I want to design an EEG paradigm for passive affective picture viewing. Can I use a block design for three different emotions, where each block represent a particular emotion? Or would a better design randomize all the images of all emotions?
$\begingroup$
$\endgroup$
3
-
2$\begingroup$ Elaborate please. I doubt this is sufficient information for people to provide you with an answer which will help you out. $\endgroup$– Steven Jeuris ♦Commented Feb 11, 2014 at 15:44
-
3$\begingroup$ Welcome to cogsci.SE. Seconding @StevenJeuris' comment, I'd find it particularly useful if you could explain what you're trying to accomplish with EEG measurement, when those measurements will be taken (before/after or continuously?), and what emotions you intend to study. $\endgroup$– Nick StaunerCommented Feb 11, 2014 at 16:14
-
3$\begingroup$ Thirded, hehe. This is a good question, but we need to know what sort of analyses you'll be doing on the data (power spectra, coherence, etc.). I can tell you that you're likely going to want to lock onto a particular event (image onset, image offset, etc.) for the results to be viable. Also, the best advice any of us can give you is to look at the methods in the existing literature for your field and mimic them as closely as possible. $\endgroup$– Chuck SherringtonCommented Feb 11, 2014 at 17:16
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
0
Disclaimer: These are my preferences, biased by my own research experience.
In general, I am not a fan of block designs. Blocking introduces difficulty in comparing both neural and behavioural activity between blocks due to effects of practice, fatigue, adaptation, learning, boredom, muscle tension and possibly others. It is of course possible to counter-balance block order across participants, but then you have to hope that within participants these effects will be roughly of the same magnitude in order to properly cancel out. Since in neuroimaging research we often don't go for large numbers of participants, I tend to prefer event-related designs.
I also sometimes look at signals which are so low in intensity (such as gamma power) that I want to avoid anything that will saturate the signal, and previous trials of the same type do generally saturate it quite a lot, sometimes to the point that I can't see it any more.
However, I also have two reasons for avoiding purely event-related designs. One is that some effects might depend on previous trials. If, for example, people don't react as much to happy stimuli if they're still recovering from the previous angry stimulus, or if they're surprised by the trial change (there's another emotion for you) and you are interested in the happy reaction itself, then you do need blocks... and you also need to discard the first few trials when a block changes. Another reason is that I might be manipulating too many variables in the task and it gets difficult for the study participants. In that case I might decide to manipulate one variable in a block-wise fashion and another in a trial-by-trial fashion. Only you know whether these restrictions apply to your design.
But even if I do go for a block design, I would never* do all the trials in one trial type at once. Instead, I would create mini-blocks, consisting of a small number of trials, for example 10-20. Then I shuffle these mini-blocks randomly per participant, without repetitions. That way you can get all the positive effects from blocking with minimal negative effects.
.* never again, that is.
