When two types of stimuli are mixed in a trial block, each stimulus type is typically responded to more slowly than when those stimuli are presented by themselves in pure blocks (a "mixing cost," Los, 1996). In word/non-word naming tasks, however, mixing two types of stimuli leads to a different, "homogenization," pattern. There is a mixing cost for the easier stimuli and a mixing benefit for the more difficult stimuli (Lupker, Brown, & Colombo, 1997). In the present research we investigated the generality of this homogenization pattern by examining picture naming and a sum-naming task involving addition problems (e.g., 10+7 = ?). In Experiments 1 and 2, the homogenization pattern was observed for both pictures and sums. In Experiments 3 and 4, qualitatively different stimulus types (words and pictures, words and sums) were mixed. The mixing cost pattern was observed. Experiments 5 (words and pictures) and 6 (words and sums), however, demonstrated that a homogenization-type pattern can be obtained even when qualitatively different stimulus types are mixed. These results indicate that theoretical mechanisms like those proposed by Los (1996) and theoretical mechanisms like those proposed by Lupker et al. (1997) are both active in reaction time experiments.