In interviews with 11 pig experts the main housing systems for pregnant sows were identified as tethering (T), individual housing in stalls (IS), group housing with stalls (GS), trickle feeding or biofix (B), electronic sow feeding (ESF), and outdoor housing with huts (O). The family pen system (Fam) was added as a reference system. The experts were asked to give a welfare score for each housing system. The 2 individual housing systems (mean scores: T=1.8; IS=2.3) scored significantly lower than more intensive indoor group housing systems (GS=5.4; B=5.3; ESF=6.2), and these scored lower than the more extensive systems (O=8.0; Fam=9.1; ANOVA, PP=0.008). The most important aspects for welfare assessment were space, substrate, feeding-related agonism and social parameters such as group size and group stability. Three different models were constructed to calculate welfare scores from the arguments given by the experts. When represented graphically the results seem comparable to the expert scores, although 2 of the 3 models differed significantly from the expert scores using analysis of variance. These results indicate that pig experts are able to perform overall welfare assessment in a rational way that allows modelling and that there is a consensus underlying welfare assessment. These outcomes provide support for the further development of a decision support system to assess farm animal welfare on a scientific basis.