Athletic pacing has been described as the way power output, work, or energy expenditure is controlled or distributed to complete an event in the fastest possible time, having used all available resources (11,16,19). Different types of pacing strategy have been described, as observed in exercise tasks of varying durations (1,12). For example, in events lasting less than a minute, an all out pacing strategy is optimal (11), but for longer events, conserving energy is important (1,33). A negative strategy, involving a slow start and gradually increasing speed, is the most conservative and least risky approach to pacing an endurance event, but it probably does not produce the best performance (2,15,36). In contrast, positive fast start strategies deplete metabolic reserves too early (36), are rarely successful (1,12), and indicate either a lack of experience or poor anticipatory mechanisms (23). Parabolic strategy, comprising moderate starting speed, slower midsection, and fast finish, often results in faster completion of endurance events (1) but athletes must make a risk-based judgment about the maximum speed they can tolerate at the beginning without compromising performance later in the task.
A variety of explanations of how pace is controlled have been put forward, all of which have included RPE as a psychophysiological cue in effort regulation (12,14,18,33,38). Perceptions of exertion or fatigue are thought to arise from integrated afferent feedback and as such represent a global conscious awareness of the physiological state of the body (32–34). Slightly different explanations have been put forward about how RPE is used to regulate pace. In the template-matching model (38), pace is modified according to how well-experienced RPE compares against expected RPE, given the remaining distance to the end point. This model is in part based upon previous findings showing how pace is modified to maintain a scalar linear increase in RPE whereby peak RPE coincides with the end point (14). In the estimated time limit model (18), a pace is settled upon according to the amount of time such a pace can be maintained, given the corresponding RPE and the estimated time required to reach the end point. Although both of these models take into account the end point, a third model has placed particular emphasis on duration-based risk evaluation whereby pacing decisions arise from the product of momentary RPE and the proportion of the task remaining, a metric referred to as hazard score (12). What is explicitly recognized in this model is the riskiness of adopting a pace with a corresponding high RPE at the beginning of an event.
In recent years, numerous factors have been found to influence pacing behavior among athletes. These have included physiological factors such as core temperature, muscle acidosis, oxygen uptake, and carbohydrate availability (39). Pace is also affected by extrinsic factors such as ambient temperature (35), wind speed (4), the presence of competitors (10), strategic decisions (27), optic flow (24,26), and the nature of performance feedback information (23). A growing area of understanding is how individual differences influence pacing behavior, and in this respect, there has been some work on cognitive development and pacing (9,22), as well as many studies on the effect of prior experience on pacing (3,15–17,23). One type of difference that has not yet been investigated is how pacing is influenced by individual perceptions of risk and an individual’s propensity to take risks. Given the emphasis of risk in the hazard score model (12), individual risk perception and risk taking traits maybe important determinants of early pace, especially in medium or long-duration events where an excessively fast start can have a detrimental effect on overall performance (1,12). Risk perception refers to an individual’s appraisal of risk, and risk taking refers to an individual’s tendency to take risks, both of which can vary between different situations (6).
It is fairly well established that decisions involving risk are not just based upon a rational objective analysis of the circumstances, referred to as risk-as-analysis, but are also influenced by emotions associated with past experience, referred to as risk-as-feelings (31). An athlete, in deciding upon an initial pace, might use the risk-as-analysis approach by drawing upon various sources of available information such as knowledge of the end point, course terrain, other competitor behavior, ambient temperature, and so on. Equally an athlete’s pacing decisions might be informed by emotional experience, perhaps associated with feelings of exertion, excitement, anxiety, or other forms of effect typically experienced before, during, or after athletic events (14,23,25,32). As such, an athlete’s emotional intelligence, as defined by an ability to recognize, understand, regulate, and use their emotions (30), may have an important role in risk-based pacing decisions. An athlete’s sensitivity to perceive or feel changes in effort, as indicated by their RPE, may also be related to their emotional intelligence, particularly their general ability to recognize and understand emotions. Because RPE has previously been likened to an emotion (32,33), it might be expected that those with greater emotional intelligence might be particularly sensitive to changes in RPE. Furthermore, given the emphasis placed on RPE as a determinant of pace (12,14,32,33,38), it might also be expected that greater RPE sensitivity results among those with greater emotional intelligence cause them to reflect upon and perhaps alter their pace more frequently.
Of further interest is how an athlete’s perception of risk, risk taking traits, and emotional intelligence influences the extent to which they are willing to deviate their pace from a previously planned pacing strategy. On the one hand, adopting a pacing strategy that differs from the one supported through previous experience does constitute a risk yet; on the other hand, being unwilling to adapt a planned pacing strategy, perhaps in response to changing environmental conditions or competitor behavior, is also risky. Understanding how actual pace differs from planned pace may provide further insight about how individual differences in risk perception and emotional intelligence influence strategic pacing behavior during events.
In this article, we attempted to investigate how individual riskiness traits influence perceived exertion, pacing, and performance. We also compared conscious pacing expectations of the participants with their actual pacing profile. Two separate experiments are presented. In experiment 1, pace, perceived exertion, and performance were measured during a 5-km laboratory simulated time trial among novice cyclists with lower and higher-risk perceptions and risk taking traits. Experiment 2 examined the effect of risk perception and emotional intelligence upon pacing, perceived exertion, and performance, but in contrast to experiment 1, this was among experienced ultramarathon runners. Although we acknowledge that 5-km time trial cycling and ultramarathon running are very different, we believe that presenting and discussing both studies in a single article is helpful in the generalization of our findings about the association between individual perceptions of risk and pacing strategy. In experiment 1, we hypothesized that those with a greater tendency to perceive situations as risky and those who tend to take less risky decisions would adopt a more conservative relative starting pace compared with lower risk perceivers and higher risk takers, respectively. In experiment 2, we hypothesized that those with a greater tendency to perceive situations as risky would adopt a slower relative starting pace compared with lower risk perceivers. We also hypothesized that RPE and pace would differ between lower and higher emotional intelligence groups, although we are unable to predict in which direction these differences will be.
EXPERIMENT 1: THE EFFECT OF RISK-PERCEPTION AND RISK-TAKING ON PACING STRATEGY AMONG NOVICE TIME-TRIAL CYCLISTS
Twenty participants (female = 5, male = 15) were recruited from the University of Essex (age, 20.9 ± 1.1 yr; stature, 176.9 ± 8.5 cm; and body mass, 79.6 ± 12.2 kg). All participants were healthy and participated in moderate physical activity of at least 30 min, three times per week. Although they could ride a bicycle, all participants were novice cyclists in the sense that they had not previously been members of a cycling club or participated in time trialing or any other forms of competitive cycling. Each participant provided written informed consent to take part in this study, which was approved by the University of Essex ethics committee.
A two-way between- and within-subjects experimental design was used. All participants completed a risk taking and risk perception questionnaire (6), the results of which were used to split participants into higher and lower risk takers and higher and lower risk perceivers (between-subjects risk factor). Pacing predictions were measured in participants before actually performing a 5-km laboratory-based cycling time trial (within-subjects prediction factor). Participants provided an RPE (7) every kilometer.
Risk perception and risk taking measurements.
Perceptions of risk and risk taking traits were measured using the revised domain-specific risk taking scale (DOSPERT) (6). The DOSPERT comprises two 30-item scales, one for risk taking and one for risk perception, the responses to which are quantified using a 7-point Likert scale. The possible range of DOSPERT scores is 30–210 with a high-risk perception score indicating greater tendencies to perceive situations as risky and a high-risk taking score indicating greater tendencies to take risks. English population normative scores for risk perception is 121 ± 7.3 and for risk taking is 116 ± 7.3 (6). All participants completed the DOSPERT before making their pacing predictions and performing the cycling time trial.
Pacing predictions and 5-km time trial cycling ergometry.
Before performing the cycling time trial, all participants were asked to predict what pace they believed they would adopt during each kilometer segment. The predictions were measured using a Microsoft Excel macro in which participants were free to alter parameters of a pacing graph. Similar to the graphs presented in Figure 1, participants were presented with an average speed line, and then for each kilometer segment, they were able to adjust the percentage of pacing deviation from average speed, either faster or slower, until they were happy with the overall shape of the pacing profile. The macro was programmed such that it would not accept a pacing profile that did not mathematically balance exactly to the average speed line, and a numerical indicator was provided so that participants could see how far away from balancing their predictions they were. Essentially the macro ensured that across the whole trial, the predictions had an equal number of percentage points above and below the average speed line. Although the pacing prediction task was fully explained to participants, no advice on pacing strategy was given. Participants were permitted to use the prediction macro with no time restrictions until they were content with the predicted pacing profile they had made. A percent deviation from average speed prediction task was considered more appropriate for novice time trial cyclists than absolute speed predictions, also serving to facilitate direct comparison with normalized actual pacing data.
Time trial cycling ergometry.
After completing the prediction task, participants were asked to lay supine in a quiet room for 15 min and resting HR was measured using a Polar 610i HR monitor (Polar; Kempele, Finland). All participants completed a 5-min warm-up at 80% HR reserve and 5-km time trial using a laboratory racing bicycle mounted on a calibrated Kingcycle air-braked cycling ergometer (EDS Portaprompt Ltd., High Wycombe, UK). Handlebar position and seat height was adjusted to suit each participant. During both the warm-up and time trial, participants were permitted to self-select gearing and cadence. Cycling speed and elapsed distance were displayed to participants using a calibrated Revolution Velocity 20 wireless cycle computer (Revolution Velocity, Manchester, UK). Given that novice cyclists were used in this study, the secondary purpose of the warm-up was of familiarization in the use of the bicycle, gears, and cycling ergometry. Once the warm-up was complete, participants were given the opportunity to ask questions about the use of the bicycle or any other experimental procedures. All participants were then instructed to complete the cycling time trial in their fastest possible time. Participants were not provided with any pacing or performance guidance, advice, or instructions. Elapsed time was recorded each kilometer from which average cycling speed was calculated.
Perceived exertion and hazard score measurements.
At every kilometer during each time trial, participants were asked to provide an overall RPE using the Borg 6–20 RPE scale (7). Before testing, each participant was familiarized with the RPE scale, which was administered in accordance with published standardized instructions (7). As previously described (12), hazard score was calculated for each kilometer segment as RPE multiplied by the percentage of distance remaining.
Lower and higher risk taking and risk perception groups were created by a median split of DOSPERT scores. Risk taking and risk perception group differences in performance, expressed as average cycling speed, were analyzed using independent sample t-tests. Between-subject variance in cycling speed was normalized by calculating kilometer-by-kilometer percentage deviation from overall average cycling speed. Two-way between-subjects ANOVA and within-subjects ANOVA were used to analyze kilometer segment differences in pace, RPE, and hazard score between the lower and higher risk taking and risk perception groups. Two-way within-subjects ANOVA was used to make comparisons between predicted and actual pace. Because pace data were normalized, the sum of all segment point always equaled zero; therefore, there were no group main effects. Significant risk perception and risk taking group pacing interactions were followed up using one-tailed independent-samples t-tests. Prediction-by-segment interactions were followed up using one-tailed paired-samples t-tests. All results are expressed as means ± 1 SD and effect sizes as eta squared (η 2). An alpha level of 0.05 was used to indicate statistical significance.
Risk perception and risk taking scores.
Total DOSPERT scores for the lower and higher risk perception groups were 96.2 ± 14.5 and 131.8 ± 7.4, respectively. DOSPERT scores for lower and higher risk taking groups were 99.5 ± 9.6 and 127.9 ± 10.0, respectively.
Performance and pacing.
There was no difference in average cycling speed between lower and higher risk perceivers (29.9 ± 3.8 vs 29.3 ± 5.4 km·h−1, t 18 = −0.3, P = 0.79, η 2 < 0.01) or lower and higher risk takers (28.9 ± 5.5 vs 30.3 ± 3.6 km·h−1, t 18 = −0.6, P = 0.53, η 2 = 0.02).
For risk perception, two-way within- and between-subjects ANOVA revealed a risk perception group-by-segment interaction for pace (F 4,72 = 2.7, P = 0.035, η 2 = 0.08) as well as a segment main effect (F 4,72 = 13.0, P < 0.001, η 2 = 0.39). The interaction indicates that the kilometer-by-kilometer changes in pace differed between lower and higher risk perception groups and post hoc one-tailed independent samples t-tests, which found a slower relative starting pace among the higher risk perception group (Fig. 1A). For risk taking, there was a segment main effect for pace (F 4,72 = 12.8, P < 0.001, η 2 = 0.39), and although there was no group-by-segment interaction (F 4,72 = 2.3, P = 0.065, η 2 = 0.07), it did approach significance (Fig. 1B). An association was found between deviation of pace from average speed during the first kilometer and risk perception (r 20 = −0.457, P = 0.022), as presented in Figure 1C, and with risk taking (r 20 = −0.426, P = 0.03), as presented in Figure 1D.
Ratings of perceived exertion and hazard score.
There was no risk perception group-by-segment interaction for RPE (F 4,72 = 0.2, P = 0.95, η 2 < 0.01) and no group main effect (F 1,18 = 2.7, P = 0.12, η 2 = 0.13), but there was a segment main effect (F 4,72 = 95, P < 0.001, η 2 = 0.84), which means that regardless of risk perception group, RPE increased during the time trial (Fig. 2A). There was no risk perception group-by-segment interaction for hazard score (F 4,72 = 0.8, P = 0.54, η 2 = 0.01) and no group main effect (F 1,18 = 2.6, P = 0.13, η 2 = 0.12), but there was a segment main effect (F 4,72 = 707, P < 0.001, η 2 = 0.97), which means that regardless of risk perception group, hazard score decreased during the time trial (Fig. 2B).
There was no risk taking group-by-segment interaction for RPE (F 4,72 = 0.1, P = 0.98, η 2 < 0.01), but there was a group main effect (F 1,18 = 5.9, P = 0.026, η 2 = 0.25) and a segment main effect (F 4,72 = 95, P < 0.001, η 2 = 0.84), which means that although the lower risk takers had lower average RPE scores (group main effect), the pattern of increasing RPE throughout the time trial was the same between lower and higher risk taking groups (segment main effect) (Fig. 2C). There was no risk taking group-by-segment interaction for hazard score (F 4,72 = 1.1, P = 0.36, η 2 < 0.01), but there was a group main effect (F 1,18 = 4.8, P = 0.042, η 2 = 0.21) and a segment main effect (F 4,72 = 720, P < 0.001, η 2 = 0.97), which means that although the lower risk takers had lower average hazard scores (group main effect), the pattern of decreasing RPE throughout the time trial was the same between lower and higher risk taking groups (segment main effect) (Fig. 2D).
Actual versus predicted pacing.
Two-way within-subjects ANOVAs were used to compare kilometer-by-kilometer changes in pace (segment factor) between predicted and actual pace (prediction factor). For all subjects combined, there was a prediction-by-segment interaction (F 4,76 = 10.4, P < 0.001, η 2 = 0.29) as well as a segment main effect (F 4,76 = 7.2, P < 0.001, η 2 = 0.24). Among lower risk perceivers, there was a prediction-by-segment interaction (F 4,36 = 3.2, P = 0.025, η 2 = 0.21) but no segment main effect (F 4,36 = 1.7, P = 0.16, η 2 = 0.20). Among higher risk perceivers, there was a prediction-by-segment interaction (F 4,36 = 8.2, P < 0.001, η 2 = 0.33) and a segment main effect (F 4,36 = 9.7, P < 0.001, η 2 = 0.30). Among lower risk takers, there was a prediction-by-segment interaction (F 4,36 = 7.8, P < 0.001, η 2 = 0.36) and a segment main effect (F 4,36 = 6.6, P < 0.001, η 2 = 0.23). Among higher risk takers, there was a prediction-by-segment interaction (F 4,36 = 3.0, P = 0.03, η 2 = 0.18) and a segment main effect (F 4,36 = 2.7, P = 0.04, η 2 = 0.28). Where interactions are reported, this indicates kilometer-by-kilometer differences between predicted and actual pace. The exact nature of these differences was examined using post hoc one-tailed paired-samples t-tests and is presented in Figure 3A–E.
EXPERIMENT 2: THE EFFECT OF RISK-PERCEPTION AND EMOTIONAL INTELLIGENCE ON PACING STRATEGY AMONG EXPERIENCED UTRAMARATHON RUNNERS
Thirty-four participants (female = 2, male = 32) were recruited for this study from the field of runners participating in the Stour Valley Path 100-km ultramarathon race (SVP100). Mean ± 1 SD age, stature, and body mass were 39.9 ± 7.6 yr, 178.4 ± 7.1 cm, and 74.8 ± 9.2 kg, respectively. All participants were experienced endurance runners who, during a 2-yr period before the study, had participated in 3.8 ± 5.1 competitive marathons, 10.1 ± 17.4 competitive runs shorter than marathon distance, and 5.6 ± 8.9 ultramarathons over an average distance of 92.9 ± 39.9 km. The large SD values for competitive running history were due to a minority of runners competing in greater than 20 events in the previous 2 yr resulting in positively skewed data distribution for marathons (skewness = 2.1, SE = 0.4), ultramarathons (skewness = 2.9, SE = 0.4), and runs shorter than marathon distance (skewness = 4.1, SE = 0.4), noting that for normally distributed data, skewness ≅1. During the 3-month period preceding the study, participants ran on average 4.5 ± 1.5 times per week covering an average weekly distance of 61.4 ± 23.0 km. Each participant provided written informed consent to take part in this study, which was approved by the University of Essex ethics committee.
Similar to experiment 1, a two-way between- and within-subjects experimental design was used. All participants completed the risk perception element of DOSPERT (6) as well as an emotional intelligence questionnaire [Schutte Emotional Intelligence Scale (SEIS)] (30). In this second experiment, emotional intelligence measurements were added because, after considering the lack of differences in RPE and hazard score between the risk groups in experiment 1, we speculated that the way individuals perceive and feel effort might be related to their general ability to recognize, understand, and regulate emotions. To minimize questionnaire fatigue, participants were not asked to complete the risk taking element of DOSPERT (6), especially because the covariance with risk perception is very high. Furthermore it was felt that risk perception, as a measure of cognition, would yield stronger explanatory data about the influence of psychological processes on pacing compared with risk taking, which is a measure of behavior. Emotional intelligence was measured to investigate whether an athlete’s ability to recognize, understand, regulate, and use their emotions has any bearing on their risk-based pacing decisions. Median split of DOSPERT and SEIS results were used to create higher and lower risk perception groups and higher and lower emotional intelligence groups (between-subjects factors). All runners were asked to predict split times for each checkpoint before running the SVP100, and the predictions were later compared with actual split times (within-subjects prediction factor). An RPE was collected at each checkpoint from all participant runners.
Risk perception and emotional intelligence measurements.
Perceptions of risk were measured using the revised domain-specific risk taking scale (DOSPERT) (6), identical to the method described in experiment 1. Emotional intelligence was measured using the SEIS (30). The SEIS comprises 33 items, the responses to which are quantified using a 5-point Likert scale. The possible range of SEIS scores is 33–195, whereby a higher score represents greater emotional intelligence. All participants completed both the DOSPERT and SEIS before running in the SVP100.
Pacing predictions, ultramarathon performance.
Before performing the SVP100 ultramarathon, participants were asked to carefully examine the course profile and predict their split times for each check point (CP). The predicted split times were used to calculate predicted average running speed for the whole race in relation to which predicted running speed for each segment was expressed as a percentage deviation. In total, there were six CP throughout the race before the finish point (CP1 = 18.7 km, CP2 = 36.5 km, CP3 = 52.3 km, CP4 = 66.2 km, CP5 = 79.0 km, CP6 = 91.9 km, and finish point = 99.0 km). The distances between CP were as follows: Start–CP1 = 18.7 km, CP1–2 = 17.8 km, CP2–3 = 15.7 km, CP3–4 = 14.0 km, CP4–5 = 12.7 km, CP5–6 = 13.0 km, CP6–finish = 7.1 km. A researcher was positioned at each CP and recorded the time at which each runner arrived. Checkpoint arrival times were later used to calculate split running times and average speeds for each segment.
Perceived exertion and hazard score measurements.
At each CP, each runner provided an RPE using the Borg 6–20 RPE scale (7). Before the race, each runner was familiarized with the RPE scale, which was administered in accordance with published standardized instructions (7). As previously described (12), hazard score was calculated for each CP as RPE multiplied by the percentage of distance remaining.
A median split of DOSPERT score and SEIS scores were used to create lower and higher risk perception and emotional intelligence groups, respectively. Between-subject variance in segment running speeds were normalized by expressing them as percentage deviations from average running speed for the whole race. Main effects, interactions, and post hoc tests for race segments, risk perception group, and emotional intelligence group for pace, RPE, and hazard score were all analyzed using exactly the same statistical methods as those described in experiment 1. All results are expressed as means ± 1 SD and effect sizes as partial eta squared (η 2). An alpha level of 0.05 was used to indicate statistical significance to test the first two hypotheses, but 0.025 was used for the third two-tailed hypothesis.
Total DOSPERT scores for the lower and higher risk perception groups were 111.5 ± 13.9 and 141.9 ± 8.0, respectively. Total SEIS scores for lower and higher emotional intelligence groups were 107.4 ± 8.9 and 127.3 ± 6.8, respectively. The ultramarathon runners had greater perceptions of risk compared with the 5-km time trial cyclists reported in experiment 1 (126.3 ± 22.5 vs 114.0 ± 21.4, t 52 = 2.0, P = 0.027, η 2 =0.01).
Risk perception group comparisons of performance and emotional intelligence.
Average completion time was 822 ± 66 min for the lower risk perceivers and 800 ± 80 min for the higher risk perceivers. There was no difference in average running speed between lower and higher risk perceivers (7.3 ± 0.7 vs 7.5 ± 0.8 km·h−1, t 32 = −0.9, P = 0.4, η 2 = 0.01). However, lower risk perceivers exhibited lower emotional intelligence compared with the higher risk perceivers (76.3 ± 10.5 vs 85.1 ± 13.7, respectively, t 32 = −2.1, P = 0.022, η 2 = 0.12).
Risk perception and emotional intelligence group comparisons of pace.
Two-way within- and between-subjects ANOVA showed a risk perception, a group-by-segment interaction for pace (F 6,192 = 2.9, P = 0.04, η 2 = 0.02), and a segment main effect (F 6,192 = 227, P < 0.001, η 2 = 0.87). The interaction indicates that segment-by-segment changes in pace differed between lower and higher risk perception groups and post hoc one-tailed independent samples t-tests, which found a slower relative starting pace among the higher risk perception group (Fig. 4A). There was no emotional intelligence group-by-segment interaction for pace (F 6,192 = 0.7, P = 0.61, η 2 < 0.01), but there was a segment main effect (F 6,192 = 211, P < 0.001, η 2 = 0.87), which means that regardless of emotional intelligence group, running speed decreased during the race (Fig. 4B). An association was found between deviation of pace from average speed during the first leg and risk perception (r 34 = −0.513, P = 0.002), as presented in Figure 4C, but not with emotional intelligence (r 34 = 0.259, P = 0.139), as presented in Figure 4D.
Risk perception and emotional intelligence group comparisons of ratings of perceived exertion and hazard score.
There was no risk perception group-by-segment interaction for RPE (F 6,156 = 1.5, P = 0.19, η 2 = 0.03) or group main effect (F 1,26 = 0.5, P = 0.50, η 2 = 0.02), but there was a segment main effect (F 6,156 = 17.0, P < 0.001, η 2 = 0.38), which means that regardless of risk perception group, RPE increased throughout the ultramarathon (Fig. 5A). There was no emotional intelligence group-by-segment interaction for RPE (F 6,156 = 0.3, P = 0.95, η 2 = 0.01) or emotional intelligence group main effect (F 1,26 = 1.3, P = 0.26, η 2 = 0.05), but there was a segment main effect (F 6,156 = 16.2, P < 0.001, η 2 = 0.38), which means that regardless of emotional intelligence group, RPE increased throughout the ultramarathon (Fig. 5B).
There was no risk perception group-by-segment interaction for hazard score (F 6,156 = 1.4, P = 0.24, η 2 < 0.01) or group main effect (F 1,26 = 1.0, P = 0.33, η 2 = 0.04), but there was a segment main effect (F 6,156 = 641.0, P < 0.001, η 2 = 0.96), which means that regardless of risk perception group, hazard score decreased throughout the ultramarathon (Fig. 5C). There was no emotional intelligence group-by-segment interaction for RPE (F 6,156 = 0.6, P = 0.69, η 2 < 0.01) or emotional intelligence group main effect (F 1,26 = 1.2, P = 0.28, η 2 = 0.05), but there was a segment main effect (F 6,156 = 626, P < 0.001, η 2 = 0.96), which means that regardless of risk perception group, hazard score decreased throughout the ultramarathon (Fig. 5D).
Actual versus predicted pacing.
Two-way within-subjects ANOVA was used to compare CP-by-CP changes in pace (segment factor) between predicted and actual pace (prediction factor). For all subjects combined, there was a prediction-by-segment interaction (F 6,198 = 34.1, P < 0.001, η 2 = 0.15) and a segment main effect (F 6,198 = 123, P < 0.001, η 2 = 0.70). Among lower risk perceivers, there was a prediction-by-segment interaction (F 6,96 = 38, P < 0.001, η 2 = 0.20) and a segment main effect (F 6,96 = 87.5, P < 0.001, η 2 = 0.72). Among higher risk perceivers, there was a prediction-by-segment interaction (F 6,96 = 8.9, P < 0.001, η 2 = 0.11) and a segment main effect (F 6,96 = 44.3, P < 0.001, η 2 = 0.68). Among the lower emotional intelligence group, there was a prediction-by-segment interaction (F 6,96 = 14.4, P < 0.001, η 2 = 0.16) and a segment main effect (F 6,96 = 73, P < 0.001, η 2 = 0.67). Among the higher emotional intelligence group, there was a prediction-by-segment interaction (F 6,96 = 23.6, P < 0.001, η 2 = 0.15) and a segment main effect (F 6,96 = 50, P < 0.001, η 2 = 0.74). Where interactions are reported, this indicates segment-by-segment differences between predicted and actual pace. The exact nature of these differences were examined using post hoc one-tailed paired-samples t-tests and are presented in Figure 6A–E.
The main finding, evident in both experiments, is that perceptions of risk are significantly associated with pacing strategy. Despite the differences in exercise mode and duration of both experiments, those with a greater perception of risk were found to adopt a more conservative initial pacing strategy, and therefore, we accept our hypothesis. In both experiments, moderate correlations were also found between starting pace and risk perception, adding confidence to these findings, although we note in Figure 4C that there were two runners with a greater perception of risk who also adopted a relatively fast starting pace.
In both experiments, the higher risk perception groups had an initial pace that was on average 8% slower than the lower risk perception groups. It is important to note that the novice cyclists in experiment 1 started at a pace that turned out to be just below their overall average speed, whereas the experienced ultraendurance runners started at a pace around 30%–40% higher than their average speed. The 5-km time trial cyclists in experiment 1 progressively increased their speed throughout the trial, whereas the ultramarathon runners progressively decreased their speed, which, considering the differences in duration of these events, is consistent with previous observations (1,12,20). Although the interaction between experience and risk perception was not directly measured in this study, it has been observed in other decision-making contexts (13) and therefore warrants further investigation. Although the lower risk perception groups adopted a faster starting pace, this did not result in a better overall performance compared with the higher risk perception groups in either the cycling time trial of experiment 1 or the ultramarathon of experiment 2. Our experiments were limited by the use of between-subjects designs that, although necessary to create different risk perception groups, made it more difficult to determine how variations in risk perception and associated pacing differences actually affected performance. In particular, it is not possible to conclude whether lower or higher risk perception is most beneficial to pacing and performance, although we did find that the more experienced ultramarathon runners did have greater perceptions of risk compared with the novice time trial cyclists. This is something that does require further experimentation to establish whether altering an individual athlete’s perception of risk, perhaps through some psychological intervention, results in different pacing decisions and performance.
In both experiments, there was no difference between lower and higher risk perception groups in the pattern of change in RPE or hazard score, even though there was a difference in pace. There are several explanations for this. The first and perhaps the most simple explanation is that, as suggested by the RPE template model (14,17,38), pace is adjusted to ensure a good match between experienced and expected RPE. Thus, contrary to our previous discussion, the faster pace adopted by lower risk perceivers was from the participants’ perspective no more risky than the slower relative pace adopted by the higher risk perception group, as indicated by similar RPE responses and hazard.
An alternative explanation is that consciously experienced RPE is in fact a result of top-down processing, whereby an individual’s own particular level of risk perception modifies sensations emanating from afferent feedback about the internal physiological state of the body before reaching conscious experience. As such, experienced perceptions of exertion already take into account risk perception orientations such that the faster pace of the lower risk perception group and the slower pace of the higher risk perception group produce the same RPE. This notion of RPE being the product of top-down processes is in fact consistent with our previous work (24,26). Furthermore, it has been suggested that top-down processes provoke different perceptual outcomes because of the affective value and subjective emotional experience that are associated with the internal sensations (8,21,28).
In experiment 2, there were no pacing, RPE, or hazard score differences between emotional intelligence groups and therefore we reject our hypothesis. However, compared with higher risk perceivers, the lower risk perceivers did have a slightly lower emotional intelligence score, indicating a decreased tendency to appraise, express, regulate, and use emotions. These results, taken together with the differences in pacing, perhaps suggest that lower risk perceivers have less of a reliance on feelings than higher risk perceivers in making pacing decisions. We stress that because there was only a slight difference in emotional intelligence between the risk perception groups, this conclusion is not convincingly supported yet does have intuitive appeal. In the context of the “risk-as-feelings” model (31), it is interesting to note that affective factors naturally carry much less certainty than the kinds of informational sources associated with “risk-as-analysis.” It is therefore surprising that higher risk perceivers have a greater tendency to appraise, express, regulate, and use emotions. A potential explanation, which needs further investigation, is that in making decisions, athletes with higher perceptions of risk draw on as many sources of information as possible including their feelings, whereas low risk perceivers may be willing to make decisions based on fewer sources of information. In a health context, emotional intelligence has been found to mediate the relationship between individual traits and health behaviors including exercise (5,29,37). Much more work is needed to understand how emotional intelligence mediates relationships between risk perception traits and athletic decision-making.
A negative pacing pattern was observed in the 5-km cycling time trial of experiment 1, whereas a positive pacing pattern was observed in the ultramarathon of experiment 2. Given the huge contrast in event duration, the respective differences in pacing strategy between the two experiments are broadly what we expected to see and consistent with previous findings (1,12). What is interesting is the large discrepancies that were observed between predicted and actual pace in each experiment, especially at the beginning and the end sections. The novice cyclists actually started over 10% slower at the beginning of the time trial compared with their predictions, but in the complete opposite direction, the ultramarathon runners performed much faster at the beginning of the run than predicted. The differences between predicted and observed pacing might be a consequence of differences between novice athletes, who have much less experience upon which to base their predictions, compared with experienced athletes, whose predictions might be more accurately based on a wealth of previous experience. Consequently, the novice time trial cyclists may have set their initial pace in the belief that it was consistent with their prediction. However, as the time trial progressed, and as they came to realize their ability, they sped up. The ultramarathon data is more difficult to account for because, as experienced runners, a good match between actual and predicted pace might be expected. The effect could have been caused by being excessively cautious with the prediction but equally could just mean that the conscious awareness of behavioral intentions do not accurately represent the performance template, perhaps because the way they are mentally represented is as feelings rather than as split times and average speeds.
Lower risk perceivers adopt a faster start than higher risk perceivers, although there is no difference in RPE or hazard score. Higher risk perceivers reported a slightly greater tendency to appraise, express, regulate, and use emotions, perhaps suggesting that they have a greater reliance on emotions in evaluating risks and making pacing decisions. Both studies highlight the need for more work in understanding how athletic decision-making is influenced by perceptions of risk and emotional intelligence, and whether risk perception modification interventions or emotion regulation training can be used to improve athletic decision-making. One question of particular interest is, “In seeking out and processing information to make decisions, are higher risk perceivers more sensitive to interoceptive feedback and their feelings compared with lower risk perceivers who might depend more on exteroceptive feedback and performance feedback?” However, the most important finding from both of our experiments is that perceptions of risk are associated with different approaches to pacing the start of an event.
The authors wish to acknowledge Carl Jeffs, Aleisha Moore, Joseph McGuiness, Terry Baker, Paige Wilson, and Kelly Murray for their help with data collection.
This study was funded internally by the University of Essex and no other source of external funding or support was used. None of the authors have professional relationships with companies or manufacturers who will benefit from the results of this study.
None of the results presented in this study constitute endorsement by the American College of Sports Medicine.