Dear Carmen,
Some reflections:
I think it depends on the aims of the study. Assuming certain stimuli are more easy to judge (e.g. those at the extreme ends of the pain continuum), then they might also be associated with shorter reaction times. If the response period is presented after some delay you might not be able to detect this effect any longer because subjects now have sufficient time to decide which rating to give for any of the stimuli. Of course you could additionally ask for a certainty rating, but this might be subjective and/or unrelated to an underlying RT effect (if there is any). Thus, if you want to test whether different stimuli affect RT the response period should probably be as close as possible.
If you want to avoid RT effects as much as possible, then temporally separating different parts of a trial would indeed be an option. However, you will nonetheless get no "pure" stimulus regressor because subjects are going to start with decision making / motor planning / ... right from the beginning, not only when a request stimulus is presented (as they know they have to give a rating). Thus corresponding activations are not just pain-related, but might reflect various other processes. Different conditions/trials might still be associated with shorter decision periods, you just can't see this effect any more. This is neglected in many papers that go with separate response periods.
One solution would be partial/half trials (just the pain stimulus, no response), as suggested by Donald. Probably the majority of the trials would have to be partial though, otherwise subjects are still going to think about the rating right from the beginning, possibly resulting in some bias. Especially in your case, if it were a more complex task one might argue that subjects are too lazy to think about it ;) except if they're really forced to.
Best,
Helmut
|