Journal Logo


Information Acquisition Differences between Experienced and Novice Time Trial Cyclists


Author Information
Medicine & Science in Sports & Exercise: September 2017 - Volume 49 - Issue 9 - p 1884-1898
doi: 10.1249/MSS.0000000000001304
  • Free


It is important for athletes to use their available energy effectively to perform optimally and avoid fatigue during exercise, so that “all energy stores are used before finishing a race, but not so much that a meaningful slowdown occurs.” (8,17,28) Pacing strategy is an essential aspect of competitive prolonged athletic performance and refers to the variation of speed during an event by regulating the rate of energy expenditure (17–20,27). Where completion time is the measure of success, pacing strategy has an influence over success in events lasting longer than 60 s (1).

Several factors are known to influence the pacing strategy that an athlete adopts including the duration of the event (8), presence of a competitor (7,56), environmental conditions (40), previous experience (34), perceptions of exertion (48), and the availability and veracity of performance feedback information (14,35). Previous models of pacing place a lot of emphasis on an athlete’s awareness of changes to the internal physiological state of their body, experienced as perceived exertion, in relation to their progress toward the endpoint as informed through various forms of feedback. According to Teleoanticipation Theory (49) and later on the Central Governor Model (39), a “central governor” anticipates exercise and presets a pacing strategy based on the endpoint or duration of exercise. In a more recent manifestation of Central Governor Model, more complex information-processing mechanisms have been proposed in which rate of change of perceived exertion is evaluated in the light of expected duration or distance of an event and modified through appropriate alternations in pace (47). The Psychobiological Model similarly supports the notion of effort-related decisions about pace in the context of event duration, but argues that such decisions are entirely conscious and that subconscious processes, such as those proposed by the Central Governor Model, are inapposite (33). The linear relationships found between RPE and the proportion of completed event are such that the RPE gradient was found to peak in coincidence with the expected endpoint (15,18,30).

In an attempt to factor for varying uncertainty about pace during endurance events, a model has been specified, whereby risk is expressed as the proportion of the remaining task multiplied by their momentary RPE, a variable the authors refer to as hazard score (9). An appealing feature of the hazard score model is that the further an athlete progresses, the lower hazard score becomes, thus explaining how athletes are sometimes able to risk performing very intense spurts of energy toward the end of an event when the risk of not completing, as a consequence of doing so, is relatively low. An alternative model proposed that pacing decisions are based on the estimated time that present power output can be maintained, as judged against the duration or length of the task (22). More recent suggestions of how pace is regulated have drawn on the decision making literature (41) and the interdependence of perception and action in attempting to account for pacing behavior in environmentally complex situations (44).

Whatever theory of pacing is subscribed to, all emphasize the knowledge of proximity to the endpoint as a key determinant of pacing strategy. However, the importance placed on endpoint knowledge in pacing models is based on experimental evidence that was collected using limited indirect observation methods where participants have been deceived about, or deprived of, progression or performance feedback information (29). A number of studies have used false feedback about distance or time to understand the importance of feedback and the use of knowledge during exercise. Studies have found that deceiving athletes about the duration of exercise, by providing false or no knowledge about the exercise endpoint, leads to increased RPE and a different pacing strategy caused by an incorrect allocation of physiological resources (3,12). Experience of using blind, true, and false performance feedback has also been found to provoke different types of learnt pacing strategies (37).

Feedback deception and blinding experimental methods have been the dominant approaches used to understand how athletes use information to pace themselves. Deductions about the significance and role of particular types of performance information are made based on what happens to the pace if that information is altered or removed. The underlying logic is that if after altering or removing a particular source of information pacing or performance worsens, then it can be inferred that that information source has an important contributory role. It has been this approach that has led to the emphasis placed on knowledge of the endpoint in various pacing models.

There are several limitations to this information-knockout approach. The first is the focus on singular sources of information and the lack of investigative sophistication in understanding how athletes interpret various sources of information in conjunction with each other. For example, the importance athletes place on speed or power information to make pacing decisions could potentially vary according to how much time or distance has elapsed, or according to environmental conditions or competitor behavior. A further, but related, limitation is that knockout and deception studies have not investigated within-trial changes in the emphasis placed on certain types of feedback. For example, potentially an athlete may be more concerned with average speed in the first half of a race and then become more interested in elapsed time or distance towards the end of an event. The final limitation is the inability to understand individual differences in feedback preferences, which could vary according to past experience or the outcome measure by which they appraise their achievement success. A threat to the validity of previous pacing models is the reliance on limited deception and blinding methods, which necessitated indirect interpretation regarding the importance of endpoint awareness as a determinant of athletic pace. It is this point that the present study intended to redress.

A more direct method of measuring what information athletes seek and use during self-paced exercise will greatly improve our understanding of pacing decisions and, to our knowledge, this have never been achieved. In one study the frequency with which children looked at elapsed time during a time-limited run was measured from a video recording and it was found that they looked at the watch more often towards the end of the run (6). Although the methods of measuring information acquisition in this study were quite basic, eye-tracking technology does provide a more sophisticated method of directly measuring what information athletes look at during self-paced exercise. Unlike previous deception and information-knockout studies, the precision with which information acquisition behavior can be measured using eye-tracking technology is able to overcome the limitations of deception studies discussed earlier. Importantly, eye-tracking enables detailed information to be gathered about how athletes acquire information in dynamic and conjunctive ways during an exercise trial, as well as how they learn to use information differently with experience to pace themselves.

The use of eye-tracking technology in sport is a powerful method (11) that has enabled researchers to develop better insights about perceptual-cognitive mechanisms of sport performance (23,32). Mobile eye-tracking technology has proven especially versatile in allowing researchers to collect data in many different sports domains where performance is dependent on the ability perceive and process complex information in often fast moving environments. In such situations, the visual is the dominant mode of sensory feedback in the perceptual-action coupling (31), a system in which attention to external cues enables the kind of adaptive movements required for the successful performance of motor tasks, such as catching or striking a ball. In the context of cycling, eye-tracking has provided useful insights about the role of visual behavior in balance and steering (50,52) but has not been used to understand information pick-up as part of the perceptual-action processes in regulating pace (44). Eye-tracking technology has also provided considerable insights about differences in perceptual-cognitive mechanisms between expert and novice performers (23,54), and this approach has great potential in developing a better understanding of information acquisition and decision-making during self-paced cycling. Generally, previous research has suggested that experts across many sports domains tend to look at task-relevant information less frequently and for less time than novices (23,26). This has a relevance to pacing theory because it raises the question of whether differences exist between expert and novice cyclists about what information feedback they consider to be task relevant, and whether differences exist in how frequently they refer to such information and for how long.

Although we acknowledge that the use of eye-tracking technology is fairly commonplace in sport domains and expertise research, the present study used eye-tracking technology in an original way to better understand information acquisition and pacing behavior in cyclists. The purpose was, for the first time, to directly measure what information cyclists look at while performing a time trial (TT), and to compare the information-acquisition strategies of novice and experienced cyclists. We hypothesized that experienced cyclists would look at fewer sources of information, and would seek out information less frequently compared to novices.



Experienced (n = 10) and novice male cyclists (n = 10) were recruited for this study from the University of Essex and local cycling clubs. Mean ± 1 SD age, stature, and body mass for the experienced cyclists was 38.6 ± 11.3 yr, 176.6 ± 6.9 cm, and 74 ± 9.4 kg for the experienced cyclists, and for the novice cyclists was 36.1 ± 9.9 yr, 178.5 ± 6.7 cm, and 80.2 ± 8.7 kg. The experienced cyclists were recruited from local cycling clubs and had participated in competitive 16.1-km TT for an average of 14.1 ± 13 yr. During the 6 months preceding the study, the experienced cyclists had on average trained each week on 4.7 ± 1.1 occasions for a total of 8.5 ± 2.1 h. The novice cyclists were recruited from the University of Essex staff and students and, although they could all ride a bicycle, they had never trained for, or participated in competitive cycling events of any kind. In an attempt to control for fitness, only physically active individuals were recruited to the novice group who had on average trained each week on 2.8 ± 0.8 occasions for a total of 4.6 ± 1.1 h across a range of different sports that did not involve 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 mixed experimental design (experience-by-segment) was used in which we compared pace, performance and visual information acquisition between novice and experienced cyclists (between-subjects experience factor) during a 16.1-km cycling TT every 4 km (within-subjects segment factor). All participants performed a 16.1-km familiarization TT (TTFAM) and then had a recovery period of 5 to 10 d before completing the 16.1-km experimental TT (TTEXP). During each TT completion time (s), speed (km·h−1), power output (W), distance (km), pedaling cadence (rpm) and HR (bpm) was measured. RPE was recorded every 4 km. Participants wore a monocular eye-tracking device for familiarization purposes during TTFAM and then to measure the type, duration and frequency of information they looked at during TTEXP.


Before each TT participants were asked to refrain from ingesting caffeine for at least 6 h, alcohol for 24 h, and food for 2 h before testing. Participants were also asked not to train or engage in heavy physical work for 24 h before testing. On the first laboratory attendance, each participant had their body mass and stature measured and was briefed as to the requirements of the trial but not the purpose of the study. Participants also completed a short training history questionnaire. After all tests had been completed, participants were debriefed about the purpose of the study.

Cycling ergometry and video simulation

All cycling tests were performed on a Velotron (3D) Racer Mate ergometer with RealVideo simulation software (Racermate, Seattle, WA). The 16.1-km TT duration was selected because this is a common format used in the UK and one which the experienced cyclists used in this study were most accustomed. All cycling tests were performed at the same time of day ± 1 h to control for circadian variation in outcome measures. Before each TT, participants performed a standardized 5-min self-paced warm-up. Participants were instructed to complete the TT in the fastest possible time. They were not provided with any information acquisition or pacing guidance.

During each TT, a RealVideo simulated cycling course was projected onto a wall in front of and slightly offset to the right of the cycling. The projected video footage was coupled in a multiplicative way to the cyclists' actual power output such that any alteration in speed was instantly represented on the screen. Notwithstanding minor projector repositioning variances, the projected screen size was 2.1 m wide by 1.5 m high with the bottom border of the projection running 1 m above and parallel to the floor. The cycle ergometer was positioned such that the handlebar stem riser was 3 m perpendicular to the plane of the screen which itself was offset to the right of the natural forward field of vision of the cyclists with a vector displacement of 8° at 3.03 m for the left border of the projection and 40° at 3.91 m for the right border (visual arc, 32°). Offsetting the screen in this way required participants to rotate their neck to look at the projected information, thus adding confidence that the eye-tracking measurements constituted deliberate attempts to acquire information, rather than information glances just because it happened to fall naturally within participants forward field of vision.

Incorporated into the projection beneath the simulated TT video, were five fields of real-time feedback information which, presented from left to right, were speed (km·h−1), elapsed distance (km), power output (W), pedaling cadence (rpm), and HR (bpm). The row of five feedback information fields were 0.375 m above and parallel to the bottom border of the projection or 1.375 m above the floor. The vector displacement of the center of each information field from the handlebar stem riser was speed (9.5°, 3.04 m), elapsed distance (18.1°, 3.16 m), power output (26.0°, 3.34 m), pedaling cadence (32.9°, 3.57 m) and HR (38.9°, 3.86 m). Elapsed time (min:sec) was displayed above the HR field (3.0°, 0.2 m). The block size of individual characters within each field was 4.5 cm high by 2.9 cm wide. Angular separation of the information fields was at its most acute 3° (elapsed time × HR) and at its least acute 8.6° (elapsed distance × speed), well beyond the manufacturer-defined eye tracker spatial resolution of 0.1° and gaze position accuracy within the nearest degree. The size and separation of the projected information blocks therefore facilitated clear differentiation in eye tracker measurements as later described. An A0-sized RPE scale was also displayed to the left of the projector screen.

Psychophysiological measures

HR was recorded during both cycling TT every 120 ms using a chest strap Polar Accurex Plus HR monitor (Polar Electro, Kempele, Finland) connected via wireless to the Velotron software. Average HR was calculated every 4 km. Participants were asked to provide an overall rating of perceived exertion every 4 km using the Borg 6 to 10 RPE scale (5). All subjects were familiarized with the RPE scale, which was administered in accordance with published standardized instructions (4).

Eye-tracking and video analysis

Participants were fitted with a SensoMotoric Instruments SMI iViewX head-mounted monocular eye-tracking device. The system consists of two cameras mounted on a cycling helmet, one that records the eye position of the participant, and a 3.6-mm wide-angle forward-looking camera that records the scene the participant is looking at. Eye position was recorded at 50 Hz, which was then downsampled to 25 frames per second for the resulting scene videos. The eye tracker was calibrated using the participant’s left eye in accordance with the manufacturer’s instructions by asking participants to fixate a series of markers spanning the area of the display. Calibration accuracy was checked sporadically and at the end of the TT by asking the participant to fixate points on the screen and information display. The equipment has a manufacturer-defined spatial resolution of 0.1° and tests demonstrated that gaze position was accurate to within the nearest degree. The system tracks eye movements using pupil and corneal reflex so that each participant’s point of regard can be superimposed onto the recorded scene, thus enabling timed measurements to be made of eye fixations.

The eye-tracking videos for TTEXP were subsequently reviewed and manually coded by the first author. Manual coding of eye-tracking data remains the state-of-the-art in active tasks, (51) and within-coder comparisons indicated that gaze location could be determined unambiguously. Reliability of similar methods has shown very good inter-rater reliability (21). Due to the relatively low sampling rate of the eye tracker, saccades could not be automatically detected, but fixations were only coded when data was within the same region for at least three frames (≅100 ms). Eye gaze was coded by recording the start and end frame of each entry into a new region of interest. This allowed us to determine the periods spent inspecting each of then eye fixation times were manually recorded in milliseconds against nine predetermined categories. Six of the categories related to information feedback that were speed, elapsed distance, power output, cadence, HR and elapsed time. Eye fixation times were also recorded for the rating of perceived exertion and the video simulation of the TT course that was projected onto the wall. A final category was created to capture all other objects of regard not corresponding to the other eight categories, for example, when participants looked at the laboratory floor or at laboratory equipment. Fixations of less than three frames, blinks, and other periods of data loss (e.g., when participants looked at extreme angles) were also included in the “other” coding category. This procedure allowed detailed coding of point of regard for the whole length of the TT.

Data processing and statistical analysis

Total gaze time and gaze frequency for each of the nine categories (speed, elapsed distance, power output, cadence, HR, elapsed time, video simulation, and other) was calculated on a participant-by-participant basis for the whole TT and for each 4-km segment. Gaze frequency, defined as the number of separate eye fixations for each category, and total gaze time, defined as the accumulated time of all eye fixations for each category, were calculated for each participant across the whole TT and for each segment. Total gaze times were then used to determine what information source that each participant looked at for longest accumulated average time (primary), second longest accumulated average time (secondary), third longest accumulated average time (tertiary), and so on, until quaternary (fourth), quinary (fifth), senary (sixth), septenary (seventh), octonary (eighth) and nonary (ninth) had all been established. To normalize absolute total gaze times for interparticipant differences in TT performance, primary to nonary fixation data were all converted from absolute time (ms) to percentage of TT completion time.

Time trial average cycling speed (performance) interactions between experienced and novice cyclists, and between the first and second TT was analyzed using two-way mixed ANOVA. Three-way mixed ANOVA were used to analyze group–trial–segment interactions in average cycling speed (pace) as well as relative fixation time and gaze frequency for the primary, secondary, and tertiary visual categories.

For both performance, pace and visual data, significant interactions were followed up using planned post hoc comparisons between segments using paired-samples t-tests for within-group comparisons and independent sample t-tests for between-group comparisons. Paired-samples t-tests were also used to compare within group comparison and RPE values. All results are expressed as mean (SD) and effect sizes as partial eta squared.


TT performance, HR, and RPE

Two-way mixed ANOVA revealed the following experience and trial factor outcomes. Average cycling speed: no group–trial interaction (F1,18 = 2.7, P = 0.082, ηp2 = 0.16) but there was a group main effect (F1,18 = 6.8, P = 0.018, ηp2 = 0.27) and a trial main effect (F1,18 = 11.2, P = 0.004, ηp2 = 0.38). Completion time: no group–trial interaction (F1,18 = 2.7, P = 0.082, ηp2 = 0.16) but there was a group main effect (F1,18 = 6.8, P = 0.018, ηp2 = 0.27) and a trial main effect (F1,18 = 11.2, P = 0.004, ηp2 = 0.38). Average power output: no group–trial interaction (F1,18 = 0.6, P = 0.440, ηp2 = 0.03), but there was a group main effect (F1,18 = 10.8, P = 0.004, ηp2 = 0.38) and a trial main effect (F1,18 = 11.6, P = 0.003, ηp2 = 0.39). Average pedaling cadence: No group–trial interaction (F1,18 = 0.1, P = 0.740, ηp2 < 0.01) or trial main effect (F1,18 = 3.6, P = 0.07, ηp2 = 0.17) but there was a group main effect (F1,18 = 12.7, P = 0.002, ηp2 = 0.414). Average HR: no group–trial interaction (F1,18 = 0.3, P = 0.086, ηp2 < 0.01), no group main effect (F1,18 < 0.1, P = 0.945, ηp2 < 0.01) and no trial main effect (F1,18 = 0.2, P = 0.646, ηp2 = 0.01). Average RPE: no group–trial interaction (F1,18 < 0.1, P = 0.929, ηp2 < 0.01), no group main effect (F1,18 = 0.4, P = 0.518, ηp2 = 0.02) and no trial main effect (F1,18 = 0.9, P = 0.361, ηp2 = 0.05). Group and trial differences in performance, HR, and RPE variables are presented in Figure 1A, with post hoc statistical outcomes indicated for significant differences between novice and experienced cyclists (independent samples t-tests) and between familiarization and experimental TT (paired samples t-tests).

Overall TT performance (A) and TT pacing by segment for familiarization (B) and TT 1 (C).

Segment comparisons of performance, HR, and RPE

There were no group–trial–segment interactions or two-way interactions for speed, completion time, power, cadence, HR, or RPE. Trial main effects were found for speed (F1,18 = 12.9, P = 0.002, ηp2 = 0.42), completion time (F1,18 = 12.9, P = 0.002, ηp2 = 0.42) and power (F1,18 = 11.5, P = 0.003, ηp2 = 0.39). Segment main effects were found for speed (F3,54 = 4.3, P = 0.009, ηp2 = 0.19), completion time (F3,54 = 4.3, P = 0.009, ηp2 = 0.19), power (F3,54 = 6.9, P = 0.001, ηp2 = 0.28), HR (F3,54 = 101, P < 0.001, ηp2 = 0.85), and RPE (F3,54 = 518, P < 0.001, ηp2 = 0.97). Group main effects were found for speed (F1,18 = 7.9, P = 0.012, ηp2 = 0.31), completion time (F1,18 = 7.9, P = 0.012, ηp2 = 0.31), power (F1,18 = 10.8, P = 0.004, ηp2 = 0.38), and cadence (F1,18 = 12.7, P = 0.002, ηp2 = 0.414). Post hoc independent samples t-tests found experienced cyclists were faster than novices during every TT segment, in both TTFAM and TTEXP. Group and segment differences in pace with post hoc outcomes are presented in Figure 1B for TTFAM and in Figure 1C for TTEXP. Mean and standard deviation data for speed, completion time, power, cadence, HR, and RPE are given in Table 1 for each group, TT and segment along with post hoc statistical test outcomes.

Mean performance, HR and RPE TT data for group, trial, and segment.

Whole TT eye-tracking outcomes: total gaze duration and gaze frequency

Novice and experienced mean total gaze duration data for primary through to nonary points of regard were calculated over the full 16.1 km for TTEXP and are presented in Figure 2A. A two-way mixed ANOVA found a group × point of regard interaction for total gaze duration (% TT duration), F8,144 = 10.9, P < 0.001, ηp2 = 0.38. Independent-samples post hoc t-tests revealed that experienced cyclists looked at primary points of regard for longer than novices during TTEXP (34.2% ± 6.1% vs 24.5% ± 4.2%, t18 = −4.2, P < 0.001, η2 = 0.49). Other experienced vs novice post hoc outcomes for total gaze time are represented in Figure 2A.

Novice and experienced total gaze duration data (A) and average gaze frequency (B) for primary (most looked at) through to nonary (least looked at) information sources calculated over the full 16.1-km distance for TT 1. (A) The type of information looked at with the corresponding number of subjects is presented alongside the data points in 2A for primary to tertiary sources but not included for quaternary to nonary sources. *P < 0.05; **P < 0.01; ***P < 0.001.

The frequency of which novice and experienced participants looked at primary through to nonary points of regard was counted overall for TTEXP and is presented in Figure 2B. A two-way mixed ANOVA found a group × point of regard interaction for gaze frequency, F8,144 = 2.2, P = 0.03, ηp2 = 0.11. Independent-samples post hoc t-tests revealed that experienced cyclists looked at information less frequently than novices (Fig. 2B).

TT segment eye-tracking outcomes: total gaze duration and frequency

Segment changes in gaze duration and gaze frequency were analyzed using two-way mixed ANOVA for primary, secondary, and tertiary points of regard. Group main effects were found for total gaze duration for the primary point of regard (F1,18 = 16, P < 0.001, ηP2 = 0.47) and the secondary point of regard (F1,18 = 6.7, P = 0.02, ηP2 = 0.27) but not the tertiary point of regard. No segment main effects or segment–group interactions were found for primary, secondary, or tertiary points of regard (Figs. 3A–C). For gaze frequency of the primary point of regard a segment–group interaction was found (F3,54 = 3.4, P = 0.02, ηP2 = 0.16) and a segment main effect (F3,54 = 2.8, P = 0.05, ηP2 = 0.13) but not a group main effect. For gaze frequency of the secondary point of regard, only a group main effect was found (F1,18 = 8.9, P = 0.008, ηP2 = 0.33) with no segment main effect or segment–group main effect. There were no gaze frequency interactions or main effects for the tertiary point of regard (Figs. 4A–C).

Experienced versus novice segment-by-segment TT 1 total gaze duration data for primary (A), secondary (B), and tertiary information sources (C). *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.
Experienced versus novice segment-by-segment TT 1 average gaze frequency for primary (A), secondary (B), and tertiary information sources (C). *P < 0.05; **P < 0.01; ***P < 0.001.

Group-by-trial-by-segment analysis for quaternary through to nonary points of regard are excluded from this article for the sake of brevity, owing to the large amount of statistical data. We also believe that the analysis of gaze data beyond the three most looked at points of regard is unlikely to yield significant insights about systematic perceptual patterns, pacing, and performance.

Primary–secondary point of regard combinations

Data is presented in Table 2 shows the combination of primary and secondary points of regard that participants looked at across the entire experimental TT and on a segment-by-segment basis. Individual participant data are present in an attempt to convey the complex, yet in some instances similar, patterns of information that participants looked at during the TT. Seven primary–secondary point of regard combinations were observed for the novice group during TTEXP, whereas the experienced cyclists exhibited only three primary–secondary point of regard combinations.

Individual gaze combinations of primary and secondary information sources.

Mann–Whitney nonparametric comparisons were made between novices and experienced cyclists in the number of primary points of regard they looked at in each segment and the number of times they switched what they primarily looked at between segments. Results showed a lower number of different primary points of regard by experienced cyclists compared to novices during TTEXP (1.7 ± 0.8 vs 2.8 ± 0.9, U = 19.5, Z = −2.41, P = 0.008). From segment to segment, the number of times participants switched to a different primary point of regard was lower among the experienced cyclists compared with novices (1.3 ± 1.4 vs 2.3 ± 0.9, U = 31, Z = −1.53, P = 0.064). Primary point of regard and switch data are given in Table 2.

A two-way mixed ANOVA found a group–segment interaction for the percent dominance of the primary point of regard in the primary–secondary combination, F3,54 = 4.4, P = 0.05, ηp2 = 0.20, a group main effect, F1,18 = 9.4, P = 0.007, ηp2 = 0.34, but no segment main effect, F3,54 = 0.4, P = 0.52, ηp2 = 0.02. Independent-samples post hoc t-tests revealed that dominance of the primary point of regard in the primary–secondary combination was greater among experienced cyclists compared with novices for the 0- to 4-km segment (63.8% ± 7.8% vs 53.6% ± 3.2%, t18 = −3.8, P < 0.001, η2 = 0.45), the 4- to 8-km segment (61.7% ± 8.0% vs 56.2% ± 4.3%, t18 = −1.9, P = 0.036, η2 = 0.17), the 8- to 12-km segment (63.4% ± 6.5% vs 56.6% ± 5.3%, t18 = −2.6, P = 0.01, η2 = 0.27) but not the final 12- to 16.1-km segment (59.8% ± 7.6% vs 60.1% ± 7.8%, t18 = 0.1, P = 0.93, η2 < 0.01). Group-by-trial-by-segment primary dominance values are given in Table 2.


This study was the first to make direct measurements of information-acquisition behavior among TT cyclists and constitutes a significant step forward in our understanding of endurance exercise pacing mechanisms. It seems that patterns of information acquisition during a self-paced cycling TT are very complex and that pacing behavior is not necessarily universally informed by the integration of endpoint awareness and perceived exertion, as previous models have argued (9,15,19,22,39,45,47,49). This is because we observed that, first, cyclists refer to different types of information according to their experience, with experienced cyclists primarily looking at speed and novices primarily looking at distance (Fig. 2A). Second, experienced cyclists appear to be more selective in their information acquisition behavior compared with novices, referring to fewer sources of information, which they look at for longer (Fig. 2A) and less frequently (Fig 2B). Third, novices increased the duration (Fig. 3A) and frequency (Fig. 4A) of looking at their primary information source during the final segment of the TT but experienced cyclists were more constant throughout the trials. Finally, with only four different combinations of primary and secondary information used by the experienced cyclists, there was better commonality in what information they looked at compared with the novices who used seven primary–secondary information combinations (Table 2). Our finding that experienced cyclists refer to task-relevant information less often is consistent with a meta-analysis of eye-tracking studies of expert performers (23), yet our findings that experienced cyclists fixate for longer than novices is not consistent with the meta-analysis (23). This maybe because, as acknowledged by the authors of the meta-analysis, the type of sport task may moderate expert-novice differences in visual behavior compared with other domains (23). Experienced cyclists also tended to stick to a primary information source throughout the TT, whereas novices switched the type of information they primarily looked at between segments much more often (Table 2). We are not suggesting that endpoint awareness is not important in pacing regulation, clearly, it is given how often it featured as either a primary or secondary point of regard in our findings (Table 2). Our argument is that previous pacing models are deficient in accounting for variations in information acquisition that we have found attributable to individual preference, expertise, or event segment. It seems that in simulated TT cycling, experienced cyclists look at speed more than distance, whereas distance feedback appears to be what novices seek out more.

An important finding of this study was that experienced and novice cyclists differed in the types of information they looked at during the experimental TT. The majority of the experienced cyclists (9 of 10 participants) tended to look at speed most across the whole TT. In contrast most novices (6 of 10 participants) looked at distance most, noting that a significant number of novices (4 of 10 participants) chose to primarily look at other information too. In addition to experienced cyclists being more consistent in what information they look at, of note is that they looked at primary information for longer and less frequently.

Although the eye-tracking data we have collected reveal a lot about how TT cyclists acquire information, it does not tell us anything about how the information is integrated and processed, or the decisions they have made. For this, other process-tracing methods, such as think-aloud protocols, may usefully compliment eye-tracking in the study of decision-making and pacing. This is because that, although eye-tracking technology provides a powerful method for measure information acquisition processes, it reveals nothing about how that information is subsequently processed. Although longer eye fixation times have been linked to greater depth of processing (16,25,42,43), rather than assuming this to be the case in future pacing studies, it would be preferable to use eye-tracking in conjunction with think aloud protocols to directly capture information processes. Nevertheless, the results of the present study so highlight differences in information acquisition between novice and experienced TT cyclists that bring to question the common information-processing mechanisms put forward by previous pacing models (9,15,17,22,30,33,39,41,44,49). In particular, the assumption in previous pacing models that the integration of endpoint awareness with perceived exertion is the primary and universal driver of pacing decisions, regardless of athletic experience or individual feedback preferences. It may be that decision making among experienced cyclists was different to novices and indeed different between individuals which resulted in a need to seek out more varied sources of information. This is consistent with the idea that individuals use information in an adaptive way according to the perceived demands of a situation or problem (24). Thus, it could be that distance information is still important to experienced cyclists but, owing to their previous experience, they are able to process and integrate such information much more quickly and thus do not need to look at it quite so often or for so long. Since the experienced participants were experienced at performing the 16.1-km TT format, it is also quite likely that their need to refer to distance information was less than novices unaccustomed to cycling such a distance. The extent to which information acquisition differences between experienced and novice cyclists are attributable to distance familiarity is something that could be tested by using the same experimental protocol but with an unfamiliar TT distance. Although it is well established that experience influences pacing strategy (18,34,37), our findings further show that information acquisition strategies accompanying pacing behavior also vary with previous experience.

As expected, the experienced cyclists completed both TT faster than the novices, with both groups exhibiting a mostly constant pace throughout. Owing to imperfect fitness matching between the novice and experienced cyclists, we cannot conclude that TT performance differences between the groups were exclusively due to experience differences. Although in future studies, greater effort should be made to measure associations between moment-by-moment change in gaze and pacing time series data (36), in this study, we have limited our analysis to detecting concomitant changes in gaze and pace at a segment-by-segment level. What our data clearly show is that, whatever type of information is preferred as the primary reference, the experienced cyclists looked at it for longer than the novices but less frequently. As previously discussed, this is broadly consistent with previous expertise literature (23). During the second TT, the experienced cyclists increased the relative amount of time they spent looking at the primary information source from 30% to 35%, showing that they became more selective in what information they referred to. The shallower curves presented in Figure 2A also show that novices tended to distribute their attention across a number of different information sources, spending more time looking at quaternary to octonary sources of information compared with the experienced cyclists. The notion that experienced cyclists are more selective in what feedback they look at is also consistent with previous expertise literature (23,32) and is supported in a number of ways. In the first three segments, the experienced cyclists on average spent between 5% and 10% longer than novices looking at the primary point of regard. It was only in the last segment of the TT from 12 to 16.1 km that the novices increase both the amount of time and the frequency with which they look at the primary information source close to that of the experienced cyclists. The increased information acquisition behavior toward the end of the TT is consistent with the behavior observed in children during a self-paced running task (6), further supporting the idea that feedback dependency is more strongly associated with proximity to the endpoint among inexperienced athletes compared with experienced athletes.

The data from our study indicate greater consistency in experienced cyclists’ approach to information acquisition both in terms in interparticipant and intraparticipant behavior. Interparticipant consistency is evident in the data showing that 9 of 10 experienced cyclists chose to primarily look at speed. Even when combinations of information sources are considered, experienced cyclists consistency chose either speed–distance (5 of 10), speed–other (2 of 10) or speed–power (2 of 10) as the combination of primary and secondary points of regard. In fact, the experienced cyclists only exhibited four different primary–secondary information combinations, whereas seven different primary–secondary combinations were observed among the novices (Table 2).

Greater intraparticipant consistency among the experienced cyclists is apparent, owing to the fact that on a segment-by-segment basis, the modal primary–secondary combinations were speed–distance and speed–other, but for the novices, it was often not possible to specify a modal combination because the primary–secondary permutations were so varied. On average, novices used 2.3 different primary information sources across the four segments compared with 1.5 for the experienced cyclists. Novices also tended to switch primary information sources between segments more frequently than the experienced cyclists as indicated in Table 2.

The primary–secondary combination data presented in Table 2 are also interesting because it highlights that distance is still an important reference source to experienced cyclists, but only secondary to and in combination with speed. In contrast, distance feedback appears to be the most dominant type of information they refer to in combination with many other types of secondary information. A lot of emphasis has been placed on the role of the endpoint in influencing pacing (2,3,9,15,18,30,33,39,45,49) support for which is found in several studies where deception or blinding methods have been used (3,12,29,37). However, our study shows that the importance placed on knowledge of the endpoint may be overstated in most pacing models and that knowledge of the endpoint may in fact be secondary to the information about speed in informing the actions of experienced cyclists. Another interesting outcome of this study is that perceived exertion did not feature in the primary–secondary information acquisition combinations for any of the participants (Table 2), and that, whether experienced or novice cyclists, all looked at least three other sources of information in preference to the 6 to 20 RPE scale (Fig. 2). That does not mean perceived exertion is not an important factor in pacing decisions as predicted by many of the previous models. It does, however, highlight methodological complexities of investigating pacing decisions in terms of acquisition and utilization of external referents, which can be easily observed using methods, such as eye tracking, and the integration of internal bodily referents, such as perceived exertion, which cannot be directly observed. This particular problem warrants innovative research using process-tracing methods of the kind described in much more detail elsewhere (36).

This eye tracking study has produced some important new data not entirely consistent with previous models of pacing about the attention to, and use of, feedback information. Nevertheless, there are a number of limitations associated with the laboratory-based nature of this experiment and the eye-tracking technology that was used. Cyclists in our study performed simulated TT on a static cycle ergometer under conditions where certain demands on the visual system were absent, for example, those associated with balancing, navigating, negotiating hazards, and avoiding collisions as reported elsewhere (50,52). Furthermore, differences between laboratory and real-world visual behavior have been reported in several studies, the most notable findings being more centralized fixations in the real world (21), a tendency to fixate on closer objects in the laboratory (21), and earlier longer object fixations in the real world (10). Therefore, it cannot be assumed that, during road-based TT, the capacity to attend to performance information will be the same as reported in this experiment because it will compete with, or be interrupted by, other demands placed on the visual system. In the future, with careful configuration of mobile eye-tracking technology, it may be possible to measure the attention to performance information in field-based studies with associated improvements in ecological validity.

Another limitation of this study relates to the link between visual information, decision-making processes and pacing behavior. Although there is some evidence that what individuals look at is associated with their choices (16,25,42,43), it is unclear whether visual attention influences choice or simply reflects a choice that has been made (43). In our study, the issue is further complicated by the difficulties of quantifying a pacing choice, because the method of detecting a meaningful change in pace from either speed or power time series data is mathematically complex (40). Even if it were possible to precisely identify moments where a decision has been made to increase or decrease pace, decisions to maintain pace would clearly be impossible to detect, because they would not be indirectly reflected in time series data. In this study, conclusions about the link between visual attention and pacing decisions are deduced from the associated changes in vision and pace observed at a segment-by-segment level. In the future, greater precision about the association between visual attention to performance information and pace could be investigated by setting up experiments where cyclists are presented with pacing dilemma where their decision to act can be pinpointed in time.

Finally, with regard to information acquisition and decision-making during endurance sport, further consideration is needed regarding fatigue-related constraints on visual behavior as predicted in Newell’s model (38) because they are often overlooked (55). A relationship between fatigue and declining visual attention was found in one interesting study where increased levels of exertion among biathletes was associated with reduced visual behavior before making a rifle shot (53). Saccadic eye movements are so fast and energetically efficient (46) that they are less likely to be responsible for such effects compared with high-order cognitive processes, such as attention allocation mechanisms which have themselves been found to become fatigued as characterized by reduced capability to suppress irrelevant external cues (13). Such factors are likely to impact information acquisition and decision making during endurance sport and warrant further investigation.


Although perhaps counterintuitive, this study challenges the degree of importance placed on knowledge of the endpoint to pacing in previous models. This is especially true for experienced cyclists for whom distance feedback was looked at secondary to, but in conjunction with, information about speed. Novice cyclists appear to have a greater dependence upon distance feedback, which they look at for shorter and more frequent periods than the experienced cyclists. Experienced cyclists are more selective in the information they refer to during a TT, and they are also more consistent in the combination of primary and secondary information they use, and more consistent between various phases of a TT. The difference in information acquisition behavior observed in this study may reflect differences in motivational regulators, with experienced cyclists perhaps focusing more strongly on performing at the fastest speed and novices focusing on completion of the distance.

This study is the first to directly measure cyclists’ information acquisition behavior during a TT, and the data show that the information athletes attend to and use during self-paced endurance tasks is much more complex than previously assumed and not necessarily dominated by knowledge of the endpoint. The limitations associated with this study are that it cannot be assumed information acquisition would be the same during a road-based TT. There are also improvements to the analysis of time series performance data that are needed to reveal hidden moments where a decision to alter pace has been made so that corresponding gaze behavior can be interrogated with greater precision. Nevertheless, this study has produced some exciting new insights about the information acquisition strategies of experienced and novice cyclists, as well as a new method for investigating visual attention and decision making during paced exercise.

This study was funded entirely by the University of Essex and no other sources of external funding were used. None of the authors have professional relationships with companies or manufacturers who will benefit from the results of this study. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. None of the results presented in this study constitute endorsement by the American College of Sports Medicine.


1. Albertus Y, Tucker R, St Clair Gibson A, Lambert EV, Hampson DB, Noakes TD. Effect of distance feedback on pacing strategy and perceived exertion during cycling. Med Sci Sports Exerc. 2005;37(3):461–8.
2. Ansley L, Robson PJ, St Clair Gibson A, Noakes TD. Anticipatory pacing strategies during supramaximal exercise lasting longer than 30 s. Med Sci Sports Exerc. 2004;36(2):309–14.
3. Billaut F, Bishop DJ, Schaerz S, Noakes TD. Influence of knowledge of sprint number on pacing during repeated-sprint exercise. Med Sci Sports Exerc. 2011;43(4):665–72.
4. Borg G. Borg’s Perceived Exertion and Pain Scales. Champaign (IL): Human Kinetics; 1998. pp. 44–9.
5. Borg G. Perceived exertion as an indicator of somatic stress. Scand J Rehabil Med. 1970;2(2):92–8.
6. Chinnasamy C, St Clair Gibson A, Micklewright D. Effect of spatial and temporal cues on athletic pacing in schoolchildren. Med Sci Sports Exerc. 2013;45(2):395–402.
7. Corbett J, Barwood MJ, Ouzounoglou A, Thelwell R, Dicks M. Influence of competition on performance and pacing during cycling exercise. Med Sci Sports Exerc. 2012;44(3):509–15.
8. DeKoning JJ, Bobbert MF, Foster C. Determination of optimal pacing strategy in track cycling with an energy flow model. J Sci Med Sport. 1999;2(3):266–77.
9. DeKoning JJ, Foster C, Bakkum A, et al. Regulation of pacing strategy during athletic competition. PLoS One. 2011;6(1):e15863.
10. Dicks M, Button C, Davids K. Examination of gaze behaviors under in situ and video simulation task constraints reveals differences in information pickup for perception and action. Atten Percept Psychophys. 2010;72(3):706–20.
11. Discombe RM, Cotterill ST. Eye tracking in sport: a guide for new and aspiring researchers. Sport Exerc Psychol Rev. 2015;11(2):49–58.
12. Eston R, Stansfield R, Westoby P, Parfitt G. Effect of deception and expected exercise duration on psychological and physiological variables during treadmill running and cycling. Psychophysiology. 2012;49(4):462–9.
13. Faber LG, Maurits NM, Lorist MM. Mental fatigue affects visual selective attention. PLoS One. 2012;7(10):e48073.
14. Faulkner J, Arnold T, Eston R. Effect of accurate and inaccurate distance feedback on performance markers and pacing strategies during running. Scand J Med Sci Sports. 2011;21(6):e176–83.
15. Faulkner J, Parfitt G, Eston R. The rating of perceived exertion during competitive running scales with time. Psychophysiology. 2008;45(6):977–85.
16. Fiedler S, Glöckner A. The dynamics of decision making in risky choice: an eye-tracking analysis. Front Psychol. 2012;3(10):1–18.
17. Foster C, De Koning JJ, Hettinga F, et al. Pattern of energy expenditure during simulated competition. Med Sci Sports Exerc. 2003;35(5):826–31.
18. Foster C, Hendrickson KJ, Peyer K, et al. Pattern of developing the performance template. Br J Sports Med. 2009;43(10):765–9.
19. Foster C, Schrager M, Snyder AC, Thompson NN. Pacing strategy and athletic performance. Sports Med. 1994;17(2):77–85.
20. Foster C, Snyder AC, Thompson NN, Green MA, Foley M, Schrager M. Effect of pacing strategy on cycle time trial performance. Med Sci Sports Exerc. 1993;25(3):383–8.
21. Foulsham T, Walker E, Kingstone A. The where, what and when of gaze allocation in the lab and the natural environment. Vision Res. 2011;51(17):1920–31.
22. Garcin M, Coquart J, Salleron J, Voy N, Matran R. Self-regulation of exercise intensity by estimated time limit scale. Eur J Appl Physiol. 2012;112(6):2303–12.
23. Gegenfurtner A, Lehtinen E, Säljö R. Expertise differences in the comprehension of visualizations: a meta-analysis of eye-tracking research in professional domains. Educ Psychol Rev. 2011;23(4):523–52.
24. Hutchinson JM, Gigerenzer G. Simple heuristics and rules of thumb: where psychologists and behavioural biologists might meet. Behav Processes. 2005;69(2):97–124.
25. Glaholt MG, Reingold EM. The time course of gaze bias in visual decision tasks. Vis Cogn. 2009;17(8):1228–43.
26. Haider H, Frensch PA. Eye movement during skill acquisition: more evidence for the information-reduction hypothesis. J Exp Psychol Learn. 1999;25(1):172–90.
27. Hettinga F, De Koning JJ, Broersen FT, Van Geffen P, Foster C. Pacing strategy and the occurrence of fatigue in 4000-m cycling time trials. Med Sci Sport Exerc. 2006;38(8):1484–91.
28. Hettinga FJ, De Koning JJ, Meijer E, Teunissen L, Foster C. Biodynamics. Effect of pacing strategy on energy expenditure during a 1500-m cycling time trial. Med Sci Sports Exerc. 2007;39(12):2212–8.
29. Jones HS, Williams EL, Bridge C, et al. Physiological and psychological effects of deception on pacing strategy and performance: a review. Sports Med. 2013;43(12):1243–57.
30. Joseph T, Johnson B, Battista R, et al. Perception of fatigue during simulated competition. Med Sci Sports Exerc. 2008;40(2):381–6.
31. Le Runigo C, Benguigui N, Bardy BG. Perception-action coupling and expertise in interceptive actions. Hum Mov Sci. 2005;24(3):429–45.
32. Mann DT, Williams AM, Ward P, Janelle CM. Perceptual-cognitive expertise in sport: a meta-analysis. J Sport Exerc Psychol. 2007;29(4):457–78.
33. Marcora SM. Do we really need a central governor to explain brain regulation of exercise performance? Eur J Appl Physiol. 2008;104(5):929–31.
34. Mauger AR, Jones AM, Williams CA. Influence of feedback and prior experience on pacing during a 4-km cycle time trial. Med Sci Sports Exerc. 2009;41(2):451–8.
35. Mauger AR, Jones AM, Williams CA. The effect of non-contingent and accurate performance feedback on pacing and time trial performance in 4-km track cycling. Br J Sports Med. 2009;45(3):225–9.
36. Micklewright D, Kegerreis S, Raglin J, Hettinga F. Will the conscious-subconscious pacing quagmire help elucidate the mechanisms of self-paced exercise? New opportunities in dual process theory and process tracing methods. Sports Med. 2017;47(7):1231–9.
37. Micklewright D, Papadopoulou E, Swart J, Noakes T. Previous experience influences pacing during 20 km time trial cycling. Br J Sports Med. 2010;44(13):952–60.
38. Newell KM. Coordination, control and skill. In: Goodman D, Wilberg RB, Franks IM, editors. Differing Perspectives in Motor Learning, Memory, and Control. Amsterdam (the Netherlands): Elsevier Science Publishing; 1985. pp. 295–317.
39. Noakes TD, St Clair Gibson A, Lambert EV. From catastrophe to complexity: a novel model of integrative central neural regulation of effort and fatigue during exercise in humans: summary and conclusions. Br J Sports Med. 2005;39(2):120–4.
40. Passfield L, Dietz K, Hopker J, Jobson S. Objective time-binning in exposure variation analysis. IMA J Manag. 2013;24(3):269–82.
41. Renfree A, Martin L, Micklewright D, St Clair Gibson A. Application of decision-making theory to the regulation of muscular work rate during self-paced competitive endurance activity. Sports Med. 2014;44(2):147–58.
42. Shi S, Wedel M, Pieters F. Information acquisition during online decision making: a model-based exploration using eye-tracking data. Manage Sci. 2013;59(5):1009–26.
43. Shimojo S, Simion C, Shimojo E, Scheier C. Gaze bias both reflects and influences preference. Nat Neurosci. 2003;6(12):1317–22.
44. Smits BL, Pepping GJ, Hettinga FJ. Pacing and decision making in sport and exercise: the roles of perception and action in the regulation of exercise intensity. Sports Med. 2014;44(6):763–75.
45. St Clair Gibson A, Lambert EV, Rauch LH, et al. The role of information processing between the brain and peripheral physiological systems in pacing and perception of effort. Sports Med. 2006;36(8):705–22.
46. Thiele A, Henning P, Kubischik M, Hoffman KP. Neural mechanisms of saccadic suppression. Science. 2002;295(5564):2460–2.
47. Tucker R. The anticipatory regulation of performance: the physiological basis for pacing strategies and the development of a perception-based model for exercise performance. Br J Sports Med. 2009;43(6):392–400.
48. Ulmer HV. Perceived exertion as a part of a feedback system and its interaction with tactical behaviour in endurance sports. In: Borg G, Ottoson G, editors. The Perception of Exertion in Physical Work. London (UK): Palgrave Macmillan; 1986. pp. 317–26.
49. Ulmer H, Schneider A, Neumahr S, Freitag W. Precision of time and velocity estimation as a base of teleoanticipation. Eur J Physiol. 1996;431(48):1–5.
50. Vansteenkiste P, Cardon G, D’Hondt E, Philippaerts R, Lenoir M. The visual control of bicycle steering: the effects of speed and path width. Accid Anal Prev. 2013;51:222–7.
51. Vansteenkiste P, Cardon G, Philippaerts R, Lenoir M. Measuring dwell time percentage from head-mounted eye-tracking data—comparison of a frame-by-frame and a fixation-by-fixation analysis. Ergonomics. 2015;58(5):712–21.
52. Vansteenkiste P, Zeuwts L, Cardon G, Philippaerts R, Lenoir M. The implications of low quality bicycle paths on gaze behavior of cyclists: a field test. Transport Res F Tr Psychol. 2014;23:81–7.
53. Vickers JN, Williams AM, Rodrigues ST, Hillis F, Coyne G. Eye movements of elite biathlon shooters during rested and fatigued states. J Sport Exerc Psychol. 1999;21:S116.
54. Williams AM, Ericsson KA. Perceptual-cognitive expertise in sport: some considerations when applying the expert performance approach. Hum Mov Sci. 2005;24(3):283–307.
55. Williams AM, Janelle CM, Davids K. Constraints on the search for visual information in sport. Int J Sport Exerc Psychol. 2004;2(3):301–18.
56. Williams EL, Jones HS, Andy Sparks S, Marchant DC, Midgley AW, Mc Naughton LR. Competitor presence reduces internal attentional focus and improves 16.1 km cycling time trial performance. J Sci Med Sport. 2015;18(4):486–91.


© 2017 American College of Sports Medicine