Infants' Categorization of Rhythms on the Basis of Periodic Temporal Structure Erin Hannon Cornell University eeh5@cornell.edu   Much of the research in statistical learning demonstrates pattern learning on the basis of sequential temporal structure, but some pattern learning may benefit from other types of temporal structure, such as the periodic structure commonly found in music. In order to perceive the periodic cycle of the musical beat (the "meter"), an individual must be sensitive to the regular placement of more salient over less salient events at particular positions in a periodic cycle. The length of the cycle is inferred from the consistent placement of more salient events (accents) at downbeat positions and less salient events at other upbeat positions, but once a listener grasps this periodic structure it may aid in learning and remembering relationships between nonadjacent events. These studies examined infants' ability to discriminate rhythms on the basis of the underlying metrical structure they imply. A simple algorithm was used to create rhythmic patterns that would reinforce a primary metrical pulse every three or four events, or, in musical terms, a triple or duple subdivision of the beat. Four different patterns were created in each meter, composed of identical numbers of events and silences. Rhythms differed only in the periodic placement of temporal grouping accents, which made some events much more salient than others depending on their position in the cycle. Seven-month old infants were habituated to a series of three different rhythms with the same metrical structure, and tested on a novel rhythm in the same meter and a novel rhythm in a different meter. Infants looked longer at the static monitor/speaker display when tested with a novel meter rhythm than with a familiar meter rhythm. When the placement of accented events was less consisten! t, this preference disappeared. These results suggest that by 7 months, infants can perceive metrical structure in rhythms that contain reliable and consistent accents at the downbeat. This may be a type of statistical learning based on detection of regularities in periodic temporal structure.   Back to main page