The role of affect in exercise adherence

Grounded in hedonic principles, ‘affectivism’ (i.e., the study of affect) is a line of thought postulating that people tend to engage in activities they consider pleasurable while avoiding those wherein they experience pain and displeasure [9, 10]. Accordingly, several studies have shown that affect and affective dependent variables (e.g., anticipated affective response to exercise; implicit attitude) demonstrate some predictive value to exercise adherence-related variables (e.g., habit, intention, frequency), and could advance current theoretical models of exercise behavior [11–14].

As an example, in the Affect and Health Behavior Framework (AHBF [15]) the affective response to exercise is highlighted as a determinant in the sustainability of such behavior. According to this framework, core affect (i.e., an elementary, non-reflective feeling, consciously available to the individual) is the central aspect of the affective response to exercise and has been suggested to be a reliable predictor of exercise adherence. To maintain ecological validity, the affective response is best measured during or immediately after the activity, since it can only be experienced in vivo [15, 16]. However, this assessment can be difficult to perform given the dynamic nature of particular forms of PA (e.g., assessing core affect during the execution of the bench press exercise). Notwithstanding, some advancement in the assessment of exercise-related affect has been made, but several methodological clarifications for proper use and interpretation are needed [17].

The affective response dynamics of a given exercise session can be evaluated through the measurement of two core dimensions of affect: affective valence (perceived pleasure-displeasure) and arousal (perceived activation) [16, 18]. For the purpose of measuring these constructs, the Feeling Scale (FS [19]) and the Felt Arousal Scale (FAS [20]) were developed and have been used extensively in exercise settings [17, 21]. Both scales can be used in conjunction and interpreted in relation to the circumplex model of affect [22] to map affective fluctuations throughout an exercise session [16, 18]. Information derived from the FS and FAS can facilitate adjustments during exercise sessions to promote a more pleasurable experience, and potentially contribute to sustainable exercise programs [23, 24].

Measuring core affect in resistance training

Exercise guidelines first recommended the measurement of affective responses (e.g., with the FS) to exercise over a decade ago [25]. Currently, the latest edition of the ACSM’s guidelines for exercise prescription and associated recommendations are still present in the form of ‘affect regulation’ (p. 455 [1]). However, there is persistent uncertainty regarding how to effectively assess and regulate affect in exercise prescription (i.e., operationally; how and when to evaluate, interpret, and make the necessary adjustments). This knowledge gap is particularly prevalent within RT research literature, and has been noted in recent works [17, 26].

One important limitation in RT studies concerns the timing of affect measurement (i.e., when [27]). In their review, Evmenenko and Teixeira [17] report that no apparent standardization of measurement timing exists in the RT research that plots core affect within the circumplex model of affect. In particular, the timing of assessment is of notable concern due to a possible ‘affective rebound’ phenomenon (i.e., improvement of the affective response after exercise termination) that has been well documented in aerobic activities [16], as well as emerging evidence from the limited studies on RT [28–30]. Inappropriate timing of assessment may lead to the measurement of confounding variables beyond core affect and promote the reporting of biased results. Thus, one critical primary issue to address is to clarify when to assess core affect during RT.

Additionally, the frequency of FS/FAS assessments required for an adequate interpretation of affective response to exercise is another concern that remains unaddressed within RT research [17]. As suggested by Haile et al. [31] as well as Zenko and Ladwig [32], measurements should be recorded at regular intervals to effectively assess the exerciser’s affective response, whilst simultaneously avoiding the burden of excessive assessment. It is proposed that this balance is influenced by several variables (e.g., exercise experience, type, health status), making the frequency of measurement an important aspect in the contextual application of these scales. However, no clear indication in the research literature or exercise guidelines has emerged regarding the frequency of assessment in RT; thus, the intention of promoting positive affect in this form of exercise is impeded. Given that the advancement on this frequency of application issue is heavily dependent on methodological clarifications regarding when to assess core affect, assessment standardization efforts must come forward to enhance further research endeavors.

It also warrants consideration that contextual factors such as exercise intensity and exercise mode (e.g., machines, free-weights, calisthenics), as well as individual factors (e.g., age, sex, exercise experience), may also be relevant for advancing understanding of core affect assessments in RT. This line of study would attend to the concerns raised by previous reports [28, 32, 33] that have highlighted the potential interaction of contextual and individual variables influencing affective responses, as well as evaluate the feasibility of using these scales across various conditions.

Present study

Given the need to reduce inconsistency and promote specific guidelines for affect assessment during RT activities, more research to improve methodological quality is warranted [17, 28, 29]. As such, the main objective of the present review was to analyze how the FS and/or the FAS have been used in RT to assess core affect. More specifically, we aimed to analyze how the FS and/or the FAS have been used in RT practices, and particularly, how the timing and frequency of assessment have been made, with the purpose of understanding these scale’s value and contextual feasibility for core affect measurement in RT. To this end, their feasibility, focused on the contextual and specific characteristics of this exercise type (intensity, expression, apparatus) and sociodemographic variables (age, sex, experience), in conjunction with the timing and frequency of assessment of both scales, will be explored.

Eligibility criteria

The present review applied the following inclusion criteria: (1) experimental, longitudinal, and cross-sectional studies; (2) published in a peer-reviewed journal or as gray literature until July 30 of 2022; (3) written in English; (4) utilizing the FS and/or the FAS in RT exercise; and (5) focused on apparently healthy individuals of any age. The exclusion criteria were as follows: (1) populations with mental disease; (2) body mass index > 34.9 Kg/m²; (3) mixed exercise programs (i.e., circuit training and similar exercise program structures); (4) instrument validation studies; and (5) review studies.

Information sources and search strategy

A wide search of the literature was conducted from February 1 of 2022, to July 30 of 2022 on the following databases: PubMed (host: MEDLINE): last search run July 2022; SportDISCUS (host: EBSCO): last search run July 2022; and PsycINFO (host: EBSCO): last search run July 2022. The PICOS strategy was applied, and the search was executed with the following entries in each individual database: (((physical AND (exercise OR activity)) AND (feeling scale OR felt arousal scale) AND ((resistance OR strength) AND training))). A sample of the PubMed search strategy can be found in S1 File.

Bibliographic references from related research and other sources were examined with the purpose of including more studies that potentially met the inclusion criteria (the last search was conducted on July 30 of 2022).

Selection process

Three independent reviewers (VB, DT, and FR) were involved in the article selection process. All reviewers were trained in the study procedure and disagreements were resolved in group discussion and consensus. At Level I, titles and abstracts of all identified records from the database search were manually screened, analyzed, and checked against eligibility criteria. Full-text publications of every study not eliminated in the previous screening were retrieved for complete review at Level II. At this stage, the authors read the full-text publications to guarantee that the inclusion criteria were met and no exclusion criteria were present. The complete search and screening process is illustrated in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study flow-chart.

https://doi.org/10.1371/journal.pone.0294529.g001

Data collection process and data items

Two reviewers (VB and DT) independently conducted the data collection process. For general description (Table 1), the following characteristics were extracted from the included studies: (1) bibliographic information (authors, year of publication, country of research); (2) study design; (3) sample size; (4) sample characteristics; (5) intervention; (6) measures; (7) statistical analyzes; and (8) outcomes of interest. A data extraction sheet was made in Excel to summarize all data of interest from the studies. For a summary of the main characteristics of interest (Table 2), the following data were collected: (1) sample size; (2) sex; (3) location; (4) age; (5) effect size and/or power calculation; (6) applied instruments; (7) affective response and arousal measurement prior training with FS/FAS; and (8) timing of measurement.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive characteristics of the studies and main outcomes.

https://doi.org/10.1371/journal.pone.0294529.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of studies’ and samples’ characteristics.

https://doi.org/10.1371/journal.pone.0294529.t002

Study risk of bias assessment

The risk of bias assessment was conducted with the recommended tools for this purpose by the Cochrane Collaboration (Risk of Bias 2 [35]). This instrument assesses the risk of bias in randomized controlled trials according to the following six domains: randomization process; deviation from intended interventions; missing outcome data; measurement of the outcome; selection of the reported results; and overall bias. Each domain was judged as ‘low risk’, ‘some concern’, or ‘high risk’. An extension of this tool for crossover trials was also used to assess the included studies with this specific design. The domains assessed are the same as its counterpart for randomized controlled trials, with the addition of the ‘bias arising from period and carryover effects’ domain. Lastly, the risk of bias in the quasi-experimental studies was evaluated with the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I [36]). With this tool, reviewers give a score of ‘low risk’; ‘moderate risk’; ‘serious risk’; ‘critical risk’; or ‘no information’ to each of the following domains: ‘confounding’; ‘selection of participants into the study’; ‘classification of interventions’; ‘deviations from intended intervention’; ‘missing data’; ‘measurement of outcomes’; and ‘selection of the reported outcomes’. Two reviewers (VB and FR) independently analyzed the included studies and disagreements were resolved by consensus. If disagreements persisted, a third reviewer (DT) was consulted to resolve the discrepancies. All reviewers were debriefed and instructed prior to the use of each risk of bias assessment tool. Lastly, the robvis online application [37] was used to create figures that present a more detailed overview of each included study’s risk of bias (S2–S4 Files).

Study selection

A total of 237 studies were identified during the database search for potential inclusion. After the removal of five duplicate records, 232 studies entered the screening process. Following a meticulous read-through of the title and abstract of every record, 206 were ultimately excluded for reasons described in Fig 1. Of the 26 resulting studies, two additional studies were excluded after full-text reviews. One study was excluded due to not following a traditional RT program [38], and another study was excluded due to only measuring affective response 15 minutes after the RT session [39]. Analysis of the bibliographical references of the selected studies revealed six additional studies that were identified to be potentially relevant. Of these studies, three met the inclusion criteria, and three were excluded [40–42], resulting in a final number of 27 studies, published or accepted for publication until July 31 of 2022, that proceeded to an in-depth analysis.

Study characteristics

A synthesis of the data collected from the 27 studies that comprise this review can be observed in Tables 1 and 2 (organized in alphabetic order according to the first author’s surname). All studies had an intervention/experimental design, including three randomized controlled trials, two randomized crossover trials, and 22 quasi-experimental studies. All studies used convenience methods of recruitment. Participants in the studies met the inclusion criteria set for this review, allowing for wider coverage of possible physical activity contexts and varying use of the FS and/or FAS in RT. In sum, the reviewed studies totaled a sample size (total N = 718); 17 studies were conducted with fewer than 30 participants (61%), eight studies had a sample between 30 and 50 individuals (31%), and two studies presented a sample size between 50 and 100 participants (8%).

Risk of bias in studies

In the three randomized controlled trials included in this review, the overall risk of bias was of ‘some concern’ for Chang and Etnier [47] due to possible deviations from intended interventions and in the selection of the reported results; the other two trials [51, 60] were deemed at ‘high risk’ of bias due to the randomization process (see S2 File). As for the crossover trials a similar trend was observed, with Tavares et al. [63] being classified as of ‘some concerns’ due to possible deviations from intended interventions and in the selection of the reported results; Richardson et al. [59] was deemed ‘high risk’ due to the randomization process (see S3 File). In the quasi-experimental studies, the overall bias was of ‘moderate risk’ for 15 studies [29, 30, 33, 45, 46, 48, 50, 52–58, 61] predominantly due to a possible bias in the measurement of outcomes, and selection of reported results. Seven studies were deemed ‘serious risk’ [26–28, 43, 44, 49, 62] also due to possible bias in the measurement of outcomes (see S4 File). Overall bias classification is presented in Table 1.

Results of individual studies

Contextual feasibility, timing, and frequency of assessment

Considering the importance of exercise intensity in affect research [15], the requirement of an instrument that adequately assesses core affect in RT is of paramount importance. In this regard, the FS and the FAS appear to address this demand. Although vastly different exercise intensities could be observed across the 27 studies included in this review, the FS and FAS successfully measured affective responses regardless of the level of intensity or research/training protocol (i.e., 1-RM, self-selection, or repetitions in reserve). Additionally, the scales have been utilized in RT with varying proximities to concentric failure and were able to measure core affect even when an exercise was performed to the point of failure. Concerning sex representation, both sexes are well represented with the majority of the studies showing no clear differences in the affective response. Both scales were also applied not only to adults but also to adolescents and the elderly, demonstrating no age-related limitations in their interpretation and application. Lastly, no issue was reported when evaluating these instruments in RT sessions using free weights and/or machines, indicating core affect can be assessed independent of variations in exercise equipment.

One of the major concerns noted in the findings of the present review is the lack of standardization on the timing of measurement. In general, the timing of the measures was not based on a theoretical foundation, an evidence base, and/or recommendations presented in previous studies of affective responses. As a result, the variation in research practice impairs the ability to make comparisons across studies, extrapolate relatable findings, and develop systematic guidelines. Current literature [15, 26, 27] suggests measures of affective responses should be obtained during or immediately after exercise, on the grounds that as time elapses between the behavior and the assessment moment, it becomes increasingly difficult to assess core affect without the interference of cognitive processes (i.e., possibly leading to an affective rebound). This suggests that measurements that are not applied during or immediately after exercise may fail to capture the true affective response to RT and present a distorted insight into the exerciser’s experience. Four studies in the review demonstrated that measuring affect during exercise is feasible [27–30], and lend support to the affective assessment recommendations proposed by Stevens et al. [15]. Alternatively, 12 studies aligned with the recommendation to assess affective response immediately after a set [26, 27, 43, 46–48, 51, 54–56, 58, 59]. Although both approaches appear somewhat comparable in terms of feasibility, some considerations warrant discussion in the interpretation of the data derived for the alternative timings of assessment. For example, stopping mid-set (or at any point whilst the muscle is under tension) can affect the exercise dynamic (e.g., interrupting the intended exercise cadence and inducing an isometric contraction when only isotonic movement is wanted), influence the target objectives (e.g., increasing fatigue due to interruption in answering the scales), and have safety implications (e.g., diverting task attention from task execution under heavy load). Therefore, the use of scales immediately after the load removal/completion of a set can avoid mid-set disruption and record accurate data prior to a potential affective rebound effect [27].

Regarding the number of affective response measurements per RT session, the issue remains vastly unexplored. The number of data collection points in the studies reviewed were only aimed to be integrated into the study protocol, and they ranged from one to 60 assessments during the exercise session with no theoretical justification or attempts to provide recommendations for feasibility. Bastos et al. [26] is the only exception, highlighting that in apparently healthy and experienced exercisers a single core affect measurement may suffice to understand the pleasure/displeasure response. The matter of affective measurement frequency can be relevant in maximizing a pleasurable experience in RT. For example, in Portugal et al. [58], high-activation pleasure was the norm for the affective response in every condition except for the leg curl exercise when performed at 80% 1-RM, where core affect shifted to the high-activation displeasure quadrant. Failure to measure core affect frequently enough could result in such unpleasant responses being masked by the overall affective response of the RT session; conversely measuring too often could result in less ecological validity, less overall feasibility (e.g., reactivity to the test; variance carry-over effects), and burden the participant [32]. As such, the identification of recommendations in this regard that consider individual (e.g., exercise experience; health status) and exercise (e.g., intensity; volume) characteristics is paramount for the advancement of the affect regulation approach to exercise promotion.

Other potentially relevant indicators

Considering the selected studies comprising this review, 15 studies used both the FS and the FAS, 12 studies used only the FS, and no study applied only the FAS. These results align with Stevens et al.’s [15] suggestion that much of the literature on affective responses to physical activity as a determinant of exercise adherence has predominantly focused on measuring core affective valence. According to the researchers that developed the FS [19] ‘the scale was designed to evaluate the core emotions: pleasure / displeasure’ (p. 305). Thus, while the FAS measures levels of activation, the FS is necessary to differentiate positive activation from negative activation [15, 65]. Likewise, the FS alone cannot dissociate different states of positive (i.e., happy from calm) or negative (i.e., distressed from bored) affect from each other without the measurement of arousal [22]. This would suggest that some of the studies included in this review may be limited in their interpretation of core affective responses by relying solely on one dimension of the affective response to RT.

Nine of the 15 studies that used both the FS and the FAS plotted the scores in a circumplex model to differentiate states of affective response. The application of the circumplex model offers a more comprehensive and fine-grained degree of affective states differentiation, acting as a map for the possible dynamic fluctuations across an RT exercise session [17, 22, 66], and allowing an idiographic and longitudinal approach to the affective response information.

When extracting and analyzing the data related to the included studies, some attention seems warranted in future research efforts on this topic. First, most of the literature presents small and convenient samples (i.e., N < 30 in 63% of the studies). Less than half of the studies (48%) conducted effect size and/or power calculations, risking confounding results due to a lack of statistical power. Second, only 13 studies (42%) reported participants received training in answering these scales, and only two (8%) reported these familiarization procedures for both researchers and participants. Moreover, verbal encouragement during data collection was reported in four studies, which can interfere with naturally occurring affective responses and subsequent comparisons across studies. This may suggest a heterogeneous approach to the training and preparation for these scales’ use, and contrast with the suggestions made by Duda [67] and Evmenenko & Teixeira [17]. The knowledge of what is being assessed (i.e., feelings experienced in the present moment), as for the definition of subjective anchoring examples with the respondents, along with the ability of the researchers to provide them through contextual questioning, are important factors to achieve consistent application results. This methodological issue reflects a bias in data collection that must be addressed in future study efforts, particularly given the perceptual nature of core affect.

Study limitations and future directions

This review is the first to explore the use of the FS/FAS to assess affective responses exclusively within RT activities; as such, it presents the most recent scientific evidence on how to use these instruments in studies undertaken with this mode of exercise. Nevertheless, this review is not without its limitations. First, the current literature on affective responses to RT is still scarce, this limits the extension of the extrapolation of the results. While the number of studies in this field of research is on the rise (of the 27 studies, 24 were published in the last 10 years, 15 in the last five years, and 10 in the last two years), more research grounded on current evidence of core affect measurements for this exercise type is warranted. Secondly, the heterogeneity of some sample characteristics as well as variability in methodological assessment approaches (e.g., timing), impairs the ability to interpret the recorded affective responses in RT. However, some consensus appears to be emerging in the most recently developed studies, and a suggestion on the timing of assessment (i.e., during or immediately after a set) is being used more consistently. Given the low number of studies that have focused on understanding how to promote quality measurement using both scales, replication studies will be paramount for this clarification and may help create a theoretically robust foundation for core affect assessment. Third, and still pertaining to the studies’ heterogeneity, no clear indication of consistent practice emerged regarding the frequency of assessment. Although highly dependent on the sample characteristics (e.g., exercise experience), protocols to be tested (common RT session vs. circuit RT), and other exercise-related variables (e.g., session length), understanding which characteristics of RT will determine the frequency of assessment is paramount for the development of future affect regulation guidelines. Finally, although this review focuses on the FS and the FAS, it is important to remember that there are other instruments that may be of interest for affective response measurement. For example, the Activation Deactivation Adjective Check List [68] represents another satisfactory option for affective response measurement in exercise, that also utilizes a circumplex model of affect [69].