I have a learning experiment with this basic design (cue -> predict using scale -> outcome -> fixation). The whole trial lasts for about 7 seconds. There are 16 different cues in one epoch and I decided that I need 20 epochs (at max) at two stages (10 for stage 1 and 10 for stage 2).
I plan to use this for an fMRI study later on but based on very few pilot (behavioral) data I see that subjects get tired and their attention deteriorates after about 20 minutes as reflected by the number of missed trials and the bad performance (predictions get worse).
Do you have any suggestions on how to make subjects more engaged?
(1) If relevant: feedback on performance, so participants are motivated to do as well as possible. Might be a at a trial-by-trial level, and/or include running averages over trials, and so on. Giving it graphically (e.g. a bar where the length indicates the percentage correct, trying to hit the 100% line) can be useful to supplement text feedback.
(2) A pleasant tone or sound for correct trials, vs a slightly annoying buzz on bad ones.
(2) Incorporate pauses between blocks of trials, allowing participants to take the time they need to rest, blink, stand-up etc, as required. This can be challenging to fit within the demands of an fMRI tasks. But even there, it can be worth breaking the task into chunks with rest periods in between, and combine them again at the analysis stage. The cost of having a fatigued and unmotivated participant is generally higher than giving a minute or two of downtime between runs.