Monday, April 02, 2007

Friday CogSci: Conditioning on perceptual cues.

Ben Backus from Penn

Dr. Backus asked the question, which pairing of perceptual stimuli can be classically conditioned? Not just learning greater acuity, but actually learning perceptual experience; "the way the world looks". Can stimuli from different modalities influence each other? See for example the illusion of simultaneiety when someone claps their hands together a fair distance away. How does perceptual learning compare with other forms (appetetive, aversive) conditioning?

Interesting that, given these illusions, Dr. Backus assumes that visual perception is "optimal" e.g. uses all sources of information. What does it mean to argue that a structured representation is "correct"?

So, is the cue arbitrary? No, some cues (unconditioned stimulus) work better than others when paired with a given conditioned stimulus. For example, sound doesn't work to flip a Necker Cube illusion, but it does work on the bouncing/ghost ball illusion. There was a lively discussion afterward and during lunch about what makes a good model species, whether evolutionary justifications can explain (not just post hoc) what cues will work and which won't.

Is learning incremental or abrupt (i.e. paradigm shift)? Incremental, although perception is abrupt ("all or nothing").

Is there a "learning to learn" effect? No.

No comments: