Using contrasts to test trends: precautions

Sometimes a study will have multiple trial types where it might be expected that the BOLD signal will increase linearly, quadratically or with some other trend over trial types. For example, there were 3 tasks with increasing difficulty, you might expect the BOLD signal to increase linearly. One approach to studying this linear trend is to first model each trial type separately and then use a contrast to test the trend. For example, say you have 3 trial types, with 1 block of each trial. Ignoring the hemodynamic delay, the hypothesize data, regressors you would use in the model and hypothesized parameter estimates obtained using that model are illustrated in Figure 1. As shown in the third panel of Figure 1 the parameter estimates increase linearly with the difficulty if our hypothesis is true.