MIT Encyclopedia of the Cognitive Sciences - Cryptome

180 Conceptual Change
concept, perhaps because they occur more frequently in a
set of experiences. A young child might weight shape
somewhat more heavily in her concept of a bath towel than
texture, while an older child might do the opposite. In its
simplest form, such a developmental change may not connect
to any other relations or beliefs, such as why texture is
now more important. An older child might disagree with a
younger one on identifying some marginal feature of bath
towels, and while we might thereby attribute this difference
to conceptual change, we might not see the concepts
as really being very different.
Changes in feature weightings and dimensional value
shifts are ubiquitous in cognitive science studies of concepts.
They are seen at all ages ranging from studies of
infant categorization to adult novice-to-expert shifts (see
INFANT COGNITION and EXPERTISE). Any time that some bit
of information is incrementally added to a knowledge base
and results in a different feature weighting, such a change
occurs. When such changes have no other obvious consequences
for how knowledge in a domain is represented, they
constitute the most minimal sense of conceptual change,
and for many who contrast “learning” with true conceptual
change, not a real case at all (Carey 1991).
2. Shifting use of different sorts of properties and relations.
Conceptual change could occur because of changes in
the kinds of feature used in representations. Infants and
young children have been said to use perceptual and not conceptual
features to represent classes of things, or perceptual
and not functional ones, or concrete and not abstract ones
(e.g., Werner and Kaplan 1963). More recently, young children
are said to use one-place predicates and not higherorder
relational ones (Gentner and Toupin 1988), or to rely
heavily on shape-based features early on in some contexts
(Smith, Jones, and Landau 1996). Similar arguments have
been made about novice to expert shifts in adults (Chi, Feltovich,
and Glaser 1981) and even about the evolution of
concepts from those in “primitive” cultures to those in more
“advanced” ones (cf. Horton 1967; see also LURIA).
Several forms of conceptual change can be captured by
shifts in what feature types are used in concepts. Despite a
wide range of proposals in this area, however, it is striking
how many have always been controversial, especially in
claims of cross-cultural differences (Cole and Means 1981).
There is no consensus on changes in the sorts of properties,
relations, or both available at different points in development,
expertise, or historical change, nor on the very real
possibility of no true changes in the availability of property
types.
Part of the problem is the need for better theories of
property types. It is difficult to make claims about perceptual
to conceptual shifts, or perceptual to functional shifts, if
the contrast between perceptual and conceptual features is
murky. Claims of changes in feature types therefore need to
attend closely to philosophical analyses of properties and
relations, which in turn need to attend more to the empirical
facts.
3. Changes in computations performed on features.
Conceptual change can also arise from new kinds of computations
performed on a constant set of features, such as from
tabulations of features based on frequency and correlational
information, to more rulelike organizations of the same features
(Sloman 1996). In other cases, there have been claims
of changes from prelogical to quasi-logical computations
over features (Inhelder and Piaget 1958), or changes from
integral to separable operations on features and dimensions
(Kemler and Smith 1978); or changes from feature frequency
tabulations to feature correlation tabulations.
Although most models tend to propose changes in computations
that apply across all areas of cognition, such transitions
can also occur in circumscribed domains of thought
even as there are no global changes in computational ability
(Chi 1992). Second, these models do not require that concepts
be interrelated in a larger structure. They are neutral in
that respect and thus allow each concept to change on its
own. In practice, this is highly implausible and may in the
end render such models inadequate because they fail to
make stronger claims about links among concepts.
There are also cases where there is no absolute change in
feature or computational types, but rather a strong change in
the ratio of types. Thus a younger child may have true conceptual
or functional features but may have ten times as
many perceptual ones in her concepts, whereas an older
child may have the opposite ratio. Similarly, a younger child
may perform logical computations on feature sets, but may
do so much more rarely and may more frequently resort to
simpler probabilistic tabulations. This variant is important
because it offers a very different characterization of the
younger child in terms of basic competencies. Younger children
are not incapable of representing certain feature types
or engaging in certain computations; rather, they do so
much less often, perhaps as a function of being much more
inexperienced in so many domains (Keil 1989).
4. Theoretic changes, where theories spawn others and
thereby create new sets of concepts. The most dramatic
kinds of conceptual change, and those occupying most discussions
in cognitive science at large, are those that view
concepts as embedded in larger explanatory structures, usually
known as “theories,” and whose changes honor DOMAIN
SPECIFICITY. Sweeping structural changes are said to occur
among whole sets of related concepts in a domain. For
example, a change in one concept in biology will naturally
lead to simultaneous changes in other biological concepts
because they as a cluster tend to complement each other
symbiotically. Within this type, three kinds of change are
normally described: (a) birth of new theories and concepts
through the death of older ones (Gopnik and Wellman
1994); (b) gradual evolution of new theories and concepts
out of old ones in a manner that eventually leaves no traces
of the earlier ones (Wiser and Carey 1983); and (c) birth of
new theories and attendant concepts in a manner that leaves
the old ones intact (Carey 1985).
One key issue in choosing among these kinds of theoretic
change is the extent to which concepts of one type are
incommensurable or contradictory with those of another
type (Kuhn 1970, 1982). Kuhn suggested that conceptual
changes in domains could lead to “paradigm shifts” in
which concepts in a prior system of beliefs might not even
be understandable in terms of the new set of beliefs, just as
concepts in that newer system might not be understandable
in terms of the older one. The ideas of paradigm shifts and