Which way is down? Visual and tactile verticality perception in expert dancers and non-experts.

Gravity provides an absolute verticality reference for all spatial perception, allowing us to move within and interact effectively with our world. Bayesian inference models explain verticality perception as a combination of online sensory cues with a prior prediction that the head is usually upright. Until now, these Bayesian models have been formulated for judgements of the perceived orientation of visual stimuli. Here, we investigated whether judgements of the verticality of tactile stimuli follow a similar pattern of Bayesian perceptual inference. We also explored whether verticality perception is affected by the postural and balance expertise of dancers. We tested both the subjective visual vertical (SVV) and the subjective tactile vertical (STV) in ballet dancers and non-dancers. A robotic arm traced downward-moving visual or tactile stimuli in separate blocks while participants held their head either upright or tilted 30° to their right. Participants reported whether these stimuli deviated to the left (clockwise) or right (anti-clockwise) of the gravitational vertical. Tilting the head biased the SVV away from the longitudinal head axis (the classical E-effect), consistent with a failure to compensate for the vestibulo-ocular counter-roll reflex. On the contrary, tilting the head biased the STV toward the longitudinal head axis (the classical A-effect), consistent with a strong upright head prior. Critically, tilting the head reduced the precision of verticality perception, particularly for ballet dancers' STV judgements. Head tilt is thought to increase vestibular noise, so ballet dancers seem to be surprisingly susceptible to degradation of vestibular inputs, giving them an inappropriately high weighting in verticality judgements.