Determining the areas of brain activity associated with cognitive processing has typically relied on the use of a subtraction paradigm, which is based on the premise that the neural processes underlying behavior are additive. If the additivity assumption is valid then brain regions associated with a semantic processing task should be the same regardless of how participants make a response. To investigate this proposition, participants underwent six PET scans, in which they made semantic or letter word judgments, responding "yes" or "no" in three different modes: mouse-clicking, spoken response, or silent thought. Analyses showed an increase in regional cerebral blood flow associated with semantic processing in the left inferior frontal cortex, anterior cingulate, and right cerebellum for all three response conditions. However, there was a significant interaction: the greatest increase was observed in the mouse-click condition and the weakest change seen with silent thought. Moreover, other areas of the brain were uniquely activated for each response mode. The results indicate that different areas of the brain were recruited for semantic processing depending on how participants had to organize their responses. Implications for the additivity assumption and methods of analysis to be used in conjunction with the subtraction technique are discussed.