For a quick overview for those just joining, I am examining the results of an experiment on latent inhibition using a video-game method by developing a tool to rapidly view and group the eye-tracking data. Some of the data I’ve already described.
I finished my Area of Interest editor so that I can open the editor, create circular areas of interest and free-form polygonal areas of interest. I can also create what I believe to be a new way of looking at eye-tracking data and that is a “scan path” interest area (but, that “new way” may reflect my ignorance of eye-tracking analysis in general).
Basically, when collapsing across subjects and time, patterns emerge that show generally how each individual’s gaze moved across areas. So, I can simply draw a curved line on the screen and call that a “path.” Then, for the trials and time-intervals of interest, I take each observation point and compare it to each point in the path to find the minimum distance of each observation point from the path. (That little exercise demonstrated to me just how slow “managed” code can be.)
To illustrate, the screenshot above shows two areas of interest. One is circular and around the sensor used in conditioning. The other is a curved line. I’ll have to work on that line a bit because it is impossible for the user to draw it smoothly using a mouse.
What I shall do is set it up so that the user can define several points through which he or she wishes the curve to pass, and then connect the points using one spline or another (probably Catmull-Rom) to achieve a smooth curved line.
The points in the image show all the gaze points for all the subjects in the Control Group during the last five conditioning trials. The points show that their gaze generally focuses on the CS and the weapon that is used to repel the US spaceship, with points falling in between indicating where their gaze goes when shifting between the two.
They could be looking at the gun because the plasma is animated, because that is where the US arrives, or because the gun is associated with the US. Until the next experiment tells me otherwise, I’m going to believe that the cause is one of the latter two reasons.
So, for every subject and every point I calculate the shortest distance of each point from the curved line. If the two groups are behaving the same, as far as directing their attention to the screen, then they should not differ on that measure. But, as the figure below shows, they do not.
The figure shows the average distance of the gaze points from the CS-US scanline diagramed above, for each second of the CS on each trial. The control group is clearly keeping their gaze closer to the scanline than the Latent Inhibition group (and it is statistically significant). Both groups attention becomes more focused towards a path that involves the CS and the US over trials. (Remember, I am showing the five seconds of the CS before the US arrives)
The data from the five extinction trials tell essentially the same story. (These show all 20-s of the CS as no US was present).
The latent inhibition group showed less attention to the path from the CS to the US arrival zone.
Next post I’ll describe the “count” data, which simply reflects a count of the number of gaze hits within an area of interest, but I need to correct a little bit of flawed logic in the program that I just found.