Chapter 2. Literature Review
2.3 Specific hypothesis and lexical acquisition
2.3.1 Previous studies on specificity and processing
Some psycholinguistic studies have examined that the effect of semantic specificity
(i.e., semantic complexity or semantic weight) of verbs on some processing aspects such
as memory for sentences (Gentner, 1981) and lexical retrieval (Breedin et al., 1998;
Gordon & Dell, 2003). Two models have been involved: the Componential Model and
the Connectionist Model. The Componential Model (Kintsch, 1974) predicts that
semantically specific verbs require more processing resources since relatively more
features need to be processed than general ones. That is, this model suggests that
semantic features have their cost in processing time. On the other hand, the
Connectionist Model, emphasizing the role of the structure of semantic representation,
predicts that specific verbs would be processed faster since additional features imply
more connections among components and a more complex network between
components, which would facilitate processing or memory.
Different levels of specificity
Unlike other studies which only acknowledge the difference between general verbs
and specific verbs or between light verbs and heavy verbs (Barde et al., 2006; Gentner,
1981; Gordon & Dell, 2003), Breedin et al. (1998)4 compared performance for
semantic complexity of verbs at two levels: light verbs vs. heavy verbs (go vs. walk)
and general verbs vs. specific verbs (e.g., clean vs. wipe)5. This distinction allows us to
be aware that there are actually at least three levels of specificity and thus a word can be
general or specific at different levels of comparison. For example, in Breedin et al.’s
(1998) study, mix was used as a stimulus at both comparisons, it was a “heavy verb”
when compared to make as a light verb whereas it was a “general verb” when compared
with stir as a specific verb. As for another example carry, which was used at the level of
comparison between general and specific verbs for two times, was categorized as
“general” when compared to deliver but also defined as “specific” when compared to
hold. Their results showed that aphasic patients had more difficulty in retrieving
“semantically simple verbs” than “semantically complex verbs”.
Specificity involving syntactic properties
On the other hand, some studies provide a definition of specificity or semantic
complexity or semantic richness involving morphological or syntactic properties, or link
4 “Light verbs” in Breedin et al.’s (1998) study are similar to semantic primitives like make, come, bring and so on.
5 The variable “specificity” manipulated in this study is more similar to the later level in Breedin et al.’s (1998) study (general verbs vs. specific verbs) in that the so-called general words in this study are not so general to be primitives in a language whereas light verbs in the former level refer to primitives.
semantic specificity to syntactic properties6. For instance, Gordon & Dell (2003)
proposed a connectionist verb-production model and argued for the implication of
“division of the labor” to the dissociation between semantically heavy and semantically
light verbs as well as that between nouns and verbs. This model suggested that syntactic
and semantic inputs share “responsibility (or ‘division of labor’) for lexical activation
according to their predictive power” (2003, p. 1). They argued that semantically light
verbs rely more on the syntactic cues and less on the semantic cues when compared to
semantically heavy verbs, just as verbs have “more complex grammatical
representations” while nouns have “richer semantic representations” (2003, p. 31). In
other words, they suggested that this “division of the labor” between semantics and
syntax can provide an account not only for the dissociation between verbs and nouns in
aphasic patients but also for the dissociation between semantically heavy verbs and
semantically light verbs. Their results from sentence production and single-word
naming simulation revealed that anomic patients had more difficulty in retrieving heavy
verbs whereas aphasics with agrammatism were more impaired in retrieving light verbs.
Similarly, Barde et al. (2006) reported that agrammatic aphasics had higher difficulty in
6 Mobayyen & de Almeida (2005) used similar term “semantic complexity” in their study, yet their focus fell in a different area. They seemed to use argument structure to define the semantic complexity in their sentence recall tests. Causatives (e.g., grow/ fertilize) were used as stimuli for semantically complex verbs, while perception verbs were used for semantically simplex verbs (e.g., smell/ re-smell). They reported that participants performed better in recalling when sentences included semantically complex verbs than when sentences included semantically simplex verbs.
light verbs in a story completion test and argued that it is the syntactic deficit that
contributes to the difficulty of light verbs.
Assessing difficulty or ease resulting from specificity
Additionally, previous studies have provided various explanations for the difficulty
of light verbs or the ease of heavy verbs on the basis of results from different tasks. As
mentioned above, Gordon & Dell (2003) suggested that the greater difficulty of light
verbs compared to heavy verbs is attributable to the greater dependency on syntactic
cues. On the other hand, Gentner (1981) argued for a connectionist account, which
suggested that more semantic components of heavy verbs provide more connections in
the network of verb meaning, and thus provide stronger “memory traces” whereas light
verbs have less connections. In addition, Breedin et al. (1998) mentioned another
possibility that the difficulty of retrieving light or general verbs is due to the relative
wideness of contexts that light verbs can apply to, which leads to instable
representations of verb meaning: Light verbs can generate a variety of meanings and
should be limited by the context where it occurs.
Both evidence for the relationship between specificity and processing and studies
on how aphasic patients perform with specific verbs or general verbs provide us with
some insights about verb learning. Understanding the memory load caused by lexical
specificity or the stableness of mental representation would allow us to re-examine the
argumentations concerning verb acquisition. However, evidence provided by studies on
adults for either approach seems not valid evidence for language acquisition. There
might be some discrepancy between results from adult processing and child language
acquisition because people may have different responses or develop different strategies
when faced with familiar and unfamiliar materials (Gentner, 1981). Additionally,
children would have an different understanding or assumption about word meaning
from adults’ since it requires time to develop full understanding about word meaning
after children produce certain words (Clark, 1993).
Taken as a whole, it is still in debate whether specific verbs are easier to process or
learn since there are different points of view to explain the phenomena of verb
specificity. Specificity can be determined by the number of semantic features or the
amount of information that is encoded in a word. That is, the more semantic features
one word has, the higher specificity it has. Therefore, specific verbs are also called as
“heavy” (Gentner, 1981) or “semantically rich” (Gordon & Dell, 2003, p. 1) or
“semantically complex” verbs (Breedin et al., 1998, p. 2). If semantic features are
separately processed suggested by “Componential Model” or “Complexity Hypothesis”
or are gradually learned as argued by Clark (1973), it would be predicted that a word
with more features would require more resources for processing and learning. On the
other hand, if specificity is viewed in terms of the contexts to which a word can be
applied, the direction would contrast to the earlier one: fewer contexts that a word can
apply to imply higher specificity. In other words, a highly specific verb would be
restricted to a limited number of contexts by its internal meaning. Thus, in the process
of retrieval of a semantic complex word, one did not have to select the possible meaning
since this word has a more “uniform representation” (Breedin et al., 1998, p. 21). In
other words, if a verb is more specific, its perceptual characteristics would be more
stable. In contrast, connectionists view specificity in terms of a network of meanings.
Higher specificity implies not only more semantic features but also more connections
and thus facilitates the processing of sentences.
These studies mentioned above provide us with various accounts for ease or
difficulty of specificity through examining the general mechanism involved in
processing specific verbs and general verbs. The following section will provide a review
on studies concerning typological differences in specificity and the role of specificity in
lexical development.