The need for moderate-intensity exercise is recognized universally as both vital for human health (19) as well as performance in sport (6). Human HR changes in proportion to exertion during exercise. Although individuals can control their level of effort by altering behavior, it is often difficult to interpret the relationships between observable bodily sensations and HR. That is, despite a variety of cardiopulmonary (e.g., respiration rate and oxygen uptake) and peripheral (e.g., skin temperature, sweat rate, and mechanical strain) sensory cues providing feedback on exertion during exercise (18), these sensations provide general rather than specific feedback on HR. Collectively, these “afferent” cues may create an internal model, refined and developed through prior experience, against which exertion is gauged (23).
It appears that no single cue dominates subjective effort perception across individuals (18). Instead, recent research supports a general model of fatigue (29) integrating multiple sources of afferent feedback. Regardless, despite the existence of multiple sensory cues available to gauge exertion, individuals experience difficulty selecting and maintaining specific exercise intensities (4,8,14).
The ability to exercise at a particular intensity appears to be trainable, however, through various applications of augmented feedback. Here, cues are conveyed through an external, rather than internal, mechanism delivered by a coach, instrument, or computer. For example, it has been shown that receiving verbal instructions during exercise can improve the selection of walking pace corresponding to moderate intensity (4) and that HR feedback from visual instruments (24), a form of biofeedback, can affect physiological parameters during treadmill walking. Moreover, several studies have used rate of perceived exertion (RPE) to develop an individual’s ability to reproduce a specific level of effort (13,16,22,28). This technique has been successfully used to reproduce average HR between entrainment and reproduction trials for short-duration exercise bouts (less than or equal to 5 min) with times ranging between estimation and reproduction from 2 to 14 d (13,16,22). However, Smutok et al. (28) observed that entrainment was dependent on exercise intensity, noting successful reproduction of effort only at intensities above 80% of maximum HR (HRmax).
Although the relationship between HR and RPE has been previously established (27), few studies have explored whether direct feedback via HR monitors (HRM) can improve subsequent self-regulation of HR in the absence of further external feedback. One exception is a study by Conley et al. (10) in which children were equipped with HRM providing feedback defining a moderate to vigorous level of exercise using both visual and audio cues over the course of six sessions. Upon completion of the intervention, participants failed to show improvement in estimating the time they had spent engaging in this level of activity versus a control. The researchers concluded that limitations in the children’s cognitive ability to estimate both effort and time and the absence of any mechanism encouraging the children to internalize effort perception relative to afferent cues might have contributed to this result.
Although instrumentation is commonly used to develop fitness in a variety of endurance disciplines, including cycling (5), swimming (25), and running (15), few studies aside from that undertaken by Conley et al. (10) have used HRM to further develop self-regulation as an intrinsic ability. Likewise, few studies have explored longer-term entrainment as suggested by Dishman (12). We have therefore chosen to explore the entrainment of self-regulation during exercise to address some of the limitations and suggestions of previous work. Specifically, the present study focused on a more realistic exercise experience, a steady-state, moderate-intensity bout with the duration that was more representative of a typical exercise session. We hypothesized that multiple sessions during which HR feedback was delivered automtaically could entrain an ability to self-regulate at the same intensity. We were also curious as to whether gender might influence entrainment because several studies have suggested its influence on perceived exertion (20,21). We used mobile technology to build ad hoc instrumentation allowing us to focus on exercise performance and not simply HR averages as typically measured by HRM. Lastly, we explored postentrainment effects over a considerably longer period than has been previously reported and explored the potential for decay of these effects over time.
Twenty healthy adults (11 women and 9 men) age 18–33 yr were recruited from an undergraduate sports science course. Effective sample size was computed using the technique described in Cohen (9) with α = 0.05 and β = 0.2. All subjects completed the Physical Activity Readiness Questionnaire (7) to assess suitability for participation in the study as well as the Godin Leisure–Time Exercise Questionnaire (17) to establish baseline physical activity levels. A summary of participant characteristics can be found in Table 1. Ethics approval was received from the Human Research Ethics Office at the University of Western Australia, and all subjects provided written informed consent.
Mobile software instrument.
We developed a novel piece of software for this study (HRTrainer) that fully automated our protocol. HRTrainer is a mobile application that runs on the iOS (Apple Inc., Cupertino, CA) family of devices. It was installed on multiple fourth-generation iPod Touch devices and paired with HR sensing hardware from Wahoo Fitness (Wahoo Fitness Inc., Atlanta, GA). The Wahoo HRM consists of a transmitter and HR belt worn around the thoracic region that wirelessly broadcasts HR information to the HRTrainer software through a hardware dongle (Wahoo Key) inserted into the iPod. This combination of hardware allows the HRTrainer software to receive continuous HR input from a participant during exercise. Throughout testing, subjects were provided with identical equipment consisting of an iPod Touch running the HRTrainer software, an armband, an HRM and Wahoo Key, and headphones.
Subjects performed an initial testing session to assess their ability to maintain a moderate HR while exercising. Subject age was entered into the HRTrainer software from which HRmax was calculated using the Inbar formula [205.8 – (0.685 × age)] (26). A formula-based approach gave us the opportunity to fully encapsulate our protocol in the application. The software then calculated the upper and lower limits of a moderate-intensity training zone representing 70% to 80% of estimated HRmax. A session number and subject identifier were also entered to uniquely identify the particular testing session for later storage and retrieval using an internal database. After successfully pairing with the HRM, HRTrainer provided the subject with written and audio instructions for the session they were about to complete. The software ensured that audio instructions were played at least once before proceeding, although they could be repeated if desired. In this initial assessment (PREFB), subjects were instructed to cycle on an indoor stationary ergometer (Monark 827E/828E; Monark Exercise AB, Varberg, Sweden) while HRTrainer provided feedback. Among other things, the instructions stated that subjects were free to change their cadence or the resistance of the ergometer at any time to meet their HR goals. The feedback consisted of four low-pitched beeps indicating that the subject’s HR was below 75% of HRmax (i.e., the middle of the target training zone). An example of this feedback was built into the audio instructions subjects received. This feedback was spaced to occur no more than once every 15 s. Upon reaching 75% of HRmax, a prerecorded audio instruction told the subjects to maintain the intensity they had just achieved for an additional 20 min. During this subsequent period, no further feedback was provided. In this way, all subjects were first guided into the center of their training zone before being assessed on their current ability to maintain HR at that level. HR was recorded at 1-s intervals during the entire exercise session. Upon completing 20 min of exercise after first achieving 75% of HRmax, the software played another prerecorded message indicating the session was over. The message also invited subjects to complete an RPE response (3) and a six-item instrument (1,11) designed to measure subjects’ instrumental and affective attitudes toward biofeedback software. These attitudes were based on the generic statement “Using mobile phone software to provide HR feedback during exercise would be …,” and response options were anchored on a bipolar 1 to 7 scale (where higher scores denoted more positive perceptions). All input was done directly within the application using graphical interfaces designed for the purpose, and the software ensured all items were completed before allowing the subjects to finish. All data were subsequently stored to an internal database, and subjects were dismissed with a final audio message thanking them for their time and reminding them not to use any other forms of HR-based feedback while enrolled in the study.
In the week immediately after the initial testing session, subjects began a 5-wk feedback intervention (FB) after either a Mon/Wed or Wed/Fri schedule for a total of 10 sessions. The exercise modality remained unchanged using the same hardware, software, facilities, and cycling ergometers. However, during these sessions, subjects exercised for precisely 20 min while the software provided continuous feedback using both upper and lower limit alarms if subjects exceeded 80% of HRmax or fell below 70% of HRmax, respectively. Examples of the “too high” and “too low” alarms were played as part of the prerecorded instructions subjects received at the beginning of each session, and the audio and written instructions provided to them by the application reflected this change. As in PREFB, subjects were free to change cadence or ergometer resistance to stay in their target HR zone using the feedback to guide them. No subjective data were collected during these 10 sessions, but HR data were recorded and stored as in PREFB. After these sessions, subjects once again repeated a protocol identical to the initial PREFB assessment at exactly 1 wk (POST1), 2 wk (POST2), and 4 wk (POST3) after their last feedback session. As in the PREFB session, feedback was only provided up to 75% of HRmax, at which point a prerecorded message instructed subjects to maintain intensity for an additional 20 min, during which no further feedback was provided. Subjective attitude ratings and HR data were collected and stored as per PREFB.
Time in zone (TIZ), defined as ratio of time spent between 70% and 80% of HRmax to the overall time of exercise, was calculated for each session. We removed the initial “warm-up” period before calculation so as not to bias the results against subjects who took longer to raise their HR. For feedback sessions, this meant removing all data before the subject first reaching 70% of HRmax, whereas for the assessment sessions, we removed all data before the subject initially reaching 75% of HRmax. The TIZ values for all 10 feedback sessions were averaged after one-way repeated-measures ANOVA revealed no significant differences across session (Greenhouse–Geisser, F(3.15, 59.90) = 0.81, P = 0.50). TIZ for PREFB, the average of 10 feedback sessions (FB), and POST1, POST2, and POST3 were subsequently compared using two-way, mixed-model ANOVA with session (i.e., five levels: PREFB, FB, POST1, POST2, and POST3) and gender as factors. RPE and instrumental and affective attitudes were likewise compared using two-way mixed-model ANOVA using sessions (PREFB, POST1, POST2, and POST3) and gender as factors.
Mean TIZ for PREFB, FB, POST1, POST2, and POST3 are presented in Figure 1. Two-way mixed-model ANOVA revealed a significant difference in TIZ across sessions using Greenhouse–Geisser correction, F(2.20,39.51) = 8.98, P < 0.01,
= 0.33. Neither the gender main effect, F(1,18) = 0.48, P = 0.50,
= 0.03, nor the interaction between gender and session, F(2.20,39.51) = 2.43, P = 0.10,
= 0.12, were significant. Post hoc comparison of TIZ means with Bonferroni correction showed no significant difference between FB, POST1, POST2, and POST3, all P > 0.35. TIZ for PREFB (0.43 ± 0.37) was significantly lower than FB (0.90 ± 0.13), POST1 (0.79 ± 0.27), and POST3 (0.83 ± 0.21), all P < 0.05, all Cohen’s d > 1. There was no significant difference for TIZ between PREFB and POST2 (0.74 ± 0.33), P = 0.40, Cohen’s d = 0.86.
Descriptive statistics for RPE and the two attitudinal measures are presented in Table 2. No significant differences in RPE were observed between sessions, F(3,54) = 0.21, P = 0.89. In addition, no main effect of gender, F(3,54) = 0.11, P = 0.74, or session-by-gender interaction, F(3,54) = 0.85, P = 0.47, was observed. Similarly, no significant differences between sessions were detected for instrumental attitudes, F(3,54) = 0.33, P = 0.80, and there was no effect of gender, F(1,18) = 1.08, P = 0.31, or gender-by-session interaction, F(3,54) = 0.56, P = 0.64. Likewise, no difference in session was detected for affective attitudes, F(3,54) = 1.56, P = 0.21, and no gender effect, F(1,18) = 0.27, P = 0.61, or gender-by-session interaction was observed, F(3,54) = 1.08, P = 0.37.
The purpose of this study was to determine whether autonomous self-regulation of exercise intensity could be entrained using automated biofeedback. Our results suggest this is indeed possible using our approach. After completing 10 sessions of feedback training, subjects significantly improved TIZ without the further use of feedback to a point where results were statistically indistinguishable from the prior feedback sessions. This effect was demonstrated in all three postintervention assessments without any evidence of decay at the 4-wk mark after the feedback training intervention. To our knowledge, this is the first successful demonstration of such a long-term result regardless of the feedback mechanism used.
In keeping with previous findings (4,8,14), subjects experienced difficulty maintaining exercise intensity during the initial (PREFB) exercise session. Despite being guided by auditory feedback into the center of the target training zone before the withdrawal of feedback, subjects only managed to remain in that zone an average of 43% of the time. However, with feedback, this number more than doubled to 90%, a significant improvement with a notably large effect size. Moreover, this improvement was consistent from the onset of the 10-session intervention, showing no significant variation across the 10 feedback sessions. Such a major improvement supports the direct use of an HRM whenever regulation of exercise intensity is desired. As with the improvement during feedback, postintervention performance improved significantly at POST1 and POST3 over PREFB. The lack of statistically significant difference between PREFB and POST2 is suggested to be due to the elevated SD of POST2 compared with POST1 and POST3. However, a large effect size was still observed. In cases where improvement occurred, a large effect was observed (Cohen’s d > 1). Collectively, these findings suggest that the sensation and interpretation of afferent cues used to self-regulate exercise intensity may be refined by external feedback.
Hampson et al. (18) have suggested that “the feedback in a bout of exercise with a longer duration may allow the coordination of afferent and efferent commands such that an appropriate plateau in intensity is achieved” (p. 944). “Feedback” in this context refers to internal/afferent feedback rather than external biofeedback. We observed the HR traces and individually defined training zones of six distinct individuals from the POST1 trial, exactly 1 wk after the cessation of feedback training (Fig. 2). Having observed a significant improvement in TIZ compared with PREFB, we too were hoping to find a consistent pattern of cardiac dynamics among our participants with potential plateaus at the middle or borders of the target training zone. However, actual performance does not appear to follow such a straightforward prediction. In all cases, we can see HR drifting throughout the zone. Sometimes this drift has a specific direction (Fig. 2E, F), whereas at other times, it appears roughly centered about a particular mean (Fig. 2B, C), although still showing large volatility. In several instances, there appears to be a very high sensitivity to the boundaries of the training zone (Fig. 2B, D, F), with abrupt adjustments in intensity seen as HR approaches what Dunbar refers to as “perceptual anchors” (13) (p. 95). The effect is particularly vivid as this change in behavior occurs within no more than 2–3 bpm of the HR at which feedback was previously administered. Although we were unable to observe this “anchoring” behavior broadly, its existence may warrant further investigation by using a narrower zone to establish a more explicit plateau.
No significant changes in RPE scores were observed during the course of the study despite significant improvements in TIZ. Previous studies have used RPE, trained on observed HR, to regulate intensity in subsequent production trials where HR was not explicitly known (13,16,22). However, our results suggest that RPE alone may not be a sufficiently effective cue to ensure repeatability of effort intensity. This brings into question the efficacy of simply using RPE to regulate intensity, at least under the conditions of our study where TIZ, a moderate training intensity, and longer-term entrainment effects are desired for steady-state exercise bouts lasting at least 20 min. These results support and extend the previous work by Smutok et al. (28) that reported RPE alone was only effective entraining higher intensity efforts above 80% of HRmax. The stability of RPE also suggests that changes in fitness or economy were not responsible for the improvements in self-regulation we observed. Conversely, there is no indication that improvement in self-regulation ability affects perceived exertion.
Attitudes toward HR biofeedback did not change with consistently high scores on both the affective and instrumental measures observed before and after the intervention. It would appear that despite the additional effort involved in wearing HRM, as well as the potential cognitive load from actually receiving feedback, participants did not find the experience cumbersome enough to affect their attitudes toward the technology. Given how potentially effective aerobic training with biofeedback is, this is a reassuring result and one that suggests the population is quite open to the use of such technology when value is perceived.
Encouraging as these findings are, they do raise further questions. We were unable to observe decay in the newly acquired skill and are left without much insight into how it might attenuate over time. It is also worth considering just how much feedback is enough to achieve the same results. Would an equal number of feedback sessions undertaken over a shorter period have the same effect? Or could we accomplish similar results with fewer sessions? We are also very curious to observe how removing the initial feedback, guiding users into their target training zone before cessation of feedback, would impact these findings. Would trainees have any trouble finding their target HR zone if starting from rest with absolutely no feedback? We also have to consider different exercise lengths and how feedback duration complements postfeedback performance. For example, is training with 20-min feedback sessions sufficient to entrain self-regulation during 60-min exercise tasks or must entrainment and performance sessions be of equal duration? Lastly, the issue of pacing during exercise bouts of known duration is worth considering. The concept of “teleoanticipation” (30) suggests different self-regulatory strategies between closed versus open-loop exercise protocols, and this has been borne out experimentally (2). Our subjects were explicitly informed that their exercise sessions would last exactly 20 min once target HR was reached. We could augment our protocol and software in the future to remove this preliminary knowledge and explore performance during different exercise lengths of unknown duration. Moreover, the partial or complete automation of research protocols, as we have explored, may in itself represent a powerful new research method. In conclusion, HR biofeedback is an effective vehicle for improving self-regulation at moderate exercise intensities while direct use of biofeedback improves compliance to a moderate HR target zone immediately.
The authors do not have any conflicts of interest to declare. This study was supported by the Australian Postgraduate Award (A. Shaykevich), the Australian Research Council (B. Jackson), and the National Health and Medical Research Council (J. R. Grove).
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis Process
. 1988; 50: 179–211.
2. Baden DA, McLean TL, Tucker R, Noakes TD, Gibson ASC. Effect of anticipation during unknown or unexpected exercise
duration on rating of perceived exertion, affect, and physiological function. Br J Sports Med
. 2005; 39: 742–6.
3. Borg G. Borg’s Perceived Exertion and Pain Scales
. Champaign, IL: Human Kinetics; 1998. p. 49.
4. Braham R, Rosenberg M, Begley B. Can we teach moderate intensity activity? Adult perception of moderate intensity walking. J Sci Med Sport
. 2012; 15 (4): 322–6.
5. Broker JP, Gregor RJ, Schmidt RA. Extrinsic feedback and the learning of kinetic patterns in cycling. J Appl Biomech
. 1993; 9 (2): 111–23.
6. Castagna C, Impellizzeri FM, Chaouachi A, Bordon C, Manzi V. Effect of training intensity distribution on aerobic fitness variables in elite soccer players: a case study. J Strength Cond Res
. 2011; 25 (1): 66–71.
7. Chisholm DM, Colliss ML, Gruber N, Kulak LL. Physical activity readiness. B C Med J
. 1975; 17 (2): 375–8.
8. Chow RJ, Wilmore JH. The regulation of exercise
intensity by ratings of perceived exertion. J Cardiac Rehabil
. 1984; 4: 382–7.
9. Cohen J. Statistical Power Analysis for the Behavioral Sciences
. Hillsdale (NJ): Psychology Press, 1988. p. 567.
10. Conley MM, Gastin PB, Brown H, Shaw C. Heart rate biofeedback
fails to enhance children’s ability to identify time spent in moderate to vigorous physical activity. J Sci Med Sport
. 2011; 14 (2): 153–8.
11. Dimmock JA, Jackson B, Clear SE, Law KH. Matching temporal frame to recipients’ time orientation in exercise
messaging: does argument quality matter? Psychol Sport Exerc
. 2013; 14 (6): 804–12.
12. Dishman RK. Prescribing exercise
intensity for healthy adults using perceived exertion. Med Sci Sports Exerc
. 1994; 26 (9): 1087–94.
13. Dunbar CC, Robertson RJ, Baun R, et al. The validity of regulating exercise
intensity by ratings of perceived exertion. Med Sci Sports Exerc
. 1992; 24 (1): 94–9.
14. Ekkekakis P, Lind E, Joens-Matre RR. Can self-reported preference for exercise
intensity predict physiologically defined self-selected exercise
intensity? Res Q Exerc Sport
. 2006; 77 (1): 81–90.
15. Eriksson M, Halvorsen KA, Gullstrand L. Immediate effect of visual and auditory feedback to control the running mechanics of well-trained athletes. J Sports Sci
. 2010; 29 (3): 253–62.
16. Eston RG, Davies BL, Williams JG. Use of perceived effort ratings to control exercise
intensity in young healthy adults. Eur J Appl Physiol Occup Physiol
. 1987; 56 (2): 222–4.
17. Godin G, Shephard RJ. A simple method to assess exercise
behavior in the community. Can J Appl Sport Sci
. 1985; 10 (3): 141–6.
18. Hampson DB, Gibson ASC, Lambert MI, Noakes TD. The influence of sensory cues on the perception of exertion during exercise
and central regulation of exercise
performance. Sports Med
. 2001; 31 (13): 935–52.
19. Haskell WL, Lee I-M, Pate RR, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc
. 2007; 39 (8): 1423–34.
20. Koltyn KF, Focht BC, Ancker JM, Pasley J. Experimentally induced pain perception in men and women in the morning and evening. Int J Dev Neurosci
. 1999; 98 (1–2): 1–11.
21. Koltyn KF, O’Connor PJ, Morgan WP. Perception of effort in female and male competitive swimmers. Int J Sports Med
. 1991; 12 (04): 427–9.
22. Marriott HE, Lamb KL. The use of ratings of perceived exertion for regulating exercise
levels in rowing ergometry. Eur J Appl Physiol Occup Physiol
. 1996; 72 (3): 267–71.
23. Mihevic PM. Sensory cues for perceived exertion: a review. Med Sci Sports Exerc
. 1980; 13 (3): 150–63.
24. Moleiro M, Cid FV. Effects of biofeedback
training on voluntary heart rate control during dynamic exercise
. Appl Psychophysiol Biofeedback
. 2001; 26 (4): 279–92.
25. Pérez P, Llana S, Brizuela G, Encarnación A. Effects of three feedback conditions on aerobic swim speeds. J Sports Sci Med
. 2009; 8 (1): 30–6.
26. Robergs R. The surprising history of the “HRmax= 220 − age” equation. J Exerc Physiol
. 2002; 5 (2): 1–10.
27. Robertson RJ. Central signals of perceived exertion during dynamic exercise
. Med Sci Sports Exerc
. 1982; 14 (5): 390–6.
28. Smutok MA, Skrinar GS, Pandolf KB. Exercise
intensity: subjective regulation by perceived exertion. Arch Phys Med Rehabil
. 1980; 61 (12): 569–74.
29. Stirling LM, Tscharner Von V, Fletcher JR, Nigg BM. Quantification of the manifestations of fatigue during treadmill running. Eur J Sport Sci
. 2012; 12 (5): 418–24.
30. Ulmer HV. Concept of an extracellular regulation of muscular metabolic rate during heavy exercise
in humans by psychophysiological feedback. Experientia
. 1996; 52 (5): 416–20.