When we look at an object, we subconsciously make multiple eye movements to different parts of the object, building up information about it until we can decide what it is. These tiny bursts of information create important activity in our brains, but it’s difficult to see this activity because eye movements happen so quickly that responses overlap, and because eye movements create lots of noise in EEG measurements. We are trying to solve both of these problems by applying techniques from fMRI to EEG activity, helping us see how information builds up in the brain like never before.
The experiment is a very simple version of the process described above: the subject is asked to move his eyes to five colored squares in a line across the screen. If he sees two or more squares with a blue bar in the middle (a “complete trial”), he must press a button. When the subject sees the first blue bar, he starts to integrate information: “this might be a complete trial…” When he sees the second blue bar, the integration is complete and he can make a decision: “this is definitely a complete trial!” So with this simple experiment, we can separate out and study the integration and completion of decision information.
The problems with analyzing this data, described above, can be dealt with by using a general linear model, or GLM. A GLM uses correlation analyses to try to tease apart responses that overlap. fMRI analyses use this technique because the changes in oxygen-rich blood flow that it detects are very slow (so the response to one trial runs into the response to the next trial), and because there are lots of blood flow changes happening that are unrelated to the thing you’re trying to study. Normally we don’t have the overlapping problem with EEG because responses happen so fast, but eye movements happen quickly enough to give us some overlap. Where other techniques fail to figure out which eye movement caused certain EEG activity, a GLM succeeds!