A novel flexible method to split feeding behaviour into bouts

Abstract

Before meal patterns can be analysed properly, a biologically relevant meal criterion must be determined in order to group short-term feeding behaviour into meals. Existing methodologies are based on modelling of the frequency distributions of intervals between feeding events but these methods cannot be used if the proper distributions cannot be clearly identified. For such cases we developed two new methods - (1) based on the analysis of the distribution of between-meal interval lengths only and (2) based on the analysis of changes in the probability of animals starting to feed with time since the last feeding event. Both methods were developed using a data set of over 700,000 records of visits to feeders obtained with broilers (Gallus gallus) aged between 2 and 5 weeks. The two methods resulted in meal criteria estimates of 20.1 and 17.5 min, respectively, which, when applied to the data set, gave statistically significant but Very small differences in meal characteristics. The new methods were tested against an independent cow (Bos taurus) data set and the resultant meal criteria compared with those predicted by an existing method. The two novel methods estimated meal criteria for cows at 27.9 and 35.5 min, compared with 28.9 min for the existing method. Again, these differences in meal criteria resulted in only very small differences in meal characteristics. Even though meal criteria were relatively similar for birds and cows, characteristics such as average daily number of meals (10.9 and 5.9). meal size (12.5 g and 7.4 kg) and meal duration (7.7 and 31.4 min) were very different. The analyses show that, if the appropriate distributions of intervals cannot be identified, meal criteria can still be estimated for species as diverse as mature cows and young birds by the novel methodologies developed here. (C) 2008 Elsevier B.V. All rights reserved.