There have been various
approaches to analysing the content and structure of concepts.
One early content approach is the defining feature approach
(originating from Aristotle) which states that concepts are
represented by lists of properties or features (perceptual
properties) which must apply to all category members and only
to category members i.e. properties must be necessary and
sufficient.
Defining features must also apply equally to all category
members; it is thought that all concepts are represented in
this way and that everyone represents concepts in the same
way. This approach also proposes that relationships between
concepts are in the form of a hierarchy, consisting of superordinate,
intermediate and subordinate category levels. The superordinate
level is the general, common, encapsulating parent category,
the intermediate level is more specific and the subordinate
level is the most specific.
As the superordinate category is the parent category, the
lower level categories are exemplars (members) of the parent
category. The intermediate and subordinate categories inherit
the properties of the superordinate category and the subordinate
categories inherit the properties of the intermediate categories.
Collins and Quillian (1969) proposed a conceptual hierarchy
with defining features, consisting of the aforementioned category
levels joined by nodes at which lists of defining features
are attached. The parent superordinate category has a list
of properties shared by the lower level categories. The lower
level categories also list additional properties (the properties
from the level above are not required to be listed).
In order to verify statements (saying whether they are true
or false) the conceptual hierarchy needs to be searched to
find the category an instance belongs to. Collins and Quillian
argued that a search through two levels of a hierarchy would
take longer than a one level search (the more levels need
to be searched, the longer the search time). They used a sentence
verification task with animal sentences which subjects had
to gauge true or false.
The results showed that sentences needing two level searches
did take longer to verify than one level searches (no level
searches were quickest). There were exceptions to this rule,
however: this was perhaps because the researchers failed to
consider that people might have greater knowledge or be more
familiar with certain words than others. In addition, some
sentences that required a search through the same amount of
levels still produced differences in verification times. Collins
and Quillian suggested that the quicker times for same level
searches occurred when the item to be verified was a more
typical member of a category.
|
