Re: Categorisation and Prototypes

From: HARNAD Stevan (harnad@cogsci.soton.ac.uk)
Date: Mon Jun 03 1996 - 20:03:12 BST


> From: "Humphreys, Rebecca" <RJH295@psy.soton.ac.uk>
> Date: Thu, 23 May 1996 11:36:01 GMT
>
> According to the Prototype theory categories are organised around
> central prototypes- the best example of the concept. Category
> membership is determined by the simularity of an objects attributes to
> the category's prototype. In contrast to the classical view of
> categorisation where features are necessary and sufficient for deciding
> what belongs in what category. Rosch gave evidence for the theory
> saying whilst a categories defining attributes existed people could
> often not define them.

She also thought they often did not exist at all.

> Some may have features impossible to define i.e
> "game" the only thing it's members have in common are vague "family
> resemblances" not features (Wittgenstein).

But what ARE family resemblances?

> They could however name more
> "typical" members of categories i.e Robin is a more typical example of
> a bird than Ostrich, and faster than the more obscure. Berlin and Kay's
> (1969) crosscultural studies showed a universality in people's colour
> ctaegorisation. People could agree on the best example of red or blue,
> and seemed to categorise colours on the basis of resemblance to focal
> colours.

But whereas redness DOES admit of degrees, birdness does not...

> The prototype theory requires a template of an idealised
> member of categories, but not all concepts have prototype
> characteristics i.e what is the "ideal" chair. The prototype of which
> is a set of features not an "ideal" member. Prototypes do not explain
> how people know some attributes of a categegory are more likely to vary
> than others. Simularity cannot be the only mechanism responsible for
> category cohesion- some categories are coherent but don't share many
> attributes.

What do you mean by "cohesion"? (I really don't know!)

> People may not be able to tell you what the features of a
> category are but may know implicitly. The prototype approach has not
> led to a model being built that can categorise, whereas the classical
> view has had some success in modelling in human categorisation. The

Although you seem to have meant to add something further, this reply is
already an A! To get a still higher mark, you'd need to go deeper into
what features are, and how closeness to a prototype is really just a
special kind of feature.

Well done! But please do use a spell-checker...



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:43 GMT