I like your idea. But,  I'm sure you need professional advice of scientists, equal professionalism such as David Hubel and Torsten Wiesel.

Hey Jaoa!

I'll answer mainly to your first question.

Are your targets randomly distributed on the screen? If not, I think assuring fixation is important to prevent your subjects from directing covert/overt attention to the important parts before the trial.

Assuring fixation is also important to avoid eye movements (overt attention shifts). If you can not assure fixation, this will, depending on your stimulus' size and color combination, also affect the performance due to the different spatial resolution and color sensitivity in the central & peripheral field of view. If you have random item placement, you can reduce the effects of eye movements by using stimulus presentation times below saccade latencies

Hi João, 

In aspects related to the preceptual attention, fixation can be a good idea (precisely, fixation duration). And in addition to the fixation duration, you might use "Pupil diameter" to measure one's cognitive load. 

To confirm the occurrence of participants' fixation in certain task, you could use graphical representations such as "heat map", especially heat map-based fixation duration.

HTH.

Samer

Ensuring a fixation always entails a dual task for the subject, unless the task itself promotes fixation. The reason is that the subject has to suppress the normal exploratory eye movements. So adding an explicit dual task is not a problem. However, I see no good reason why you need to ensure fixation for your task. 

I agree with Joren, that it's not clear your task and question really require participants to fixate.  If you have a strong hypothesis that fixation is critical, perhaps you complement your large sample Mturk study with a small controlled lab study that manipulates the fixation requirement.  

Unnamed people spoke negatively about the professionalism of David Hubel and Torsten Wiesel. It's modern life.

I'm not sure fixation matters here either. In an in-lab experiment of this type, people would frequently just provide instructions relating to where fixation should be but not necessarily use eye-tracking equipment to verify it (this isn't the kind of study that would normally call for an eye-tracker I don't think). I don't see that there's any reason to think that in-lab participants follow instructions any better than online participants. I test with both in-lab participants and on-line paid subject pools and I usually get better data online (in terms of not having to chuck participants out for being boxheads, at least).