An experimental design for functional MRI (fMRI) is presented whose conceptual units of analysis are behavioral trials, in contrast to blocks of trials. This type of design is referred to as a trial-based (TB) fMRI design. It is explained how TB designs can afford the ability to: (1) randomize the presentation of behavioral trials and (2) utilize intertrial variance in uncontrolled behavioral measures to examine their functional correlates. A particular type of TB design that involves modeling trial-evoked fMRI responses with one or more shifted impulse response functions is described. This design is capable of discriminating functional changes occurring during temporally separated behavioral subcomponents within trials. An example of such a design is implemented and its statistical specificity, functional sensitivity, and functional specificity are tested.
Zarahn et al. (1997) studied this question.