Thursday, August 16, 2012

quick note

Something I've been meaning to write a note on for a while. Even as I write this, I don't have the idea quite in my head.

What is it about apparent contrast that is interesting, as opposed to sensitivity? By most measures, they're the same thing. If I ask someone to do an experiment where they have to discriminate between one contrast and another, I'm assuming that they are doing some mental comparison between two apparent contrasts, or memories of two apparent contrasts. By varying the physical difference between the two contrasts, I can then quantify the person's performance on this task, e.g. the difference at which they can no longer measurably discriminate. These are the sorts of measurements that are usually made in psychophysics.

These performance measures are understood to reflect the subjective, phenomenal properties of interest. But they involve a nearly unsolvable confound: we can't tell what part of the discrimination is due to relative internal response strength and what part is due to internal noise.

So to me, measuring apparent contrast is a way at getting around this problem. You get to measure, directly, the internal response to the stimulus. The new problem, then, is in quantifying what you've measured. The reverse correlation experiment that I did, I realize now, was cognizant of all this, but it was subliminal for me. The experiment is not measuring performance, but it is very similar to an experiment that would be measuring discrimination performance. In this experiment, the stimuli are always easily discriminable, so there are no limits to measure. The subject is asked to discriminate between the strengths of the two stimuli, but I measure no interval or reliability of this discrimination. This is because there is no objective stimulus strength.

The purpose of the experiment is to find out what constitutes stimulus strength when strength is defined, to an observer, as luminance contrast. What I get back is (no matter how I measure it) a description of what components count more or less than other components in making decisions about stimulus strength. I then test a bunch of plausible models to see whether they might also count components in similar ways. Lucky for me, only a particular type of model works, so I can make a sort of conclusion from the study.

So, apparent contrast is the way things look, and then behaviors can be carried out on the basis of how things look. Most visual psychophysics directly analyzes the behaviors that are based on appearances. I've tried to directly analyze the appearances themselves. Did I succeed?

No comments:

Post a Comment