The performance of athletes is often influenced by the presence of an audience. This pressure situation, which is common for competition, cannot be trained for on available simulators. Therefore, a novel rowing simulator with virtual reality technology was developed and evaluated. Ten participants of different skill levels were rowing 3 × 1000 m with positive, neutral, and negative virtual audience tribunes in blocks of random order. The analysis of movement variables, physiological response, questionnaires, and interviews was used to detect differences between audience blocks and adjoining non-audience blocks. Although some participants responded in part to the investigated movement and physiological variables, no reliable effect of any of the audience types could be shown. Interestingly, self-reported measures indicated in general a high degree of realism and presence in the scenario. The explanation for the low incidence of behavioural change is that the virtual audience did not create enough pressure on the participants, although no definite conclusion can be drawn owing to the small sample size.