Perceptual Priors Revealed by Iterated Reproduction.
Probability distributions over external states (priors) are essential to the interpretation of sensory signals. In many areas of perception and cognition, humans appear to combine current observations with internal beliefs about the environment (the prior) in a process approximating statistical inference. Priors for cultural artifacts such as music and language remain largely uncharacterized, but critically constrain cultural transmission, because only those signals with high probability under the prior can be reliably reproduced and communicated.
Extending previous research on iterated learning, we developed a method to estimate priors for rhythm via iterated reproduction of random temporal sequences. Listeners were asked to reproduce random “seed” rhythms; their reproductions were fed back as the stimulus, and over time became dominated by internal biases, such that the prior could be estimated by applying the procedure multiple times. We measured priors on simple rhythms in residents of the United States as well as members of the Tsimané, an Amazonian society with very limited exposure to Western music. We found that priors in US participants showed peaks at rhythms whose time intervals were related by small integer ratios. The modes of the prior were limited to small integer rhythms prevalent in Western music, and were similar for musicians and non-musicians, suggesting that priors are shaped primarily by passive exposure to the music of a culture. Priors in Tsimané participants also exhibited modes at integer ratios, but were otherwise qualitatively different from priors in US participants, in ways that are consistent with the structures prevalent in their music. Results were similar for different modes of reproduction (finger tapping versus rhythmic vocalization of a repeated syllable), but did not extend to the reproduction of spoken phrases. Our results are consistent with the claim that rhythm perception exhibits universal cognitive constraints favoring small integer ratios, but indicate that any such constraints are strongly modulated by experience. Our method holds promise for characterizing priors in a range of other domains in both audition and vision, including spatial memory, phonetics, and melody.