Abstract
We studied how children 8-9 years old respond to simultaneous training of two different perceptual cues. Previous research without training has shown that children at this age will fail to take an appropriate precision-weighted average of an auditory and visual cue (e.g. Gori, Sandini & Burr, 2012). We wanted to know if children at 8-9 years are fundamentally incapable of precision-weighted averaging or if they just require training. We put six children in a virtual reality environment with a curved wall in front of them and a 'magic wand'. They were played a spatialized sound, visually shown a probability distribution sitting on top of the wall, or both. They had to use these cues to try and point to a hidden target on the top part of the wall. The width of the visual probability distribution varied from trial to trial. They were always shown the location of the actual target afterwards. Data were analysed by Bayesian model comparison. The 'optimal' model, which takes a reliability-weighted average of the two single cues on each trial, has the most support on balance. Its main competitor is an extremely-similar model with a small systematic mis-weighting of the two cues. Models that rely on one cue or just switch back and forth between the two cues are unlikely explanations (Bayes factors < .05) for all but one child. A model that does not account for trial-to-trial changes in reliability of the visual cue is also a poor fit overall. We conclude that children 8-9 years old are actually capable of quickly learning to weight and combine perceptual cues, perhaps with a small mis-weighting, at least under these circumstances. We are investigating how this depends on the nature of the feedback and the type of visual cue.
Meeting abstract presented at VSS 2016