Schyns 2.6 - 3

From: Stent, Hannah (hms195@soton.ac.uk)
Date: Sat Feb 28 1998 - 10:59:03 GMT


Schyns et al supply evidence in section 2.6 to support their belief in
novel functional features. They view these as the 'synthesis of
elements from raw data'. They state that it is however challenging to
find empirical support for this for 2 reasons:

> "The first difficulty is pre-existence: how do we show empirically that
> a "created" functional feature did not exist prior to the
> categorization problem? The second difficulty is reduction. How can we
> insure that a "created" functional feature does not result from the
> combination of pre-existing functional features?"

Ideally this could be answered by demonstrating that a functional
feature 'fx' was not initially present amongst the features but
becomes a set member due to a new categorisation. However this also
poses a problem:

> "Unfortunately, a nonexistent feature is behaviorally equivalent to
> an existing feature with an "attentional weight" of 0. This makes it
> difficult to tease apart feature weighting from feature creation
> based on simple, direct tests of the existence of a feature in
> memory."

Thus indirect test must be used to assess the assumptions of the
fixed feature theories. These assumptions include 1) that objects are
characterised by pre-specified, unambigious and non-decomposable
features. And 2) that learning always selects and combines the fixed
features to form that form the categories.
Schyns states:

> "An important implication of these two assumptions is that category
> learning is only strategic. That is, learning weighs features of the
> fixed set, but it does not change the perceptual analysis and the
> perceptual appearance of the input"

However Schyns and Rodet have provided experimental evidence that
learning can change perceptual analysis which is an obvious blow to
the fixed feature theories. Their study (mentioned earlier in the
paper)involved subjects in categorisation conditions who should have
seen 'xy' exemplars as 'feature conjunctions', but saw them as
'mutually exclusive'.
They state that:

> "Feature creation as opposed to feature weighting is preferable if
> category learning induces mutually exclusive perceptual analysis of
> an objectively identical object property, when the experimental
> design would predict identical perceptual analysis if the
> subjects used fixed features."

To consider the problem of feature reduction is more simple. Priming
tasks can be used to indicate that the suggested combination of
features that make up the novel feature are not involved in the
perceptual encoding of the new feature.

The feature weighting theory is said to be particularly hard to
disprove because:

> "it is used a posteriori to interpret patterns of data. Feature
> weighting is a form of curve fitting with free parameters (the
> weights assigned to features). Feature weighting therefore covers not
> one, but a potential infinity of models of categorization, and can
> potentially accomodate any pattern of experimental data if its
> features are not pre-specified."

Surely the fixed feature theorists must be forced to pre-specify and
hypothesise their results prior to experimentation? impossible?

A final problem with the fixed feature theories is that concept
learning programs fail to explain how the features are generated.

In section 2.7 schyn considers the advantages of new feature
learning. Firstly it places constraints on what can count as a
feature.

> "Unlike purely formal models of similarity and categorization,
> our approach places constraints on what can count as features:
> Features will be incorporated into a system to the extent that they
> distinguish between object categories; features should not be limited
> to the finite set of a priori features designed by a particular
> researcher for a particular domain."

But can in not be said that all psychology experiments consider a
particular domain, and it is from detailed analysis of specific areas
that information can be incorporated to form a whole?
I agree that these features must be pre-specified though.

Schyns suggests that the fixed feature approach has numerous useless
features:

> "To the extent that each new feature accommodates at least the
> categorization for which the feature was created, the repertoire
> should be free of useless features. A fixed feature approach is
> necessarily much less parsimonious: Many spurious features must exist
> in the feature repertoire to foresee new categorizations. Moreover,
> most features of the fixed would never be used--they would keep
> waiting for their "Godot category." Fixed features necessarily have
> suboptimal fit outside the scope of the stimuli they were designed to
> represent. A flexible set of features tuned to specific
> categorizations reduces the necessity of complex categorization
> rules."

I dont think that Schyns can really state that the fixed theory
features are useless. Although it is unlikely, all the features the
fixed theory has suggested may be useful sometime in the future as
new categorisations are discovered. It cannot be proved that the
fixed feature theorists are not actually showing great foresight.
A flexible set of features is of course the most attractive option.
However a lack of complexity should not always make a theory more
plausible.

Schyns also considers what is the most natural process for a person:

> "Concept learning theories have frequently stressed the importance of
> learning categories by discovering complex rules that integrate
> several distinct stimulus features Concept learning certainly does
> sometimes require such integration. However, these problems have
> effortful, strategic solutions. They are rather unnatural; people are
> not particularly adept at explicitly combining psychologically
> separated sources of information."

Surely if this was the way that categorisation was done then people
would be adept at it through constant practice?
The flexible approach to learning categories does however sound more
logical and it can explain certain developmental phenomena such as
the narrowing of lexical categories throughout childhood:

> "Our alternative is that new categorizations can be based on
> relatively few, specially tailored features.In the flexible feature
> approach, categorizations can induce a decomposition of features into
> subfeatures. Consider the contrast between glasses and cans. Early in
> conceptual development, these objects may be indistinguishable
> because their memory representations corresponds to a single,
> undifferentiated feature. Now assume that the organism needs to
> distinguish between these objects. This can be achieved by
> decomposing the undifferentiated feature into two specific features
> tailored to glasses and cans."

Stent, Hannah
hms195@soton.ac.uk



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:20 GMT