We describe an implementation of the general linear model that facilitates the characterization of evoked hemodynamic responses to sensorimotor or cognitive processing, when the exact form of these responses is not known. The importance of this approach is that one can test for differential responses among tasks that may elude more conventional analyses. In particular, we suppose that an evoked response has early and late components and that a differential response may involve (i) both components to the same degree, as in a conventional "activation" or (ii) differential expression of the early and late components in two tasks, as might be seen in differential adaptation, or differences associated with the tasks (e.g., requiring and not requiring sustained attention). Using this approach we were able to demonstrate that the anterior cingulate differentiates, in terms of its response, between two motor tasks that did and did not require sustained attention. This differential response was observed even though there was no classical "activation" (i.e., there was no difference in the mean activity associated with the two conditions). It is suggested that these demonstration results point to the possibility of making greater use of the temporal resolution afforded by fast fMRI techniques.