The progress of a project aimed at developing a decision support system (DS
S) for overall welfare assessment is described. We chose the pregnant sow a
s a 'case' to develop the assessment method. First, we designed and tested
a very simple prototype DSS in order to examine several basic ideas as to h
ow welfare can be assessed on a scientific basis. In this prototype welfare
assessment was based on the needs of animals. The second version of the DS
S, which extended the approach taken in the first, included a wider range o
f housing systems and an improved knowledge base. It used a rigid formal pr
ocedure to construct the model, moving directly from scientific statements
into if-then rules. However, this extended version resulted in counter-intu
itive scores for the seven main housing systems for pregnant sows. In the t
hird version of the DSS the complexity of the model was reduced, if-then ru
les being excluded. It was based on more general biological considerations
(needs and sub-needs) and produced scores which accorded better with expert
opinion. Weighting was managed in a simplified, but procedural way. For im
proved welfare assessment we suggest that the method of 'reflective equilib
rium' should be used.