Abstract

The locomotor pattern of a goldfish may be described by 17 locomotor variables whose time series are known to exhibit distinct patterns of autocorrelation. The present model reduces the set of 17 variables to a set of seven predictor variables which optimize the prediction of future locomotor behavior. These seven variables are then grouped into two clusters by a dendogram analysis, and it is shown that the clusters are also characterized by their strength of prediction. A locomotor control model is then formulated where one cluster is interpreted to be a level of activity factor and the other is hypothesized to be a turning behavior factor