributions to the structure of the NEO-PI. Personality and Indivi- dual Differences, 17, 741-753. Smith, T. W,, & Silliams, PG. (1992). Personality and health: Advantages and limitations of the five- factor model. Journal of Personality, 60, 395-423. Stephens, T, & Caspersen, C. J. (1994), The demography of physical activity. In C, Bouchard, R. J. Shephard, & 'T, Stephens (Eds.), Physical activity, fitness, and health. Champaign, IL: Human Kinetics. ‘Tokar, D. M., Fischer, A. R., Snell, A. B, & Harik-Williams, N. (1999). Efficient assessment of the five factor model of personality: Structural validity analyses of the NEO-FFL, Measurement and Evaluation in Counseling and Development, 32, 14-30. White, K. M,, Terry, D. J., & Hogg, M. A. (1994). Safer sex behavior: The role of attitudes, norms, and control factors. Journal of Applied Social Psycbology, 24, 2164-2192 Received 30 May 2001; revised version 3 September 2001Copyright © 2003 EBSCO Publishing