Abstract
Humans are highly effective at dealing with noisy, probabilistic information. One hallmark of this is Cue Combination: combining two independent noisy sensory estimates to increase precision beyond the best single estimate. Surprisingly, this is not achieved until 10-12 years of age (Gori et al, Curr Biol 2008; Nardini et al PNAS 2010), despite other multisensory skills appearing in infancy. It is unclear if lack of cue combination in children is due to maturation or experience with specific cues. The "experience" account predicts that adults learning new cues would fail to combine them for many years. To test this, we asked 12 adults in a virtual reality environment to judge distance to a whale hidden under the sea, using both vision and a novel audio cue to distance akin to echoes used in human echolocation (Thaler & Goodale, WIREs Cogn Sci 2016). Within two hour-long sessions, participants successfully learned to judge distance using the novel audio cue (correlation coefficients > .80, all ps < .001). In subsequent sessions we compared variable error in trials with only one cue (visual, novel audio) and those with both cues together. We found significantly lower variable error with both cues versus the best single cue, as expected under cue combination (averaging). This pattern of results persisted even when the novel cue was provided using a new auditory frequency, and when we changed the cues' relative reliabilities. Further, when relative reliabilities changed we also observed a re-weighting of cues. This supports the idea that participants combined cues, rather than having learned a decision rule with specific stimuli. In conclusion, our results suggest that humans develop general-purpose cue combination abilities as they mature. The discovery that people can immediately integrate new signals into their existing repertoire further suggests an optimistic outlook on substituting or augmenting human senses.
Meeting abstract presented at VSS 2018