A systematic narrative review of extrinsic strategies to improve affective responses to exercise

Background Extrinsic strategies affect the exercise experience but fall outside the frequency, intensity, time, and type (i.e., dose-determining) principles. To our knowledge, no systematic review has focused on extrinsic strategies to influence the affective responses to exercise. The objective was to identify extrinsic strategies that seek to influence affective responses during exercise and other motivationally relevant variables including post-exercise momentary affective valence, remembered and forecasted pleasure, and enjoyment. Methods For inclusion, eligible articles reported peer-reviewed original research, used acute bouts of exercise, and used a dimensional approach for measuring affective responses or measured enjoyment post-exercise. Web of Science, PubMed, and PsychINFO databases were last searched on 10th September 2021. Quality assessment was completed following the Effective Public Health Practice Project approach. Results were presented using a narrative synthesis. Results 125 studies were included with sample descriptions, study design (extrinsic strategies, mode, type, intensity, and duration), measurement details, and results summarised for each study. Conclusions 71% of studies were categorised as Weak according to the quality assessment tool with sampling practices (self-referred participants) and poor reporting of participant withdrawals/drop-outs the predominant reasons for Weak ratings. A wide variety of extrinsic strategies were reported with music, music videos, immersive virtual reality, outdoor exercise, caffeine, high-to-low pattern of exercise intensity, self-selected exercise intensity, and manipulation of self-efficacy offering promise as suitable strategies to positively change how people feel during exercise. Systematic Review Registration https://osf.io/jbh8v/.


Introduction
Rates of physical inactivity in many areas of the world continue to cause physical and mental health issues. The negative effects of physical inactivity have long been established and there is no debate that increasing population-level physical activity would have significant and positive effects at individual and societal levels. However, Hallal et al. (1) explained that "The traditional public health approach based on evidence and exhortation has-to some extent-been unsuccessful so far" (p. 254). Consequently, new approaches to increasing physical activity might be required.
Recent increased interest in the role of affect in behavioural sciences has prompted some acknowledgement that a new "era of affectivism" is emerging (2). Within exercise psychology, there is growing awareness of the role that affective responses might play in promoting adherence (3). For example, Rhodes and Kates (4) performed a systematic review that indicated that the affective responses experienced during exercise, but not after, predicted future exercise behaviour. Recent theoretical developments such as affective reflective theory of physical activity and exercise [ART (5)], and the theory of effort minimization in physical activity [TEMPA (6)] adopt a dual-process approach to understanding physical activity behaviour. Moreover, those models include affective responses as deterministic automatic pre-cursors to behavioural decisions about exercise. Owing to the renewed focus on affect from traditional and emerging theoretical perspectives, researchers are seeking to develop strategies and interventions that can enhance affective responses to exercise. Evidence suggests that enhancing how people feel while they move may be most meaningful (4).
Physical activity is a broad term capturing "any bodily movement produced by skeletal muscles that results in energy expenditure" (7), with exercise and sport considered types of physical activity. Sport and exercise both include bodily movement, but a key difference between physical activity and exercise is the "planned, structured, and repetitive bodily movements" (7), and the distinguishing feature of sport is the competition regulated by a regulatory agency (8). This review focuses on affect enhancing strategies that have been developed in an exercise context. Exercise can be characterised by four principles: Frequency, Intensity, Time, and Type (FITT). There are numerous strategies that have been developed with a focus on these dose-determining intrinsic characteristics [see Jones and Zenko (9)]. In addition to those intrinsic characteristics, are extrinsic factors that relate to the broader exercise environment and encompass anything related to exercise that is outside of the FITT principles (i.e., factors that will not change the dose of exercise) (9). For example, the intensity (70% maxHR), time (60 min), and type (running) are intrinsic factors but whether that activity is conducted outside or inside, or with or without music would be the extrinsic factors. This review seeks to capture studies examining the role that extrinsic strategies could play in changing affective responses to exercise. Intensity of exercise is a crucial determinant of affective response [e.g., (10)] and has previously been captured in a review by Ekkekakis et al. (11). However, a review including extrinsic strategies designed to improve affective responses to exercise has not yet been conducted and would assist researchers in understanding the current state of the field and opportunities for future advancements. Because this is, to our knowledge, the first review of its kind, we take a broad approach and include any factor outside of directly changing overall workload. We also take a broad and inclusive approach regarding participant inclusion criteria; this review includes diverse populations (e.g., trained individuals, untrained individuals, older adults).
The terms core affect, moods, and emotions are often used in exercise literature to capture how people feel about their experience. However, concerns over the suitability of how these terms and concepts have been implemented to capture how people feel during and after exercise have been raised [see Ekkekakis (12)]. Suitability of the use of those terms concerns the conceptual understanding of what each of them refer to. The definition of those terms in the present review follows previously proposed descriptions [e.g., (12,13)]. Briefly, core affect is an always present state conceptualised with an arousal and valence dimension and is the "most elementary consciously accessible affective feeling" (14). We adopt Russell's 1980 cirumplex model of affect as a conceptual model and a dimensional measurement approach based on its apparent superiority in allowing broad domain coverage of affective responses, compared to the distinct-states approach, which may allow for the measurement of distinct feeling states during exercise (e.g., calmness, vigor) but may also fail to assess all theoretically possible changes in affect [see (12)]. In-task assessments of affect appear more useful with regards to predicting future behaviour with a review article stating that "positive changes in basic affective responses during moderate intensity exercise was reliably linked to future physical activity behavior" (4). Owing to the issues with unsuitable measurement approaches in some studies, the present review seeks to capture those studies that align: (1) relevant affective phenomena, (2) appropriate conceptual models, (3) strong psychometric properties, and (4) assessment at appropriate time points. Affective outcomes measured using a dimensional approach will be prioritized (12) and the review will separately examine (a) in-task, moment-tomoment affective valence (4), (b) post-task, momentary affective valence, and (c) post-task, remembered and forecasted affective responses [e.g., remembered pleasure of the physical activity or exercise session, predicted pleasure of future physical activity or exercise sessions, or enjoyment (9)].
As interest in the affective responses to exercise grows, so do the attempts by researchers to influence those responses. One of the most heavily researched extrinsic strategies is music. A recent meta-analysis by Terry et al. (15) on the effects of music in sport and exercise included 139 studies from the past 100 years. The broad ranging review captured psychological, physiological, psychophysical, and performance outcomes that demonstrated the depth of research on this popular strategy. The effects of outdoor exercise (e.g., Green Exercise) is garnering increased interest and the systematic review by Lahart et al. (16) included 28 trials examining the physical (e.g., cortisol, heart rate) and mental health (e.g., depression) effects of green exercise. There are also other approaches that have received less attention to date but offer innovative and emerging approaches to positively influence how people feel during exercise (e.g., virtual reality, mindfulness). The present review seeks to capture the effects of these strategies in a focused way that is designed to assist researchers and practitioners specifically interested in influencing affective responses to exercise.
The purposes of this review are to identify extrinsic strategies that aim to: influence affective responses during exercise; influence post-task momentary affective valence; influence post-task remembered and forecasted pleasure; influence enjoyment. This review will take inventory of extrinsic strategies that have been used to alter the affective experience of exercise. Experimental studies that manipulate the exercise experience with the aim of influencing affective responses using extrinsic strategies are the focus of this review. Although studies that may indirectly influence intensity (e.g., by allowing people to listen to music) or involve intensity (e.g., by comparing self-selected intensity to a matched, prescribed intensity) are included in this review, studies that solely manipulate the intensity or workload of exercise are not within the scope of this review. Similarly, studies that compare types (or modes) of exercise (e.g., running compared to swimming) are not within the scope of this review.

Method
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The PRISMA flow diagram is presented in Figure 1.

Eligibility criteria
A study was considered for the review if the following eligibility criteria were met: (a) they were peer-reviewed, original research and published in the English language, (b) acute bouts of physical activity were studied (e.g., planned exercise), (c) a dimensional approach to measuring affective responses was used to assess in-task core affect, or enjoyment was measured following physical activity. In line with the aim of the review, all eligible studies included an extrinsic strategy that aimed to influence affective responses during the activity.

Screening
Both authors collected and screened articles for eligibility and a consensus was reached in cases of disagreements. Initial screening included article titles and studies were excluded if both authors agreed the study was not suitable for inclusion based the described criteria. Authors then screened abstracts for inclusion and subsequently screened full-text articles. Data were extracted to an Excel workbook.

Data abstraction and analysis
Each author extracted data from half of the eligible full-text articles. To check agreement and completeness, a random sample of half of the articles were examined by the other author. Extracted data included: (a) author names, year of publication, article citation information, (b) participant characteristics, (c) study design, (d) Primary manipulation, (e) type/mode of activity, (f) affective constructs, measures, collection timepoints, (g) intentions following exercise (if applicable), (h) results. Quality assessment of individual studies was completed following the approach of the Effective Public Health Practice Project (17). Specifically, (a) selection bias, (b) design, (c) confounders, (d) blinding, (e) data collection methods, and (f) withdrawals and dropouts were evaluated. Although all studies that met inclusion criteria are included in the synthesis, they were identified as either "Strong", "Moderate", or "Weak". The overall percentage of Strong, Moderate, and Weak studies are reported. This tool was selected owing to its application to many public health topics.
The extracted data are summarized and synthesised to discuss strategies for facilitating pleasant exercise. Following the guidelines by Popay et al. (18), the review focuses on (a) interpreting the literature to date and discussing the applicability of the extant literature, (b) describing the effectiveness of the previously studied strategies for enhancing the pleasure of physical activity and exercise, (c) exploring the relationships within the data that may explain study results, and (d) assessing the strength of the research evidence and generalizability to understudied populations and contexts.

Results
The number of studies included was 125 (see Figure 1). Each study was assigned one or more strategy name(s) and those strategies are discussed. The strategies have been grouped under the following headings: audio-visual; indoor and outdoor exercise; nutrition; pattern of exercise intensity; prescribed vs. self-selected exercise; mindfulness; self-efficacy; social environment; gamification and exergaming; other factors. Summary data for each study is presented in Table 1.

Audio-visual strategies
Audio-visual strategies represent the largest proportion of studies in this review with music being the most frequently included strategy (33 out of 125 studies). Music has long been an adjunct to exercise but more recently other audio strategies such as podcasts have been tested experimentally.

Music
One of the oldest studies included in this review focused on music and demonstrated the capacity to influence Feeling Scale [FS (140);] scores (which range from very bad to very good). Brownley et al. (40) reported that fast tempo music positively increased pleasure during low and moderate intensity treadmill exercise in 16 untrained participants. Music tempo has been a central feature of several studies [e.g., (20,84)]. The study by Karageorghis and Jones (84) demonstrated that all music tempo conditions (slow [95-100 bpm], medium [115-120 bpm], fast [135-140 bpm], and very fast [155-160 bpm]) elicited greater pleasure than no music during treadmill exercise, and that medium tempo music elicited greater pleasure than the other music tempi conditions across intensities ranging from 40% to 90% max heart rate reserve. Music tempo has also been examined during interval training with results indicating limited effects of music during this mode of exercise [e.g., (82,85)]. However, affective responses reported post-interval training have shown more consistent positive changes as a consequence of listening to music [e.g., (123)].
Beyond studies examining the internal components of music such as tempo, the effects of music have been examined more broadly. Campbell and White (43) reported greater enjoyment during a 20 min, moderate intensity treadmill walking task when listening to music in a study of 148 undergraduate students. Elliott et al. (55) found that "motivational" music led to greater pleasure than not listening to music during cycle ergometry at 60%-80% maximum heart rate. In comparison to no-music, and a tempo-matched metronome condition, Lim et al. (93) found that music applied synchronously and asynchronously was reported as more pleasant. Karageorghis et al. (86) reported that females recorded less negative affect than males during circuit training tasks while listening to music that was either motivational or motivationally neutral. Bigliassi et al. (33) compared music to a podcast condition and a no-audio control condition while participants walked a self-selected pace on a 400 m outdoor athletic track. They found the music condition led to the highest in-task FS scores and post-exercise Physical Activity Enjoyment Scale (PACES) (87) scores, and the podcast was also more enjoyable than the control condition. Hutchinson et al. (77) combined self-selected music with affect-regulated exercise and reported that participants found the music condition more pleasant while exercising at a higher intensity (d = 1.12). Further, participants recalled the music session as more pleasant (d = .72).
While there are often positive affective responses reported in music and exercise studies, some studies suggest this effect is not universal. Dyrlund and Winninger (54) tested the effects of most-preferred or least-preferred music on enjoyment of treadmill walking at a range of exercise intensities but found no differences with 200 university students. Chen et al. (44)  The FS was administered during exercise There was no main effect for conditions. There was a significant interaction effect whereby pleasure decreased between 9 and 10 min during the condition where participants expected to stop at 10 min but were asked to continue for another 10  The FS administered to capture how the participants "were feeling on average in the past (five/ten) minutes". Pleasure was higher in the mindfulness and distraction/ podcast conditions compared to associative focus. Mindfulness and distraction/podcast were not different.
The affective slope was also more positive in the mindfulness and distraction/podcast groups Mixed-design; imposed and self-selected exercise, one group was aware the intensities were matched; Cycle ergometer for 20 min The FS was administered during and after exercise, and a mean was created for in-task responses. "Session Affective Response" was measured 15 min after exercise and seemed to function as a measure of remembered pleasure Session affective response was greater than the in-task values for the self-selected exercise intensity, but not the imposed condition treadmill at 10% of max capacity above and below the ventilatory threshold The FS was administered before, during, and after exercise. FS scores during exercise were averaged Music videos were more pleasant regardless of intensity.
Music was also more pleasant than control. Post-task FS scores for Music videos and music-only were higher than control during below ventilatory threshold exercise compared to pre-exercise. Post-task FS scores were higher than in-task FS scores for music videos and music-only. Within-subjects; Primes, music video, music; 8 min of treadmill walking at 65% heart rate reserve The FS was administered during exercise. Remembered pleasure (VAS) and forecasted pleasure (EVS) measured after exercise The music video + primes condition resulted in more pleasure than the music video and control conditions, but not more pleasure than the music condition. Remembered pleasure was higher in the music video + primes condition than in the music, music and video, and control conditions. Remembered pleasure was also higher in the music condition compared to control, and higher in music video compared to control. Forecasted pleasure was higher in the music video + primes condition than the music and control conditions, and in the music compared to the control condition, and the music video compared to control condition Between-subjects; Self-paced exercise; walking at a moderate intensity or a self-paced intensity (but not to exceed 76% of max heart rate) The FS was administered during and after exercise, and also randomly during the day (to control for baseline affective valence) Self-paced exercise resulted in more positive affective responses, and affective responses predicted more exercise behaviour Frontiers in Sports and Active Living examined the effects of listening to music compared to no-music in a small sample (n = 12) of participants with Down's Syndrome and reported no differences in enjoyment following treadmill walking. Feiss et al. (58) reported no effects on pleasure-displeasure when listening to fast or slow tempo music compared to a control condition for isometric tasks (e.g., wall sit).

Music and video
Music has been employed as an extrinsic strategy alongside, and in comparison to, other strategies. Jones et al. (81) showed that music as a solo strategy and alongside a point-of-view video travelling through parkland was more pleasant than the no stimuli and the video-only condition during exercise above and below the ventilatory threshold. The 2015 study by Hutchinson et al. (75) showed a slightly different pattern of results with music videos resulting in the most positive affective responses during treadmill exercise, although music by itself was also more pleasant and enjoyable than a control condition. Bird et al. (34) demonstrated a similar pattern of results with pleasure reported as higher for music video conditions compared to music-only, and with both experimental conditions more pleasant than control. Hutchinson et al. (76) extended the music video experiments into a clinical context with diabetic patients and found that music video and music-only were effective at positively enhancing pleasure and enjoyment during flexibility, aerobic, and resistance exercise. Miller et al. (98) demonstrated utility for music and video as independent strategies to enhance pleasure during recumbent cycle ergometer compared to no stimuli in a sample of 25 college students. In 2019, Jones and Ekkekakis (80) examined the effects of music videos delivered using immersive head-mounted displays (i.e., Samsung Gear VR) compared to a wall-mounted television and a control condition. In their study, 21 adults with overweight and low fitness reported highest FS scores in the highly immersive music video condition during 20 min of recumbent cycle ergometry. Similarly, Bird et al. (36) reported a 360°video accompanied by music elicited the most pleasure and enjoyment in 18 university students during a 10 min cycle ergometry task.

Virtual reality
The delivery of virtual reality (VR) stimuli has only recently shifted to head-mounted displays and some studies have examined virtual reality using screen projections [e.g., (29)]. In this section, virtual reality relates to computer generated simulations of avatars or environments. A study adopting the projection approach with 109 children with overweight did not find any significant differences in enjoyment in comparisons between screen-projected VR and a no-VR condition following treadmill exercise (29). Similarly, Neumann and Moffitt (103) did not find any differences in a screen-projected VR study examining the effects of running with virtual runners or neutral images. Contrastingly, Zeng et al. (137) found greater enjoyment during a VR condition delivered via a head-mounted display compared to a no-VR condition during a 20 min cycle ergometry task with 12 college students. Bird et al. (35) combined virtual reality cycling with music and found more pleasure and enjoyment for the VR, and VR with music conditions compared to control conditions in a sample of 24 undergraduate students. Also using a head-mounted display, Ijaz et al. (79) showed that an "Open World" condition affording participants an opportunity to move through Google Maps Street View images in time with the cycling cadence was more enjoyable than watching static images from the same platform.

Indoor and outdoor exercise
Examining differences between indoor and outdoor exercise is a popular approach for researchers, with different designs being employed to understand how the outdoors can influence feelings during exercise. A similar pattern of results has been shown in a number of studies with outdoor exercise being reported as more pleasant and/or enjoyable than indoor exercise (41,52,60,89,104,128). The participants varied from physically active university students (60) to women over 40 with obesity (89) but the pattern of results was similar. However, Rogerson et al. (117) did not find any differences in enjoyment when asking participants to cycle on an ergometer placed in a field at a university campus, and cycle ergometry in a visually sterile laboratory. Similarly, Frühauf and colleagues (62) found no differences between indoor and outdoor exercise in a small group of in-patients with mild-to-moderate depression. Frühauf et al's study was one of few that included a clinical population but appeared underpowered to find statistical significance given the effect sizes (indoor exercise, d = 1.44; outdoor exercise, d = .64).
Studies have also examined the efficacy of digital representations of outdoor environments. Calogiuiri et al. (42) examined the effects of exercising outdoors compared to indoor exercise accompanied by virtual footage of outdoors and found that outdoor exercise was more enjoyable. White et al. (133) compared simulations of outdoor environments while cycling on a stationary bike for 15 min in a sample of 37 post-menopausal women; findings showed that green and blue exercise environments (e.g., countryside or coastal setting) were more pleasant than a control condition. Rather than comparing indoor and outdoor exercise, Crust et al. (51) found that a countryside walk was more enjoyable than walking in a green urban environment for a sample from British recreational countryside walking groups.

Nutrition
Studies including nutrition strategies have often focused on caffeine or carbohydrate supplementation and indicate that supplementation can positively influence affective responses.

Caffeine
Using a mixed design and 16 healthy active men (8 endurance trained), Astorino and colleagues (22)

Carbohydrate
Backhouse et al. (25) observed no effects of carbohydrate ingestion on pleasure measured before, during, and after cycling exercise in 17 young male soccer players; this was in contrast to earlier findings among 9 endurance trained males, in which carbohydrate consumption enhanced pleasure in response to prolonged cycling [120 min at 70% VO 2 max (27)]. Both studies used the FS (140) to measure affective valence before, during, and after exercise and the reason for the conflicting results is unclear. In another study of 15 healthy individuals (no signs or symptoms of cardiovascular, renal, or metabolic disease),

Other nutritional factors
Backhouse et al. (24) studied 15 young (mean age: 21 ± 0.5 years) and aerobically fit (VO 2 max: 65.0 ± 1.2 ml/kg/min) men and found that pleasure was greater when allowing fluid replacement; fluid replacement resulted in greater postexercise pleasure, with large effect sizes, in response to a prolonged bout of treadmill exercise (90 min at 70% VO 2 max). In another study focused on fluid, but this time mouth rinsing with either a pink or a clear non-caloric and artificially sweetened solution, Brown et al. (39) observed that mouth rinsing with a pink solution enhanced pleasure in response to running at a self-selected pace for 30 min (compared to a clear, non-caloric artificially sweetened solution). Shaver et al. (120) observed no differences in affective valence or energetic or tense arousal, measured with the FS or the Activation-Deactivation Adjective Checklist (141), between conditions with drinking or rinsing water. In this study, 19 recreationally active females ran 15 km time trials outdoors, while either drinking or rinsing water.

Pattern of exercise intensity
Researchers have also investigated the impact of the pattern of exercise intensity on affective responses. Zenko et al. (138) studied both men and women using a between-subjects design. Participants exercised for 15 min on a cycle ergometer and either increased intensity (from 0 Watts to 120% of the Watts corresponding to their ventilatory threshold) or decreased intensity (from 120% of the Watts corresponding to their ventilatory threshold to 0 Watts) and the FS was administered before, during, and after exercise. Remembered pleasure (i.e., a retrospective global evaluation of the session) was measuring using a bipolar visual analogue scale (VAS), and forecasted pleasure (i.e., predictions about how future repeated bouts would feel) was measured using the empirically spaced anchors from the Empirical Valence Scale (EVS) (141). Decreasing intensity resulted in more pleasure during and after exercise, more remembered pleasure, more positive forecasted pleasure, and more enjoyment. In a later pre-registered study, Hutchinson et al. (78) largely replicated these findings in resistance exercise. Using a within-subjects design, participants completed three resistance exercise circuits that either increased in load (from 55% to 65% to 75% of 1 repetitionmaximum) or decreased in load (from 75% to 65% to 55% of 1 repetition-maximum). Affective valence was assessed using the FS and enjoyment was assessed using the PACES-8 (142). The EVS was used to assess remembered pleasure shortly after exercise and an ad-hoc measure was used to assess remembered pleasure 24 h after exercising, via text message. The decreasing intensity resulted in more pleasure during exercise, more enjoyment, and more remembered pleasure. Also see Hutchinson et al. (143) for a more recent study, not included in the systematic search.
Malik et al. (94) completed a within-subjects design, this time using high-intensity interval exercise on a cycle ergometer; 16 adolescent boys and girls performed exercise at an intensity that either increased, decreased, or stayed the same. Enjoyment of exercise was measured using the Exercise Enjoyment Scale (144) during the HIIT protocol, and PACES after exercise; the FS was used to assess affective valence before, during, and after exercise. Decreasing exercise intensity resulted in more pleasure and enjoyment toward the end of the session. However, there was no effect of the exercise condition on enjoyment when measured with PACES. In another interval study, Cortis et al. (46) used a within-subjects design in which 18 participants performed interval cycling exercise in either an "ascending", "descending", or "mixed pyramid" pattern. The ascending pattern increased in intensity from 50% to 75% to 100% of peak power output (PPO) across three 3 min work bouts, with 3 min recovery periods at 25% PPO. The descending pattern reversed this, with bouts decreasing in intensity from 100% of PPO to 75% to 50%. The "mixed pyramid" pattern, the initial workload was 75% PPO, then 100% PPO, and ended with 50% PPO. Enjoyment was assessed using the Exercise Enjoyment Scale and there was no difference between conditions.
Although Watson et al. (131) did not directly alter the pattern or order of exercise intensity or training load, they studied the effects of the order of exercises. Using a within-subjects design and 17 elite hockey players, participants completed resistance exercises in either a predetermined or self-selected order. The order of exercise made no difference on enjoyment.

Prescribed vs. self-selected exercise characteristics
Several researchers have investigated the effects of allowing participants to self-select aspects of their exercise protocol; this has often focused on allowing participants to choose their exercise intensity, rather than prescribing exercise intensity [e.g., (118)]. Importantly, not all studies controlled for intensity (i.e., sometimes intensity was not directly manipulated, but was different between groups or conditions as an indirect effect of the exercise prescription). Haile et al. (67) tested the effects of self-selected exercise intensity in a mixed design, where the experimental group was unaware that the imposed workload was identical to the selfselected workload and the control group was fully aware. Affective valence was measured using the FS during exercise and "session affective response" was measured 15 min after exercise and seemed to function as an indicator of remembered pleasure. There was no impact on mean affective responses during exercise or remembered pleasure between prescribed or self-selected workloads or between groups (aware vs. unaware). This finding is consistent with a later finding by Oliveira et al. (105), but somewhat conflicting with an earlier report by Vazou-Ekkekakis and Ekkekakis (130), who found that self-selected exercise intensity (vs. matched prescribed intensity) enhanced energetic arousal and interest/enjoyment, but not FS scores. Hamlyn-Williams et al. (68) did find that affective responses were enhanced during treadmill exercise at a self-selected intensity among 27 adolescent females, compared to prescribed intensity.
Other reports by Williams and colleagues (134,135) were also mixed, with one study indicating that self-paced exercise resulted in more positive affective valence (134) and one study indicating that allowing sedentary and low-active women to self-select their own walking pace (vs. a matched pace set by the experimenter) did not impact affective valence (135). Zenko et al. (139) allowed participants to self-select their exercise intensity on a cycle ergometer, although there was an emphasis on pleasure. In the experimental condition of a within-subjects design, participants were prompted to choose an intensity that maximized their pleasure. Remembered pleasure and enjoyment was higher when participants were prompted to maximize their pleasure, and pleasure was more positive over time during exercise when compared to the control condition. However, there may have been demand effects in this study as the experimenters did not control for participant-experimenter interaction (i.e., the level of attention may have been a confounding factor). Future investigators should try to keep interaction between participants and experimenters more consistent.
Self-selected intensity was not the only characteristic investigated; Emanuel et al. (56) allowed participants to choose their number of repetitions in a resistance exercise context (choice of between 8 and 12 repetitions vs. a fixed 10 repetitions using a within-subject design) and found no difference in enjoyment.

Mindfulness
Researchers have studied the effects of mindfulness on the affective experience of exercise and this has also been paired with other factors [e.g., music (48)]. There have been mixed results; some studies report positive affective responses during exercise and increased enjoyment in response to mindfulness conditions (47), but other results indicate that although mindfulness increases forecasted pleasure, it makes no difference on affective responses during exercise (48). In the context of yoga, mindfulness has resulted in more pleasure (pre-to-post exercise) compared to a focus on changing one's appearance. Researchers also found that focusing on changing one's appearance resulted in lower remembered and forecasted pleasure compared to both mindfulness and yoga-poses conditions (50). Other researchers have found positive effects of both mindfulness and distraction (podcasts) among 78 insufficiently active walkers (64).

Self-Efficacy
Researchers have generally found that self-efficacy manipulations (e.g., providing false or true feedback after completing an exercise task) can positively impact affective outcomes, including enjoyment (74,127). Although, Hu et al. (73) found that self-efficacy feedback influenced enjoyment only after a maximal exercise test (not sub-maximal intensity). Further, although McAuley et al. (95) found that participants reported more pleasure after self-efficacy inducing feedback, this was not statistically significant.

Social environment
The effects of the social environment in which exercise takes place has been explored by several researchers although the manipulations are highly variable and include a wide array of specific strategies (e.g., being observed by others, exercising in a private or public setting, exercising with a buddy). Plante et al. (111) found that exercising with a workout partner (either a stranger or a friend) did not make an impact on enjoyment when cycling on a stationary ergometer, although enjoyment was higher when walking around campus with a friend, compared to indoor exercise. The researchers noted that "exercising 'alone' was actually in a busy university fitness club facility" (p. 97), providing threats to the validity of the study.
Winchester et al. (136) found that affective valence was higher among men who ran at 60% of their peak running speed for 20 min when being observed by "highly attractive" (p. 217) observers compared to a control condition. Session affect was higher after being observed by the female-observer (vs. the control condition). Thus, exercising with an observer appeared to be beneficial for this small sample of 10 men. However, the "observers" acted as sham research assistants and spoke to the participants during exercise and did not solely watch.
Focht & Hausenblas (61) investigated the affective impact of exercising in a public environment (i.e., a university fitness center) vs. a private, non-mirrored laboratory setting. Pleasure initially decreased in the public environment but rebounded; compared to baseline, exercising in the public environment resulted in moderate improvements in pleasure during the last in-task FS measurement (d = .44) and large improvements postexercise (d = .86). Pleasure also increased in private settings, with medium-to-large effects (d = .55, d = 1.23).
Galway et al. (63) recently investigated the impact of functionvs. appearance-focused cues from exercise instructors on body image, enjoyment, and intentions to exercise in the context of group exercise classes lasting about 45 min; the researchers found no difference in enjoyment, measured with the PACES. Feltz et al. (59) found that planking with an exercise "buddy" for as long as possible did not make an impact on either exercise enjoyment or exercise intentions. Interestingly, other studies have included the effects of the presence of other people at different locations. Exercising with other people who were ostensibly exercising over the internet at different locations at a university made no difference on affective responses (101). In contrast, exercising in a competition against five computerized players appeared to be more enjoyable than a conventional stationary exercise session, although this may have been due to playing a game, rather than competing against other players (100).

Gamification and exergaming
Gamification has become an increasingly popular strategy to empirically examine. In two experiments, van der Kooij et al. (129) found that enjoyment was enhanced after gamified treadmill and balance exercise (i.e., allowing participants to virtually play a gardening game by using their feet and allowing participants to avoid virtual obstacles by changing their centre of mass). Researchers have found promising effects of video games among university students (65). Paired with virtual reality, exergaming has been found to increase enjoyment of moderateto-vigorous intensity cycling among a sample of 12 college students, with large effects (d = .89) (137).
Among participants who have previously experienced a stroke, video games that included stepping and marching in self-paced or game-paced conditions did not seem to make a difference in enjoyment (53). McDonough et al. (96,97) have provided some mixed results. Among 47 adolescents from minoritized backgrounds, exergaming (Just Dance 2018 on Xbox One Kinect) was more enjoyable when in a small group compared to a large group (96). In another within-subjects design, exergaming in single player or double player mode did not make an impact on enjoyment among 20 retired Olympic athletes (97). It is important to note that in both of these studies, exergaming was not compared to a control (i.e., exercising without gamification), indicating that the factor of exergaming itself was not manipulated.

Other strategies
Finally, several other extrinsic strategies have been manipulated by researchers although most of these have not been studied in more than a few studies. These factors include priming (19,113), cognitive load (23), transcranial direct brain stimulation (28)

Quality assessment
Of the 125 studies, 89 were categorised as Weak, 35 as Moderate, and 1 as Strong. Further details of the individual components that contributed to the global ratings are provided in Figure 2. In addition to the quality check, a check regarding a priori power calculations was conducted to assess the extent to which the number of participants recruited was appropriate. 72 of the 125 studies (58%) did not include an a priori power calculation for the sample size.

Discussion
This review aimed to collate studies that examined manipulations designed to influence affective responses during and immediately after exercise. 125 suitable studies were identified and they spanned a range of extrinsic strategies.

Extrinsic strategies
The body of evidence for listening to music during exercise appears broadly consistent and supportive of a capacity to positively influence pleasure. Music tempo appears to be a salient factor for consideration when deciding the music to be used. The majority of studies on this popular topic have been conducted with healthy, young adults and the consequence of this data for different populations are unknown. For example, Chen et alia's (44) study with participants with Down's syndrome revealed no difference in affective responses between music and no music conditions. This study appears underpowered but is demonstrative that further studies are needed to verify the effects of music during exercise in different populations. The use of video alongside music appears to confer an additive effect wherein affective responses are enhanced. Additional research on the specific types of video footage that are most beneficial is required, but initial work on music videos appears promising and Hutchinson et alia's (76) work in a clinical population showed positive results. For virtual reality, the emergence of widely available head-mounted displays appears to have changed the findings for virtual reality studies. Those designs including head-mounted displays report more positive affective responses to the VR stimuli than was previously reported in screen-projected VR [e.g., (35)]. The greater immersion created by the head-mounted displays appears a relevant factor. The participants of the VR studies were typically young and active which limits to applicability of these findings to other groups.
Outdoor exercise appears to consistently show more positive affective responses than indoor exercise and this has been shown across different ages and activity levels. However, there appears a gap in research for work with clinical populations with only Frühauf et al. (62) exploring the strategy in a clinical sample. For some people, outdoor exercise is not easily achievable (particularly in green spaces) and initial attempts to create digital or virtual recreations of outdoor exercise have shown mixed results. Direct comparisons to outdoor exercise do not appear favourable for virtual green exercise (42), but when comparing to indoor exercise without virtual nature, the results are more positive (133). Additional work on virtual nature during exercise appears warranted given the incapacity of many people to access green spaces (145).
A strategy that has received recent increased attention is the pattern of exercise intensity. That is, does it matter whether an exercise session starts or ends with a high intensity, or higher levels of pleasure? The findings of Zenko et al. (138) and Hutchinson et al. (78,143) aligned with earlier behavioral science research focused on the peak-, end-, and trend-rules, which indicate that overall summary evaluations are highly influenced by the most intense moment, last moment, and overall trend (e.g., increasing or decreasing pleasure) of an experience (146,147). Further, these results are aligned with prior research from exercise psychology demonstrating that the peak-and end-of an experience explains a large percentage of the variance in one's overall memory of the experience (148).
Although not matching the inclusion criteria of this systematic review, a study by Stuntz et al. (149) found that, among collegiate athletes, increasing exertion resulted in more positive feeling states when increasing exertion resulted in more feelings of accomplishment. While it is clear that affective responses are sometimes meaningfully predictive of summary evaluations, including enjoyment, remembered pleasure, and forecasted pleasure, it is also clear that these summary evaluations are biased by memory and other cognitive filters and appraisals (9). Thus, there is need to further determine the conditions, contexts, and populations that may benefit from manipulating the pattern of exercise intensity. At present, it does not appear that the benefits are universal.
Lastly, it is important to note that several studies did not manipulate or control for intensity directly, and intensity may have differed between groups as an indirect effect of some extrinsic strategy [e.g., exergaming (97)]. As an example, for McDonough et al. (97), the physiological outcomes of exergaming alone or in pairs were relevant and therefore not standardized. However, there was greater energy expenditure and more steps when engaging with the exergame alone compared to with a partner. Since exercise intensity is a highly reliable determinant of affective responses (10,11), one must acknowledge that any study that unintentionally included different intensities between groups or conditions may have contained an influential confound.

Quality check
The data presented in Figure 2 displays a mixed picture regarding the quality of studies. Beyond the global ratings of 71% Weak, 28% Moderate, and 1% Strong studies, 98% of studies were categorised as Weak for selection bias and this represents a significant issue for this field. A weak categorisation reflects selfreferred participants, as opposed to random selection from a comprehensive list of individuals (Strong). Participants in the included studies would often be those who responded to adverts or be students at the authors' institution. These issues with selection bias limit the relevance of this field in contributing to wider discussions regarding exercise behaviour. However, if a study aims to influence affective responses of, for example, low active individuals with overweight from the general population, there is no list that participants can be drawn from. Therefore, there will be some instances where authors have made attempts to include representative participants, but the item used in the chosen quality check tool did not account for this.
The quality check results for study design and confounders indicated Strong in most studies. A key factor in the categorisation of Strong for Study Design was the use of randomisation for condition/group allocation. Many of the designs adopted a crossover RCT approach that we categorised as Strong. Following this, the effects of confounders was generally low as many studies included participants as their own controls. Moreover, in studies with separate groups, authors often demonstrated no significant differences between groups on important variables such as age.
The majority of studies were categorized as Moderate for Blinding as procedures were often not stated. The studies with Strong ratings were typically nutrition studies with effective placebo designs regarding drinks or caffeine. Many of the studies included in the review make blinding of the conditions difficult (e.g., a participant will know whether they are listening to music or not). However, researchers could make greater efforts to blind the true nature of the study and report their procedures for doing this. For example, Tritter et al. (127) provided participants with a letter stating the purpose of the study was to compare the effects of static or dynamic stretching, thereby attempting to hide the true nature of the study relating to affective responses to selfefficacy manipulations. We recommend that researchers make explicit statements to indicate whether outcome assessors were aware of the intervention or exposure status of participants and whether study participants were aware of the research questions. This would allow for more accurate assessment related to blinding.
A consequence of our inclusion/exclusion criteria was that we only included studies with certain pre-identified approaches to measurement. The majority of the studies in this design were categorised as having Strong data collection methods owing to the use of valid and reliable tools. The poor reporting of withdrawals and drop-outs led to two-thirds of studies categorised as Weak. The main drivers for 71% of studies receiving a global categorisation of Weak were issues with selection bias and poor reporting of withdrawals and dropout; one of these issues is easily remedied with improved reporting while the other is more challenging.
It seems common in exercise and sport science to not report dropouts and withdrawals unless participants dropped out or withdrew during a study. In other words, if zero participants dropped out or withdrew, it can often be assumed that the dropout and withdrawal rate was 0. However, according to the quality assessment tool used in this study, a weak rating must be assigned if the withdrawal and drop-out rates were not explicitly described. We recommend that researchers explicitly report dropout and withdrawal rates for all studies; this would also include explicit statements indicating if the follow-up and adherence rate was 100%. Overall, typical reporting practices might have inflated the number of "weak" and "moderate" quality ratings in this literature.
In addition to thorough and explicit reporting of the criteria highlighted in the quality assessment tool, the field would benefit from greater focus on important issues such as: a priori power analysis or sample size justification; clear research questions; preregistered methods; and theoretically and psychometrically sound measurement practices. Only 53 of the studies included in this review included some form of sample size justification, including a priori power analysis. Since underpowered studies may be less replicable than adequately powered studies, researchers must work to recruit more sufficient samples to ensure reliable effect sizes and conclusions. The percentage of studies with some level of sample size justification (about 42%) was higher than previously reported for sport and exercise science more broadly [i.e., only 22.67%; (150)]. Nonetheless, more than half of the studies included in this review had no sample size justification, indicating that this is an area for improvement.

Recommendations
A wide range of extrinsic strategies are presented in the review and provide practitioners and researchers with options for changing how people feel during, and after, exercise. However, the evidence supporting some of the strategies is weak in terms of quality of the study and also weak in terms of number of studies to support that evidence. Researchers have attempted many strategies to improve affective responses to exercise but several of these strategies have only been studied one or a few times (e.g., cognitive load, transcranial direct brain stimulation, facial expression, goal setting, manipulations of anticipated affect, imagery). The diverse array of research strategies, frequency of unjustified sample sizes, varying levels of research quality, and various research contexts highlights a need for replication attempts with many of these extrinsic strategies. In a recent study, for example, the findings of Pottratz et al. (113) were tested and failed to replicate (151). We speculate that this would not be a unique result if more replication attempts occurred; instead, more replication attempts in more research contexts (e.g., different cultures, ages, environments) might highlight highly variable effects.
The first practical recommendation is that new and innovative strategies are tested in robust research designs and with high quality reporting; this would include preregistration of study methods whenever possible (152) and full sample size justification (153). More high-quality design and reporting would also include stronger participant selection processes and transparent reporting of withdrawals and drop-outs. The second recommendation is for efforts to be directed towards replication of results especially of, but not limited to, strategies with one or two studies supporting their use (e.g., attentional focus, imagery). When choosing strategies for study, researchers should justify their choices and provide a rationale for why and how they believe the strategy will have an impact on affective responses. The third recommendation is to forge stronger links between practitioners and researchers to develop greater awareness of how strategies can be delivered outside of the laboratory in practical settings. The fourth recommendation is to conduct longitudinal studies examining the effectiveness of new strategies, and in those where only efficacy has been demonstrated (154). The fifth recommendation is for researchers and practitioners to engage with participants and clients in co-production of strategies [e.g., (155)] to help create meaningful and more positive experiences during exercise. The sixth recommendation is for studies to be conducted with participants in greater immediate need. The majority of studies included healthy, active participants but perhaps a significant impact this field could make is shaping the exercise experience for people who are overweight, obese, inactive, or among clinical populations. By changing experiences of people in those groups, our field might contribute to meaningful changes in positive outcomes with greater impact at individual and societal levels. Finally, the options for strategies to positively improve the exercise experience are limited only by our imagination. Sport and exercise psychologists already employ a number of strategies with their clients (e.g., self-talk, thought stopping) and these could be examined further with the aim of shaping how people feel during exercise. More broadly, a number of behaviour change techniques are employed in other healthrelated domains to positively change behaviour and these might also offer some direction for further experimental testing in an exercise context [see (156)]. As new technologies continue to develop (e.g., artificial intelligence), these offer avenues of exploration for researchers seeking to positively change how people feel during exercise.

Limitations
Many exercise trials are performed in the presence of additional potential extrinsic strategies although these are not captured within the study designs. For example, music might be the primary independent variable, but the researcher might also interact more with participants, and this could also be a source of affective influence during exercise. Further, there are studies that have combined strategies and the designs and analytical approaches do not always permit separation of the (possible) differential effects. These issues can limit the certainty of the results and the interpretations presented herein.

Deviations from preregistration
While we did our best to adhere to the preregistered protocol in a good-faith manner, there were several deviations. We did not perform a backward or forward search of systematic reviews discovered in the initial search due to the large number of records initially retrieved and impracticality. We also performed multiple rounds systematic searches in order to be inclusive of more current literature. Finally, reasons for exclusion are not shown on the PRISMA diagram due to the exceedingly large number of reasons for exclusion (i.e., many articles excluded for multiple reasons).

Conclusion
We have reviewed numerous strategies and sought to focus on studies with high measurement quality. There are strategies that have been shown to positively influence how people feel during exercise and these provide options for practitioners to implement. From the included studies, music, music videos, immersive virtual reality, outdoor exercise, caffeine, high-tolow pattern of exercise intensity, self-selected exercise intensity, and manipulation of self-efficacy all offer implementable strategies to change how people feel during exercise. We have proposed several ways to improve the quality of studies in this field with the aim of enhancing the impact that these strategies might have.

Data availability statement
This systematic review did not include the collection of any original data.