Learning is an important way of behavioural adaptation to improve future interactions between animals and their environment by individual acquisition and processing of information. Modern animal husbandry is increasingly characterised by automation in food and water supply, care, and even cow milking. Within this framework, cognitive capabilities gain importance in farm animals. Most of the systems that are offered, however, integrate the interactions between animals and technology only on a low level. One of the reasons for this could be an incomplete knowledge on farm animal cognition. The aim of our study, therefore, was to develop an effective and suitable method to test farm animals for their learning capabilities and cognitive abilities. In addition, the set-up was designed in such a way that it is possible to examine learning behaviours in larger, socially structured groups of animals. As a model, the African dwarf goat was used.

A general-purpose computer system for behavioral conditioning experiments, called "Feldermonitor", was developed. It consists of a 17" computer monitor with the screen covered by a clear acryl glass divided into four sectors each containing a switch in the middle. When an animal chooses the switch with a specific conditioned visual pattern, the computer delivers a portion of water (20-30 ml) into a bowl. By this way the animals are forced to retreat after the correct choice, drink the water, and then move back to the screen. Each goat is registered individually by a transponder system. Each pressing of a switch triggers the data collection. Automatically registered data are: animal number, time, position of the S+ stimulus, position of the S- stimuli, current state of the screen (active or black) and correct or incorrect choice. The data collected during an experimental day where written to disk.

This set-up allows flexible stimulus presentation and rewarding for different kinds of learning behavior of animals under group housing conditions. Using this device, up to now 160 African dwarf goats were tested in 14 groups for simultaneous visual discrimination tasks with four stimulus patterns (simple geometric figures) presented simultaneously. Each test consisted of a period of 14 days.

The learning ability of each goat and the mean learning ability of a group of goats can be modelled by using a non-linear model with four parameters. These parameters can be interpreted very well in the context of learning. A learning curve can be obtained for each goat. A future long-term experiment will include the genetic improvement of learning success by means of selection. The selection criterion will be derived from learning curves.

© 2000 Noldus Information Technology b.v.