Exercising with a moderately more capable partner can increase how long people keep going, but the effect is inconsistent and group training is not clearly better than training alone
Interpretation
Lab trials from one research group show large persistence gains with a superior virtual partner (58 young women cycled 21.9 vs 10.6 min alone; Irwin 2012), and a p-curve of 13 studies suggests the effect is not a product of p-hacking (Hill 2021). But in 82 participants over 12 sessions the gain appeared only in men (Max 2016), 85 active middle-aged adults did not choose higher intensities with a partner (Samendinger 2019), and in 176 students only a moderately better partner helped (Samendinger 2021). Broader evidence: social influence on exercise is small to moderate (Carron 1996), and a 2026 meta-analysis of 71 studies found group-based programmes only slightly, non-significantly better than individual ones (Kritz 2026). Competitors also push effort up: 15 trained male cyclists rode 16.1 km faster against an avatar (Williams 2015). Mostly students and lab tasks; no triathlete data.
Practical implication
Use a slightly stronger partner or group for sessions where you want more effort or persistence, and protect easy and steady sessions from the same pull by riding them alone or with a pace-holding partner. Extrapolated from lab studies in students and general exercisers, not triathletes; no evidence that group training improves fitness or race results.
Sources
8 sources, independently verified.
Stationary cycling at 65% of heart rate reserve to voluntary stop on six days, randomised to alone, coaction with a virtual partner, or a conjunctive team with a superior virtual partner. Mean persistence across sessions: conjunctive 21.89 min, coaction 19.77 min, alone 10.6 min. Young women, likely untrained students; persistence at fixed intensity, not performance.
12 cycle-ergometer sessions at 75% of maximum heart rate, alone, with a consistently superior virtual partner, or with a superior partner who showed fatigue. Men with the fatiguing partner rode 12 min 32 s longer than men alone; women showed no change over time and no condition differences.
Six-day cycle-ergometer protocol alternating 30-minute continuous and interval sessions (targets 75% and 90% of maximum heart rate) with self-selected power; no partner, superior non-teammate virtual partner, or superior teammate. All increased output over baseline with no differences between conditions: no Köhler effect on intensity. Enjoyment and self-efficacy related to effort in interval sessions.
Plank persistence alone or with a software-generated partner of low, moderate or high ability discrepancy. Overall condition effect not significant (P = 0.093); in a planned contrast only the moderately better partner beat exercising alone (P = 0.014). Authors note recent software-partner studies showed a diminished effect.
P-curve analysis of 13 exercise studies found evidential value: the Köhler effect is likely not a by-product of p-hacking. Does not estimate effect size or real-world transfer; authors are from the research group behind many of the original studies.
Social influence (important others, family, class leaders, co-exercisers, social and task cohesion) had a generally small to moderate positive effect on exercise behaviour, cognitions and affect (effect sizes 0.20-0.50); task cohesion and adherence was among the moderate-to-large effects (0.50-0.80). General exercisers, not athletes.
Group-based conditions showed small, non-significant advantages over individual interventions for physical activity (g = 0.086, n = 22,042), psychosocial outcomes (g = 0.292) and health outcomes (g = 0.125); functional outcomes favoured groups after outlier removal (g = 0.164, P = 0.015). 'True groups' using group-dynamics principles had a larger effect on health outcomes. Both approaches may be effective. Not athletes.
16.1 km ergometer time trials: racing an avatar presented as a similar-ability rival (actually their own fastest ride) gave 27.8 min vs 28.4 min with no display and 28.7 min against their own avatar, with higher power, speed and heart rate. Gain attributed to more external attentional focus. Single lab effort, men only.