Predictable Timing Enhances the Use of Predictable Content

Abstract

As events unfold, they carry structure in both their timing and content, with these dimensions often interrelated. Most research has treated these dimensions separately, although processing them jointly may provide important computational benefits. It is an open question whether the brain processes a sequence's timing and predictable content structure independently, or integrates these two sources of information to guide prediction. We hypothesised that predictable timing would allow listeners to better exploit structure in the content, with this effect increasing as content structure becomes stronger. To test this, we presented tone sequences with pitch trajectories following a 1/f^{beta} structure, where {beta} determines how strongly the content of each tone persists, while independently varying the temporal periodicity of the sequence. Performance was quantified using response errors relative to outcome predictions of a Gaussian Process Regression model and a model-free measure of how strongly preceding tones influenced responses. Response error decreased and reliance on preceding tones increased with higher {beta} and periodic timing. Moreover, the influence of preceding tones showed an interaction between timing and content, such that the effect of periodicity grew with increasing {beta}. The response error showed no such interaction, although the effect of periodic timing was reliable only for the most structured sequences. Our findings suggest an interdependency between content and timing, where predictable timing helps listeners make greater use of predictable content.