candidate is sufficiently similar either to the prototype or to a set of encoded examples.
Similarity in that case is based on matches and mismatches of independent features
(Leddo, Abelson, Gross, 1984; Medin & Ortony, 1989; Medin 1989).
With reference to the first type of probabilistic views - prototypes theory- although the
general idea that concepts are organized around prototypes remains popular, at a more
specific empirical level, prototype theories have not fared very well. Prototype theories
imply constraints that are not observed in human categorization (Medin &
Schwanenflugel, 1981; Kemler-Nelson, 1984). They also predict insensitivity to
information that people readily use (Estes, 1986), and fail to reflect the context sensitivity
that is evident in human categorization (Roth & Shoben, 1983; Medin & Shoben, 1988).
Thus, there is a shift again to the role of knowledge structures and theories in
categorization. The Semantic Feature Hypothesis which follows next shows a change at
the level of analysis.
2.2.3.2 The formation of concepts
2.2.3.2.1. The Semantic Feature Hypothesis
Clark (1973) proposed the Semantic Feature Hypothesis. She argued that the meanings of
the words are analysed into smaller elements-semantic features. For example, the meaning
of the word father could be broken down to male, adult etc. According to that theory, the
child gradually learns the meaning of a word by learning the set of features that
characterizes a word's meaning. To identify the category of a word, Clark hypothesized
that children simply attach a single semantic feature to a word they hear, and the semantic
feature is usually perceptual in nature. An example she uses, is that when children hear the
word dog they choose a semantic feature like four legged as being the definition of a word
dog. Therefore, the child would treat the word dog as meaning four legged, and be
unaware that the word has other additional meanings.
The theory was able to explain the overextensions (when children apply a word to
inappropriate referents) and underextensions (when children restrict a word’s use to a
subset of the adult’s category in children’s speech). Such mistakes are predicted by the
Semantic Feature Theory. Thus, in the example with the word dog this would be expected
34