We examined the 25-indicator CAPL using factor analytic techniques to examine the validity of evidence based on factor structure and to determine the relative weighting of each domain of physical literacy. Using confirmatory factor analyses to test the
a priori specified CAPL model and exploratory post-hoc modifications based on theory, group consensus, and statistical criteria, we found support for a revised and conceptually concise 14-indicator version of CAPL that maximized content representation while reducing construct irrelevant variance and participant burden. Aligning with the currently accepted Canadian Consensus Statement of physical literacy [
5,
6], we found that a four-domain-correlated CAPL model of (1) Physical Competence, (2) Daily Behaviour, (3) Motivation and Confidence, And (4) Knowledge and Understanding provided a good fit; however, caution is warranted when interpreting the correlations between domains given the large sample size. Additionally, these four domains of CAPL could be subsumed under an overall physical literacy factor. Finally, based on the results from the higher-order confirmatory factor analysis, we suggest revising the CAPL weighting procedure for each domain such that three domains (Physical Competence, Daily Behaviour, and Motivation and Confidence) are equally weighted (i.e., 30 points each) whereas one domain (Knowledge and Understanding) has a weaker (i.e., 10 points) relative contribution to total physical literacy (see Fig.
1d).
Refining CAPL
Four out of the seven original indicators comprising the domain of Physical Competence were removed because they did not load as strongly as the other indicators. In the first iterations of CAPL, BMI z scores and waist circumference were proposed to serve as indicators of “body composition”, subsumed under a domain labelled “physical fitness” [
15]. As a result of the Delphi process, the CAPL model was revised and the domain of “Physical Fitness” was replaced with “Physical Competence”; yet, the indicators of BMI z score and waist circumference were retained under the Physical Competence domain, with a call for future research to assess their contribution. Our finding that the BMI z score was a weak indicator of Physical Competence corroborates previous findings [
10], whereas the finding that waist circumference was a weak indicator was a novel finding. Conceptually, because BMI z score and waist circumference were originally proposed as indicators of “body composition” subsumed under “physical fitness”, we are not surprised that they represent weak indicators of the broader domain of Physical Competence. As such, we deemed the removal of these two measures appropriate and suggest that they may be better used as a separate outcome. Two additional indicators (i.e., sit and reach and grip strength) were removed given their weak factor loadings on Physical Competence. It could be argued that the CAMSA, PACER, and plank are better indicators of physical competence and, in turn, of physical literacy, because all tests have been shown to mimic realistic movements and durations of children’s experiences during physical activity. For example, the PACER mirrors the sporadic patterns of starting and stopping by children [
23]; the CAMSA enables children to perform simple movements and complex skills in reaction to a changing environment [
14]; and the plank assesses torso endurance, which is vital for effective and efficient use of the upper and lower extremities and spinal stability [
13]. In contrast, the static flexibility and muscular strength assessments used herein may be less relevant to the physical competence of children of this age, who perform primarily dynamic rather than static movements. It is nonetheless possible that if a few alternative indicators of flexibility and muscle strength were also measured, there could have been additional latent variables that would better characterize these aspects of physical competence. Overall, our results confirm emerging evidence that the domain of Physical Competence pertaining to physical literacy is best characterized by indicators that capture whole body movement, including both skill and fitness components.
For conceptual reasons discussed in the Daily Behaviour domain’s results around the distinctiveness between physical activity and sedentary behaviours (including screen time), a decision was made to remove screen time as an indicator of physical literacy. Additionally, the Daily Behaviour domain is comprised of one subjective (i.e., self-report physical activity) and one direct (i.e., pedometer step counts) indicator. The subjective indicator had a stronger factor loading than that of the pedometer step counts. It is unlikely that this finding stems from the large amount of missing data since the results were similar when the model was rerun with only complete data; nevertheless, there were gender differences between boys and girls who provided valid pedometer data that could have impacted the results. Alternatively, the discrepant factor loadings could be related to what each indicator is actually measuring. In the past, researchers have found that self-reported physical activity and scores from more direct measures may be weakly correlated [
36] and tap into somewhat different components of physical activity [
35]. For example, children can self-report physical activities that direct measures of physical activity do not capture (such as swimming and cycling, which children report frequently) [
35], complex upper body movements, and non-load-bearing activities [
37]. Additionally, within the CAPL, the
number of days in a
typical week engaged in physical activity was assessed via self-report, whereas the
number of steps taken over a
specific week was assessed via pedometer. These discrepancies between measurements could be causing the differential magnitudes of the factor loadings. Nonetheless, both indicators served as moderately robust indicators of Daily Behaviour and were retained.
An unexpected finding from the Daily Behaviour domain was the large cross-loading from knowledge of physical activity guidelines onto daily behaviour in comparison to its hypothesized domain of Knowledge and Understanding. There are a few possible explanations for this finding. First, it is possible that the instructional stem for the self-report physical activity indicator (which asks the children to report how many days they have engaged in moderate and vigorous physical activity for at least 60 min) actually provided the answer to the subsequent question about how long children should engage in moderate to vigorous physical activity each day (i.e., the correct answer is 60 min). Second, as identified by CAPL administrators through informal feedback, the response option for the physical activity guideline question may have been too easy given that the correct response was the highest value listed. This response option is problematic because it is easy to guess and might have been cued by the self-report physical activity indicator, which includes “60 min” in the instructional stem. In other words, it is possible that the two questions are sharing variance given that children can link the responses of one to the stem of the other. Future research is needed to determine if altering the response options for the knowledge of physical activity guidelines can reduce this problematic cross-loading. Finally, it is worth noting that we did not remove knowledge of physical activity guidelines as an indicator of the Knowledge and Understanding domain and place it as an indicator of Daily Behaviour. Our decision to retain it as an indicator of knowledge and understanding and allow for the cross-loading was based on the conceptual content of the indicator. Because the indicator asks children to report how long kids should be active, we did not feel it was a good indicator of how long the children themselves were actually active.
Five out of 10 indicators from the Knowledge and Understanding domain were removed. Conceptually, many original indicators that have not undergone previous validation do not align with current physical literacy research. For example, knowledge of what it means to be healthy could be an indicator of health literacy rather than physical literacy. Removing self-report screen time from the Daily Behaviour domain and knowledge of screen time guidelines from the Knowledge and Understanding domain can be justified for similar reasons; they may not be conceptually linked to physical literacy given that sedentary behaviours are distinct from physical activity behaviours [
33,
34]. Additionally, as noted in the “CAPL Measures” section, we identified an error in the CAPL manual such that an originally conceived indicator for motivation and confidence was erroneously placed within the Knowledge and Understanding domain (i.e., activity preferences). Not surprisingly, asking children their activity preferences was a weak indicator of their knowledge and understanding, and so was removed from that domain. Supplemental analyses indicated that the indicator did indeed load within the Motivation and Confidence domain, although it resulted in a slight deterioration in model fit. Given that the indicator is dichotomous and that the other indicators of Motivation and Confidence provided a superior model fit, combined with an eye toward reducing the participant burden of CAPL, we decided to retain the Motivation and Confidence domain without this indicator.
Knowledge about the safety gear required for physical activity may be a weak indicator of knowledge and understanding, given that the response format did not provide standardized response options. Indeed, this question had an unusual format where children first selected activities they personally engaged in and then, of those, they circled activities requiring safety gear. Consequently, the answers could vary widely.
The question around how to “improve fitness” was removed because a similar question (i.e., “improve sport skills”) was deemed to add more unique content representation whereas “improve fitness” may already be captured by the definition of cardiorespiratory fitness. We retained some indicators of Knowledge and Understanding that had weak factor loadings so as to retain content representation in light of the number of indicators removed. We believe that despite their weaker factor loadings, “improve sport skills” and “knowledge of physical activity guidelines” were querying unique information and that with further redevelopment of the remaining Knowledge and Understanding indicators, they may have stronger factor loadings. The results herein were used to inform the redevelopment of the knowledge and understanding indicators. Notably, the response options for the knowledge of physical activity guidelines indicator were changed, and another fill-in-the-blank question was added to the physical activity comprehension indicator (see Longmuir et al. [
38] for further details).
Lastly, one indicator was removed from the Motivation and Confidence domain. “Activity compared to others” was removed to assist in reducing participant burden and reduce construct irrelevant variance attributable to similarity with another indicator. Otherwise, the Motivation and Confidence domain had strong indicators.
Practical implications
The results of this study will inform the creation of CAPL-2 [
38]. It is important to note that although we have eliminated indicators that had weak factor loadings on targeted CAPL domains (e.g., waist circumference, grip strength), these indicators in and of themselves may not be bad indicators of what they are purported to asses (e.g., grip strength and sit-and-reach scores can be valid indicators of muscle strength and flexibility, respectively [
39,
40]). Therefore, researchers may wish to include these assessments depending on their research question. Additionally, when collecting longitudinal data with the CAPL, it would be prudent to use the same version of CAPL as previously administered such that change over time can be examined unambiguously. Lastly, as noted in the “Background” section, researchers around the world have debated whether or not physical activity should be included within the definition of physical literacy or whether it would be better conceptualized as an outcome of physical literacy. Within the CAPL, we believe that valuing and taking responsibility for engaging in physical activity situates physical activity as an inherent component of physical literacy. Nonetheless, a strength of the CAPL is that researchers are free to use (or not use) domains that are of interest to them, given their theoretical perspective. As such, it is possible to use the CAPL and omit the Daily Behaviour domain to assess physical literacy, as described in the CAPL Manual (available at
www.capl-ecsfp.ca).
A key finding of this study, when taken in the context of past versions, conceptualizations, and operationalizations of the CAPL, is that physical literacy cannot be reduced only to fitness or motor skill assessments. Indeed, original conceptualizations placed strong emphasis on indicators of physical fitness; yet over the past decade, with emerging validation and physical literacy research, it has become apparent that physical competence encompasses more than fitness and body composition and that motivation and confidence are equally important for physical literacy.
Strengths, limitations and future directions
A major strength of this study was the large sample size, which included children from 11 regions across Canada, enabling the assessment of all CAPL indicators. Notwithstanding these strengths, limitations of this study should be acknowledged. There was a large amount of missing data on the pedometer daily step count scores, and there were gender differences in the amount of missing data for pedometers. Although modern procedures were used to handle missing data, it is not known to what extent the missing data influenced the final conclusions. Nonetheless, sensitivity analysis of the final higher-order model without missing data (i.e., using listwise deletion
n = 5073) revealed a similar pattern and magnitude of results to those we report herein. Second, although our decisions to remove or retain indicators were made based on statistical criteria, theory, and considerations of content representation, many of these decisions were ultimately subjective in nature. It is possible that alternative models would have fit the data well. Moreover, although we used confirmatory factor analyses, the post-hoc modifications were more exploratory in that they were based on discussion, consideration of theory, and statistical criteria. It is paramount that researchers continue to examine the factor structure of the CAPL found herein using new samples in different contexts. Researchers may wish to test alternative models that align with other definitions of physical literacy theory [
4]. Additionally, researchers may wish to extend validity evidence based on internal structure by examining grade and gender invariance to determine if CAPL scores are being measured in the same way across genders and grade levels. Finally, the measures in the Knowledge and Understanding and the Motivation and Confidence domains are self-reported and therefore susceptible to social desirability or recall bias.