A conceptual framework is presented for studying trivariate inference, in which the influence of a primary independent factor, x, on a dependent variable, y, depends on a secondary factor, z. In two experiments, this was operationalized as the contingency between mating strategies, x, and mating success, y, for targets stemming from different tribes, z. In Experiment 1,, was manipulated to produce four contingency types: control, orthogonal, suppressor, and spurious correlations, Performance variation was not restricted by encoding capacity; explicit and implicit measures were more accurate at the trivariate level than at the bivariate level of assessment. Aside from normal regression effects, neither task complexity nor competition among influence factors led to impaired performance. Impairment was evident mainly in a radical discounting effect for spurious correlations due to representational conflicts. Experiment 2 demonstrates that the difficulties with spurious correlations can be overcome when temporal cues disambiguate the joint influence of x and z on y. (C) 2001 Elsevier Science.