An automatic milking system (AMS) uses a robot to attach the teatcups and can be continuously available for cows to present themselves to be milked with minimal human intervention. Milking in an AMS is therefore a continuous rather than a batch process, and the conventional practice of cleaning a milking system after each batch of cows has been milked is not applicable.

Milk flow through an AMS is different in scale and in temporal distribution from that in a conventional system. These differences in flow patterns may be significant for hygiene. It has been suggested by Verheij (1992), for example, that milk will flush away much of the residue of the previous milking and that microbial growth will be retarded because of the bacteriostatic properties of fresh milk. Ordolff & Bölling (1992) reported experiments in which batches of milk were passed through a milking installation that was said to simulate an AMS. Their main conclusion was that milk residues left in the system for up to 60 min did not increase the bacterial contamination of the next batch of milk. However, the practical relevance of this conclusion is not clear because the design of the simulated AMS was not specified, so it can not be compared to a practical AMS. Moreover, mixtures of fresh and UHT milk were used, with unknown consequences. The purpose of the experiment reported in this paper was to strengthen the evidence on which to base recommendations for effective cleaning procedures for AMS.

The bacterial count of milk that had been passed through a system was used as the indicator of the cleanliness of the system. It was apparent from previous work by Ordolff & Bölling (1992) that this would produce results that varied widely. The experiment was therefore designed to be as simple as possible. In particular, a simple, easily defined laboratory rig was used rather than an AMS.