Chunking and Perceptual Unitization.
Fisher's model of visual search illustrates that a target shape is
harder to identify when it is placed among other shapes with which it
shares similar features. However training can significantly increase
the speed taken to identify the target shape. Czerwinski et al explain
this using the notion of perceptual unitization. This supports Schyns
et al's feature creation theory as it suggests similar features will be
chunked together forming a catagory for items that share these common
features which can then be distinguished from distractor items.
Peutzow and Goldstone's experiment illustrates how individuals will
break down an object into it's component features or parts with which
they have had previous experience. Unitization therefore requires that
stimuli is discretized before being unitized, and the way in which this
process is carried out depends on our past experience with the stimulus
features. However in some situations we may recognise something as a
whole, and therefore we need not discretize everything we see.
New names may be given to catagories by identifying common features of
members of the catagory. Or, as Wisniewski and Medin illustrated,
individuals may alter their verbal description of an object in order
for it to fit a catagory label, based on links between abstract
background knowledge and concrete object information. Schyns et al
however stress that for new features to be created and learnt raw
stimulus properties are needed. Raw stimulus properties are needed
that have not already been interpreted, so that limitations are not
placed on the new features that can be created. Inductive operators
can produce an infinite number of features but they will be restrained
by previous object interpretation made by symbolic features. In adding
new features to a system it does not mean however that the appearance
of the object will be altered. Although, it may mean that different
inferences about the object's properties may be made, which in turn may
be affected by an individual's experience with such features.
The role of theories in object parsing.
There are an infinite number of ways in which objects can be
represented with features, and this therefore makes it difficult to
explain hoe children gain a featural object description from a limited
data set. Some argue there is an innate conceptual core that could
bias the features infants will notice in objects. The fact that
newborn infants will be more responsive to a picture consisting of
organised features of the human face than a picture of random lines,
supports this idea by suggesting that the infant has some innate
tendancy to notice the features of the human face.
The early role of perception in object parsing.
Experimental research with children has revealed that catagory
inductions are guided by a bias for the shape of objects. Children
will therefore ignore other differences between the objects'
properties. Children's parsings are inconsistent with those of adults,
and therefore research is needed to explain how children learn to later
decompose a catagory of objects into their relevant object features.
The features that the child attends to may be structurally different to
those that an adult attends to when looking at the same objects. This
then may hold children back from learning new catagories by attending
to features of an object that are less salient.
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:20 GMT