One aspect to consider when investigating bicycling performance during alcohol intoxication is how the bicycle is handled by the bicyclist, i.e. the stability of the individual when bicycling. The literature on bicycling stability (balance) reveals that speed is important (Cain et al., 2016). At lower speeds, novice and skilled bicyclists show similar balance performances, but at higher speeds, skilled bicyclists use more lean control and less steer control. (Hartung et al., 2015) studied bicycling performance and found severe coordinative faults when participants bicycled under the influence of alcohol. In addition, they found differences between men and women in terms of the BAC level at which severe faults occurred.

The literature suggests that there is a complex interplay between alcohol intoxication and cognitive performance (Adminaité-Fodor & Jost, 2020; Cash et al., 2015; Greenstein et al., 2010; Mintzer, 2007; Park et al., 2011; Weiss & Marksteiner, 2007). The results suggest that cognitive functions are usually affected by alcohol intoxication, but not always (Hoffman et al., 2015; Spinola et al., 2017). Several cognitive tests have been administered, and differences have been found between age groups (Boissoneault et al., 2014), gender (Hoffman et al., 2015) and even naturalistic settings (Lyvers & Tobias-Webb, 2010). Individuals take riskier decisions when intoxicated (George et al., 2005), and it has been argued that self-regulation impacts cognitive executive function, such that alcohol intoxication might affect risk taking (Brooks et al., 2019).

The present experiment focused on ability, even though reduced ability almost certainly increases risk. It is predicted that the results for gambling tasks, such as that used by George et al. (2005), would be true for bicycling on a treadmill as well. However, the present study focused on the ascending curve of BAC levels (up to the peak of 0.8‰ breath alcohol content, BrAC), and it was difficult to predict when a reduction in ability ratings would present itself. (Hartung et al., 2015) found deficits at relatively low BAC levels. In addition, as the Feenstra et al. (2010) study revealed, positive attitudes towards alcohol might be related to the ability ratings made by participants.

A Swedish questionnaire study with 1 769 participants, aged 18–65 years, revealed that approximately 50% of the participants had bicycled when intoxicated with alcohol at least once during January to August 2018 (Wallén_Warner et al., 2018); see also (Huemer, 2018). Intention, defined as the willingness to try to perform the behaviour, was therefore included in the study.

The experiment focused on stability measurements and their relation to alcohol intoxication levels. The primary aim of this study was to investigate how bicycling stability decreased as an effect of higher intoxication levels. A secondary aim was to investigate if and how cognitive performance and ability ratings was related to different levels of intoxication and stability.

The experiment was conducted in a controlled setting on a large treadmill, enabling control of speed and physical effort. To relate the treadmill results with reality to some extent, bicycling on the treadmill was combined with self-reports of the participants’ intentions to bicycle in real traffic when intoxicated with alcohol. It was predicted that participants would have a strong intention to bicycle in real traffic, despite being so intoxicated that their stability, executive functions, and bicycling abilities were affected.

A total of 29 participants were enrolled in the study, see Table 1 for details. Of these 29 participants, 18 were given alcohol during the experiment (intoxicated group), while 11 participants completed the bicycling test sober (sober control group). Participants were randomised after screening. The objective was to minimise age difference and balance for gender.

The first column shows data for the intoxicated group, who were given increasing amounts of alcohol while bicycling; the second column shows data for the sober control group, who bicycled sober.

The participants were recruited via a Facebook advertisement and by personal contacts. Prospective participants were directed to a dedicated project webpage, where the required criteria for participation (age gender, health etc.) were checked as the first step for inclusion in the study. The participant screening criteria included: a minimum age of 20 years old, no history of alcohol or substance addiction, and good health for bicycling. The participant criteria also included living within the town of Falun (to limit taxi costs after the end of the experiment). Only non-pregnant women could participate in the experiment and women were offered a pregnancy test upon request. Other requirements included previous experience of bicycling and drinking alcohol.

The selection of participants included several steps, starting with the initial screening mentioned above. The screening procedure was repeated when each participant arrived at the test facility. The Alcohol Use Disorders Identification Test (AUDIT) (Saunders et al., 1993) scale was also completed and checked prior to the experiment. The participants needed to be able to stay on the facilities for approximately 4 hours, so that their post-experiment safety and sobriety could be monitored.

The participants were reimbursed for their participation with approximately €150 or €75 (the sober control group did not have to wait at the facility until their BrAC level was low). All participants were also offered a cost-free taxi ride home. The study was approved by the Swedish Ethical Review Authority (approval number 2019-01268).

The experiment was carried out in the sports laboratory at Lugnet's Sports Science Institute, in co-operation with Dala Sports Academy, in Falun, during July, October/November 2019 and May 2022. The complete set-up and context were the same during all three periods. The participants bicycled on a motor-driven treadmill (Saturn 450/300 RS; H/P/Cosmos Sports & Medical, Nussdorf, Germany), which had a surface area of 13.5 m² and could be tilted to increase physical workload (if necessary). The treadmill’s general specifications comprised a running surface of 4.5 × 3.0 m, a speed range of 0–40 km/h, an elevation of -5% to +25% (-2.8 to 14.0 degrees), a 30 kW motor system (40.8 HP), and a re-enforced thick rubber running belt that is also suitable for use with ski rollers, ski poles, track shoes and bicycles.

The participants bicycled on a mid-range mountain bike (Biltema Yosemite X-dirt bicycle), with relatively large wheel dimensions for stability (27.5 × 2.8 inch wheels). The brakes were disconnected on the test bicycle. The bicycle was equipped with a VBOX data acquisition system mounted on the handlebar (Racelogic 3i, Buckingham, UK) and gyro sensor (Racelogic IMU02, Buckingham, UK), see Figure 1. The participants wore a pulse sensor (to monitor their physical effort) that was connected to the treadmill system. The participants also wore a safety harness, which was attached to the gantry above the treadmill. The harness attachment point was connected to a kill-switch, to enhance safety. All participants wore a bicycle helmet.

This experiment used a mixed design. The between group variable was intoxicated versus sober, to investigate the effects of alcohol intoxication on participants’ bicycling performance. The second independent variable was the breath alcohol concentration (BrAC) level, measured 5 times. The dependent variables were stability (yaw and roll rate), cognitive performance (n-back score), subjective level of intoxication (CR10 rating) and ability assessment (rating). Speed and physical exertion might affect the results as well; the experimental set-up controlled for these aspects by keeping the speed at 20 km/h and physical exertion constant by adjusting the gradient of the treadmill between 0.2 and 2.0 degrees (mean = 0.9 degrees; SD = 0.47) to maintain a uniform level of physical workload.

The 29 participants were first given study information and instructions, and then signed an informed consent form. A health declaration was also completed by the participants. In addition, the participants were asked about any food allergies, as snacks were provided. All participants were measured and screened upon arrival (weight, identity check, 0‰ BrAC level and pregnancy). The participants also filled in a pre-test questionnaire, covering questions about their age and gender, as well as their bicycling, physical activity and drinking habits.

The bicycling test consisted of five sessions (time 1–time 5). Each session comprised: (1) a 10-minute bike ride on the treadmill, (2) a mouth wash to reduce possible effects of alcohol still left in the mouth, (3) a BrAC assessment, (4) a subjective intoxication rating, (5) the 20 item Borg Scale of Perceived Exertion rating, (6) an n-back task, (7) a subjective ability rating, and (8), time for refreshments, drinking alcohol, and rest. In the bicycling part, the experiment leader stabilised the participant by holding the gantry until the target speed was reached. The same procedure was used before the treadmill was stopped.

The alcohol dosage was calculated according to Mumenthaler et al. (1999) based on the participant’s gender and weight (the mean quantity of 40% spirits administered was 197 ml (SD = 55.2 ml); range = 140–340 ml). The participants in the intoxicated group could choose between whisky, white rum, vodka, or gin. The spirits could be mixed with soft drinks to make them more palatable. After the first sober 10-minute bicycle ride (time 1), participants in the intoxicated group were given 75% of the estimated alcohol dosage calculated to reach the target of BrAC = 0.8‰. The remaining 25% were given after the second 10-minute bicycle ride (time 2). Since the algorithm typically underestimated the alcohol dosage, a third drink was administered in later sessions if the target level of BrAC = 0.08‰ had not been reached. Participants were never told about their level of intoxication.

When the participants had completed the five bicycling sessions on the treadmill, they completed the post-test questionnaire. The participants then remained in the research laboratory facility until they became sober (several hours later) after which they were offered a taxi home. The ‘sober’ level was negotiated to some extent, not all participants’ BrAC levels were below 0.2‰ when they left the laboratory. The sober control group were free to leave as soon as all post-test evaluations were completed.

All participants answered/completed several questions/tests before, during and after each 10 minute bicycle ride. A summary of these results is provided in Table 2.

The pre-test questionnaire results revealed that both groups drank alcohol approximately two to four times a month (mean of 2.83 and 2.63 times per month respectively) and worked out two to three times a week (mean of 4.16 and 3.82 times per week respectively). Intoxicated participants bicycled significantly more often (mean of 3.94 and 2.45 times per week respectively) and bicycled significantly more often when intoxicated during the snow-free period of 2018 (mean of 1.89 and 1.18 times a month respectively). Hence, the intoxicated group might have been somewhat more familiar with bicycling and bicycling while intoxicated. These differences between the groups will probably not affect the results given the design used, i.e., the between-groups design where two groups bicycled while experiencing different levels of intoxication.

A 2 (group) by 5 (test times) analysis of variance (ANOVA) showed that subjective exertion ratings (measured via the Borg Scale of Perceived Exertion) were unaffected over time or between groups (mean for the intoxicated group = 11.67 and mean for the sober control group = 12.04). Hence, the adjustment of the tread mills gradient was successful, meaning that the other measurements (stability, executive function, and ability rating) should be unaffected by different degrees of experienced physical effort. Overall, the results indicate that the two groups were comparable in all aspects studied.

The results revealed that women’s and men’s BrAC levels were very similar over the sessions. The 2 (women versus men) by 5 (number of times BrAC levels were measured) ANOVA revealed one significant effect (only intoxicated participants were compared), namely that the BrAC level increased over time (F(1, 16) = 341, p < 0.001, Mse = 0.05). This was expected given the experimental design. In subsequent analyses, we therefore examined the intoxicated group as a whole, without included gender as a factor. Figure 2 reveals a constant rise in BrAC level until assessment time 4, where intoxicated participants started to reach their peak.

Mean values and SDs for all variability measures are presented in Table 3. The roll rate measures for the last 7 sober control participants were affected by unknown technical problems. Roll rate measures were therefore only calculated for 4 sober control participants.

A 2 (group) by 5 (times) mixed ANOVA on the SD of yaw rate and SD of roll rate revealed no main effects; however, there was a significant interaction effect for both measures. Figure 3 and Figure 4 show that the intoxicated group became increasingly unstable, while the sober group became marginally more stable over time.

Pairwise comparisons of yaw rate SD revealed that intoxicated participants' stability increased from time 1 to time 2 (not significantly, p = 0.076), but later decreased from time 2 and onwards (except between time 4 and time 5). The mean increase in yaw rate SD from time 2 to time 3 was 0.554 (p > 0.001), and 0.561 from time 3 to time 4 (p > 0.001).

Pairwise comparisons of roll rate SD revealed that intoxicated participants' stability increased from time 1 to time 2, (mean increase 0.498, p < 0.02) but later decreased from time 2 and onwards (except between time 4 and time 5). The mean increase in roll rate SD from time 2 to time 3 was 0.385 (p > 0.001), and 0.423 from time 3 to time 4 (p > 0.001).

Table 4

		Group	Time	Group x Time
Roll rate RR	5°/s	ns	ns	ns
	10°/s	ns	ns	F (4, 80) = 5.62, p < 0.001, MSe = 2723
	15°/s	ns	ns	F (4, 80) = 7.02, p < 0.001, MSe = 2232
	20°/s	ns	ns	F (4, 80) = 7.00, p < 0.001, MSe = 920
	25°/s	ns	ns	F (4, 80) = 5.08, p < 0.001, MSe = 307
	SD	ns	ns	F(4, 108) = 4.16, p < 0.005, MSe = 0.25
Yaw rate RR	5°/s	ns	ns	F(4, 108) = 2.98, p < 0.03, MSe = 1444
	10°/s	ns	ns	F(4, 108) = 6.64, p < 0.001, MSe = 3069
	15°/s	ns	F(4, 108) = 4.58, p < 0.01, MSe = 3092	F(4, 108) = 8.84, p < 0.001, MSe = 3092
	20°/s	ns	F(4, 108) = 5.48, p < 0.001, MSe = 1843	F(4, 108) = 8.34, p < 0.001, MSe = 1843
	25°/s	ns	F(4, 108) = 3.83, p < 0.01, MSe = 888	F(4, 108) = 6.94, p < 0.001, MSe = 888
	SD	ns	ns	F(4, 108) = 4.72, p < 0.01, Mse = 0.56
N-Back (d-prime)		ns	F(4, 108) = 3.51, p < 0.01, MSe = 1.93	ns
Ability rating		ns	ns	F(4, 108) = 14.51, p < 0.001, Mse = 2.81

ns = non-significant

A total of 10 analyses were performed for the yaw rate and roll rate RR measures, i.e. 2 for each of the five gap sizes (Table 3). The RR measures revealed a clear pattern. First, the 2 (group) by 5 (times) mixed ANOVA on yaw rate RR as well as on roll rate RR with 5 and 10°/s gap sizes revealed no main effects but interaction effects for 3 of the 4 analyses (see Table 4). Hence, these four analyses revealed that intoxication did not have the same effect on small gap sizes.

Second, the 2 (group) by 5 (times) mixed ANOVA on yaw rate RR at 15°/s, 20°/s and 25°/s gap sizes revealed interaction effects and main effects of time (see Figure 5 as an example). The 2 (group) by 5 (times) mixed ANOVA on roll rate RR at 15°/s, 20°/s and 25°/s gap sizes revealed interaction effects as presented in Figure 6 (as an example). Pairwise comparisons are presented only for the 25°/s gap size roll rate RR and yaw rate RR analyses. Similar patterns were observed for all three yaw and all roll rate analyses at 15°/s, 20°/s and 25°/s gap sizes, respectively. The sober group became marginally more stable over time.

Just as for the yaw and roll rate SD measures, pairwise comparisons of the yaw and roll rate RR measures revealed that intoxicated participants became significantly more unstable over time (from time 2 and onwards, except between time 4 and 5). The mean increase in yaw rate RR with 25°/s gap size from time 2 to time 3 was 25.11 (p < 0.001), and 31.56 (p < 0.001) from time 3 to time 4. Corresponding increases for roll rate were 11.67 (p < 0.02) from time 2 to time 3, and 17.78 (p < 0.04) from time 3 to time 4, respectively.

The n-back test results were analysed in terms of d-prime values based on the relation between hits and false alarms (see Macmillan and Creelman (1991) for details). The 2 (group) by 5 (times) mixed ANOVA revealed a main effect of time (see Table 4). No significant effects were found for group or for the interaction between group and time. Pairwise comparisons revealed no significant effects at times 1, 2 and 3 but significant differences between the sober group and the intoxicated group at time 4 (mean diff = 1.26, p < 0.04) and 5 (mean diff = 1.16, p < 0.04), see Figure 7). Hits, false alarms, correct rejections, and misses were also analysed, and the findings point in the same direction as the d-prime analysis but to a lesser extent.

The 2 (group) by 5 (times) mixed ANOVA showed a significant interaction between group and time. Pairwise comparisons revealed that the sober group rated themselves as more able over time whereas the intoxicated group rated themselves as less capable. Significant differences were found between the groups at time 4 (mean diff = 3.32, p < 0.02) and at time 5 (mean diff = 4.46, p < 0.001). Figure 8 suggests that the intoxicated group showed an understanding of their condition, i.e. lower abilities when intoxicated.

Correlation analyses were carried out for the intoxicated participants (since all sober participants had an intoxication level of BrAC 0.00‰). In total 12 analyses where made, namely the correlation between BrAC and the subjective rating of intoxication, between BrAC and the ability ratings, and between the subjective intoxication ratings and the ability ratings, for each of the four assessment times of interest (times 2–5, at time 1 all participants were sober). None of the correlation results were significant. First, the participants could not predict their level of intoxication. Second, the correlation between how intoxicated they were or how intoxicated they experienced that they were nor correlated with the decrease in ability rating. The insignificant results, together with the ability rating results on a group level in the previous section, indicate that the intoxicated group rated, as a group, that their bicycling ability decreased due to alcohol intoxication, but this was not true on an individual level based on the correlation analysis.

Only data from the 15 participants in the experiment group during July were included in these analyses as intention were measured with regard to the upcoming three months, and it was likely to snow in the three months following the October/November experimental session. The 15 participants were aged 25–32 years (average of 29 years), and 6 were women. Most participants experienced moderate levels of alcohol intoxication during the tests, according to the modified CR10 scale (mean score = 2.64, range = 0.50–5.00).

The results showed that the participants' intention to bicycle in real traffic at the highest level of intoxication they experienced during the test was fairly neutral (M = 3.93, SD = 2.49; where 1 = likely; 7 = not at all likely) while their intention to bicycle at an even higher, hypothetical, level of intoxication was lower (M = 5.20, SD = 2.11; where 1 = likely; 7 = not at all likely). A Wilcoxon signed-ranks test indicated that the difference in intention was significant (Z = -2.39, p < 0.05).