Development and validation of a scale for the tendency to exercise in response to mood, eating, and body image cues: the Reactive Exercise Scale (RES)

Earlier work on engaging in physical exercise when experiencing negative affect demonstrated robust associations with eating disorder (ED) behaviors and attitudes; however, measurement of the behavior was primitive, relying on one yes/no question that cannot capture much variability. We report on the development of a self-report measure, the Reactive Exercise Scale (RES), that disentangles the tendency to engage in exercise in response to negative mood cues from the tendency to engage in exercise in response to eating and body image cues, which themselves may be associated with negative mood. The measure also assesses exercising in response to positive mood cues. Exploratory factor analysis (EFA) guided item and factor selection. Confirmatory factor analysis (CFA) in an independent sample tested a 3-factor solution—exercising in response to negative mood cues, eating and body image cues, and positive mood cues. Correlations with exercise attitudes, eating disorder and body image attitudes, mood, and personality were used to evaluate construct validity. Results supported the 3-factor structure and indicated that exercising in response to negative mood cues may not uniquely relate to most aspects of ED psychopathology when accounting for eating and body image cues, which themselves are associated with negative mood. The RES captures the tendency to exercise in response to negative mood, positive mood, and eating and body image cues. Together, these constructs allow researchers to examine the unique relations of negative mood cued exercise with ED constructs, while accounting for appearance-related motives for which exercise may also be used. Level III: evidence obtained from well-designed cohort or case–control analytic studies.


Introduction
Exercise has mood-regulating and stress-reducing effects [10,58,59]. It is associated with increases in positive affect and decreases in negative affect [5,23], including reducing anxiety [2,49]. Exercise interventions have beneficial physical and psychological effects in individuals with psychiatric disorders [47,55]. Among individuals with eating disorders (EDs), although the primary motive of exercise is to control shape and weight [15], exercise is often reportedly used to regulate mood [39], and evidence supports pre-to postexercise decreases in negative emotion in individuals with EDs [19,41]. Thus, the mood-regulating effects of exercise can be understood as one of its benefits, however, exercise behavior as it occurs among individuals with EDs has been the subject of scrutiny for its possible role as a problematic behavior targeted for reduction in treatment (e.g., [12,36,37,45]).
In fact, exercising in response to negative affect, which may reflect using exercise to regulate emotions, is both common and robustly associated with ED psychopathology. For instance, 58% of college students reported having ever engaged in exercise because they were experiencing negative mood, and these students were much more likely to endorse current ED thoughts and behaviors [16]. Distinguishing participants by whether they ever exercised in response 1 3 to negative mood resulted in large effects on measures of ED behaviors and psychopathology, in some cases being the only unique predictor of these outcomes when included among existing exercise dimensions with known relations to ED constructs (e.g., obligatory exercise). Importantly, in the cited study, exercise in response to negative mood was assessed using a single dichotomous item, precluding thorough analysis of the relations between ED constructs and the dimension underlying this dichotomous item. The goal of the studies reported here was to develop a measure of this dimension-the tendency to exercise in response to negative mood cues. In addition, we aimed to include other similar dimensions that might be important for clarifying the relations between the tendency to exercise in response to negative mood cues and ED constructs.
Most concepts of exercise in association with ED psychopathology have focused on the extent to which individuals feel compelled or obligated to engage in exercise, and a variety of terms have been used to describe these aspects of exercise [38]. For instance, obligatory exercise involves experiencing negative emotion, such as feelings of guilt, depression, and anxiety, as a consequence of not exercising [39]. Exercising is understood to alleviate or prevent such negative emotional consequences among individuals who engage in obligatory exercise, indicating that it may also function to regulate emotion [54,61]. Obligatory exercise is associated with a number of ED variables, illustrating its importance as a way of conceptualizing problematic exercise in EDs. Obligatory exercise is undertaken more frequently and intensely than non-obligatory exercise, and individuals who engage in obligatory exercise spend more time thinking about exercise [52]. A systematic review concluded that obligatory exercise is associated with heightened eating and weight psychopathology [39]. In fact, obligatory exercise was reported by 20-80% of individuals with EDs [50] and was associated with poorer treatment outcome, illness chronicity, and severity of psychopathology [15,18]. Importantly, why individuals' emotions are related to their exercise behavior has not been central to investigations of exercise and ED psychopathology.
Individuals may desire to exercise, and feel guilty if they do not exercise, for a variety of reasons, including reasons related to their eating behaviors and body shape and weight, as well as reasons unrelated to eating, shape, and weight (e.g., athletic performance) [30]. However, because ED behaviors and body shape and weight concerns are themselves associated with negative emotion (e.g., [27,35]), investigations of the relations between negative emotion, exercise, and ED psychopathology should account for eating behavior-and body image-driven exercise. In the absence of parsing these effects, evidence might indicate relations between negative emotion, exercise, and ED psychopathology that are better explained by the eating and body image exercise motives held by individuals with ED psychopathology who exercise, which themselves are associated with negative mood. Thus, the goal of the following studies was to develop a measure-the Reactive Exercise Scale(RES)-for assessing the dimension of negative mood cued exercise that also assesses eating cued and body image cued exercise so that the unique relations of negative mood cued exercise and ED psychopathology can be examined. In addition, we aimed to assess positive mood cued exercise, due to the increases in positive mood associated with exercise (e.g., [44,53]) and the increased likelihood of exercising when experiencing positive life events that might elicit positive mood [23]. Including positive mood allowed us to test whether the valence of mood leading to exercise is specific to the risk of ED psychopathology (i.e., to establish whether general emotion reactivity or specific negative emotion reactivity carries risk for ED psychopathology) and to establish the construct validity of the scale. Exercise might be obligatory for a number of reasons, and those reasons may not all be related to ED psychopathology equally. Thus, we devised the RES to capture three domains that might cue such exercise. This work has the potential to support research on the emotion regulatory aspects of exercise and the associations of such exercise with ED psychopathology, clarifying the extent to which exercise itself may be a treatment target and how it might best be altered to support adaptive behavior.

Study 1
In Study 1, we developed an initial item pool and conducted an exploratory factor analysis (EFA) to gather preliminary evidence of the convergent and discriminant validity of the RES factors in an undergraduate student sample. We wrote items intending to characterize tendencies to engage in exercise in response to certain distinct moods, thoughts, and perceptual cues. These included negative mood cued exercise, eating cued exercise, body image cued exercise, and positive mood cued exercise. Broadly, we hypothesized that: (a) negative mood cued exercise would be positively associated with general psychopathology and emotion regulation, but not with other aspects of exercise (e.g., fitness and health motivations); (b) eating cued and body image cued exercise would be positively associated with ED psychopathology, but not with aspects of emotion regulation; and (c) positive mood cued exercise would be negatively associated with general psychopathology, and not associated with emotion regulation.

Procedure
The local institutional review board approved this study. Undergraduate students from a Northern Plains university 1 3 participated in an online study (hosted by Qualtrics) for credit toward course requirements that was advertised as a survey study of eating and exercise. After providing informed consent, participants completed a series of selfreport measures.

Reactive Exercise Scale (RES)
The initial version of the RES contained 42 items rated on a visual analog scale, with endpoints anchored by "Not at all" and "Extremely". The instructions asked participants to "Please read each of the following statements. Using the scale provided, select the point on the line to indicate how likely it is that you would exercise given the situation described in the statement." Items are listed in Table 1. We wrote the initial RES items using a top-down approach to capture four domains that might cue exercise behavior. They were: (a) exercising in response to negative mood cues; (b) exercising in response to eating cues; (c) exercising in response to body image cues; and (d) exercising in response to positive mood cues. Items were phrased generally to capture cues that we hypothesized would prompt exercise in each domain, given our focus on reactivity to such cues. We used literature on the constructs of positive and negative affect and ED psychopathology to inform our choice of wording and selection of cues.

Difficulties in emotion regulation scale (DERS)
The DERS [25] is a 36-item self-report measure of dimensions of emotion regulation, measured on six facets: nonacceptance of emotional responses (α = 0.92), difficulty engaging in goal-directed behavior (α = 0.89); impulse control difficulties (α = 0.86); lack of emotional awareness (α = 0.84); limited access to emotion regulation strategies (α = 0.91); and lack of emotional clarity (α = 0.77), and a total score (α = 0.94). Each item is rated on a scale of 1 (Almost never) to 5 (Almost always), with higher scores indicative of greater difficulties in emotion regulation.

Positive and negative affect schedule (PANAS-X)
The PANAS-X [60] contains measures of negative and positive affect in addition to several more specific facets. We included the two 10-item affect scales to measure positive affect (α = 0.89) and negative affect (α = 0.89) and used the "in general" timeframe in the instructions. We also included 13 additional items that allowed for the measurement of sadness (α = 0.90), guilt (α = 0.91), and fatigue (α = 0.88). Each item is rated on a 5-point scale ranging from 1 (Very slightly or not at all) to 5 (Extremely). Higher scores are indicative of more intense affect.

International personality item pool (IPIP)
The 142 items from the IPIP [24] approximate Cloninger et al. [11] Temperament and Character Inventory dimensions of novelty seeking (α = 0.82), harm avoidance (α = 0.80), and reward dependence (α = 0.83) and Carver and White's [8] Behavioral Inhibition and Activation Scales. The behavioral inhibition system (BIS) corresponds to motivation to avoid aversive outcomes, and the behavioral activation system (BAS) corresponds to motivation to approach goal-oriented outcomes. This measure contains four subscales, one for measuring the BIS dimension (α = 0.82), and three meas-

Obligatory exercise questionnaire (OEQ)
The OEQ Pasman and Thomspon [43] is a 20-item selfreport questionnaire consisting of statements about the extent to which an individual feels obligated or compelled to exercise. Statements are rated on a 4-point scale ranging from 1 (Never) to 4 (Always), with higher total summed scores indicating more obligatory exercise (α = 0.90).

Statistical analyses
We conducted an EFA using principal axis factoring with varimax rotation in SPSS (IBM 2019) to maximize variance in factor loadings and aid in achieving an interpretable solution in this early stage despite the likelihood that factors would be correlated (e.g., [31]) to identify dimensions present in the RES. Pairwise exclusion was used for missing values, which represented 6.3% of observations. We visually examined the scree plot ( Fig. 1) and performed a parallel analysis [28] to determine the optimal number of factors to extract. Items that loaded above 0.60 with cross-loadings below 0.32 were automatically retained [13,56]. We used partial correlations-controlling for RES subscales, sex, and BMI, due to the moderate correlations between RES scales and the known relations of sex and BMI to the measured phenomena-to assess convergent and discriminant validity, using p < 0.05 as the statistical significance threshold to maximize detected relationships. We aimed to recruit 300 participants for EFA, as recommended by Tabachnick and Fidell [56].

Results
Sample size per RES item ranged from 269 to 296. The Kaiser-Meyer-Olkin measure of sampling adequacy was 0.959, above the recommended value of 0.6 [13], and Bartlett's test of sphericity was significant (χ 2 = 12,359.51; df = 861; p < 0.001), indicating that the data were appropriate for an EFA. Examination of the scree plot, initial eigenvalues, and results of a parallel analysis all indicated that three factors should be extracted (Fig. 1). Together, these three factors explained 67.09% of the inter-item variance. We retained five items (items 4, 8, 9, 14, and 22; Table 1) despite crossloadings greater than 0.32 [56], because we judged their content to be central to the construct being measured or important for covering the breadth of the construct (i.e., they were critical for content validity). In total, we retained 25 of the 42 items.
In Table 2, we provide means, standard deviations, bivariate Pearson correlations, and internal consistencies for the three extracted factors. The first extracted factor is represented by 12 items and indicates the tendency to engage in exercise in response to negative mood cues. The second extracted factor is represented by 10 items and indicates the tendency to engage in exercise in response to eating and body image cues. The third extracted factor is represented by three items and indicates the tendency to engage in exercise in response to positive mood cues. Because the eating and body image cues resided on the same factor, we also examined a four-factor solution. However, extracting a fourth factor did not separate eating items from the body image items. Instead, there was no discernible pattern to the item-factor configuration, further indicating that the three-factor solution fit better.
In Table 3, we provide preliminary evidence of convergent and discriminant validity. The negative mood factor was positively associated with the OEQ total score, REI mood subscale, DERS impulse and total scores, and PANAS sadness facet. It was negatively associated with the REI weight control subscale and. not associated with behavioral activation, positive affect, novelty seeking,  reward dependence, or fitness and health. The eating and body image factor was positively associated with REI weight control, BIS/BAS anxiety, PANAS negative affect and guilt facet, and IPIP novelty seeking. There were no negative assocations. It was not associated with behavioral activation, positive affect, novelty seeking, reward dependence, and REI fitness and health motives. Finally, the positive mood factor was positively associated with OEQ obligatory exercise, PANAS positive affect, and BAS drive and was the only factor associated with the REI fitness and health subscales. The positive mood factor was negatively associated with DERS impulse and total scores; PANAS negative affect, guilt, and sadness; and IPIP harm avoidance. It was not associated with behavioral inhibition, novelty seeking, reward dependence, mood, or weight control.

Discussion
The results of Study 1 preliminarily indicate that a 25-item version of the RES is composed of three dimensions representing the tendencies to exercise in response to negative mood cues, eating and body image cues, and positive mood cues. Each dimension's distinctiveness is indicated by unique patterns of associations with other variables of interest, demonstrating initial convergent and discriminant validity. These scales appear to have excellent internal consistency reliability, indicating low idiosyncratic error. Despite using an orthogonal rotation, factors were moderately related, which is common [42].

Study 2
In Study 2, we tested the fit of the 3-factor solution from Study 1 in an independent undergraduate student sample. We included measures of ED psychopathology and other clinical syndromes (i.e., depression, self-injury, anxiety, and alcohol use). We hypothesized that, when controlling for sex, BMI, and the two other RES subscales, (a) negative mood-cued exercise would be positively associated with measures of general psychopathology but not ED psychopathology; (b) eating and body image-cued exercise would be positively associated with ED psychopathology; and (c) positive mood-cued exercise would be negatively associated with all psychopathology.

Procedure
The local institutional review board approved this study. Undergraduate students from a Northern Plains university participated in an online study (hosted on Qualtrics) for credit toward course requirements advertised as a survey study about exercise, eating behavior, and mental health. After providing informed consent, participants completed a variety of self-report measures.

Participants
Participants were 300 undergraduates, with a minimum age of 18 years. Their mean (SD) age was 19

RES
The 25 RES items retained from Study 1 comprised the version of the RES used in this study.

Eating disorder examination-questionnaire (EDE-Q)
The EDE-Q [20] is a 28-item self-report questionnaire that assesses the frequency of eating behaviors over the prior 28

Eating pathology symptoms inventory (ESPI)
The

Clinical impairment assessment (CIA)
The CIA [6] is a 16-item self-report measure assessing psychosocial impairment due to ED symptoms over the prior 28 days using a 4-point scale from 0 (Not at all) to 4 (A lot). Participants rate the extent to which eating problems and weight and shape concerns have caused them impairment in various aspects of their life, with higher scores indicative of greater impairment. Cronbach's alpha was 0.96.

Center for epidemiologic studies depression scale-revised (CESD-R)
The CESD-R [17] is a 20-item self-report measure of depression, based on the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (APA, 2000) criteria. Items assess the frequency of depression symptoms over the prior week, ranging from 0 (Not at all) to 4 (A lot). Scores on the CESD-R range from 0 to 60, with higher scores indicating greater presence of depressive symptoms. Cronbach's alpha was 0.95.

Alcohol use disorders identification test (AUDIT)
The AUDIT [48] is a 10-item measure assessing alcohol consumption, drinking behaviors, and alcohol-related problems. Items measure typical frequency, typical quantity per occasion, and heavy episodic drinking (six or more drinks per occasion) frequency. Participants indicate the frequency of each item, with response options coded from 0 to 4 and higher scores indicating more problems with alcohol (α = 0.81).

State-trait inventory for cognitive and somatic anxiety (STICSA)
The STICSA [46] is a self-report measure assessing cognitive and somatic symptoms of anxiety as pertaining to one's current feelings of anxiety (state) and one's general, overall feelings of anxiety (trait). Participants rate each item on a 4-point scale, ranging from 1 (Not at all) to 4 (Very much so). Cronbach's alpha was 0.94 for the total trait measure, which is what we used here.

Brief experiential avoidance questionnaire (BEAQ)
The BEAQ [22] is a brief measure of the Multidimensional Experiential Avoidance Questionnaire (MEAQ). It is a 15-item self-report questionnaire assessing experiential avoidance. Each item is scored on a 6-point scale ranging from 1 (Strongly disagree) to 6 (Strongly agree). Higher scores are indicative of greater levels of experiential avoidance. Cronbach's alpha was 0.87.

Inventory of statements about self-injury (ISAS)
The ISAS [32] assesses lifetime frequency of non-suicidal self-injurious (NSSI) behaviors performed "intentionally (i.e., on purpose) and without suicidal intent (i.e., not for suicidal reasons)." The behaviors assessed include banging/ hitting self, biting, burning, carving, cutting, wound picking, needle-sticking, pinching, hair-pulling, rubbing skin against rough surfaces, severe scratching, and swallowing chemicals. Participants estimated the number of times in their lives they performed each behavior, which we dichotomized as ever present versus absent.

Statistical analyses
We conducted a confirmatory factor analysis (CFA) using Mplus (Version 8.3; [40]) to test the fit of the 3-factor solution using the MLR estimator. We assessed fit using the root mean square error of approximation (RMSEA, < 0.08 is fair; [7]), comparative fit index (CFI, > 0.95 is excellent; [29]), Tucker-Lewis index (TLI, > 0.90 is acceptable; [3]), and the standardized root mean square residual (SRMR, < 0.06 is ideal; [29]). Kyriazos [34] summarized sample size recommendations for CFA as typically ranging from 10 to 20 participants per item. We recruited 300 particpants, which is in the lower end of this range, given our expectation of a three-factor structure. Missing data represented 0.2% of observations and were handled using full information maximum likelihood. We calculated partial correlations between the RES scales and other study measures while controlling for the other RES scales and participant sex and BMI. We conducted independent samples t-tests to examine whether scores on each subscale differed by sex. Finally, we conducted a multiple linear regression to account for EDE-Q Global Score, a broad measure of ED psychopathology, using the RES scales, sex, and BMI (due to the moderate correlations between RES scales and the known relations of sex and BMI to the measured phenomena) to examine which, if any, RES scales were uniquely associated with this important indicator of ED psychopathology while accounting for their shared variance.

Results
After freeing the covariances of similarly worded items, the solution fit well (χ 2 (263) = 485.33, p < 0.001; RMSEA = 0.053 [95%CI: 0.046, 0.061]; CFI = 0.95; TLI = 0.94; SRMR = 0.05). The factors indicate the tendency to exercise in response to negative mood cues, in response to eating and body image cues, and in response to positive mood cues (Fig. 2). Results of correlations between RES scales and other measures are displayed in Tables 3, 4. The results of the multiple linear regression on the EDE-Q Global Score using all three RES scales indicated that, while controlling for sex and BMI, only the exercising in response to eating and body image cues factor was uniquely associated with EDE-Q Global Score (B = 0.51, t = 7.83, p < 0.001).

The tendency to exercise in response to negative mood cues
Exercising in response to negative mood cues was positively associated with the EDE-Q item assessing driven exercise, the EPSI excessive exercise and muscle building subscales, and the MBSRQ fitness and health orientation subscales. It demonstrated no negative associations, and it was not associated with ED psychopathology subscales that do not assess exercise (e.g., dietary restraint, shape concern, restriction, impairment) or with any non-ED psychopathology measures.

The tendency to exercise in response to eating and body image cues
Exercising in response to eating and body image cues was positively associated with all EDE-Q subscales, driven exercise, and the presence of objectively large binge-eating episodes and purging; EPSI restriction, cognitive restraint, purging, negative attitudes toward obesity, body dissatisfaction, and excessive exercise subscales; ED-related impairment; MBSRQ appearance orientation; depression; and experiential avoidance. Scores on this dimension were negatively associated with body image satisfaction, appearance evaluation, and health evaluation. Finally, this dimension was not associated with binge eating measured by the EPSI, objectively large overeating episodes, muscle building, fitness evaluation and orientation, health orientation, and illness orientation, presence of self-injury, anxiety, and alcohol use.

The tendency to exercise in response to positive mood cues
Exercising in response to positive mood cues was positively associated with EPSI excessive exercise; and MBSRQ body satisfaction and appearance evaluation, health and fitness evaluation and orientation, and illness orientation. The positive correlation with illness orientation suggests that exercising in response to positive mood is associated with increased alertness about, or reactivity to, being sick. Exercising in response to positive mood cues was negatively associated  Table 1. Error covariance was freely estimated for similarly worded items with EPSI body dissatisfaction and BEAQ experiential avoidance. No significant correlations were found between the tendency to exercise in response to positive mood cues and ED psychopathology or anxiety, depression, alcohol use problems, self-injury, age, or BMI.

Study 2 discussion
The results of Study 2 provided support for the three-factor structure from Study 1, albeit with minor modifications to address correlated errors between similarly worded items. Convergent and discriminant evidence indicates that the negative mood cue subscale measures the tendency to engage in exercise in response to negative mood cues that are not specific to ED psychopathology, but are associated with other Values are partial correlations, except where otherwise indicated, controlling for sex, BMI, and the other RES scales themselves. BMI body mass index, EDE-Q eating disorder examination-questionnaire, OE objectively large overeating episodes, OBE objectively large binge-eating episodes, CIA clinical impairment assessment, EPSI eating pathology symptom inventory, MBSRQ multidimensional body-self relations questionnaire, CESD-R center for epidemiological studies depression scale-revised, AUDIT alcohol use disorders identification test, STICSA state trait inventory for cognitive and somatic anxiety, BEAQ brief experiential avoidance questionnaire a Odds Ratios: Behavior coded as 1 = present and 0 = absent b Controlling for other RES scales and sex only *** p < .001; ** p < .01; * p < .05 aspects of mental health, generally involving low mood and mood regulation. However, the negative mood factor did not demonstrate many associations with psychopathology, contrary to our prediction. Conversely, the eating and body image cues subscale appears to measure the tendency to engage in exercise that is related to ED psychopathology. Finally, the positive mood cue subscale appears to measure exercise that is generally associated with favorable mental health and not associated with psychopathology, ED or otherwise. Importantly, all three subscales demonstrate unique relations with the EPSI Excessive Exercise scale, and both the negative mood cue and the eating and body image cue subscales had unique relations with the EDE-Q question assessing driven exercise. Differences by sex indicated that male participants report stronger tendencies to exercise in response to mood cues but weaker tendencies to exercise in response to eating and body image cues than female participants, the latter of which is consistent with sex differences commonly observed on dimensions of ED psychopathology.

General discussion
Earlier work on the tendency to engage in exercise when experiencing negative mood demonstrated that it was robustly associated with ED behaviors and attitudes [16]. However, given the benefits of exercise for mood and the common overlap in motives for exercise in EDs, the extent to which exercise in response to negative mood may be pathological is unclear. Thus, we aimed to devise a measure, the RES, to disentangle exercising in response to negative mood cues from exercising in response to eating and body image cues, which themselves may be associated with negative mood. We also included a dimension of exercising in response to positive mood cues, in part to aid in establishing discriminative validity (i.e., that exercising in response to negative mood cues is distinct from exercising in response to positive mood cues), but also due to the known relations of exercise with positive mood. EFA guided item and factor selection, and CFA in an independent sample tested the three-factor solution fit: exercising in response to negative mood cues, exercising in response to eating and body image cues, and exercising in response to positive mood cues. We expected eating and body image cues to be distinct, but this was not supported. Exercising in response to negative mood cues was positively associated with difficulties with emotion regulation, sadness, and fitness and health orientations. It was also associated with driven exercise, excessive exercise, and muscle building. It was not associated with problems with alcohol, depression, self-injury, or anxiety, which was counter to our expectations. Thus, while the RES appears to capture the tendency to exercise in the presence of negative mood cues, and this tendency is stronger the more one has difficulty regulating emotion or experiences sadness, this tendency does not relate to frank psychopathology. Future research should examine whether distress tolerance helps explain this pattern of findings, as individuals who tend to exercise in response to negative mood cues may have higher than typical distress tolerance, protecting them from general psychopathology. Consistent with our hypotheses, exercising in response to eating and body image cues was positively associated with nearly every ED-related variable we included and other measures of psychopathology, and was negatively associated with health evaluation. As hypothesized, exercising in response to positive mood cues was negatively associated with measures of psychopathology and positively associated with health, fitness, and appearance evaluation. Overall, these results support the construct validity of the RES subscales and indicate that exercising in response to negative mood cues may not uniquely relate to ED psychopathology, or perhaps even other forms of psychopathology, when accounting for eating and body image cues, which themselves may be associated with negative mood.
Notably, both studies indicated the RES subscales are uniquely related to measures of obligatory exercise (i.e., the OEQ, EPSI Excessive Exercise, and less so with EDE-Q driven exercise). While this could be interpreted as the RES scales demonstrating a lack of discriminant validity with obligatory exercise, we think these findings reflect a strength of the RES. Specifically, individuals may desire to exercise, and feel guilty if they do not exercise, for a variety of reasons. Certainly, some of those reasons are part of ED psychopathology and rightly point to such exercise as problematic in its own right. However, individuals who adopt forms of exercise as a valued, generally health-promoting, non-professional activity may set goals for their progress and improvement to which they commit. Like other valued activities to which one commits, individuals may prioritize the activity, look forward to engaging in it, and be disappointed in general and perhaps even with themselves (i.e., feel guilty) when they fail to meet their own expectations. That disappointment may provide motivation for continued valued engagement. In fact, guilt may be understood as an adaptive emotional experience when the opportunity for reparative actions are available [14,57]. Thus, whether such a pattern of exercise is problematic requires a detailed understanding of individuals' values and commitments and the extent to which their pursuit of exercise-related goals interferes with them, however, the findings of these studies with the RES seem to make clear that it is specifically exercising in response to eating and body image cues that uniquely relates to most aspects of ED psychopathology. This important finding can help contextualize work on the emotion-regulation functions of exercise in EDs [33].

3
The RES is a promising new tool for testing relations between exercise and ED psychopathology and how exercise cued by mood is implicated in these disorders. Indeed, while the ED literature presently contains a variety of measures for exercise [26], the results of our studies indicate that the RES approaches characterizing the quality of exercise behavior uniquely, and this may be advantageous for future work. For instance, as research on the emotion regulation functions of exercise in ED increases (e.g., [33]), it will be important to disentangle exercise undertaken in response to negative mood from exercise undertaken in response to eating and body image cues, which itself may often be accompanied by negative mood, a task the RES appears able to accomplish.

Strengths and limits
We tested the validity of the RES in homogenous samples in terms of sex, age, race, and ethnicity, limiting generalizability to more diverse samples. Using undergraduate students in both RES studies also limits generalizability to other populations. We have not yet tested the properties of the RES in individuals with EDs, so caution should be exercised when generalizing to them. Our aim in developing the RES was to first ascertain reactive exercise in undergraduates with a range of ED psychopathology prior to recruiting a clinical sample, in which restriction of range on variables of interest is likely. Future research should aim to examine the RES in clinical samples of individuals with EDs, as well as individuals with other forms of psychopathology (e.g., depressive disorders), to examine factorial invariance. Future work should also attempt to measure exercising in response to positive moods associated with ED attitudes and behaviors (e.g., the tendency to exercise following successful restriction), which might be more likely among individuals with EDs. We did not examine the RES' test-retest reliability or stability. Future research should examine its test-retest reliability over a multi-week (e.g., 4-week) interval, as the constructs assessed by the RES are theorized to be stable over such a period. Examining how RES subscale scores change over longer periods (e.g., from adolescence to adulthood) is an important avenue for future research. Finally, participants may not encounter visual analog scales, such as the one used for the RES, as commonly as 4-7-point ordinal response scales; however, the ubiquitous use of online survey technology has led to the easy programming and scoring of these scales to address shortcomings of ordinal response scales (e.g., the tendency of participants to avoid extreme options on typical ordinal scales; [4]). We encourage interested readers to contact us about how to program this scale in Qualtrics, as the use of a visual analog scale makes paper and pencil scoring of this measure labor-intensive. Strengths of these studies include the use of two large, independent samples to establish the factor structure of the RES in separate exploratory and confirmatory steps, along with the inclusion of a large variety of variables to test convergent and discriminant relations.

What is already known on this subject?
Exercise behavior is often disrupted in individuals with EDs and itself may be an important ED behavior. The aspects of exercise that explain its relation to ED features, and what differentiates exercise as it occurs within and outside EDs, remains an important topic of study.

What this study adds?
This study contributes a novel measure of the tendency to engage in exercise in response to specific cues: negative emotion, eating/body image, and positive emotion. Together, these subscales may aid in identifying the specific components of exercise that relate to ED features from those that may occur both within and outside of EDs.
Funding These studies were not funded.

Data availability
The data that support the findings of this study are available from the corresponding author upon request.

Conflict of interest
The authors have not disclosed any competing interests.

Ethical approval
The authors have no conflicts of interest to disclose. The studies reported here were not funded. All participants provided informed consent to participate prior to beginning study procedures. The University of North Dakota institutional review board approved this research.