The way in which the data are combined affects the interpretation of short-term feeding behavior

Short-term feeding behavior of pigs has been analyzed using random process models and log-normal models. Both were successful despite very different u nderlying assumptions relating to the theory of control. Feeder visits of g rowing pigs, housed individually from 17 to 52 kg live weight, were recorde d electronically over a continuous period of 35 days. For the combined data , intervals between visits to the feeder greater than 30 min could be descr ibed well by the negative exponential model. The starting probability of a visit was constant at around 0.3, suggesting randomness. Disaggregating the data for individual pigs or for individual weeks did not change this concl usion. Intervals in the day were of a different nature to those at night, a nd disaggregation of the data into these two periods revealed that the nega tive exponential model was not satisfactory for either period. The starting probability for both periods increased with time since the last visit. Thi s is consistent with the idea of satiety. Therefore, the apparent randomnes s in the data pooled across the day and night is an artefact caused by pool ing itself, and is not in conflict with the satiety concept. The implicatio ns of data handling are discussed with reference to studies of the physiolo gical control of food intake. (C) 2000 Elsevier Science Inc. All rights res erved.