Minimising Errors in Splitting Behaviour Into Bouts

Abstract
Some problems in using log survivor functions to split behaviour into bouts are outlined. It is argued that it is usually best to choose that bout criterion which leads to the fewest within and between bout intervals being assigned to the wrong category. A way of doing this is illustrated and other possible criteria are considered. The influence of misassignment on data analysis is discussed and it is argued that where this is substantial analysis in terms of bouts may not be useful.