Validation of the exercise-related cognitive errors questionnaire short form

Abstract

Exercise-related cognitive error represent the extent to which individuals view their exercise engagement through a negative and biased lens. Three datasets were examined to develop a short form of the original 16-item exercise-related cognitive errors questionnaire (E-CEQ) and evaluate evidence of validity. Exploratory factor analysis on datasets 1 (N = 394), 2 (N = 177), and 3 (N = 1027) suggested that a seven-item, one-factor model fit the data. Findings suggested that the ECEQ short form had a unidimensional factor structure that did not vary based on age or gender. As evidence of criterion-related validity, similar magnitude correlations were observed for the E-CEQ short-form (ECEQ-SF) and the original E-CEQ with key exercise variables in datasets 1 and 2 (| rs | ranged from .20 to .76). The ECEQ-SF captures the extent to which individuals view their perceived exercise barriers through a cognitively errored lens.

Keywords

physical activity cognitive bias social cognition factor analysis

Introduction

Exercise is recommended by public health authorities for its numerous health benefits, yet many Canadians do not meet these guidelines (Colley et al., 2011). While social cognitive theories tend to assume that human beings often think in a rational manner (Ekkekakis and Zenko, 2016), they also acknowledge that behavioral disengagement can result from “inaccurate, biased, or even irrational” beliefs (Ajzen and Fishbein, 2005: p. 193). Cognitive errors are one mechanism theorized to cause inaccurate perceptions that lead to behavioral disengagement (Beck, 1976), like being insufficiently active (Locke and Brawley, 2016). Overly negative or biased perceptions are associated with a variety of negative health-related outcomes such as low adherence during cardiac rehabilitation (Anderson and Emery, 2014), poorer sleep quality including insomnia (Akram et al., 2023; Harris et al., 2015), poor fatigue management (Spence et al., 2005), increased likelihood of alcoholism relapse (Rinck et al., 2018), disturbed eating attitudes (Jakatdar et al., 2006), and poorer food regulation (Paslakis et al., 2021). The study of cognitive errors specific to exercise engagement suggests that some individuals have biased and inaccurate perceptions about exercise that cause them to struggle with maintaining regular exercise (Locke and Brawley, 2017; Yuan et al., 2025).

Exercise-related cognitive errors (ECEs) represent a biased lens that causes individuals to exaggerate the challenge associated with exercise, such that perceptions become inaccurate (Locke and Brawley, 2016). ECEs magnify the challenge of overcoming one’s barriers to physical activity leading them to feel that they are not able to exercise despite evidence to the contrary. Barriers are a key influence of behavioral maintenance (Schwarzer, 2008). By magnifying barrier perceptions, ECEs have the potential to disrupt planned activity and may be a useful concept for understanding adherence to physical activity.

The Exercise-Related Cognitive Errors Questionnaire (E-CEQ; Locke and Brawley, 2016) was developed as a quantitative measure of ECEs, which allows researchers to investigate correlates of ECEs and whether ECEs are susceptible to change. The E-CEQ has demonstrated utility in predicting exercise behaviour after controlling for past exercise behaviour and in predicting the consistency of exercise engagement (Locke and Brawley, 2016, 2017). These studies also demonstrated that ECEs were related to social cognitions important for exercise engagement (e.g., self-regulatory efficacy, decisional struggle). Further applications have shown that ECEs are modifiable through reframing counselling amongst both symptomatic and asymptomatic individuals (Locke et al., 2019). These participants also reported increased exercise behaviour following cognitive error reframing.

Despite this promising evidence for the utility of the E-CEQ, opportunities for refinement exist. The original E-CEQ uses 16 vignette items to operationalize situations that could cause an individual to struggle to decide to exercise if viewed through an ECE. The use of short-form versions of measures is encouraged where possible to reduce participant burden and maximize completed responses (Edwards et al., 2004; Nakash et al., 2006). A shorter measure would also allow for the reduction of underperforming items with low response variability. This is of particular importance in larger trials with high numbers of measures, as higher numbers of items may produce greater numbers of incomplete responses (Rogelberg and Stanton, 2007). If a short form of the E-CEQ can demonstrate strong psychometric properties and retain its predictive utility, then researchers will have an alternative tool to capture ECEs with potentially improved response rates and reduced participant burden. The purpose of this study was to develop and examine the psychometric properties of the E-CEQ short form (ECEQ-SF).

Conceptual background

The original E-CEQ and ECEQ-SF were developed to operationalize cognitive errors specific to exercise engagement. The concept of cognitive errors comes from the counselling literature in the understanding of depressive psychology (Beck 1976), describing how individuals with depression to systematically misinterpret the meaning of events perpetuates depressive symptomatology (Lefebvre, 1981). A cognitive error is described as a “verbal statement that suggests ways of evaluating information that reflect errors or biases away from the average or normative evaluation of the same material” (Milman and Drapeau, 2012: p. 129). Established cognitive error definitions by Drapeau and Perry (2010) were operationalized to guide item development. There were three cognitive error factors operationalized in the original E-CEQ (Locke and Brawley, 2016). The catastrophizing cognitive error occurs when “one predicts that the future outcome of some situation will be negative without considering more likely outcomes, which may be less negative” (Drapeau and Perry, 2010: p. 18). All-or-nothing occurs when “the individual views a situation as fitting into one of only two opposing categories, rather than as a mixture or on a continuum between the two” (p.30). Mental filter occurs when “the individual pays undue and complete attention to only one aspect of an individual or situation without acknowledging the other sides of the issue which would yield a whole picture” (p. 39).

Importantly, ECEs have been demonstrated to occur in non-clinical samples without psychopathology (e.g., Locke and Brawley, 2017, 2019; Yuan et al., 2025). As such, individuals that are not experiencing depression or anxiety can magnify the degree of difficulty perceived in overcoming given barriers to exercise if viewed through a cognitive error (Locke et al., 2019). Viewing exercise through a cognitively errored lens could cause individuals to exaggerate the challenge associated with exercise or to unilaterally focus on a barrier to the extent that the benefits of exercise are discounted or completely ignored.

Understanding the potential influence of ECEs on exercise perceptions is important given that perceptions are subjective judgements that are mentally constructed by the individual. These judgements may or may not align with the “objective” levels of challenge associated with the obstacles one might face (Glasgow, 2008). For example, not having enough money to buy a gym pass will strongly influence the perceived challenge of exercising at a gym. In contrast, believing there is not enough time to exercise may be misaligned with reality, particularly given recent statistics demonstrating the average American has over 5 hours of daily free time (Sturm and Cohen, 2019). As subjective judgements, exercise perceptions are subject to misinterpretation and bias. Thus, cognitive errors may create a malalignment between “objective” and perceived levels of challenge to exercise.

The purpose of the present study was to develop a short form of the original E-CEQ (Locke and Brawley, 2016) across three datasets that demonstrate: (a) sufficient content coverage across cognitive error domains; (b) adequate internal reliability; (c) a factor structure that provides goodness of fit indexes meeting acceptable standards across multiple occasions; (d) invariance across age and gender; and (e) convergent validity to provide evidence of whether the ECEQ-SF correlates at a similar magnitude as the original E-CEQ across exercise, mental health, and social cognitive variables. These criteria align with recommendations relevant for current short-form development (Marsh et al., 2005; Smith et al., 2000).

The relative convergent validity between the original and short form E-CEQ will be assessed using datasets 1 and 2 across key exercise, social cognitive, and mental health variables (dataset 1: exercise engagement, self-regulatory efficacy; dataset 2: symptoms of anxiety and depression, motivation, self-compassion, and physical activity enjoyment). These are all antecedents that help us understand exercise engagement. For example, Deci and Ryan (2008) suggest that intrinsic motivation is essential for sustained health behaviour change. Self-compassion involves being caring and compassionate towards oneself in the face of perceived difficulty or personal inadequacy (Neff, 2023) and is a psychological resource helpful for overcoming exercise barriers or lapses (e.g., Simpson et al., 2021). Within a social cognitive paradigm, theorists like Bandura (1997) have acknowledged that individuals can misconstrue self-relevant information causing inaccurate social cognitions. ECEs could be indicative of the misconstrual process biasing these exercise self-perceptions. Finally, Beck’s (1976) cognitive error model suggests that those with emotional disturbances (e.g., depression, anxiety) are more likely to make cognitive errors across situations, thus, we would anticipate that those with higher symptoms of depression and anxiety would be more likely to make ECEs. Examining ECEs’ relationship to variables related to exercise regulation is an important step in demonstrating criterion-related validity.

Methods

Content representativeness and item selection

Short forms should be designed to adequately cover the content of the original measures (e.g., Soto and John, 2017). There were three primary factors driving item selection. First, items were selected that operationalized the most common and relevant perceived exercise challenges (e.g., Bandura, 2006; Glasgow, 2008; Herazo-Beltrán et al., 2017; Hoare et al., 2017; Patay et al., 2015). For example, being too tired to exercise occurs more frequently in the general population than does “an aggravated medical condition”. Second, redundant items were removed. For example, items 2, 4 and 10 in the original E-CEQ represented different scenarios related being too tired or fatigued to exercise. Selection among redundant items was determined by examining the item response breakdown statistics for salience. For example, 36 percent of participants rated item 10 on the E-CEQ at the scale minimum (i.e., a 1 out of 9) and the median response was 2, suggesting the item was not very salient and had low variability. Third, items retained for the ECEQ-SF should be evenly dispersed among the three cognitive error factors so that at least two items originally representing each cognitive error factor were retained (e.g., selecting item 2 over item 4). In the end, 7 items were retained for the short form.

Single- versus multi-factorial model

A common misconception of short-form construction is that their purpose must necessarily be to substitute decision making from the long-form and therefore must reproduce the exact hierarchical factor structure (Ziegler et al., 2014). While short-form measures are often designed to reproduce the same factor structures as their parent measures, researchers may decide a priori to only preserve the higher-level factors of the model (Smith et al., 2000). While multifactorial models will be compared, a one-factor model of the ECEQ-SF was proposed, which deviates from the three-factor hierarchical model from the original E-CEQ (Locke and Brawley, 2016). Short forms that retain few items per sub-factor should only seek to use the overall factor score (Clark and Watson, 2019; Soto and John, 2017). Conceptually, a single-factor model would represent the extent to which an individual generally views barriers to exercise through a biased lens, similar to interpreting the hierarchical overall cognitive error factor from the original E-CEQ.

Participants and procedures

Informed consent and ethics approval were obtained from a university’s research ethics board for all collected data, including for the secondary analysis. Three datasets were examined in this study, whereby dataset 1 was collected for this validation study, while datasets 2 and 3 were secondary data from different studies. For dataset 1, adults, at least 18 years of age, were recruited via in-person presentations to university classrooms, via posters put up around the community and campus, and via online postings (e.g. Facebook groups, Kijiji, craigslist) to complete a 15-min online survey responding to different exercise-related survey measures. Dataset 2 represented a secondary analysis of baseline data from an ongoing community-based diabetes prevention program. Participants were adults aged 18-65, were either diagnosed with prediabetes or at high risk for developing type 2 diabetes and were without comorbidities or contraindications to exercise. For more details about the trial’s design and protocol, see Bourne et al. (2019). Dataset 3 represented a secondary analysis of pooled data from a series of three inter-related studies examining the relationship between ECEs and different automatic evaluations of exercise. Participants were either recruited through a university participant pool and received course credit for participation, or through Prolific (https://www.prolific.co/about) and received £5 for participation. Prolific is an online platform for recruiting research participants and has been shown to have similar drop-out rates and high scale reliability compared to other online recruiting platforms (Peer et al., 2017).

Measures

E-CEQ

The 16-item E-CEQ assesses cognitive errors that occur in relation to exercise and within an exercise context (Locke and Brawley, 2016). Items are short vignettes (i.e., depicting a cognitively distorted response to an exercise situation). An example item reads, “You know the health benefits of exercise, but think to yourself, “Exercising will take a lot of time from all the other fun things I could be doing.” Items were rated on a nine-point Likert scale ranging from 1 (not at all like I would think) to 9 (exactly like I would think). Datasets 1 and 2 used the full version of the E-CEQ, while dataset 3 analyzed items only from the newly created ECEQ-SF. Internal consistency across the three subscales in datasets 1 and 2 ranged from α = .78 to .88.

Self-regulatory efficacy to manage cognitive errors

Participants were instructed to think about their confidence to manage their exercise while experiencing the unhelpful thoughts depicted in the E-CEQ. An example item reads, “How confident are you that you can prevent your unhelpful thoughts from interfering with deciding to exercise as planned?” Responses were scaled from 0% (not at all confident) to 100% (completely confident). Internal consistency was α = .98.

Moderate to vigorous physical activity

MVPA was measured by asking participants to report the number of bouts of moderate and vigorous activity lasting at least 30 minutes that they engaged in over the past 7 days using the modified Godin Leisure-Time Exercise Questionnaire (GLTEQ; Godin and Shephard, 1985; Amireault and Godin, 2012). The number of bouts were summed and used for the analysis.

Symptoms of anxiety was measured using the seven-item Generalized Anxiety Disorder scale (Spitzer et al., 2006). Participants were asked to report how often different anxiety-related symptoms had occurred over the past 2 weeks. Items are rated on a four-point scale ranging from 0 (not at all) to 3 (nearly every day). The internal consistency was α = .92.

Symptoms of depression was assessed using the nine-item Patient Health Questionnaire (Spitzer et al., 1999). Participants were asked to report how often different depression-related symptoms had occurred over the past 2 weeks. Items are rated on a four-point scale ranging from 0 (not at all) to 3 (nearly every day). The internal consistency was α = .84.

Self-compassion was measured using the Self-Compassion Scale (SCS; Neff, 2023). This 26-item scale assesses six self-compassion sub-factors (i.e., self-kindness, mindfulness, over-identification, self-judgement, common humanity, and isolation). An example item reads, “I try to be understanding and patient towards those aspects of my personality I don’t like.” Items were rated on 5-point Likert scale ranging from 1 (almost never) to 5 (almost always). This scale has demonstrated several forms of validity (Neff and Tóth-Király, 2020). The overall scale mean was used by averaging scores on the six sub-factors (subscales ranged from α = .76 to .87).

Physical activity enjoyment was measured using the Physical Activity Enjoyment Scale (Kendzierski and DeCarlo, 1991). The scale asks 18 bipolar statements with seven points between them (e.g.., “I enjoy it” to “I hate it”). Items were summed to obtain a unidimensional measure of enjoyment (α = .94).

Motivation was measured using the 19-item Behavioral Regulation in Exercise Questionnaire (BREQ-2; Markland and Tobin, 2004). The scale assessed five forms of motivation: intrinsic, identified, introjected, external, and amotivation. Only intrinsic motivation (α = .87) and amotivation (α = .72) were examined in this study to demonstrate whether ECEs were related to these different forms of motivation in their expected directions.

Demographics

Across the datasets, age, gender, ethnicity, marital status, employment status, and education were collected.

Analytical plan

All analyses were run using Mplus Version 7.31 (Muthén & Muthén, 1998–2012) Missing data were handled with full estimation maximum likelihood (FIML), the default method implemented in Mplus with the MLR estimator (Graham, 2009). Missing data were low across dataset 1 (0.3%), dataset 2 (3.39%), and dataset 3 (0.5%). Assessment of model fit was based on multiple indicators as per recommendations (cf., Byrne, 2010; Hu and Bentler, 1999). Fit indices included the RMSEA, CFI, and TLI, where RMSEA values closer to 0 suggest better fit, whereas values closer to 1 represent better fit for the CFI and TLI (Byrne, 2010; Hu and Bentler, 1999).

Exploratory factor analysis (EFA) was conducted across all 3 datasets using Mplus’s robust maximum likelihood estimator (MLR) with a geomin oblique (epsilon = .5) rotation following recommendations by Marsh et al. (2009). MLR estimation is robust to violations of normality (Muthén & Muthén, 1998–2012). Parallel analysis and computation of model fit indices (i.e., chi square test of exact fit (χ²), Root Mean Square Error of Approximation (RMSEA; and its associated confidence interval [CI]), the comparative fit index (CFI), and the Tucker-Lewis Index (TLI)) were used to examine the number factors and model fit. Parallel analysis is a Monte Carlo simulation that generates random data to compare eigenvalues obtained from the sample data to eigenvalues obtained from completely random data. The number of factors retained is specified when the eigenvalues obtained from the data are larger than the eigenvalues obtained at random.

Invariance testing

Configural, metric, and scalar invariance was examined separately in dataset 3 across gender (1 = man, 2 = woman) and age (1 = age 25 and under, 2 = age 26 and over) using. The following sequence of steps were used to test invariance: (i) fit of the overall model was assessed, (ii) fit for group 1 was assessed (i.e., for men or for age 25 and under), (iii) fit for group 2 was assessed, (iv) configural invariance was assessed via chi square difference test and CFI change between the configural model and the overall model, (v) metric invariance was assessed between the metric model and configural model, (vi) scalar invariance was assessed between the scalar model and the configural model. Configural invariance determines whether the number of factors and factor loading are similar across groups. Metric invariance determines whether item intercepts are similar across groups. Scalar invariance determines whether the means are similar across groups. For comparing nested models in steps v and vi, a non-significant chi-square would suggest model fit did not get significantly poorer. Since chi-square is sample-size dependent, we also relied on ΔCFI, ΔTLI, and ΔRMSEA. A change of 0.010 or less in CFI or TFI or a change in RMSEA of 0.015 or less would suggest there was not a significant decrement in fit between invariance models (Chen, 2007). If a model was not scalar invariant, the degree of difference between intercepts of the two groups will be estimated (e.g., whether mean of group 1 is higher than group 2). Differences between group means will be estimated by setting the latent factor mean of group 1 to zero (i.e., reference group) and freely estimating the latent factor mean of group 2.

Convergent validity analyses

Prior to running these analyses, confirmatory factor analysis was run using Mplus to verify fit to the factor structure of the original E-CEQ. Correlations were run using to examine the relative strength of the relationships between the original E-CEQ and the ECEQ-SF with a variety of outcome variables in datasets 1 and 2. Fisher’s z-to-r transformation was used to examine whether there were significant differences in the magnitude of correlations between the two versions of the measure.

Results

Demographics

Participants in dataset 1 (N = 394) reported a mean age of 38.9 years (SD = 17.3), were predominantly women (73%; 24% men, 1% transgender, 2% unknown), of European ethnic descent (72%; 15% Asian, 7% other ethnicity, 6% unknown/chose not to answer), single (46%) or married (38%), obtained some level of post-secondary education (79%), and were employed (54%).

Participants in dataset 2 (N = 177) reported a mean age of 42.3 years (SD = 7.1) were predominantly women (75%; 24% men, 1% unknown), White (89%; 5% Asian, 5% other, 1% unknown/chose not to answer), single (17%) or married (58%), obtained some level of post-secondary education (79%), and were employed full- or part-time (55%).

Participants in dataset 3 (N = 1027) reported a mean age of 22.93 years (SD = 5.72) and with balanced gender distribution (51.0% men, 49.0% women). Regarding ethnicity, 49.2% were White, 23.0% Asian, 7.4% Black, 11.6% biracial or other, and 8.8% were missing or not disclosed. Regarding education level, 45.5% had a high school degree and/or were currently in an undergraduate degree program, 38.1% had an undergraduate or graduate degree, 12.6% had a college diploma, and 3.8% were missing or not disclosed.

Exploratory factor analysis

The EFA results across all three datasets yielded similar results in support of a 1-factor model of the ECEQ-SF. Parallel analysis suggested 1-factor models across all three datasets. Residual correlations across all items were below r = .1. Fit indices are presented in Table 1. While fit indices suggested slightly better fit for the 2- and 3-factor models, a review of the factor loadings yielded uninterpretable factors, low item loadings, and/or multiple cross-loadings which would be unideal for a measure with so few items. Across all three datasets, fit indices suggest the data fit a 1-factor model (i.e., CFI and TLI exceeded .90, RMSEA ≈0.7). The 1-factor models across all three datasets yielded similar significant factor loadings (Table 2). There was acceptable internal consistency reliability across all three datasets (Cronbach’s α’s = .83, .87, and .90). See the supplemental file for the full measure. Overall, the 7-item, 1-factor ECEQ-SF fit the data.

Table 1.

Measurement models examined for the ECEQ-SF and original E-CEQ.

Data set	# Of factors	Fit indices
		Chi-square analyses			RMSEA analysis		CFI	TLI
		χ²	df	p	RMSEA	90% C.I.	CFI	TLI
ECEQ-SF 7-item EFA
1	1	42.55	14	<.001	.072	.048–.097	.957	.935
	2	7.610	8	.472	.000	.000–.057	1.00	1.00
	3	—	—	—	—	—	—	—
2	1	32.60	14	.003	.087	.048–.126	.944	.916
	2	13.21	8	.105	.061	.000–.117	.984	.959
	3	—	—	—	—	—	—	—
3	1	83.14	14	<.001	.069	.055–.084	.971	.956
	2	41.80	8	<.001	.064	.046–.084	.985	.962
	3	48.29	3	<.05	.041	.008–.076	.998	.984
Original 16-item E-CEQ CFA
1	3	446.498	98	.001	.095	.086–.104	.883	.857
2	3	325.205	98	.001	.114	.101–.128	.856	.824

Note: ECEQ-SF = Exercise-related cognitive errors questionnaire (ECEQ) short form (SF); ECE = exercise-related cognitive error; EFA = exploratory factor analysis; CFA = confirmatory factor analysis.

Table 2.

Items and factor loadings for the final ECEQ-SF model.

Items	Sample 1λ(δ)	Sample 2λ(δ)	Sample 3λ(δ)
1. You are considering starting to cycle with a local club. Every time you consider going to the club to join, you think to yourself, “I have not biked in years, I am going to get way too tired to even be able to finish the ride.”	.54* (.71)	.67* (.55)	.51* (.74)
2. You are having a pretty busy week. You plan to exercise tonight, but when you get home from work you think to yourself, “I cannot justify exercising because I have so many other things to do.”	.69* (.53)	.77* (.41)	.64* (.59)
3. Because your exercise class is cancelled this week, you think to yourself, “I am going to take the week off because I have no exercise class”	.47* (.76)	.76* (.43)	.69* (.52)
4. You know the health benefits of exercise, but think to yourself, “Exercising will take a lot of time from all the other fun things I could be doing.”	.73* (.47)	.59* (.66)	.79* (.38)
5. You have been exercising for a few weeks. However, you’re getting frustrated because you are not seeing changes and the exercises are not getting easier. You think to yourself, “This is way too hard and no fun”. And you decide to stop going to the gym	.62* (.61)	.76* (.42)	.80* (.36)
6. You have been feeling down and depressed all day. You think to yourself, “I should just stay home instead of going to the gym today.”	.90* (.36)	.71* (.50)	.76* (.42)
7. You feel awkward and lost in the first gym/fitness class you attend. You think to yourself, “I feel so uncomfortable that I don’t want to go back.”	.68* (.54)	.69* (.53)	.76* (.43)
Cronbach’s α	.83	.87	.88
ω	.84	.88	.88

Note: ECEQ-SF = Exercise-related cognitive errors questionnaire short form; λ: standardized factor loading; δ = uniqueness. *p < .05.

Measurement invariance across gender

Prior to testing measurement invariance, fit for the overall model, the men-only model and the women-only was assessed. Across all three steps, the models provided good fit to the data (see Table 3). The configural model fit the data, suggesting a similar number of factors and loadings across gender (RMSEA CI: .056-.086, CFI = .968, TFL = .952). There was not a meaningful decrement in model fit between the metric and configural model supporting metric invariance and suggesting that the factor loadings did not vary by gender (Δ χ² = 4.96, p = .55, ΔCFI = −.002, ΔTLI = +.006, ΔRMSEA = −.005). There was a significant decrement in model fit between the scalar and configural model suggesting there was not scalar invariance and indicating that item intercepts varied by gender (Δ χ² = 74.71, p < .001, ΔCFI = −.026, ΔTLI = −.012, ΔRMSEA = +.008). The standardized difference between latent means for men and women groups was 0.563 (S.E. = .074, Cohen’s d = 0.48), which suggested that women’s scores on the ECEQ-SF were 0.563 units higher than men’s. Overall, findings suggest that the factor structure and item intercepts of the ECEQ-SF did not differ between women and men, however, women reported slightly higher overall exercise-related cognitive error scores than men.

Table 3.

Gender and age invariance for the short form measurement models.

Step			Fit indices
	Chi-square analyses		RMSEA analysis		CFI	TLI
	χ2	df	RMSEA	90% C.I.			CM	Δ χ2 (Δdf)	p	ΔCFI	ΔTLI ΔRMSEA
Invariance across gender
1-Overall model	83.14*	14	.069	.055–.084	.971	.956
2–Men only	50.24*	14	.070	.050–.092	.969	.954
3-Women only	50.01*	14	.072	.051–.093	.966	.949
4-Configural invariance	100.25*	28	.071	.056–.086	.968	.952
5-Metric invariance	109.67*	34	.066	.052–.080	.966	.958	5–4	4.96 (6)	.550	.002	.006 -.005
6-Scalar invariance	168.94*	40	.079	.067–.092	.942	.940	6–4	74.71 (12)	<.001	.026	.012 +.008
Invariance across age
1-Overall model	83.14*	14	.069	.055–.084	.971	.956
2–Age 25 and under	59.58*	14	.066	.049–.084	.973	.959
3-Age 26 and over	25.63*	14	.056	.018–.090	.981	.971
4-Configural invariance	82.25*	28	.062	.047–.078	.975	.963
5-Metric invariance	98.54*	34	.061	.047–.076	.971	.964	5–4	15.61 (6)	<.05	.004	.001 -.001
6-Scalar invariance	116.95*	40	.062	.049–.075	.965	.963	6–4	34.52 (12)	<.001	.010	.000 .000

Note: CM = comparison model; Δχ² = change in chi square; ΔCFI = change in CFI; ΔTLI = change in TLI.

Measurement invariance across age

Prior to testing measurement invariance, fit for the overall model, the model for age 25 and under, and the model for age 26 was assessed. Across all three steps, the models provided good fit to the data (see Table 3). The configural model fit the data, suggesting a similar number of factors and loadings across gender (RMSEA CI: .047-.078, CFI = .975, TFL = .963). There was not a meaningful decrement in model fit between the metric and configural model supporting metric invariance and suggesting that the factor loadings did not vary by age (Δχ² = 15.61, p < .05, ΔCFI = −.004, ΔTLI = +.001, ΔRMSEA = −.001). There was not a meaningful decrement in model fit between the scalar and configural model suggesting scalar invariance and indicating that item intercepts did not vary by age (Δχ² = 34.52, p < .001, ΔCFI = −.010, ΔTLI = .000, ΔRMSEA = .000). Overall, findings suggest that the factor structure, item intercepts, and overall exercise-related cognitive error scores on the ECEQ-SF did not differ across age.

Convergent validity

See Table 4 for results of correlations between exercise-related variables and the two versions of the E-CEQ. As hypothesized, across both the E-CEQ and ECEQ-SF, exercise-related cognitive errors were negatively associated with self-regulatory efficacy, MVPA, intrinsic motivation, physical activity enjoyment, and self-compassion. Exercise-related cognitive errors were positively associated with symptoms of anxiety, symptoms of depression, and amotivation. All correlations were significant (p < .001) and moderate to large in size. There were similar magnitude correlations observed between the second-order overall score of the original E-CEQ and the ECEQ-SF across datasets 1 and 2. According to Fisher’s Z-to-r transformation, the magnitude of the correlations did not significantly decrease in the newly-developed short form, suggesting the reduced number of items did not negatively impact the predictive utility of the measure.

Table 4.

Correlations between exercise-related variables with the original E-CEQ and ECEQ-SF.

	Original E-CEQ	ECEQ-SF	Fisher r-to-z
Dependent variable	r	r
Dataset 1
Self-regulatory efficacy	−.64***	−.76***	Z = 3.33, p < .001
Weekly bouts of MVPA	−.20***	−.23***	Z = 0.44, p = .66
Dataset 2
Symptoms of anxiety	.39***	.35***	Z = .43, p = .67
Symptoms of depression	.40***	.37***	Z = 0.33, p = .74
Amotivation	.45***	.40***	Z = 0.57, p = .56
Intrinsic motivation	−.48***	−.50***	Z = 0.25, p = .80
Physical activity enjoyment	−.41***	−.42***	Z = 0.11, p = .91
Self-compassion	−.44***	−.40***	Z = 0.45, p = .65

Note: ECEQ-SF = Exercise-related cognitive errors questionnaire (ECEQ) short form (SF); MVPA = moderate to vigorous physical activity. Small r = .1; medium, r = .3; large, r = .5; ***p < .001.

Discussion

The purpose of this paper was to develop a short form of the original E-CEQ to operationalize cognitive errors about engaging in exercise. Across all three datasets, the 7-item 1-factor ECEQ-SF model fit the data. The relationships between the original E-CEQ and the ECEQ-SF across a variety of exercise behaviour, mental health, and social cognitive variables were similar in magnitude and directionality, providing evidence of criterion-related validity. Researchers aiming to use the ECEQ-SF should interpret the overall 7-item score as representing the extent to which individuals view their perceived barriers to exercise engagement through a cognitively errored lens. It will be particularly useful for gathering data in large samples with multiple assessments. It should not be used to parse out the relative strength of different cognitive errors; for this, the original E-CEQ or Drapeau and Perry’s (2010) rating guide should be used.

Findings from the invariance testing of the ECEQ short form suggest a unidimensional factor structure that did not vary based on age or gender. However, the latent means varied by gender (but not age) whereby women reported moderately higher scores on the ECEQ-SF than men. This aligns with numerous chronic pain studies suggesting that women consistently report higher catastrophizing scores (a type of cognitive error) than men (Keogh, 2022). Recent research has demonstrated attentional differences between men and women, with women directing more attention towards pain-related body expressions than men (Keogh et al., 2021). Others have argued from a social learning perspective (Bussey and Bandura, 1999) that women may be socialised to more readily accept and express their challenges compared to men (Cerbara et al., 2022). While the current study is not designed to explain gender differences on the ECEQ-SF, it does illuminate questions for future research. Researchers should use caution when examining and interpreting gender differences using the ECEQ-SF.

Past research investigating gender invariance across measures of exercise barriers has been mixed. While Koehn and Amirabdollahian (2021) observed gender invariance in their measure, others have not (Liu et al., 2014). Gender non-invariance, suggesting women report higher ECEQ-SF scores than men, supports past findings demonstrating women report more time and interpersonal constraints to physical activity than men (Casper et al., 2011; Louw et al., 2012). While the ECEQ-SF operationalizes cognitively errored responses to common physical activity challenges, questions remain about whether the process of making cognitive errors or the type of barriers contribute to the observed latent mean gender differences.

Across all variables, both the original ECEQ and ECEQ-SF were correlated in the expected direction and with similar magnitude relationships. These findings provide evidence of criterion-related validity, indicating that both forms represent cognitive errors within the exercise context. The evidence also suggests that the short form did not appreciably attenuate the magnitude of the relationships to key exercise-related variables. Indeed, fewer items will not necessarily reduce the psychometric properties of a measure, beyond Cronbach’s alpha values which are biased to produce higher alpha values for scales with a greater number of items (Streiner et al., 2015). In the present study, internal reliability scores from the short-form were not appreciably smaller than those reported on the original long-form (Locke and Brawley, 2016).

ECEs increase the challenge of deciding to exercise and should be positively related to and interact with variables known to discourage exercise (e.g., exercise avoidance, Locke and Berry, 2021; Blumenkrans and Locke, 2024). They should also be negatively related to variables that encourage exercise (e.g., self-regulatory efficacy). Our results bore this out, as both forms were related to variables known to be important in the regulation of exercise in the theoretically expected directions. The original E-CEQ has been previously shown to be positively related to self-reported MVPA and self-regulatory efficacy (Locke and Brawley, 2016, 2017; Yuan et al., 2025). The present study supported these findings and built on them by demonstrating relationships to other social cognitive variables: namely motivation, physical activity enjoyment, and self-compassion. The growing number of associations between ECEs with established determinants of physical activity (e.g., motivation, Hankonen et al., 2017; self-efficacy, Williams and Rhodes, 2016; self-compassion, Phillips and Hine, 2021) strengthen the notion that ECEs may also be a physical activity determinant.

This study was the first to examine the relationship between ECEs and symptoms of anxiety and depression. Beck’s (1976) original definition of cognitive errors suggested that they are biased perceptions that can exacerbate depressive and anxious symptomatology. Lefebvre’s (1981) CEQ was validated as a measure of cognitive errors specific to depression and observed a large correlation between the CEQ and depressive symptoms (r = .61, p < .001). The ECE-Q was developed as a measure of biased perceptions about exercise barriers which may occur regardless of mental health status. We observed moderate-sized correlations (.35 ≤ r ≤ .40) between both versions of the E-CEQ and symptoms of depression and anxiety. These correlations were much smaller than correlations with depression and anxiety observed for the depression-specific CEQ measure. We believe this could suggest that both E-CEQ measures operationalize biased exercise perceptions that are not explicitly linked to the presence of symptoms of mental health. However, it is not surprising that both E-CEQ measures correlate with mental health variables. Individuals who report depressive or anxious symptoms will be more likely to make cognitive errors across a variety of real-world situations (e.g., social relationships, job performance, and perceived successes and failures), including exercise participation (Shickel et al., 2020). For this reason, individuals reporting depressive or anxious symptoms may be more likely to report ECEs; however, this does not preclude asymptomatic individuals from reporting ECEs.

Some limitations should be taken into account when interpreting the findings. First, the reliance on cross-sectional data precludes prospective analyses that would provide informative validity evidence (i.e., test-retest reliability, examining predictive utility across time). Future research should examine whether the ECEQ-SF yields similar associations as the full E-CEQ over time. Two of the three samples were comprised predominantly of younger adults. While dataset 2 was comprised of few young adults and may be more broadly generalizable to the adult population, all participants were at high risk for developing type 2 diabetes. Further examination is required to better generalize the utility of the ECEQ-SF to general adult and older adult samples. Future research should also seek to investigate gender differences in the latent scores of the ECEQ-SF. Qualitative approaches could be used to determine whether gender differences are attributable to the underlying cognitive error factors or the barriers being operationalized. Novel approaches, like extended alignment, could also be used to flexibly address scalar invariance (Marsh et al., 2009).

Through this study, we developed a 7-item short-form of the E-CEQ that demonstrated acceptable psychometric properties. The ECEQ-SF should be interpreted as capturing the extent to which individuals view their exercise engagement through a cognitively errored lens. Findings also suggested that the ECEQ-SF had a unidimensional factor structure that did not vary based on age or gender. Cross-sectional relationships demonstrated the convergent utility of the short form; however, future prospective and experimental research is needed to better understand these relationships. The short form’s intended use should be for examining individuals’ propensity to view exercise through a negative lens rather than for identifying the influence of specific cognitive errors.

Supplemental Material

Supplemental Material - Validation of the exercise-related cognitive errors questionnaire short form

Supplemental Material for Validation of the exercise-related cognitive errors questionnaire short form by Sean R Locke, James Sessford, Mary E Jung in Health Psychology Open

Footnotes

ORCID iDs

Sean R Locke

Mary E Jung

Ethical considerations

This study was approved by Brock University’s Health Science Research Ethics Board (#21-031).

Consent to participate

All participants provided informed consent to participate.

Consent for publication

Consent to publish this paper has been provided by all authors and research participants, as outlines in the consent form.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by an Insight Development Grant from the Social Sciences and Humanities Research Council of Canada to the corresponding author (#430-2018-00671).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data will be made available upon reasonable request to the corresponding author.*

Supplemental Material

Supplemental material for this article is available online.

References

Ajzen

Fishbein

(2005) The influence of attitudes on behavior. In: Albarracín

Johnson

Zanna

(eds) The Handbook of Attitudes. Mahwah, NJ: Erlbaum, pp. 173–221.

Akram

Barclay

Milkins

, et al. (2023) Sleep-related attentional and interpretive-bias in insomnia: a systematic review and meta-analysis. Sleep Medicine Reviews 67: 101713.

Amireault

Godin

(2012) Physical activity guidelines for health: how valid are the godin-shephard leisure-time physical activity Questionnaire and the international physical activity questionnaire-short form? Poster Session Presented at the Annual Meeting of the International Society for Behavioural Nutrition and Physical Activity. ISBNPA.

Anderson

Emery

(2014) Irrational health beliefs predict adherence to cardiac rehabilitation: a pilot study. Health Psychology: Official Journal of the Division of Health Psychology, American Psychological Association 33(12): 1614–1617.

Bandura

(1997) Self-efficacy: The Exercise of Control. New York: Freeman.

Bandura

(2006) Guide for constructing self-efficacy scales. Self-efficacy beliefs of adolescents 5(1): 307–337.

Beck

(1976) Cognitive Therapy and the Emotional Disorders. New York: International Universities Press.

Blumenkrans

Locke

(2024) Predicting physical activity for people with multiple sclerosis: the role of exercise-related cognitive errors. Rehabilitation Psychology 69(1): 29–35.

Bourne

Little

Beauchamp

, et al. (2019) Brief exercise counseling and high-intensity interval training on physical activity adherence and Cardiometabolic Health in individuals at risk of type 2 diabetes: protocol for a randomized controlled trial. JMIR research protocols 8(3): e11226.

10.

Bussey

Bandura

(1999) Social cognitive theory of gender development and differentiation. Psychological Review 106: 676–713.

11.

Byrne

(2010) Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming (2nd edition). New York: Taylor and Frances Group.

12.

Casper

Bocarro

Kanters

, et al. (2011) Measurement properties of constraints to sport participation: a psychometric examination with adolescents. Leisure Sciences 33(2): 127–146.

13.

Cerbara

Ciancimino

Tintori

(2022) Are we still a sexist society? Primary socialisation and adherence to gender roles in childhood. International Journal of Environmental Research and Public Health 19(6): 3408.

14.

Chen

(2007) Sensitivity of goodness of it indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal 14: 464–504.

15.

Clark

Watson

(2019) Constructing validity: new developments in creating objective measuring instruments. Psychological Assessment 31(12): 1412–1427.

16.

Colley

Garriguet

Janssen

, et al. (2011) Physical Activity of Canadian Adults: Accelerometer Results from the 2007 to 2009 Canadian Health Measures Survey. Ottawa: Statistics Canada, pp. 7–14.

17.

Deci

Ryan

(2008) Self-determination theory: a macrotheory of human motivation, development, and health. Canadian Psychology / Psychologie Canadienne 49(3): 182–185.

18.

Drapeau

Perry

(2010) The Cognitive Errors Rating Scales. (3rd edition). Quebec: McGill University.

19.

Edwards

Roberts

Sandercock

, et al. (2004) Follow-up by mail in clinical trials: does questionnaire length matter? Controlled Clinical Trials 25(1): 31–52.

20.

Ekkekakis

Zenko

(2016) Escape from cognitivism: exercise as hedonic experience. In: Raab

Wylleman

Seiler

, et al. (eds) Sport and Exercise Psychology Research from Theory to Practice. London: Academic, pp. 389–414.

21.

Glasgow

(2008) Health Behavior Constructs: Theory, Measurement, & Research. Retrieved from National Cancer Institute. https://cancercontrol.cancer.gov/brp/research/constructs/perceived-barriers

22.

Godin

Shephard

(1985) A simple method to assess exercise behavior in the community. Canadian Journal of Applied Sport Sciences 10(3): 141–146.

23.

Graham

(2009) Missing data analysis: Making it work in the real world. Annual review of psychology. 2009 Jan 10;60(1):549–76.

24.

Hankonen

Heino

Kujala

, et al. (2017) What explains the socioeconomic status gap in activity? Educational differences in determinants of physical activity and screentime. BMC Public Health 17: 1–15.

25.

Harris

Spiegelhalder

Espie

, et al. (2015) Sleep-related attentional bias in insomnia: a state-of-the-science review. Clinical Psychology Review 42: 16–27.

26.

Herazo-Beltrán

Pinillos

Vidarte

, et al. (2017) Predictors of perceived barriers to physical activity in the general adult population: a cross-sectional study. Brazilian Journal of Physical Therapy 21(1): 44–50.

27.

Hoare

Stavreski

Jennings

, et al. (2017) Exploring motivation and barriers to physical activity among active and inactive Australian adults. Sports 5(3): 47.

28.

Bentler

(1999) Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal 6(1): 1–55.

29.

Jakatdar

Cash

Engle

(2006) Body-image thought processes: the development and initial validation of the assessment of body-image cognitive distortions. Body Image 3(4): 325–333.

30.

Kendzierski

DeCarlo

(1991) Physical activity enjoyment scale: two validation studies. Journal of Sport & Exercise Psychology 13(1): 50–64.

31.

Keogh

(2022) Sex and gender differences in pain: past, present, and future. Pain 163(S1): S108–S116.

32.

Keogh

Attridge

Walsh

, et al. (2021) Attentional biases towards body expressions of pain in men and women. The Journal of Pain 22(12): 1696–1708.

33.

Koehn

Amirabdollahian

(2021) Reliability, validity, and gender invariance of the exercise benefits/barriers scale: an emerging evidence for a more concise research tool. International Journal of Environmental Research and Public Health 18(7): 3516.

34.

Lefebvre

(1981) Cognitive distortion and cognitive errors in depressed psychiatric and low back pain patients. Journal of Consulting and Clinical Psychology 49(4): 517–525.

35.

Liu

Chung

Chen

(2014) Constraints of recreational sport participation: measurement invariance and latent mean differences across sex and physical activity status. Perceptual and Motor Skills 119(2): 363–376.

36.

Locke

Berry

(2021) Examining the relationship between exercise-related cognitive errors, exercise schema, and implicit associations. Journal of Sport & Exercise Psychology 43(4): 345–352.

37.

Locke

Brawley

(2016) Development and initial validity of the exercise-related cognitive errors questionnaire. Psychology of Sport and Exercise 23: 82–89.

38.

Locke

Brawley

(2017) Perceptions of exercise consistency: relation to exercise-related cognitive errors and cognitions. Journal of Health Psychology 22(5): 684–694.

39.

Locke

McKay

Jung

(2019) “I’m just too busy to exercise”: reframing the negative thoughts associated with exercise-related cognitive errors. Psychology of Sport and Exercise 43: 279–287.

40.

Louw

Van Biljon

Mugandani

(2012) Exercise motivation and barriers among men and women of different age groups psychology. African Journal for Physical, Health Education, Recreation and Dance 18(41): 759–768.

41.

Markland

Tobin

(2004) A modification of the behavioural regulation in exercise questionnaire to include an assessment of amotivation. Journal of Sport & Exercise Psychology 26: 191–196.

42.

Marsh

Ellis

Parada

, et al. (2005) A short version of the Self Description Questionnaire II: operationalizing criteria for short-form evaluation with new applications of confirmatory factor analyses. Psychological Assessment 17(1): 81–102.

43.

Marsh

Muthén

Asparouhov

, et al. (2009) Exploratory structural equation modeling, integrating CFA and EFA: Application to students' evaluations of university teaching. Structural equation modeling: A multidisciplinary journal. 2009 Jul 1416;(3):439–76.

44.

Milman

Drapeau

(2012) Cognitive errors in cognitive behavioural therapy: a survey of researchers and practitioners and an assessment of the face validity of the cognitive error scale. Journal of Cognitive and Behavioral Psychotherapies 12(2): 125–138.

45.

Muthén LK and Muthén BO (1998-2012) Mplus User’s Guide. Seventh Edition. Los Angeles, CA: Muthén

& Muthén Neff

KD.

The development and validation of a scale to measure self-compassion. Self and identity. 2003 Jul 12;(3):223–50.

46.

Nakash

Hutton

Jørstad-Stein

, et al. (2006) Maximising response to postal questionnaires–a systematic review of randomised trials in health research. BMC Medical Research Methodology 6(1): 1–9.

47.

Neff

(2023) Self-compassion: theory, method, research, and intervention. Annual Review of Psychology 74(1): 193–218.

48.

Neff

Tóth-Király

(2020) Self-compassion scale (SCS). In: Handbook of Assessment in Mindfulness. Springer.

49.

Paslakis

Scholz-Hehn

Sommer

, et al. (2021) Implicit bias to food and body cues in eating disorders: a systematic review. Eating and Weight Disorders 26: 1303–1321.

50.

Patay

Patton

Parker

, et al. (2015) Understanding motivators and barriers to physical activity. The Physical Educator 72(3): 496.

51.

Peer

Brandimarte

Samat

, et al. (2017) Beyond the Turk: alternative platforms for crowdsourcing behavioral research. Journal of Experimental Social Psychology 70: 153–163.

52.

Phillips

Hine

(2021) Self-compassion, physical health, and health behaviour: a meta-analysis. Health Psychology Review 15(1): 113–139.

53.

Rinck

Wiers

Becker

, et al. (2018) Relapse prevention in abstinent alcoholics by cognitive bias modification: Clinical effects of combining approach bias modification and attention bias modification. American Psychological Association; 2018 Dec.

54.

Rogelberg

Stanton

(2007) Introduction understanding and dealing with organizational survey nonresponse. Organizational Research Methods 10(2): 195–209.

55.

Schwarzer

(2008) Modeling health behaviour change: how to predict and modify the adoption and maintenance of health behaviors. Applied Psychology 57(1): 1–29.

56.

Shickel

Siegel

Heesacker

, et al. (2020) Automatic detection and classification of cognitive distortions in mental health text. In: 2020 IEEE 20th International Conference on Bioinformatics and Bioengineering (BIBE). IEEE, pp. 275–280.

57.

Simpson

Semenchuk

Strachan

(2021) Put MY mask on first: mothers’ reactions to prioritizing health behaviours as a function of self-compassion and fear of self-compassion. Journal of Health Psychology 27(5): 1–8.

58.

Smith

McCarthy

Anderson

(2000) On the sins of short-form development. Psychological Assessment 12(1): 102–111.

59.

Soto

John

(2017) Short and extra-short forms of the big five inventory–2: the BFI-2-S and BFI-2-XS. Journal of Research in Personality 68: 69–81.

60.

Spence

Moss-Morris

Chalder

(2005) The Behavioural Responses to Illness Questionnaire (BRIQ): a new predictive measure of medically unexplained symptoms following acute infection. Psychological Medicine 35: 583–593.

61.

Spitzer

Kroenke

, et al. (1999) Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study. JAMA 282(18): 1737–1744.

62.

Spitzer

Kroenke

Williams

, et al. (2006) A brief measure for assessing generalized anxiety disorder: the GAD-7. Archives of Internal Medicine 166(10): 1092–1097.

63.

Streiner

Norman

Cairney

(2015) Health Measurement Scales – A Practical Guide to Their Development and Use. (5th edition). Oxford: Oxford University Press.

64.

Sturm

Cohen

(2019) Peer reviewed: free time and physical activity among Americans 15 Years or older: cross-sectional analysis of the American time use survey. Preventing Chronic Disease 16(E133): 1–8.

65.

Williams

Rhodes

(2016) The confounded self-efficacy construct: conceptual analysis and recommendations for future research. Health Psychology Review 10(2): 113–128.

66.

Yuan

Kuang

, et al. (2025) Relationship between trait mindfulness and physical activity among emerging adults: the mediating roles of exercise-related cognitive errors and trait self-control. Complementary Therapies in Clinical Practice 59: 101953.

67.

Ziegler

Kemper

Kruyen

(2014) Short scales–Five misunderstandings and ways to overcome them. Journal of Individual Differences 34(4): 2151–2299.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.17 MB