Analysis of the feeding behavior of pigs using different models

Abstract

Short-term feeding behavior is conventionally analysed using random process models. The assumption underlying these models have recently been questioned and this article describes the application of both random, and more biologically based, models to the feeding behavior of pigs. Feeder visits of 16 growing pigs, housed individually from 17 to 52 kg live weight, were recorded electronically over a continuous period of 35 days. Daily food intake increased linearly with time, but there was considerable individuality in the degree of order. Pigs made between 18.8 and 80.3 (mean 47.9) daily visits to the feeder. Intervals between visits could be described by two log-normal distributions. Two Gaussian density functions were fitted to the distribution of the log-transformed intervals. For the combined data from all animals the within- and between-meal intervals were 11.2 s and 100.1 min, respectively. A model with three Gaussian functions gave an improved fit to the interval distribution. The within and between meal intervals were then estimated to be 4.2 s and 93.9 min, respectively. The middle distribution of intervals ranged from 0.5 to 38.1 min. The intervals were also described by random process models; again, a three-process model gave an improved fit compared to a two-process model. The mean estimated number of meals per day from the three Gaussian model was 14.3, and from the three process random model, 16.3. A biological interpretation of the three types of interval suggests that: (1) pigs eat in meals separated by long intervals; (2) meals consist of clusters of eating bouts separated by shorter intervals, sometimes associated with drinking; (3) within each eating bout short intervals occur as pigs constantly move in and out of the feeder. It remains unclear what underlies the observed patterns of eating.