> I have seen a study from UCL, they took ONSET only for target images
I don't know which study you refer to, but there are indeed instances in which one would model only one condition, although the experiment consists of two conditions. The most common case is a fast event-related Go/No-go task or similar paradigms in which one condition is very frequent, thus the corresponding predictor / expected time course would be (more or less) constant over time, making it pointless to explicitely model the condition. You would then only model the infrequent condition and possibly additional regressors for errors.
Best,
Helmut
|