Schyns Comms 19-21 palmeri perruchet phillips

From: Chan Dorothy (dwyc195@psy.soton.ac.uk)
Date: Mon Mar 02 1998 - 11:55:50 GMT


> Implicit learning shapes new conscious percepts and representations.
> Feature creation as a by-product of attentional processing

> Pierre Perruchet and Annie Vinter

> In traditional category learning studies, new categories emerge from
> new combinations of elementary features, which compose themselves a
> fixed repertoire. SG&T show cogently that low level features
> themselves can change with experience, thus altering the immediate
> appearance of objects.

Throughout our lifetime, we are developing our experiences. When we
encounter new events or people, our perceptions of the situation or
context that we are in change. The features and categories which
we have already acquired as we go along help us to make sense of the
world around us.

> However, we believe that one aspect of SG&T's proposal may ultimately
> limit its implications. SG&T claim repeatedly that feature creation
> depends on categorization requirements, and that people create
> features to subserve the representation and categorization of objects.
> We have no special concern with this proposal insofar as it intends to
> describe the ultimate function of features in adaptive behavior.
> However,..... because feature creation, according to SG&T, implies
> concurrent category knowledge, extending the scope of learning to
> features paradoxically prevents genuine category discovery. We
> subscribe to the view that new features can be created, but we intend
> to show that the process is functionally independent of categorization
> requirements. Our proposal is that feature creation allows formation
> of new categories instead of requesting information about categories.

If the process of creating new features are independent of
categorization requirements, then why is there a difference between
new and old features? It is from categorising the events or objects
or whatever that are present in your situation that you can
distinguish what you don't yet know (the new features?) from what you
already know (the known categories) because there is a "need" to
make sense of the external world and it can only be done by this
process. Moreover, on what basis can we achieve the categorisation?
 We have to know the content or the information of the "things" that
we are categorising in order to do so. Perhaps what we claim to be
new features are the stuff that are different from the fixed "old"
categories but which of them might become relevant for us at some of
our life. Therefore, it is necessary to "label" them.

> We have proposed a model that relies on simple and ubiquitous
> attentional and memory processes (e.g. Perruchet and Vinter, 1997;
> Perruchet, Vinter, & Gallego, 1997). In this model, the key point is
> that a given tends to be repeated only when it is structurally
> relevant, as a mandatory consequence of the rule-governed structure of
> the material. This entails that irrelevant units, the occurrence of
> which is infrequent, will be forgotten, whereas the initial selection
> of meaningful units will be progressively reinforced by their
> repetition. With repeated exposure, subjective units become strong
> enough to shape perceptual processes, and alter the immediate
> appearance of objects.

Thus, does that mean, it is dependent on the type of context that we
are in, which determine the process of categorisation? So, the
process for each one of us is different then? But how can we
communicate with each other? Why do we know what is a square or
round object. Are they just terms for conventional use?

> Formal Models and Feature Creation

> By Thomas J. Palmeri Vanderbilt University

> Feature creation is but one (albeit largely ignored) aspect of
> categorization. Even if new features are created (by some presently
> unspecified mechanism), issues of how these are used features to make
> a categorization decision must still be addressed: How are the
> features of a presented object compared with information stored in
> memory? What information about a category is stored in memory? What
> kind of categorization decision process is used?

Indeed there are numerous features for all objects and if the process
of categorisationis to provide easier ways to categorise the world,
but paradoxically, we have to take up the "load" to remember all
those features! Although our memory capacity is vast but how can we
successfully draw out the "right feature" at the right feature
decision moment? Can the "prototype theory" be of any use? In
that, the prototypic member sets out the path for allocating
weightings for certain features? But how does this comparison
process work?

> Context-dependent feature discovery is evidence that the coordination
> of function is a basic cognitive capacity

> W. A. Phillips

> Object concepts are distinguished by what distinguishes objects not by
> non-fixedness etc. Though this may sound tautological, it is not,
> because what distinguishes objects is specified in relation to the
> environmental input by the classes of objects met, and in relation to
> the internal cognitive operations of object perception by the sets of
> stimuli to which discriminative object recognition responses
> generalize.

Not only our senses are in some way linked e.g. what your ears hear
aids to locate the focus of your eyes but what you acquire from one
context can increase your understanding of the next context that you
are facing with. Although, contexts do not always linked, but they
do help to consolidate your knowledge about things that are around
you. So enriching your prior experience and therefore would affect
your perception of the world. I surely don't think you can live on
your "old" experiencealone but you are currently facing and perhaps
your expectations for your future.

As can be shown:
> Information to guide feature discovery need not be restricted to
> top-down sources I see no need to restrict the information that guides
> feature discovery to that provided by top-down sources, which in any
> case are not well specified. What matters is not that the guiding
> information come from some high level strategic controller but that it
> arise from an input that is distinct from and which (at the level
> concerned) is kept distinct from the input whose analysis it guides.

> Relations between the features must also be discovered and used S, G &
> T imply this but provide no explicit account of how feature discovery
> can be combined with discovery of the relations between those
> features.

How at the end of the day do we say that "this is a feature?"
Feature is the term which we develop to define objects or even
events or to specify that "something" is a characteristic of
"another thing" or is it beng created for organisation purpose?!
How can it be stored in our brain?



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