The functions of observational learning questionnaire (FOLQ)

Psychology of Sport and Exercise 6 (2005) 517–537 www.elsevier.com/locate/psychsport

The functions of observational learning questionnaire (FOLQ) Jennifer Cumminga, Shannon E. Clarkb, Diane M. Ste-Marieb,*, Penny McCullaghc, Craig Halld a

School of sport and exercise sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK b School of Human Kinetics, University of Ottawa, Ottawa, Ont., Canada K1N 6N5 c Department of Kinesiology and Physical Education, University of California, Hayward, 25800 Carlos Bee Blvd., Hayward, CA, USA 94542 d School of Kinesiology, University of Western Ontario, London, Ont., Canada N6A 3K6 Received 20 January 2004; accepted 22 March 2004 Available online 16 December 2004

Abstract Objectives: The main aim of the present investigation was to examine how athletes use observational learning (OL) through the development of a valid and reliable questionnaire. A second purpose was to determine how the functions of OL that emerged compared to the functions of imagery that have already been determined [Paivio, A. (1985). Cognitive and motivational functions of imagery in human performance. Canadian Journal of Applied Sports Sciences, 10, pp. 22S–28S] general analytical framework for imagery. Design: Four samples of questionnaire data, presented in three studies. Methods: Male and female athletes in a variety of sports ranging from recreational to the elite level completed the questionnaire. Study 1 consisted of 400 athletes (197 male and 203 female) with a mean age of 21.26 (SDZ2.88). For Study 2, 953 athletes (462 male, 483 female, eight unreported), with a mean age of 22.37 (SDZ5.15) completed the questionnaire. Finally, Study 3 consisted of 200 athletes (77 male, 123 female) with a mean age of 19.62 years (SDZ2.17). Results: Study 1 consisted of computing a principal component analysis of the Functions of Observational Learning Questionnaire (FOLQ). From this, the 17-item FOLQ emerged that contained three factors (skill, strategy, and performance). In Study 2, a confirmatory factor analysis was computed that confirmed the items and the factor structure of the questionnaire. Finally, Study 3 confirmed the concurrent validity and the test–retest reliability of the questionnaire, along with examining group differences in terms of OL usage by athletes.

* Corresponding author. E-mail address: [email protected] (D.M. Ste-Marie). 1469-0292/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.psychsport.2004.03.006

518

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Conclusions: Athletes use OL for both cognitive functions (skill and strategy) and motivational functions (optimal arousal and mental performance state). It seems that athletes use OL primarily for cognitive functions, whereas, imagery is mainly used by athletes for motivational functions. Overall, the results indicate that the FOLQ may be a useful tool for examining research questions surrounding OL. q 2004 Elsevier Ltd. All rights reserved. Keywords: Observational learning; Imagery; Athletes; Questionnaire development; Functions; Motivation; Cognitions

Motivated by inconsistent conclusions and variable effect sizes found in the reviews of the imagery literature available at the time, Paivio (1985) developed a general analytical framework to explain how motor skills can be improved through mental practice. Paivio’s main purpose was to determine when and why imagery techniques were not effective by analyzing the functional roles through which imagery can affect performance. According to the framework, imagery plays both a motivational and cognitive role in mediating behaviour, and each one operates at either a specific or general level. As a result, imagery can serve a cognitive specific function (i.e. learning and performance of skills), a cognitive general function (CG; i.e. learning and performance of strategies, routines, and game plans), a motivational specific function (MS; i.e. obtainment of goal-related behaviour), and a motivational general function (MG; physiological arousal and affect). Following this 2!2 classification, Salmon, Hall, and Haslam (1994) designed a questionnaire to assess soccer players’ use of the different imagery functions, and found that soccer players employed all four in training and game play, but reported using imagery more for motivational rather than cognitive purposes. Hall, Mack, Paivio, and Hausenblas (1998) sought to replicate these findings across different types of sports and competitive levels though the development of the Sport Imagery Questionnaire (SIQ), and found that the motivational general function of imagery could be further subdivided into motivational general-arousal imagery (MG-A) and motivational general-mastery imagery (MG-M). The MG-A function of imagery entails using imagery to regulate arousal levels whereas the MG-M function involves using imagery to stay focused, confident and mentally tough. Although Paivio focused primarily on imagery, Munroe and her colleagues (Munroe, Hall, & Weinberg, 2004) have since pointed out that this framework can be applied to other types of mental skills. To date, researchers have successfully applied the framework to self talk (Gammage, Hardy, & Hall, 2001; Hardy, Gammage, & Hall, 2001) and goal setting (Munroe et al., 2004). Perhaps even more similar to imagery, however, is observational learning (OL). In fact, early definitions of mental practice tended to group both skills together. For example, Marteniuk (1976) defined mental practice as ‘improvement in performance that results from an individual either thinking about a skill or watching someone else perform it’. Over the years, numerous other writers have suggested that OL and imagery might be similar, yet distinct cognitive processes, with the characteristic distinguishing between the two being the presence or absence of an external stimulus for the individual (e.g Bandura, 1986; Feltz & Landers, 1983; McCullagh & Weiss, 2001; McCullagh, Weiss, & Ross, 1989; Ryan & Simons, 1983). In a modeling situation, the observer watches a live or videotaped model of either someone else or themselves (self-modeling) executing a behavior. Comparatively, there is typically no external stimulus used in imagery, rather, the observer must instead create the image of the desired behavior based on memory and past behavior. Despite this difference, it is likely that Paivio’s general analytical framework can also be applied to OL. Indeed, the idea that OL can serve both a cognitive and motivational function is not new and has

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

519

been alluded to in the modeling literature (e.g. Feltz & Landers, 1983; McCullagh, Weiss, & Ross, 1989). For instance, it is already well recognized that demonstrations are an effective tool for providing information about how to perform a skill and thus offer invaluable information to the performer. Specifically, information concerning both the end goal of the movement to be achieved and the movement pattern needed to attain this outcome is conveyed through observing demonstrations. In other words, athletes can acquire information on both the process and outcome of the skill to be performed. Martin, Mortiz, and Hall (1999) provided evidence to support this notion by finding that control group participants (i.e. did not observe a demonstration) were not able to perform the skill with the desired form as compared to participants who had observed a demonstration. Others have also found that the observation of a model can lead to improvements in form (e.g. Sidaway & Hand, 1993; Whiting, Bijlard, & den Brinker, 1987), as well as other important aspects of performance such as movement pattern recall and error recognition (McCullagh, Burch, & Siegel, 1990), the symbolic coding of physical activities into memorable words and images (Caroll & Bandura, 1982, 1985, 1987, 1990), and the timing of movement sequences (Adams, 1986; McCullagh & Caird, 1990). According to Bandura’s (1997) social cognitive approach for explaining how behaviour is acquired, watching demonstrations will benefit learning and performance by operating through a four-stage process that involves attention, retention, production, and motivation (for a review, see McCullagh & Weiss, 2001). First, the individual must pay careful attention to the model being observed. The extent to which this is achieved will not only depend on specific characteristics of the model, but also on the observer’s cognitive ability, arousal level, and expectations. Second, the individual must commit the observed act to memory. Different methods of retaining observed information include mental practice techniques (e.g. imagery), using analogies (e.g. telling a figure skater to lift upwards into a jump like a pole vaulter), and having the individual verbally repeat the main points aloud (Weinberg & Gould, 2003). Third, the individual must learn how to produce the movement by coordinating their muscle actions with their thoughts. Physical educators and coaches enable this process by providing their athletes with ample practice time, and appropriate lead up skills and progressions (Weinberg & Gould, 2003). Finally, and most importantly, the individual must be motivated to attend to, remember, and practice the observed behaviour in order to perform the skill accurately. Based on the above theoretical and empirical evidence, therefore, it is possible that athletes use cognitive types of OL to gain information about the acquisition and performance of motor skills and to gain information about the acquisition and performance of strategies, game plans, and routines. For example, two separate studies indicated that OL can be used to enhance the recall and recognition of dance sequences (Downey, Neil, & Rapagna, 1996; Laugier & Cadopi, 1996). Furthermore, Weeks (1992) has suggested that OL can enhance an athlete’s ability to respond to constantly changing environments. Evidence in support of Weeks’ statement reveals that observing slides, films or videos of game plays can lead to improvements on perceptual and performance tasks (for a review, see Williams & Grant, 1999). In addition, Christina, Barresi, and Shaffner (1990) reported a case study of a football linebacker, which suggested that observation of a video can lead to improvements in response selection without sacrificing speed of responses. While research and practice has typically emphasized the use of OL to modify the performance of skills and strategies, it is also recognized that modeling can have an effect on psychological responses such as the motivation to change or perform a behavior, coping with fear and anxiety, and cognitions such as self-confidence and self-efficacy (Schunk, 1987; Starek & McCullagh, 1999; Weiss, Ebbeck, & Wiese-Bjornstal, 1993). Most of this research originates from Bandura’s (1986) contention that OL is

520

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

a major source of self-efficacy, either through mastery experiences (i.e. seeing yourself perform the desired skill) or vicarious experiences (i.e. seeing others perform the desired skill). For example, Weiss, McCullagh, Smith, and Berlant (1998) found that OL was an effective technique for improving swimming skills, increasing self-efficacy, and regulating anxiety in children fearful of water. Furthermore, Starek and McCullagh (1999) found that adult beginner swimmers reported increased self-efficacy beliefs when they viewed a model. These findings encourage us to pursue the motivational functions of OL. Despite the wide use of OL in practical situations, we have little information as to why athletes might use this skill as a behavior change technique. Instead, the vast majority of our knowledge of OL has come from experimental lab-based research. But, as pointed out by Weinberg, Butt, and Knight (2001), athletes can be a rich source of data for both researchers and practitioners. Moreover, it is commonly accepted that obtaining descriptive information is the first stage of any program of study, and will eventually form the basis for future explanation, prediction, and control (Weinberg & Gould, 2003). By asking athletes how they make use of demonstrations in training and competition, therefore, we will benefit by having a better understanding of the functions that OL can serve in sport. In turn, a better understanding of OL should lead to more effective intervention techniques. The available empirical and theoretical evidence indicates that athletes use OL for both cognitive and motivational functions. Presently there is no instrument that assesses these two main functions of OL. Therefore, the main aim of the present investigation was to develop a questionnaire that measures athletes’ cognitive and motivational use of OL. Paivio’s (1985) general analytical framework was employed as the theoretical foundation for the questionnaire development reported in the following three studies. In addition, given the similarities between OL and imagery use, direct comparisons were made between the two forms of mental practice.

Study 1 Various researchers (e.g. Feltz & Landers, 1983; McCullagh et al., 1989) have proposed that OL can serve both a cognitive and motivational function. The purpose of Study 1 was to development of the Functions of OL Questionnaire (FOLQ) as a measure of both the cognitive and motivational functions of OL. Participants A sample of 400 Canadian athletes (197 male and 203 female) representing a broad range of sports and competitive levels were recruited to participate in Study 1. More specifically, the athletes participated in one of 28 sports at the recreational (nZ123), club (nZ109), provincial (nZ33), varsity (nZ111), and elite (i.e. national or international, nZ24) levels. Approximately 60% of the athletes competed in interactive sports (i.e. basketball, football), and the remaining 40% of the sample competed in independent sports (i.e. athletics, wrestling). Ages ranged from 18 to 44 years, with a mean age of 21.26 (SDZ2.88). As a result, recommendations were fulfilled for research conducting an item analysis of a questionnaire to have a large, heterogeneous sample (Nunnally, 1978; Nunnally & Bernstein, 1994).

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

521

Instrument development Using the SIQ (Hall et al., 1998) as a starting point, a pool of items was developed to form the basis of the FOLQ and represent each of the five proposed functions of OL. It is important to note that although the items from the SIQ were used for the initial step in the questionnaire development, each item was deliberated upon to ensure that it was within the context of the theoretical and empirical information available from the OL literature. To this end, the 30 items from the SIQ (Hall et al., 1998) were modified to reflect an athletes’ use of OL, rather than imagery. Similar to procedures used for the development of the SIQ, four research experts and 10 athletes then assessed the content validity of this initial pool of items by examining the content, format, wording of the items, and possible usage within the athletic population. Based on their recommendations, changes were made to the content and wording of several items, with all 30 items retained in the FOLQ that was completed by the sample participants. The distribution of items across each of the five functions was as follows: seven cognitive specific items; six cognitive general items; five motivational specific items; six motivational general-arousal; and six motivational general-mastery. The introduction to the FOLQ provided all participants with a definition of OL. The exact wording of this definition was as follows: Demonstration, either by having a person watch another team mate execute a skill, by watching a videotape of a skill, or even by watching yourself on videotape is a common means of communicating information about how to perform a skill or game play The participants were then instructed to rate on a 7-point scale (1Zrarely and 7Zoften) how often they utilize OL for the function described in the given statement. In addition to responding to the items on the FOLQ, the participants were also asked to provide relevant demographic information concerning age, gender, type of sport, and competitive level. Procedure The participants were contacted directly by a trained research assistant immediately following a practice or in different locations, such as at work or school. All participants were informed of the nature of the study, and those who agreed to participate were given a letter of information, a consent form and the questionnaire. Participants were asked to complete each item of the questionnaire as honestly as possible. Completed questionnaires were then returned directly to the investigators. The questionnaire required approximately 15 min to complete. Data analysis Following data screening procedures for missing cases and outliers, the psychometric properties for the initial pool of items developed for the questionnaire were examined using the following statistical techniques. First, individual item characteristics and violations to the assumption of normality were evaluated through calculation of the mean, standard deviation, skewness, and kurtosis for each item. Next, a principal component analysis with a varimax rotation method was conducted to reduce items to several meaningful factors (Hair, Anderson, Tatham, & Black, 1998; Tabachnick & Fidell, 2001). Finally, the internal consistency of each factor identified in the principal component analysis was examined through calculation of Cronbach alphas.

522

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Results and discussion Data screening Possible outliers were detected by calculating Mahalanobis distance values using a criterion of p!0.001 (Tabachnick & Fidell, 2001). The Mahalanobis distance values were evaluated as a c2 with degrees of freedom equal to the number of variables (i.e. 30). Accordingly, a case was determined to be an outlier if the Mahalanobis distance value was greater than 59.7. As a result, 21 cases were deleted from the original sample. In addition, two cases were deleted due to missing data resulting in a final sample size of 377. Individual item characteristics Means for the individual items ranged from 2.68 to 5.68. Similar to the initial stages in the development of the SIQ (Hall et al., 1998), examination of the standard deviations was used to examine response variability. Because the standard deviations for each item were greater than G1.00, response variability was deemed to be satisfactory. With respect to the assumptions of normality, the majority of items were distributed within the tolerance levels except for item 20, ‘I use OL to make corrections to physical skill’, which had a skewness of 3.18 and a kurtosis of 42.89. To improve normality of this item, it was transformed using a log to base 10 procedure. Skewness and kurtosis was then reassessed, resulting in values of K2.12 and 8.69, respectively. Given that problems with normality were still evident with this item, it was then eliminated from further analysis. The means, standard deviations, skewness, and kurtosis for each item are presented in Table 1. Principal component analysis A principal component analysis with a varimax rotation method was conducted to separate items into their respective scales (Hair, Anderson, Tatham, & Black, 1998; Tabachnick & Fidell, 2001). To this end, 29 items were entered and we specified that factors with an eigenvalue over 1.0 were to be extracted. A criterion level of 0.35 was set for items loading on a factor (Tabachnick & Fidell, 2001). Five factors emerged with eigenvalues ranging from 1.01 to 11.07, accounting for 65.23% of the variance. The results indicated that items reflecting both cognitive and motiavational functions of OL separated cleanly onto different factors. However, several of the motivational OL items (nZ8) loaded on more than one factor, and were therefore dropped from subsequent runs of the principal component analysis (item 3Z‘I use OL to understand what it is like to give 100% effort’, item 5Z‘I use OL to help me control my emotions’, item 9Z‘I use OL to understand how to appear self-confident in front of my opponents’, item 10Z‘I use OL to assist me in behaving like a champion (e.g. congratulating my opponents)’, item 12Z‘I use OL to learn how to respond to difficult situations’, item 16Z‘OL assists me in knowing what it is like to win a medal’, item 17Z‘I use OL to know how to handle the stress and anxiety associated with my sport’, and item 18Z‘I use OL to learn how to stick to my game/event plan)’. Furthermore, one item had a communality of less than 0.5, and was also dropped (item 26Z‘OL assists me in setting winning goals’). On the next run, 20 items were entered, and factors with an eigenvalue over 1.0 were again extracted. Three factors emerged with eigenvalues ranging from 1.71 to 7.20, and accounted for 62.93% of

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

523

Table 1 Item characteristics for Study 1 Item

M

SD

Skewness

Kurtosis

I use OL to understand what it is like to win a championship I use OL to make up new plans/strategies in my head. I use OL to understand what it is like to give 100% effort I use OL to help me properly perform a physical skill. I use OL to help me control my emotions I use OL to improve my skills. I use OL to form alternative plans/strategies. I use OL to assist in handling the arousal and excitement associated with my sport I use OL to understand how to appear self-confident in front of my opponents I use OL to assist me in behaving like a champion (e.g. congratulating my opponents) I use OL to develop game plans and routines. I use OL to learn how to respond to difficult situations I use OL to change how I perform a skill OL assists me in knowing how to win I use OL to understand how to perfectly perform a skill. OL assists me in knowing what it is like to win a medal I use OL to know how to handle the stress and anxiety associated with my sport I use OL to learn how to stick to my game/event plan I use OL to understand how to get psyched up. I use OL to make corrections to physical skills I use OL to determine how a strategy will work in an event/game. I use OL to help me learn new skills. I use OL to understand what it takes to be mentally tough. I use OL to learn how to cope with anxiety. I use OL to know how to respond to the excitement associated with performing. OL assists me in setting winning goals I use OL to learn how to be focused during a challenging situation. I use OL to help me fine tune my skills. I use OL to help me improve my game/event strategies. I use OL to assist me in staying positive in tough situations.

3.65 4.85 4.15 5.57 2.68 5.56 4.65 3.13

1.80 1.42 1.80 1.34 1.52 1.28 1.46 1.63

0.06 K0.63 K0.22 K1.06 0.58 K1.07 K0.63 0.48

K1.12 0.05 K1.01 0.82 K0.56 1.13 K0.05 K0.58

3.65

1.73

0.06

K0.95

3.69

1.83

0.10

K1.12

4.83 4.03 5.41 3.69 5.32 3.08 3.12

1.53 1.63 1.32 1.72 1.48 1.77 1.54

K0.60 K0.21 K0.90 K0.01 K0.73 0.41 0.33

K0.26 K0.89 0.65 K1.01 K0.25 K0.99 K0.60

3.75 3.53 5.50 4.64 5.68 3.35 2.75 3.09

1.57 1.77 1.62 1.55 1.27 1.60 1.48 1.53

0.01 0.16 3.18 K0.62 K1.05 0.24 0.65 0.33

K0.71 K1.02 42.89 K0.12 0.88 K0.88 K0.31 K0.70

3.87 3.81 5.49 4.77 3.65

1.72 1.60 1.36 1.52 1.61

0.04 K0.03 K1.03 K0.64 0.06

K0.89 K0.73 0.85 K0.05 K0.82

Note. The participants responded on a Likert scale from 1 to 7. Boldface identifies the items that were retained in the 17-item FOLQ.

the variance. Upon examination of the communalities, three items did not reach 0.5, and were therefore dropped from the analysis (item 1Z‘I use OL to understand what it is like to win a championship, item 8Z‘I use OL to assist in handling the arousal and excitement associated with my sport ‘, and item 14Z‘OL assists me in knowing how to win’). The remaining items loaded cleanly upon three factors representing a motivational function (performance state) and two cognitive functions (skill and strategy). On the final run, 17 items were entered with the same specifications that factors with an eigenvalue over 1.0 were to be extracted. The same three factors emerged with eigenvalues ranging from 1.70 to 6.95,

524

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Table 2 Summary of items and factor loadings for varimax orthogonal three-factor solution Item

Factor 1: performance function of OL 1. I use OL to understand how to get psyched up 2. I use OL to understand what it takes to be mentally tough 3. I use OL to learn how to cope with anxiety 4. I use OL to know how to respond to the excitement associated with performing 5. I use OL to learn how to be focused during a challenging situation 6. I use OL to assist me in staying positive in tough situations (e.g. a player short, sore ankle, etc.) Factor 2: Skill function of OL 1. I use OL to help me properly perform a physical skill 2. I use OL to improve my skills 3. I use OL to change how I perform a skill 4. I use OL to understand how to perfectly perform a skill 5. I use OL to help me learn new skills 6. I use OL to help me fine tune my skills Factor 3: Strategy function of OL 1. I use OL to make up new plans/strategies in my head 2. I use OL to form alternative plans or strategies 3. I use OL to develop game plans and routines 4. I use OL to determine how a strategy will work in an event/game 5. I use OL to help me improve my game/event strategies

Factor loading

Communality

1 (Performance)

2 (Skill)

3 (Strategy)

0.77 0.86

0.006 0.004

0.12 0.01

0.6 0.75

0.85 0.84

0.01 0.01

0.01 0.13

0.73 0.72

0.74

0.18

0.19

0.65

0.72

0.11

0.22

0.58

0.01

0.83

0.15

0.71

0.01 0.003 0.15

0.7 0.8 0.81

0.29 0.17 0.01

0.58 0.67 0.68

0.19 0.01

0.74 0.8

0.18 0.19

0.61 0.68

0.11

0.18

0.72

0.56

0.12 0.14 0.24

0.12 0.21 0.21

0.82 0.76 0.67

0.7 0.64 0.54

0.23

0.25

0.75

0.67

Boldface indicates the highest factor loadings.

and accounted for 66.02% of the variance. All items achieved a minimal communality of 0.5, loaded cleanly onto the factors that represented their constructs and all factor loadings were above the criterion level of 0.35. The items, factor loadings, and communalities for this final solution are reported in Table 2. Internal consistency and bivariate correlations Internal consistency was then calculated for the 17 items that were retained, and represented each of the three subscales, using a Cronbach alpha (Cronbach, 1951). The criterion level for the definition of a scale was set at an alpha coefficient of 0.70 (Nunnally, 1978; Nunnally & Bernstein, 1994). All scales had acceptable internal reliability: performance stateZ0.90, skillZ0.89, and strategyZ0.84.

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

525

In addition, bivariate correlations revealed significant and moderate relationships between the three subscales ranging from 0.25 to 0.47 (p!0.001). Similar to imagery use, the results from this first study indicated that athletes were using OL for both cognitive and motivational functions. The motivational function of OL, however, does not appear to operate at both a specific and a general level. While Hall et al. (1998) found that athletes reported three distinct motivational functions (motivational specific, motivational general-arousal and motivational general-mastery); all six motivational specific items were eliminated through the principal component analysis in this study, suggesting that athletes do not use OL to understand how to achieve outcome goals such as winning a medal. In addition, the performance function of OL does not appear to sub-divide into both an arousal and mastery function. Rather, three motivational general-mastery and three motivational general-arousal items were eliminated in Study 1, with the remaining six motivational general-mastery and arousal items loading on only one factor. This factor consists of items that involve using OL to regulate arousal levels (e.g. I use OL to understand how to get psyched up) and to maintain a certain mental state (e.g. I use OL to learn how to be focused during a challenging situation). Therefore, it seems that athletes use OL to improve motor skill acquisition and performance, execute and develop strategies, and to optimize performance through the regulation of arousal levels and mental states. Given that these findings do not follow the 2!2 classification outlined by Paivio (1985), the following three categories will instead be used to classify the three functions of OL: (1) motor skill acquisition and performance (skill function); (2) strategy development and execution (strategy function); and (3) optimal arousal levels and mental states for performance (performance function).

Study 2 Based on the results of Study 1, the FOLQ measures three functions of OL: skill, strategy, and performance. The purpose of Study 2 was to determine the validity of the three-factor structure of the FOLQ by using multiple samples confirmatory factor analyses. Participants Participants in Study 2 were 953 Canadian athletes (462 male, 483 female, eight unreported) who were again recruited from a large range of sports and competitive levels. More specifically, 26% of the athletes competed in independent sports and the remaining 74% of the athletes competed in interactive sports. Athletes also represented a wide range of competitive levels including recreational level (nZ338), club level (nZ161), provincial level (nZ65), varsity level (nZ302) or elite level (nZ70). The participants ranged in age from 14 to 58 years, with a mean age of 22.37 (SDZ5.15). Instrument refinement The FOLQ was refined based on a preliminary principal component analysis conducted for Study 1. An item was retained for use in the second Study if it had a factor loading above the criterion value of 0.35. No new items were added for Study 2. As a result, the revised version of the FOLQ contained 17 items that assessed the three subscales that emerged from the PCA analysis in Study 1: skill, strategy, and performance.

526

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Procedure The procedures were identical to those used in Study 1. Data analysis The data was first screened for outliers, missing data, and departures from univariate normality. Next, two subsamples were created by randomly dividing the data. Preliminary confirmatory factor analyses were then conducted to determine multivariate normality and interfactor correlations for both samples. In addition, the internal consistency of each factor was examined through calculation of Cronbach alphas. Upon establishment of multivariate normality, the existence of a relationship between the factors, and internal consistency, confirmatory factor analyses were then performed using AMOS 4.01 (Arbuckle, 1999) with Maximum Likelihood (ML) estimation procedures. Results and discussion Data screening Mahalanobis distance values were again calculated to identify multivariate outliers using a criterion of p!0.001 (Tabachnick & Fidell, 2001). For this sample, a case was determined to be an outlier if a Mahalanobis distance value of 40.79 or greater was obtained. As a result, 42 cases were identified as outliers and were deleted from the sample. In addition, 21 cases were deleted due to missing data, resulting in a final sample size of 890. Univariate and multivariate normality Given that confirmatory factor analysis techniques are sensitive to distributional characteristics of the data (Hair et al., 1998), possible departures from both univariate and multivariate normality were examined. With respect to univariate normality, examination of skewness and kurtosis values revealed that responses to the majority of the items on the FOLQ were distributed within the tolerance levels of assumptions of normality. One item (CS item 5ZI use OL to help me learn new skills), however, was identified as being slightly skewed (K1.11), and was transformed using a log to base 10 procedure. Reexamination of the skewness values revealed that the transformation procedure improved normality of this item (i.e. skewness !j1.00j). These transformed scores, therefore, were used in further analyses. The means, standard deviations, skewness, and kurtosis for each of the individual items are presented in Table 3. Upon establishment of univariate normality, two new samples were created by randomly dividing the data set. An initial run of the confirmatory factor analysis for both samples revealed significant multivariate kurtosis coefficients (sample 1Z48.10, p!0.001; sample 2Z48.29, p!0.001), suggesting that problems with multivariate normality existed. We resolved this issue by applying a ‘bootstrapping’ approach to further analyses such that final estimates and confidence estimates were derived from multiple model estimations (i.e. 1000). As a result, a bootstrapping approach does not rely on assumptions about the statistical distribution of parameters, and is not influenced by deviations in multivariate kurtosis (Hair et al., 1998).

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

527

Table 3 Item characteristics for Study 2 (nZ890) Item Skill 1 2 3 4 5 6 Strategy 1 2 3 4 5 Performance 1 2 3 4 5 6

Mean

Standard deviation

Skewness

Kurtosis

5.42 5.52 5.24 5.31 5.59 5.33

1.43 1.34 1.40 1.50 1.40 1.41

K0.99 K0.95 K0.83 K0.90 K1.11 K0.92

0.53 0.60 0.36 0.32 0.88 0.39

4.63 4.53 4.54 4.53 4.78

1.59 1.60 1.63 1.66 1.58

K0.51 K0.42 K0.50 K0.47 K0.62

K0.44 K0.58 K0.60 K0.61 K0.24

3.23 3.29 2.75 3.22 3.59 3.41

1.73 1.71 1.56 1.73 1.71 1.75

0.45 0.35 0.62 0.34 0.16 0.34

K0.76 K0.84 K0.51 K0.94 K0.98 K0.89

Internal consistency and interfactor correlations Internal consistency was calculated for the items representing each of FOLQ subscales using a Cronbach alpha (Cronbach, 1951). The criterion level for the definition of a scale was again set at an alpha coefficient of 0.70 (Nunnally, 1978; Nunnally & Berstein, 1994). All subscales showed an acceptable internal reliability with alpha coefficients ranging from 0.85 to 0.92 for both samples. Interfactor correlations from the confirmatory factor analyses revealed that significant and moderate relationships existed between the three subscales, ranging from 0.30 to 0.61 in magnitude. Confirmatory factor analyses The results of the confirmatory factor analyses for both samples are presented in Table 4. The overall goodness of fit of the model was tested using the chi-square likelihood ratio statistic (c2). The c2 statistic Table 4 Goodness of fit indices

Sample 1 Sample 2 *p!.001.

c2

df

SRMR

RMSEA

CFI

TLI

252.57* 291.37*

116 116

0.03 0.04

0.05 0.06

0.97 0.96

0.96 0.95

528

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

is generally regarded as a measure of badness of fit models, such that a small c2 corresponds to a good fit and a large c2 corresponds to a poor fit (Jo¨reskog & So¨rbom, 1993). Typically a non-significant c2 indicates a good model fit, yet it is rarely obtained in practice (Hayduck, 1996). It was not surprising to find, therefore, that the c2 statistic was significant across both samples (p!0.001). For this reason, supplementary fit indices are reported. To this end, a 2-index presentation strategy was used to assess model fit based on recommendations from Hu and Bentler (1999). Specifically, two types of fit indices are reported; the standardized root mean square residual (SRMR; Bentler, 1995) and a supplemental incremental fit index (e.g. Tucker Lewis Index, Comparative Fit Index, or Root Mean Square Error of Approximation). The SRMR is a measure of absolute fit index, and is used to calculate the average difference between the sample variances and covariances and the estimated population variances and covariance (Tabachnick & Fidell, 2001). A value of close to 0.08 indicates a good fit (Hu & Bentler, 1999). In Study 2, the SRMR values were deemed to be acceptable for both samples (!0.08). The Tucker Lewis Index (TLI; Bollen, 1989), also known as the Bentler-Bonett non-normed fit index (NNFI), evaluates the estimated model by comparing it to an independence model. The Comparative Fit Index (CFI; jjjjBentler, 1990) also tests how much better the model fits compared to the independence model, but uses a different approach from the TLI (Jo¨reskog & So¨rbom, 1993). For both of these fit indices, Hu and Bentler (1999) suggest that a cutoff value of close to 0.95 indicates a relatively good fit. In accordance to these guidelines, both the TLI and CFI values indicated a good model fit for both samples in Study 2 (O0.95). Finally, the Root Mean Square Error of Approximation (RMSEA; Steiger, 1990) assesses how well the model approximates the data by determining the lack of fit of the model to a population covariance matrix, expressed as the discrepancy per degree of freedom (Browne & Cudeck, 1993). A cutoff value of close to 0.06 for RMSEA indicates a good fit (Hu & Bentler, 1999). The RMSEA value in the present study indicated a good fit between the proposed model and the data for both samples (!0.06). Examination of the standardized factor loadings revealed that values ranged from 0.67 to 0.86, indicating that each item had a meaningful contribution to its respective subscale. In addition, no values exceeded or were close to 1.00 indicating that all factor loadings were within acceptable limits and no offending estimates existed in the data (Hair et al., 1998). Furthermore, the error uniqueness values indicated moderate levels of error for each item. The standardized factor loadings and error uniqueness values for each item is reported in Table 5. In this confirmatory study, the same three functions of OL emerged as those in Study 1, therefore, confirming that athletes use OL for skill, strategy, and performance functions. Given that the factor structure of the FOLQ has been confirmed, a third study was deemed necessary in order to assess the concurrent validity and the reliability of the FOLQ, as well as, to examine any group differences that emerge.

Study 3 The purpose of Study 3 was three-fold. Due to the fact that the FOLQ and SIQ were both developed from the same conceptual framework (Paivio, 1985), but are designed to measure distinct constructs, the first purpose of Study 3 was to examine the degree of similarity between the two questionnaires. A second purpose was to examine the test–retest reliability of the FOLQ by administering

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

529

Table 5 Standardized factor loadings and error uniqueness values Item

Skill 1 2 3 4 5 6 Strategy 1 2 3 4 5 Performance 1 2 3 4 5 6

Sample 1 (nZ477)

Sample 2 (nZ421)

Factor loading

Error

Factor loading

Error

0.76 0.83 0.75 0.79 0.74 0.73

0.58 0.69 0.57 0.63 0.54 0.53

0.81 0.86 0.77 0.81 0.81 0.80

0.65 0.75 0.59 0.65 0.66 0.64

0.69 0.74 0.75 0.69 0.82

0.48 0.55 0.57 0.48 0.66

0.67 0.74 0.74 0.70 0.78

0.44 0.55 0.55 0.49 0.61

0.69 0.74 0.71 0.78 0.78 0.78

0.48 0.55 0.51 0.61 0.61 0.62

0.73 0.81 0.80 0.82 0.83 0.81

0.53 0.65 0.64 0.68 0.68 0.65

the questionnaire on two separate occasions. Finally, a third purpose of Study 3 was to examine the influence of variables such as gender, competitive level (e.g. elite versus non-elite performers), and sport type (e.g. independent versus interactive) on the use of OL and imagery use. Participants Participants in Study 3 were 200 Canadian athletes (77 male, 123 female) with a mean age of 19.62 years (SDZ2.17). The participants competed in a broad range of levels that were categorized as being either non-elite (i.e. recreational, club, nZ118) or elite (i.e. provincial, varsity; nZ81). In addition, the participants’ sport type was categorized as being either independent (nZ88) or interactive (nZ110) in nature. Instruments The revised version of the FOLQ that contained the 17 items that emerged from the PCA in Study 1, and was established to be a good model fit in Study 2, was administered to the participants. In addition, the participants completed the SIQ (Hall et al., 1998) and supplied relevant demographic data including age, gender, and competitive level. Sport imagery questionnaire (SIQ). The SIQ (Hall et al., 1998) is a 30-item self-report instrument that asks athletes to rate on a 7-point scale (1Zrarely and 7Zoften) how often they utilize five functions of

530

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

imagery: cognitive specific (e.g. imaging perfectly executed sport skills), cognitive general (e.g. imaging plans/strategies), motivation specific (e.g. imaging specific goals and outcomes), motivation generalarousal (e.g. imaging the excitement of performing), and motivation general-mastery (e.g. imaging staying focused and working through problems). Accordingly, the SIQ consists of five subscales with six items representing each subscale. Procedure Similar to those procedures used in Study 1 and Study 2, athletes were recruited to participate in Study 3 by a trained research assistant. All participants were informed of the nature of the study, and those who agreed to participate were given a letter of information and a consent form. The data collection then followed two phases. In phase one, participants were asked to complete each item of the FOLQ and SIQ as honestly as possible. Completed questionnaires were then returned directly to the investigators. The questionnaire required approximately 30 min to complete. Approximately a week following the first phase of data collection, a random sample of 23% of the athletes (nZ46) was asked to complete the FOLQ for a second time. Data analysis The internal consistency of each factor was examined through calculation of Cronbach alphas (Cronbach, 1951). Descriptive statistics were calculated for each of the subscale of the FOLQ, and a repeated measures ANOVA was then used to examine whether significant differences existed in the athlete’s use of the different functions of OL. Concurrent validity was then established through calculation of bivariate correlations and hierarchical regression procedures. Test–retest reliability was assessed by intraclass correlation coefficients. Finally, separate multivariate analyses of variances (MANOVAs) were used to examine whether any group differences existed in OL and imagery use according to gender, competition level, or sport type.

Results and discussion Internal consistency Internal consistency was calculated for the items representing each of FOLQ and SIQ subscales using a Cronbach alpha (Cronbach, 1951). The criterion level for the definition of a scale was again set at an alpha coefficient of 0.70 (Nunnally, 1978; Nunnally & Bernstein, 1994). All subscales showed an acceptable internal reliability with alpha coefficients ranging from 0.84 to 0.88 for the FOLQ subscales and from 0.77 to 0.90 for the SIQ subscales. Descriptive statistics Means and standard deviations were calculated for each subscale of the FOLQ and SIQ, and are presented in Table 6 by gender, competitive level, and sport type. A repeated measures ANOVA revealed significant differences in the athletes’ use of OL (F(2,378)Z2.14.55, p!0.001, h2Z0.53).

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

531

Table 6 Means and standard deviations for FOLQ Subscales by gender and competitive level for study 3 Gender

Skill Strategy Performance

Competitive level

Sport type

Male M (SD)

Female M (SD)

Nonelite M (SD)

Elite M (SD)

Coactive M (SD)

Interactive M (SD)

5.01 (1.12) 4.10 (1.42) 3.00 (1.29)

5.31 (1.76) 4.01 (1.29) 2.96 (1.27)

5.32 (1.66) 4.18 (1.37) 2.97 (1.21)

5.08 (1.35) 4.00 (1.28) 3.01 (1.36)

5.51 (1.86) 3.96 (1.39) 3.24 (1.36)

4.96* (1.20) 4.13 (1.26) 2.78* (1.16)

*Significant difference (p!0.05).

A Tukey post hoc test indicated that athletes used the skill function of OL (MZ5.22, SDZ1.55) significantly more than the strategy function of OL (MZ4.08, SDZ1.34), which in turn, was used significantly more than the performance function of OL (MZ2.98, SDZ1.27). A second repeated measures ANOVA also revealed significant differences in the athletes’ use of imagery (F(4,732)Z 22.12, p!0.001, h2Z0.11). A Tukey post hoc test revealed that athletes used both MG-M imagery (MZ 4.72, SDZ1.37) and CS imagery (MZ4.69, SDZ1.89) significantly more than CG imagery (MZ4.47, SDZ1.20), which in turn was used significantly more than both MS imagery (MZ4.19, SDZ1.50) and MG-A imagery (MZ4.18, SDZ1.39). These findings are consistent with other studies using the SIQ (e.g. Cumming and Hall, 2002a; Hall et al., 1998; Vadocz, Hall, and Moritz, 1997). Concurrent validity Concurrent validity was assessed via two different statistical techniques. First, bivariate correlations indicated that significant and moderate relationships existed between the subscales of the FOLQ and SIQ. Specifically, r values ranged from 0.29 (p!0.001) to 0.54 (p!0.001) signifying that the subscales of the FOLQ and SIQ are related but represent separate and distinct constructs. Second, a series of hierarchical regression procedures were used to examine how accurately the five imagery subscales predicted their FOLQ counterparts (Bryant, 2000). For each regression, the SIQ subscale(s) assessing the function of imagery most closely resembling the OL construct was entered on the first step, and the remaining SIQ subscales were entered as a block on the second step. The rationale behind using this order was that the SIQ subscale(s) entered on the first step (i.e. cognitive specific imagery) would be the best predictor of it’s associated FOLQ subscale (i.e. skill function of OL). The skill function of OL Cognitive specific imagery was entered on the first step, and cognitive general imagery, motivational specific imagery, motivational general-arousal imagery, and motivational generalmastery imagery were all entered as a block on the second step. A significant model was found (F(1,182)Z36.04, p!0.001), with cognitive specific imagery (R2chaZ0.17) accounting for a significant proportion of the variance. Inspection of the beta weight for this variable indicated that the use of cognitive specific imagery (bZ0.41, p!0.001) was positively related to the use of the skill function of OL (Table 7).

532

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Table 7 Bivariate correlations between the subscales of the FOLQ and the SIQ

CS imagery CG imagery MS imagery MG-A imagery MG-M imagery

Skill

Strategy

Performance

0.40* 0.34* 0.23* 0.29* 0.33*

0.36* 0.51* 0.36* 0.36* 0.40*

0.38* 0.43* 0.38* 0.54* 0.48*

*Indicates p!0.001.

The strategy function of OL Cognitive general imagery was entered on the first step, and cognitive specific imagery, motivational specific imagery, motivational general-arousal imagery, and motivational general-mastery imagery were all entered as a block on the second step. A significant model was found (F(1,178)Z14.35, p!0.001) that included both cognitive general imagery (R2chaZ0.27) and cognitive specific imagery (R2chaZ 0.02). Inspection of the beta weight for these variables indicated that the use of cognitive general imagery (bZ0.59, p!0.001) was positively related to the use of the strategy function of OL, but cognitive specific imagery (bZK0.25, p!0.001) was negatively related. It must be noted, however, that cognitive general imagery accounted for most of the variance (27%) with cognitive specific imagery contributing very little to the model (2%). The performance function of OL Both motivational general-arousal imagery and motivational general-mastery were entered together on the first step, with cognitive specific imagery, cognitive general imagery, and motivational specific imagery were entered as a block on the second step. A significant model emerged (F(2,182)Z37.97, p!0.001), with only motivational general-mastery imagery emerging as a significant predictor (R2chaZ0.30), which again indicated that increased use of this function (bZ0.41, p!0.001) was related to the use of the performance function of OL. In sum, the results from this analysis revealed moderate relationships between the athletes’ reported use of the imagery and OL functions. Given that the FOLQ was developed based on the SIQ (Hall et al., 1998), it was deemed necessary to confirm the concurrent validity of the FOLQ, hereby ensuring that the FOLQ was in fact assessing a similar, but distinct construct. Therefore, it can be concluded that the FOLQ does measure a distinct construct from that of imagery as measured by the SIQ. Test–retest reliability To establish test–retest reliability, intraclass correlation coefficients (ICCs) were calculated using a two-way random effect model (Haggard, 1958; Ntoumanis, 2001). ICC values above 0.70 are considered acceptable for measurements in behavioural sciences (Vincent, 1999). In the present study, moderate values were found between the first and second administration of the FOLQ, with ICCs scores of 0.88 for the skill function of OL, 0.80 for the strategy function of OL, and 0.79 for the performance function of OL. This provides support for the FOLQ’s temporal reliability.

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

533

Group differences Separate multivariate analyses of variances (MANOVAs) were computed in order to determine whether any group differences existed in the athletes’ reported use of the functions of both OL and imagery. For each MANOVA, either gender, competitive level (i.e. nonelite or elite) or type of sport (i.e. independent or interactive) served as the independent variables, and the three subscales of the FOLQ (i.e. CS, CG, and MG) or SIQ (i.e. CS, CG, MS, MG-A, and MG-M) served as the dependent variable. Gender No significant multivariate effect was found for gender for both OL (Pillai’s TraceZ0.01, F(3, 186)Z 0.70, pZns) and imagery use (Pillai’s TraceZ0.03, F(5, 178)Z0.96, pZns). That is, gender did not influence the athletes’ reported use of either OL or imagery. These results correspond with previous research using the SIQ (Hall et al., 1998), in that no differences were found between female and male athletes’ use of imagery. Competitive level No significant multivariate effect was found for competitive level for both OL (Pillai’s TraceZ0.01 F(3, 185)Z0.86, pZns) or imagery use (Pillai’s TraceZ0.03, F(5, 177)Z0.90, pZns). Although these findings indicate that, regardless of the athletes’ competitive level, they use OL and imagery with the same frequency, it does not exclude the possibility that athletes may be using the same function for different outcomes. For example, a novice athlete may use the strategy function to learn how to execute a particular game strategy, however, an expert athlete may use the strategy function to further develop or enhance their use of a familiar game strategy. Investigation of the use of the various OL functions to achieve different outcomes would be an interesting area for future research. Type of sport No significant multivariate effect was found for sport type for imagery use (Pillai’s TraceZ0.01, F(5, 176)Z0.50, pZns). In contrast, significant effects were found for OL (Pillai’s TraceZ0.10, F(3, 184)Z 6.46, p!0.01, h2Z0.10), with an observed power of 96.8%. Univariate analyses indicated that a significant effect was found for the skill function of OL (F(1,186)Z6.10, pZ0.01, h2Z0.03), which indicated that the skill function of OL is used more by athletes who compete in independent sports (MZ5.51, SDZ1.86), when compared to athletes who compete in interactive sports (MZ4.96, SDZ1.20). Furthermore, a significant effect was also found for the performance function of OL (F(1,186)Z9.59, pZ0.01, h2Z0.03), which again indicated that the performance function of OL is used more by athletes who compete in independent sports (MZ3.24, SDZ1.36), when compared to athletes who compete in interactive sports (MZ2.78, SDZ1.16). It is important to note that the effect sizes resulting from these analyses are quite low and, therefore, it is not possible to draw any concrete conclusions from these results. The results do suggest, however, that investigating the impact of sport type on athletes’ OL usage is a fruitful area for future research. Study 3 further replicated the findings of Studies 1 and 2 showing again that there exist three functions of OL. In addition, it highlighted an interesting difference between OL and imagery use among athletes. One of the most consistent findings in the imagery literature has been that athletes use imagery more for motivational rather than cognitive purposes (e.g. Cumming & Hall, 2002a,b; Hall et al., 1998; Vadocz et al., 1997), and our Study 3 SIQ results also support this finding. In comparison, the results of

534

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

the present study revealed that athletes use the performance function of OL the least. These results are not really surprising, however, when one considers how OL has traditionally been utilised. Coaches and athletes tend to use OL techniques mainly for skill acquisition. For example, when athletes view a teammate, their coach likely directs them to pay attention to the relevant information provided by the model in terms of skill execution and/or team strategies, rather than using it for purposes of motivation. General discussion The purpose of the present three studies was to examine the possible functions of OL through the development of a valid and reliable questionnaire. Due to the perceived similarities between imagery and OL, five potential functions of OL were identified based on the analytical framework of imagery use by athletes first proposed by Paivio (1985) and later modified by Hall and colleagues (1998). The items from the SIQ were modified to reflect OL, and formed the basis of the FOLQ. The results of a principal component analysis in Study 1 identified three functions of OL. The confirmatory factor analysis conducted in Study 2 again indicated that athletes were using OL for three separate and distinct functions: (1) skill; (2) strategy; and (3) performance. The third study further confirmed the validity of the FOLQ by determining the distinct nature of the FOLQ from the SIQ. In addition, temporal stability for the FOLQ was evident, indicating that it is a reliable measure. A noted finding of all three studies is that athletes use OL for cognitive functions, with limited use of motivational functions. This trend of focusing on skill and strategy functions is also clearly evidenced in the modeling literature. The emphasis of modeling research, for example, has been on determining the best type of model to enhance skill acquisition (Adams, 1986; Landers & Landers, 1973; McCullagh & Meyer, 1997; McCullagh, 1986, 1987), and on how often and in what sequence the demonstrations should be given (Richardson & Lee, 1999; Sidaway & Hand, 1993). These types of studies have only assessed the skill improvement of the participants, thus neglecting the various motivational components that may be at play. Recent literature, however, is showing that OL can be an effective means for modifying psychological responses to skill acquisition. For example, several studies with adults and children have shown that OL increases a learner’s feeling of self-efficacy towards performing a particular action or behavior (Gould & Weiss, 1981; Starek & McCullagh, 1999). No studies, however, have been conducted with athletes. Thus, even though the majority of the athletes in the present studies report using modeling for motivational functions significantly less than for cognitive functions, it was still apparent that it is being used to some extent by the athletes. This finding, in combination with the recent modeling research, suggests that modeling should not only be promoted as a skill acquisition technique, but also as a technique for improving psychological responses and performance enhancement in competition. As such, investigation into how modeling can be employed to modify motivational aspects of performance would be a rewarding area for future research. In particular, future qualitative research may be required in order to gain a better understanding of how athletes use OL for motivational functions. The FOLQ was developed by following the recommended statistical procedures for the development of a valid and reliable questionnaire (Tabachnick & Fidell, 2001). For example, the construct validity and concurrent validity of the scale were assessed. As well, a multi-sample CFA was computed in order to more stringently test the strength and generalizability of the proposed factor structure. Noteworthy is that one of the more recent trends in questionnaire development is to begin with a qualitative component, usually in the form of interviews, in order to gain sufficient knowledge to develop the questionnaire

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

535

items. We do recognize that not including a qualitative component may be a limitation, however, considering that this study was already operating within Hall et al. (1998) framework, it did not seem necessary. Regardless, the results of these studies indicate that the FOLQ is a valid tool that can be used to examine the various functions of modeling. There are numerous possibilities for using the FOLQ in future studies. For example, it would be worthwhile to examine the use of OL with a more homogeneous sample of athletes in terms of competitive level and type of sport. It is likely that a study of this sort would find that differences exist between novice athletes and more competitive athletes in terms of their use of OL. In addition, it would be beneficial to further examine the differences among OL use in athletes across different sports. In conclusion, the present study indicates that OL is primarily used by athletes for cognitive functions (skill and strategy), but it also provides support for the notion that OL can be used for motivational functions (arousal level and mental state). These three functions of OL are very similar to three of the five functions found for imagery, providing support for the notion that imagery and OL have strong cognitive processing links. Future research should focus on confirming whether these similarities also exist at the outcome level. It may be possible to do this by employing the framework developed by Martin et al. (1999) that consists of three categories of behavior that can be modified by imagery use. Specifically, Martin et al. (1999) found that imagery can be used in the acquisition and performance of skills and strategies, to modify cognitions and emotions, and to regulate arousal and anxiety. It would be interesting to discover how the functions of OL fit into this framework. Finally, it is conceivable that with increased research in this area, OL will soon be considered by the sport psychology discipline as a useful performance enhancing technique, alongside the use of accepted imagery techniques.

Acknowledgements The authors would like to acknowledge the help of Sanna Nordin and Brie Jedlic for their help with data collection and data entry. We would also like to thank Nikos Ntoumanis for his statistical advice.

References Adams, J. (1986). Use of the model’s knowledge of results to increase the observers performance. Journal of Human Movement Studies, 12, 89–98. Arbuckle, J.L., (1999). AMOS version 4.01 user guide. Chicago: Smallwaters Corporation. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood cliffs, NJ: Prentice Hall. Bandura, A. (1997). Self-efficacy: The exercise of control. New York, USA: W.H. Freeman. Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238–246. Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software Inc. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Browne, M. W., & Cudeck, R. (1993). In K. A. Bollen, & J. S. Long (Eds.), Alternative ways of assessing model fit. Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage. Bryant, F. B. (2000). In L. G. Grimm, & P. R. Yarnold’s (Eds.), Assessing the validity of measurement. Reading and understanding more about multivariate statistics. Washington, DC: American Psychological Association. Carroll, W. R., & Bandura, A. (1982). The role of visual monitoring in observational learning of action patterns: Making the unobservable observable. Journal of Motor Behavior, 14, 153–167.

536

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

Carroll, W. R., & Bandura, A. (1985). The role of timing of visual monitoring and motor rehearsal in observational learning of action patterns. Journal of Motor Behavior, 17, 269–281. Carroll, W. R., & Bandura, A. (1987). Translating cognition into action: The role of visual guidance in observational learning. Journal of Motor Behavior, 19, 385–398. Carroll, W. R., & Bandura, A. (1990). Representational guidance of action production in observational learning: A causal analysis. Journal of Motor Behavior, 22, 85–97. Christina, R. W., Barresi, J. V., & Shaffner, P. (1990). The development of response selection accuracy in a football line-backer using video training. The Sport Psychologist, 4, 11–14. Cronbach, L. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297–334. Cumming, J., & Hall, C. (2002a). Athletes’ use of imagery in the off-season. The Sport Psychologist, 16, 160–172. Cumming, J., & Hall, C. (2002b). Deliberate imagery practice: Examining the development of imagery skills in competitive athletes. Journal of Sports Sciences, 20, 137–145. Downey, P. J., Neil, G. I., & Rapagna, S. (1996). Evaluating modeling effects in dance. Impulse, 4, 48–64. Feltz, D. L., & Landers, D. M. (1983). The effects of mental practice on motor skill learning and performance: A meta-analysis. Journal of Sport Psychology, 5, 25–57. Gammage, K., Hardy, J., & Hall, C. (2001). A description of self talk in exercise. Psychology of Sport and Exercise, 2, 233–247. Gould, D., & Weiss, M. (1981). The effects of model similarity and model task on self-efficacy and muscular endurance. Journal of Sport Psychology, 3, 17–29. Haggard, E. A. (1958). Intraclass correlation and the analysis of variance. New York: The Dryden Press. Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis. Upper Saddle River, NJ: Prentice Hall. Hall, C., Mack, D., Paivio, A., & Hausenblas, H. (1998). Imagery use by athletes: Development of the Sport Imagery Questionnaire. International Journal of Sport Psychology, 29, 73–89. Hardy, J., Gammage, K., & Hall, C. (2001). A descriptive study of athlete self-talk. The Sport Psychologist, 15, 306–318. Hayduck, L. A. (1996). LISREL issues, debates and strategies. Baltimore: John Hopkins University. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modelling, 6, 1–55. Jo¨reskog, K. G., & So¨rbom, D. (1993). LISREL 8 user’s reference guide. Chicago: Scientific Software. Landers, D., & Landers, D. (1973). Teacher versus peer models: Effects of model’s presence and performance level on motor behavior. Journal of Motor Behavior, 5, 129–139. Laugier, C., & Cadopi, M. (1996). Representational guidance of performance in adult novices: Effect of concrete vs. discrete movement. International Journal of Sport Psychology, 27, 91–108. Marteniuk, R. G. (1976). Information processing in motor skills. New York: Holt, Rinehart and Winston. Martin, K. A., Moritz, S. E., & Hall, C. R. (1999). Imagery use in sport: A literature review and applied model. The Sport Psychologist, 13, 245–268. McCullagh, P. (1986). Model status as a determinant of attention in observational learning and performance. Journal of Sport Psychology, 8, 319–331. McCullagh, P. (1987). Model similarity effects on motor performance. Journal of Sport Psychology, 9, 249–260. McCullagh, P., Burch, C.D., Siegel, D.I., (1990). Correct and self-modelling and the role of feedback in motor skill acquisition. Paper presented at the annual meeting of the Northern American Society for the Psychology of Sport and Physical Activity, Houston, TX. McCullagh, P., & Caird, J. K. (1990). Correct and learning models and the use of model knowledge of results in the acquisition and retention of a motor skill. Journal of Human Movement Studies, 18, 107–116. McCullagh, P., & Meyer, K. N. (1997). Learning versus correct models: Influence of model type on the learning of a freeweight squat lift. Research Quarterly for Exercise and Sport, 68, 56–61. McCullagh, P., & Weiss, M. R. (2001). In R. N. Singer, H. A. Hausenblas, & C. M. Janelle (Eds.), Modeling: Considerations for motor skill performance and psychological responses. Handbook of research on sport psychology. New York: Wiley. McCullagh, P., Weiss, M. R., & Ross, D. (1989). In K. B. Pandolf (Ed.), Modeling consideration in motor skill acquisition and performance: An integrated approach. Exercise and sport science reviews (pp. 475–513). Baltimore: Williams and Wilkins. Munroe Chandler, K., Hall, C., & Weinberg, R. (2004). An in-depth analysis of the types of goals athletes set. Journal of Sport Behavior, 27, 58–74.

J. Cumming et al. / Psychology of Sport and Exercise 6 (2005) 517–537

537

Nunnally, J. C. (1978). (2nd ed.) Psychometric theory. New York: McGraw Hill. Nunnally, J. C., & Bernstein, I. H. (1994). (3rd ed.) Psychometric theory. New York: McGraw Hill. Ntoumanis, N. (2001). A step-by-step guide to SPSS for sport and exercise studies. London: Routledge. Paivio, A. (1985). Cognitive and motivational functions of imagery in human performance. Canadian Journal of Applied Sports Sciences, 10, 22S–28S. Richardson, J. R., & Lee, T. D. (1999). The effects of proactive and retroactive demonstrations on learning signed letters. Acta Psychologica, 101, 79–90. Ryan, E. D., & Simons, J. (1983). What is learned in mental practice of motor skills: A test of the cognitive-motor hypothesis. Journal of Sport Psychology, 5, 419–426. Salmon, J., Hall, C., & Haslam, I. (1994). The use of imagery by soccer players. Journal of Applied Sport Psychology, 6, 116– 133. Schunk, D. (1987). Peer models and children’s behavioral change. Review of Educational Research, 57, 149–174. Sidaway, B., & Hand, M. J. (1993). Frequency of modeling effects on the acquisition and retention of a motor skill. Research Quarterly for Exercise and Sport, 64, 122–125. Starek, J., & McCullagh, P. (1999). The effect of self-modeling on the performance of beginning swimmers. The Sport Psychologist, 13, 269–287. Steiger, J. H. (1990). Structural model evaluation: An interval estimation approach. Multivariate Behavioral Research, 25, 173– 180. Tabachnick, B. G., & Fidell, L. S. (2001). (4th ed) Using multivariate statistics. New York, NY: Harper Collins. Vadocz, E., Hall, C., & Moritz, S. (1997). The relationship between competitive anxiety and imagery use. Journal of Applied Sport Psychology, 9, 241–253. Vincent, W. J. (1999). Statistics in kinesiology. Champaign, IL: Human Kinetics. Weeks, D. L. (1992). A comparison of modeling modalities in the observational learning of an externally paced task. Research Quarterly for Exercise and Sport, 63, 373–380. Weinberg, R., Butt, J., & Knight, B. (2001). High school coaches’ perceptions of the process of goal setting. The Sport Psychologist, 15, 20–47. Weinberg, R. S., & Gould, D. (2003). Foundations of sport and exercise psychology. Champaign, IL: Human Kinetics. Weiss, M. R., Ebbeck, V., & Wiese-Bjornstal, D. M. (1993). Developmental and psychological factors related to children’s observational learning of physical skills. Pediatric Exercise Science, 5, 301–317. Weiss, M. R., McCullagh, P., Smith, A., & Berlant, A. (1998). Observational learning and the fearful child: Influence of peer models on swimming skill performance and psychological responses. Research Quarterly for Exercise and Sport, 69, 380– 394. Whiting, H. T. A., Bijlard, M. J., & den Brinker, B. P. L. M. (1987). The effect of the availability of a dynamic model on the acquisition of a complex cyclical system. Quarterly Journal of Experimental Psychology, 39A, 43–59. Williams, A. M., & Grant, A. (1999). Training perceptual skill in sport. International Journal of Sport Psychology, 30, 194– 220.