Abstract
Livestock herding dogs are crucial contributors to Australian agriculture. However, there is a dearth of empirical studies of the behavioural interactions between dog and livestock during herding. A statistical approach that may reveal cause and effect in such interactions is lag sequential analysis. Using 48 video recordings of livestock herding dogs and sheep in a yard trial competition, event-based (time between behaviours is irrelevant) and time-based (time between behaviours is defined) lag sequential analyses identified several significant behavioural interactions (adjusted residuals greater than 2.58; the maximum likelihood-ratio chi-squared statistic for all eight contingency tables identified all sequences as highly significant (p < 0.001)). These sequences were: The dog ceasing all movement followed by the sheep also ceasing movement; the dog chasing the sheep and a group of sheep escaping the main flock; a single sheep escaping the flock and the dog chasing; sheep initiating movement followed by the dog following; foot-stamping followed by the dog ceasing all movement; and, foot-stamping by the sheep and the dog lip-licking in response. Log linear regression identified significant relationships among undesirable behaviours in sheep and both observed trial duration (p = 0.001) and trial score (p = 0.009). No differences in the herding styles of dogs were identified between sex of dog and frequency of sheep escape behaviours (p = 0.355) nor the sex of dog and competition level (p = 0.116). The identification of trial score as a predictor of efficient performance confirms the benefits of incorporating extant objective measures to assess livestock herding dogs.