Duration judgments with conflicting audiovisual cues

Research suggests that when observers encounter substantial audiovisual duration conflicts, they do not rely on causal inference. Previous studies indicated optimal integration of auditory and visual cues when conflicts were minimal. However, this investigation examined a broader spectrum of duration discrepancies between a test and a standard stimulus.

The standard stimulus showed differing auditory durations compared to the test stimulus by up to 250 milliseconds. Both high and low levels of auditory noise were assessed. Perceptions of auditory duration consistently aligned more closely with visual durations, particularly in high-noise conditions. This shift correlated with the magnitude of cue-conflict, which contradicts the causal inference hypothesis.

Various models were evaluated to explain the findings. A heuristic model, wherein observers probabilistically choose between visual and auditory cues, was favored by most participants, though the differences in performance among models were minimal. Across the tested conflict range, models such as forced fusion, causal inference, and probabilistic cue switching yielded similar results, with near-linear shifts in perception as a function of cue-conflict.

Simulations further demonstrated that under the recorded sensory noise levels, discerning between forced fusion and causal inference becomes implausible with substantial conflicts. Overall, the results imply that under the tested conditions, observers may not depend on causal inference when evaluating auditory durations. However, significant noise in the encoding of auditory duration may hinder the differentiation between competing computational theories.