The entropy of behaviour under stress (360G-Wellcome-109314_Z_15_A)

Stress is both widespread and difficult to detect, with massive health and economic impact on both humans and the millions of animals we care for. Individual differences in response to stressors show behaviour can become more repetitive and stereotypic or more unpredictable and chaotic, so universal behavioural indicators of stress have been elusive. We propose that entropy, a key concept in information theory which captures how chaotic or disordered a sequence of data is, will be a useful approach to apply to behavioural signals, with stress causing extremes in entropy on either side of the spectrum. We propose to test this theory by:   (i)            Developing methods to extract measures of behavioural entropy from automatically collected accelerometer signals;  (ii)           Studying the change in behavioural entropy under acute stress, with individuals selected to provide a variety of response valences;  (iii)          Investigating the effects of individual behavioural differences and personality as predictors of the directional change in entropy.  The overall aim of this research will be to develop a robust and versatile behavioural indicator of stress, with far-reaching applications to understand the neurological changes stress causes, and improve detection of poor animal welfare.