Conclusion
- There aren't a lot of cognitive models of how people categorise,
and we use that of Shepard, Hovland and Jenkins; it's a two category
task.
- We do it all with neurons, using compensatory learning.
- We need to integrate singleton and triplet features.
- This gives us a reasonable model of the task, but
not fully accurate as a cognitive model.
- We propose that it is a model if one part of the categorisation
system.
- I predict that this will accurately duplicate behaviour when people
passively learn a category.
