Oct 15 2012
-Typicality of constituency didn’t produce a good prediction of resulting structures
- prototype theory predicts linear separability and prediction but empirical
evidence has not supported this predication, not playing a crucial role
Conceptual confirmation should be able to confirm inference about those categ.
Using prototype theory however when we formalize it and apply it . it should give
us an account of compositionality
but the prototypes do not contribute to our understanding of the composition
Shared assumptions got 2 theories (classical and prototype theory) and both
highly problematic maybe both theories are failing because they have both flawed
assumptions which might be causing them to both fail.
Presupposition an assumption that is shared across more than 1 theory and 1
Some of the Assumptions
-structural assumption Concepts are feature lists (classics)
-Functional assumption the main job of a concept is to label/describe world
This is a presupposition.
Both of these assumptions are very flawed and needs to be rejected.
-Structural/ feature list problem doesn’t talk about relations between features ex.
2 features of bird, small or large if its small it’s likely to sing if its large its not
likely to sing. This is correlational information known as the gestalt
(organizational structure) THERE IS AN organizational structure more than the
feature ( there’s a system of relations that is something other than the conjunctive
joining of the list of features or “ the whole is more than the parts its made of’)
Reality isn’t a conjunctive list.
-The idea of labelling is that it just “points and describes an object/concept” but
people are trying to do smth else in their concepts like explain how the object
One of the reasons why you look for structure is so you can look for its
Theory of Concepts: -what ppl are trying to do when they form a concept when we are trying to make
theories trying to explain it not just describes
concepts are micro-theories
Problem with correlations is that there’s too many of them but people are also
interested in constitutive relations.
Theory vs. Data
-Not simply the fact that data produces data but theory also helps you look
where to find data, select the data, what interpretations you’ll draw,
generalizations to be made. Data does not simply generate theoy
-this view may override massive perceptual similarity if they have a better
- select similarities that will allow us to explain the phenomenon.
advantage of theory it’ll give us a new way of making selections micro theory
Might begin to explain how conceptualization works
-May explain what typicality gradients
-this theory can also predict fuzzy boundaries with respect to the world (without
having to rely on prototype)
-People will give features at different levels of abstraction
-Multiple levels of description/features
-explanation: people are trying to find links