A stochastic model for time-ordered dependencies in continuous scale repetitive judgments.

Publisher Website
Google Scholar
Abstract
Previous data showed that Ss performing a task in which they were required to bisect, on the basis of kinesthetic cures, a perceived angular extent (by turning a knob) tended to produce an initial drift in constant error. To study more clearly the serial dependencies in this task, 20 Ss were required to perform long series of these adjustments. Preliminary analysis showed the time series generated in this manner to be nonrandom. To procure a more complete picture of the time-ordered dependencies in the data, the time series generated by 10 of the 20 Ss were subjected to autocorrelation and power spectrum analysis. These techniques revealed that neighboring observations tended to be most alike, with influence declining as the temporal separation became greater. A stochastic model was found for this pattern of serial dependency which could be described as a Markoff (first-order autoregressive) process. This means that for the best prediction of a setting only the previous setting need be known, the history of earlier performance making no direct contribution to the prediction. On this basis, it is suggested that remote past experience in kinesthetic judgments is of less consequence than commonly assumed.
Keywords

This publication has 1 reference indexed in Scilit: