Chapter 2. Literature Review
2.3 Specific hypothesis and lexical acquisition
2.3.2 Typological pattern in specificity and lexical development
Some studies have discussed the notion of specificity in a cross-linguistic context.
One of these was conducted by Tardif (2006a, 2006b), who suggested that languages
differ in the tendency of specificity of nouns and verbs and argued that Chinese verbs
have typologically higher specificity whereas nouns are less specific. She noticed that
Chinese speakers tend to use more specific and distinct verbs to indicate distinct actions
while English speakers tend to use general-purpose verbs, occurring with prepositions
or nouns that are used to specify referents. For example, in English, carry could refer to
various ways of transporting objects with one’s body, such as carry a backpack, carry a
baby, and carry a serving dish. On the other hand, in Chinese, different verbs are used
for different ways in which objects are carried, e.g. , bēi ‘carry on the back’, pěng ‘carry
upon hands’, bào ‘carry with arms’, duān ‘carry as if serving food’, līn ‘carry with one
hand’, and ná ‘grasp/take’. Additionally, she also pointed out that specific verbs are
available in English though English speakers tend to use general words. However, she
did not further explain how frequency interacts with language-specific properties.
Additionally, similar evidence was also shown in some Mayan languages such as
Tzeltal and Tzotzil (Brown, 2001; Haviland, 1992). Through examining verbs in Tzeltal,
Brown (2001) proposed the “verb specificity hypothesis”, suggesting that the pattern of
specificity in different word classes varies across languages. Specifically, different word
classes in a particular language fall in different positions on the continuum of specificity.
In the end of higher specificity English has common nouns whereas Tzeltal has
transitive and positional verb roots. Tzeltal, for instance, has a variety of eating verbs,
which distinguish between the kinds of food that an agent eats. In contrast to English,
common nouns in Tzeltal are more general than transitive and positional verb roots. In
addition, Haviland (1992) reported that Tzotzil verbs often encode what body parts
engage in an action. Like Mandarin and Tzeltal, Tzotzil has different verb roots for
carrying something on the back (kuch) and carrying something in arms (pet).
Though making similar observation of typological patterns of verb specificity,
Tardif and Brown have made different interpretations and predictions on how these
typological properties affect the mechanism of word learning (Brown, 2001, 2007;
Tardif, 2006a). In addition, Bowerman (2005) viewed the typological differences in
specificity in terms of different patterns of boundaries between categories. The
following section will provide a brief review and discussion on their hypotheses and
approaches to the relationship between specificity and lexical development.
The implication of the specificity hypothesis in the acquisition of lexicon
Though being based on similar observation that specificity pattern of syntactic
category is different across languages, Brown (2007) and Tardif’s (2006a, 2006b)
arguments toward the learning mechanism are different from each other. Generally
speaking, Tardif attempted to explain the ease of learning Mandarin verbs, while Brown
put more emphasis on the difference in mapping patterns.
Approach 1: Specificity as a predictor of the ease of learning
To put it more specifically, Tardif not only pointed out that Mandarin verbs are
specific and but further linked this to the fact that the proportion of Mandarin verbs in
early vocabulary is much higher than that of English verbs. In other words, she seemed
to argue that specific verbs are easier to learn because of perceptual availableness.
Though not explicitly expressed, the contrast between English nouns and Mandarin
nouns was also mentioned to support her proposal. She pointed out English nouns are
specific, while Mandarin often has a root word for a group of nouns. For instance,
English has two distinct words, rooster and hum whereas the equivalents in Mandarin,
mǔjī ‘rooster’ and gōngjī ‘hum’, share a word root jī ‘chicken’. However, the role of
specificity of English nouns or the generality of Mandarin nouns remains unclear. An
alternative account is that the morphology of Mandarin nouns might provide a cue that
allows children to observe the similarity between objects that share the same root. In
addition, little is known about how specificity influences noun learning since basic
levels vary across languages. To sum up, though making attempts to employ the notion
of specificity to account for the ease of word learning across word classes, Tardif (2006a,
2006b) did not provide explanations for the role of specificity in noun learning.
Approach 2: Different degree of specificity implying different extending strategy
On the other hand, although Brown (2001, 2007) contrasted the semantic
specificity of early transitive verbs in Tzeltal children with the generality of early verbs
in English children, she did not employ the specificity of early Tzeltal verbs to explain
the ease of verb learning as Tardif did to explain the ease of Mandarin verbs. Though
recognizing that the light verb hypothesis (Casenhiser & Goldberg, 2005; Clark, 1973;
Goldberg, 2006)7, which is based on the observation from English verbs, fails to explain
the process of Tzeltal learning, Brown (2007) did not propose an opposite hypothesis of
the light verb hypothesis. Instead, to solve the paradox that English early verbs are
general and early Tzeltal verbs or Mandarin verbs are specific, she argued it is not that a
specific verb is easier nor that generality facilitates word learning; it may be that
typological differences in verb specificity contribute to different word extending or
learning strategies. Specifically, children who are exposed to a language with highly
specific verbs -- such as Tzeltal-- would avoid generation or extension after acquiring a
verb until positive evidence is available. In contrast, children acquiring a language with
many general-purpose verbs -- like English -- would suppose that verbs are “tricky”
ones then they tend to use verbs that are general enough and let nouns narrow down the
referents of events (2007, p. 181). This argument seems similar to Choi and
Bowerman’s (Bowerman, 1996; Choi & Bowerman, 1991), who pointed out that
children as young as two-year-old are sensitive to language-specific semantic
distinctions.
Taken all together, Tardif (2006a, 2006b) made an opposite argument of the light
7 Clark did not predict what kind of word would be acquired earlier but predicted some general features would be mastered first. Based on the assumption that general features are acquired earlier and other features are mastered later to narrow down the meaning of a word, it would be predicted that children acquiring Mandarin would have an incomplete understanding meaning of specific verbs.
verb hypothesis while Brown (2007) made attempts to conciliate prediction by light
verb hypothesis and counterexamples found in languages like Tzeltal and Chinese. The
light verb hypothesis predicts that light verbs are easier to learn because of fewer
semantic features to be mastered and because of higher frequency. Tardif (2006a, p. 491)
mentioned that exploring verb semantics in English and Mandarin would be
“informative as to why Mandarin appears to break the rule.” In her reasoning, Tardif
made attempts to illustrate that it is not that nouns are easier than verbs nor that verbs
are more difficult; rather, it is that specificity makes the difference. In the case of
Mandarin, nouns used in daily life are more general whereas verbs are specific, and thus
early vocabulary in Mandarin-speaking children consists of more verbs and less nouns
than that in English-speaking ones. In contrast, to explain Tzeltal children’s better
performance in learning verbs, Brown (2007) provided a different account that did not
violate the assumption that nouns are easier to learn. Instead, she argued that Tzeltal
verbs incorporate information of nouns and are more like nouns -- or more “nouny” in
her term -- and thus more privileged than verbs in other languages. She suggested that
semantic specificity of Tzeltal verbs “is indeed a possibly crucial ingredient in Tzeltal
children’s early transitive verb learning” because it provides “concreteness”, which
makes verbs more “nouny”, and “redundancy”, which indicates that information was
carried both in the verb and Object NP (Brown, 2007, p.172). Arguing that Tzeltal verbs
are more nouny and thus easier to learn is still based on the assumption that nouns are
easier; therefore Tzeltal children’s early production of specific verbs cannot be regarded
as counterexamples that “break the rule” that nouns is easier to learn than verbs as
Tardif argued. Brown, instead, suggested that different pattern of specificity would
contribute to different preferences for mapping strategies.
Additionally, Brown’s argument that children’s assumption about word meaning
reflects the typological pattern of specificity in the input language is similar to
Bowerman and colleagues’ approaches that suggested that toddlers are sensitive to the
language-specific pattern of encoding patterns for spatial words or to the semantic
partitioning of spatial words (Bowerman, 1996; Choi & Bowerman, 1991; Choi,
McDonough, Bowerman, & Mandler, 1999). In the following paragraphs, it will be
discussed how specificity can be understood in terms of semantic partitioning or
categorization as well as lexicalization.
Approach 3: Specificity in terms of conflation pattern vs. semantic category
As motioned above, specificity has been defined as how much information is
encoded in a word as Brown found that Tzeltal transitive verbs often encode object,
which facilitates the omission of an object argument. Though not using the term verb
specificity, Choi and Bowerman (1991) have also made similar observation on
cross-linguistic differences among how a word encodes information and how it affects
word learning. They argued that the pattern of information encoded in motion words
varies across languages, and analyzed the pattern of lexicalization of motion events in
English and Korean in terms of Talmy’s framework (1985) in which a motion event is
analyzed into semantic components, such as MOTION, FIGURE, GROUND, and PATH.
They found that English and Korean have different conflation patterns for motion events.
In English, motion verbs often encode MANNER or CAUSE while PATH is usually
conflated in prepositions. In addition, English does not mark distinction between caused
motions and spontaneous motions in its verb conflations. On the other hand, Korean
motion verbs encodePATH in transitive sentences expressing caused events, while in
intransitives motion verbs expressing spontaneous events Korean “encodes PATH, and
optionally MANNER and CAUSE with separate constituents” (Choi & Bowerman, 1991, p.
88). From observations in how children ranging from 14 to 24 months talk about motion
events, they found that English-speaking children and Korea-speaking children as
young as 17-month to 20-month are sensitive to language-specific patterns in lexical
conflation. Children acquiring English use path particles like up, down, in, or on for
both spontaneous and caused motion events whereas children acquiring Korean “keep
words for spontaneous and caused motion strictly separately” (p. 83).
This typological difference in encoding patterns can also be understood as a
difference in patterns of semantic partitioning or categorization. Although languages are
supposed to mark perceptually salient distinctions and constrained by human’s
perceptual capability to some degree, semantic boundaries between categories are found
to vary across languages (Bowerman & Choi, 2001; Malt, Sloman, Gennari, Shi, &
Wang, 1999). In their later studies, Bowerman and colleagues (Bowerman, 1996; Choi
et al., 1999) put more emphasis on the notion of “semantic categories” than “the pattern
of lexicalization” when discussing language-specific patterns of spatial verbs or
particles. Choi et al. (1999) reported that Korean spatial verbs make distinction between
tight-fit relation versus loose-fit relation whereas English does not mark this distinction
but distinguishes between the relation of support (put on) and containment (put in). The
Korean word, kkita ‘fit tightly/ interlock’ only can be used for tight-fit relations
regardless of support or containment relation while the distinction regarding tight-fit
relation is “indifferent” to English. In their experimental studies, they found children as
young as 18-month to 23-month are sensitive to language-specific semantic boundaries
for the spatial words. In other words, young children can observe the language-specific
regularity of lexical distribution for categorization.
More recently, Bowerman (2005) explored cross-linguistic differences in verb
meaning in terms of object categories from evidence in a variety of verbs such as
dressing verbs, carrying verbs and consumption verbs, and also examined how the
differences influence lexical development through findings from opening, cutting, and
breaking verbs. While some researchers (e.g., Tardif 2006a, 2006b; Brown, 2001; and
Ma et al., 2009) suggested that carrying verbs in different languages, such as Mandarin
or Tzeltal, differ in specificity, Bowerman (2005, p. 209) tried to explain different ways
how object classes (“covert object categories” in her term) “are woven into” verb
meaning (p. 209). She argued that while some object categories have names, some
object categories seem not consciously noticed by speakers but displayed in how verbs
or other word classes are used. Based on evidence from dressing verbs, carrying verbs,
and consumption verbs, cross-linguistic differences were shown in how “[verbs] impose
restrictions on the kinds of object involved in the event” (2005, p. 214). Dressing verbs
and carrying verbs display covet categories for body parts, while consumption verbs
show categories for things people consume. Take dressing verbs for example. English
speakers use the verb put on regardless of body parts involved or types of clothing items.
However, Tswana, a Bantu language of Botswana, has distinct verbs for putting clothing
on extremities (head, hands, arms, feet: gòrwálà) and the central region of the body
(gòàpàrà) (Schaefer, 1985, as reported in Bowerman 2005). On the other hand, Korean
has specialized verbs for putting clothing on the head (ssuta), feet (sinta), and wrist or
waist (chata). As for Japanese dressing verbs, distinctions are made between the head
(kabura), the upper torso (kiru), and the lower torso down through the feet (haku). In
addition, Yoruba, a Niger-Congo language of Nigeria, only makes distinction between
putting clothing on the head (dē) and other religions of the body (wō) (Schaefer, 1985,
as reported in Bowerman 2005). In her argument, Bowerman linked this discrepancy in
dressing verbs, carrying verbs or verbs of consumption among different languages to
different patterns of categorization of objects. Specifically, a verb with higher specificity,
in this framework, is interpreted as “a small (differentiated, concrete) [event] category
involving objects or specific types” as opposite to a “big (global, abstract) event
category” (2005, p. 225), which is referred to by a general verb in the specificity
hypothesis. Regarding lexical development, Bowerman (2005) reported that children are
able to produce appropriate verbs on the basis of how their target languages partition
events (or covert object categories) by the age of two. She further argued that “category
size”, which is identified as specificity mentioned by Tardif (2006b, 2006b) or Brown
(2001), does not determine the ease of learning. This conclusion seems similar to
Brown’s proposal that either specificity or generality dose not necessarily lead to ease of
verb learning but contributes to the difference in the tendency of mapping strategies.
Specificity or category size in lexical learning: Experimental evidence
Some studies have examined the effect of variation among exemplars for a
category, a relationship, or a label on conceptual development and language
development. One of them was conduced by Casasola (2005), habituating
10-month-olds and 14-month-olds to two or six examples for the support relation.
Through examining whether children generalized this relation to a novel object, she
found that the spatial category was formed in 14-month-olds who were habituated to
few (i.e., two) exemplars, yet children habituated to more (i.e. six) exemplars did not
form the abstract category. Less variation between objects might help children to attend
more to the similarity in relationship. In addition, some studies manipulate specificity
through providing children with different degrees of variation among exemplars for a
certain novel word. In Ma and Wong’s (2008) study, English-speaking three-year-olds
were presented with either a “narrow” verb that was mapped onto two actions with less
variation, or a “broad” verb that was mapped onto two actions with more variation.
Their results showed that children presented with exemplars with less variation
performed better when required to choose the target action from two video clips. It was
also found that children who were exposed to a narrow verb were less likely to extend
the novel word to an action carried out with a different manner than children who were
exposed to a broad verb, yet the difference was not significant in agent extension or
object extension. Ma and Wong (2008, p. 329) thus concluded that “meaning specificity
facilitates verb fast-mapping, but hinders verb extension to new manner variations.”
However, though Ma and Wong’s (2008) findings shed light on the effect of the
broadness of variation in exemplars on verb learning, it remains unclear how the
contrast between specific words or the semantic boundary between categories affects
children’s understanding about verb meaning8. In addition, although some efforts have
been made to explore the effect of labeling patterns on the structuring of categories or
generalization through providing children with different number of labels for the same
visual stimuli (e.g., Landau & Shipley, 2001), most experimental studies focus on the
process of noun learning while few studies have explored the role of labeling in forming
a semantic category of a verb.