Previous research has shown that there is a strong relationship between church involvement and preference for the Belgian Christian Democratic Party (CVP/PSC). When religious preference is controled, the relationship between social class and preference for the CVP/PSC seems to disappear completely. Indeed, this party is very heterogeneous as regards the social origins of its adherents. The situation is completely different for preference for the Socialist Party (BSP/PSB) which has a strong relationship between party preference, on the one hand, church involvement and social class, on the other. The unchurched and the workers manifest more preference for the BSP/PSB than do other people. According to Hill the degree of involvement in the «zuil» (pillar organizations), the so-called «organizational variable», is also significant in the explanation of voting behavior. This could not be confirmed, however, due to a lack of data. Two hypotheses are formulated: 1) pillar involvement in addition to social class and church involvement is significant in the preference for the BSP; 2) there is no connection between preference for the CVP and social class so that only church and pillar involvement remain. These two models were tested by means of loglinear analysis and seem to fit the data. The intention of this article, however, is primarily didactic: to show the possibilities and limitations of the analysis of contingency tables by means of Goodmans loglinear analysis procedure through the use of asimple and relevant example.