AN EXPERIMENT IS DESCRIBED WHICH INVESTIGATED DIFFERENCES IN RELIGIOSITY. IT WAS FOUND THAT RELIGIOUS PARTICIPATION, A TRICHOTOMOUS RESPONSE IN THE FORM OF A TWO-ITEM GUTTMAN SCALE, IS AFFECTED LINEARLY BY SEX AND BY MARITAL STATUS, BUT THE EFFECT OF COLOR IS NONLINEAR. THE PAPER DESCRIBES TECHNIQUES THAT ALLOW THE SOCIOLOGICAL ANALYST TO IMPOSE SIMULTANEOUSLY THE TWO LINEAR CONSTRAINTS AND A MONLINEAR CONSTRAINT OF EXTREME RESPONSE CATEGORIES VERSUS THE MIDDLE ONE TO REPRESENT THE COLOR EFFECT. SUCH TECHNIQUES ARE USEFUL AS ANALYSTS CONSIDER THE POSSIBILITIES OF SPECIFYING STRUCTURED MODELS FOR CATEGORIES OF DATA. TABULAR DATA, REFERENCES, AND AN APPENDIX ON FITTING THE MODELS ARE INCLUDED. (AUTHOR ABSTRACT MODIFIED--MJW)