The relationship between interference and facilitation effects in the Stroop task is poorly understood yet central to its implications. At question is the modal view that they arise from a single mechanism--the congruency of color and word. Two developments have challenged that view: (a) the belief that facilitation effects are fractionally small compared with interference effects, or nonexistent altogether; and (b) the finding that interference and facilitation effects are inversely correlated. Statistical simulations, reanalysis of past data, and two new experiments indicate that facilitation is robust and substantial when congruency is deconfounded from lexicality, and that the inverse correlations are mostly spurious. Instead, interference and facilitation are uncorrelated, or at most weakly but inversely related. Resolution of response conflict and lexical convergence can explain either finding. Modeling and interpretation of the Stroop task must distinguish between nonspecific lexicality-based effects and specific color-word congruency effects. (Contains 1 footnote, 4 tables, and 1 figure.)