Abstract
Vergence size constancy is thought to be entirely (Taylor, 1941; Mon-Williams, Tresilian, Plooy, Wann, & Broerse, 1997) or largely (Sperandio, Kaderali, Chouinard, Frey, & Goodale, 2013) responsible for the Taylor illusion. Observers report that an after-image of their hand appears to grow or shrink if they move their hand back-and-forth in complete darkness whilst tracking their hand with their gaze. However, given observers known their hand position, this phenomenon could equally reflect cognitive biases in size estimation. In order to isolate vergence size constancy we asked subjects to view a target whose vergence changed from 50cm to 25cm over 5 seconds (same rate of change as the observer’s hand in Sperandio et al., 2013) and judge whether the target increased or decreased in size? During each trial the physical size of the target also increased or decreased by a variable amount, and we tested whether the change in vergence biased the observers’ size judgements. We tested 11 observers, and estimated the population bias using a hierarchical Bayesian model. Our results suggest that vergence had no effect on size judgements. The bias we observed (-0.2% of angular size) was (1) in the wrong direction for size constancy, (2) not statistically significant, (3) four times more likely under the null hypothesis (no effect of vergence on perceived size) according to the estimated Bayes factor, with (4) any true size constancy effect being smaller than the smallest effect size of interest (using detection threshold of the most sensitive observer as criterion for inferiority test). Chen, Sperandio, Henry, & Goodale (2019) suggest that real-world size-constancy could either be attributed to (1) recurrent processing in V1, or (2) feedback from higher-order visual areas to V1. Our results suggest that recurrent processing of the vergence signal in V1 is not responsible for real-world size-constancy.