Cross-dimensional magnitude interactions reflect statistical correlations among physical dimensions: Evidence from space-time interaction.

Magnitudes of different physical dimensions have been assumed to be processed by a common metric in order to account for interactions between different dimensions (e.g., space, time). This paper tested a different hypothesis, that these cross-dimensional interactions reflect people's experience of statistical correlations among physical dimensions. In the experiment, we manipulated the correlation between space (length) and time (duration). A stimulus consisting of two vertical bars that demarcated a variable stimulus length was presented for a variable stimulus duration; participants were to reproduce either the stimulus length or the stimulus duration. Critically, to reproduce a stimulus length, participants held down the spacebar to grow or shrink (in a blocked design) a length to the stimulus length such that space (i.e. reproduced length) positively or negatively co-varied with time. Reproduced lengths did not vary as a function of stimulus duration under positive space-time correlation but decreased as a function of stimulus duration under negative space-time correlation; reproduced durations increased as a function of stimulus length under positive space-time correlation but this space-on-time effect appeared to be attenuated under negative space-time correlation. These findings are consistent with a Bayesian inference account whereby cross-dimensional interactions reflect people's prior belief/knowledge of cross-dimensional statistical correlation, which itself tunes to recent input.