In this section issues in language acquisition related to the present study will be discussed such as Theory of Conservativism, age effects and task effects. In addition, the present study assumes the Semantic Feature Hypothesis proposed by Clark (1971), and Vendler’s (1967) classification of verb types. The main notions of the Semantic Feature Hypothesis will be summarized briefly and then Vendler’s analysis will be reviewed in the following section, too.
1.2.1 Theory of Conservativism
To argue against Baker’s (1979) view that lexical structures are arbitrarily
represented and thus have to be learned piecemeal, Pinker (1989) put forward the Theory of Conservativism. He claims that lexical structures are not completely arbitrary, and are learnable by rule. According to Pinker, there are three levels of semantic representations:
i) The lexical specification of each individual verb
ii) The narrow-range (NR) subclasses, which are the sets of verbs that share particular semantic meanings, and whose argument structures can in fact alternate
iii) The broad-range (BR) classes, which are larger sets of verbs sharing general semantic meanings that all verbs participating in a particular syntactic alternations have in common; the semantic commonalities are presumably necessary for the syntactic alternations to occur
The three levels are exemplified by English dative constructions. The verbs bring and pull both show up in the form of NP1 V NP2 to NP3 but only bring but not pull can alternate with the form of NP1 V NP3 NP2:
(5) a. John brought the present to Mary.
b. John brought Mary the present.
(6) a. John pulled the chair to Mary.
b.*John pulled Mary the chair.
As shown in (5), bring can appear in the prepositional dative form as well as the double object dative form. However, pull can only show up in the form of the prepositional dative as in (6). The three levels of representations for them are as follows:
(7) bring a. The verb level: X acts on Y and then Y goes with X to Z
b. The BR level: cause to go
c. The NR level: continuous force, no specific manner of motion, that is, one can bring by many different kinds of manner, including pulling
(8) pull
a. The verb level: X acts on Y and then Y goes to Z alone b. The BR level: cause to go
c. The NR level: continuous force, specific manner of motion, that is, one can only pull by the manner of pulling
Pinker argues that the alternations are primarily semantically driven. Once children
learn the semantic contrast between bring and pull, they will then apply it to the syntactic alternations straightforwardly. For instance, children will learn the verb level representations of bring and pull and other verbs at first, then distinguish the kinds of bring verbs from those of pull verbs by the existence of a specific manner or not and finally apply the contrast to the dative alternations.
The implication for the theory is that children will not use the syntactic alternations (the BR level) before they acquire the semantic contrast between different kinds of verbs (the NR level) since the contrast is the element necessary for the syntactic alternations to occur. In this sense, children are conservative learners.
However, once they master the contrast, they will immediately apply it to the construction. According to the theory, we predict that children have to learn the contrast between the element exhibited and that prohibited in the ka construction before they apply the contrast to the ka construction because only when they possess the semantic contrast can they apply the contrast to the ka construction.
1.2.2 Age Effects
It seems to be a natural tendency that age plays a crucial role in children’s language acquisition. Results of many empirical studies have illustrated that children’s language comprehension and production improve with age (Lenneberg 1967, Erbaugh 1982, Cheung 1992, Fahn 1993). Lenneberg (1967) found that children at the age of one could produce one word which may convey more than one meaning. At the age of three, children could produce two words with well-established semantic relations. And, most of the four-year-olds’ utterances were comprehensible. In a longitudinal study of children’s acquisition of the Mandarin ba construction, Erbaugh (1982) found that the ba construction appeared in children’s speech as early as they were 2;3 years old. However, the use of this
construction was not mastered by children until the age of 3;6. Cheung (1992) conducted a statistical analysis on the longitudinal data provided in Tse, Tang, Shi, and Li’s study (1991) and indicated that children’s error ratio principally decreased with age. Fahn (1993) examined children’s acquisition of the ba construction and found that age five was a crucial cutoff point for certain constraints while age six was a demarcation point for the other constraints2. Therefore, in this study age will be one of the issues under investigation.
1.2.3 Task Effects
The other focus of this study is on task effects. In order to examine children’s linguistic competence and performance, both comprehension and production tasks will be designed in the present study. Comprehension tasks are generally considered as much easier than production tasks because it is assumed that children have to acquire the necessary linguistic knowledge before they can actually use it (Miller 1963, Oviatt 1980, Brown 1987, Gerken and Shady 1996, Wu and Wen 1997). In an investigation of children’s phonological development, Miller (1963) pointed out that a subject clearly perceived the contrast between English s and th, yet she could not produce the contrast herself. Brown (1987) also stated that children may be incapable of producing a sentence with an embedded relative clause in it but they may be able to understand it. Wu and Wen (1997) found that their subjects’
performances on the grammaticality judgment task was better than on the translation task. To verify this view, task effects will be discussed in this research.
2 Fahn (1993) examined five constraints of the ba constraint: (1) The Progressive constraint: a progressive marker is impossible to occur in the ba construction; (2) the Verb Selection constraint:
certain types of verbs cannot show up in the ba construction; (3) the Modifier constraint: certain verbs require modifiers to co-occur with the ba construction; (4) the Compound Verb constraint: certain compound verbs cannot show up in the ba construction; (5) the Definiteness constraint: the ba-NP cannot be indefinite.
1.2.4 The Semantic Feature Hypothesis
Based on a componential or feature analysis of words, Clark (1971) develops the theory of the Semantic Feature Hypothesis (SFH). Two properties of the componential analysis of words are emphasized in the theory:
i) The meanings of words can be represented as a set of components arranged in a hierarchy.
ii) Feature representations are given positive and negative values.
For example, the temporal prepositions before, after and while can be represented in a feature hierarchy as follows:
(9) Feature Representations of when, before and after
The first feature specification that comes into play is [+Time] which includes the words referring to some aspect of time. The next feature, Simultaneous, which is dominated by [+Time], has two values: [+Simultaneous] and [-Simultaneous].
Finally, the feature [+/-Prior] only occurs in words specified as [-Simultaneous].
These features are organized in a hierarchical order, with [+Time] dominating [+/-Simultaneous], and [-Simultaneous] dominating [+/-Prior].
According to the hypothesis, it is predicted that superordinate features in a feature hierarchy would be acquired first since these features are usually common to several different words and thus are general concepts, for instance, [+/-Time] or [+/-Simultaneous]. Also, plus values would be mastered before minus values
because the plus values are usually simpler or unmarked in a pair of relational terms.
With regard to the temporal prepositions above, a distinction between before and after would be acquired later than the general distinction between before/after and when.
Moreover, the precedence of positive over negative features expects that when would be mastered prior to before/after, and before prior to after.
The psychological validity of SFH has been proved in children’s acquisition of relational terms. Clark (1971) examined the acquisitional process of before, after and when and found that when was mastered first, followed by before and finally after.
Also, it was shown in Donaldson and Balfour’s (1968), and Donaldson and Wales’s (1970) studies that at the first stage, children seemed to treat more and less as synonyms. They performed well on questions about more but they consistently misinterpreted less as if it means more. Clark claimed that the results were explicable with SFH if it is supposed that children had mastered the feature [+Amount]
which was shared by more and less but had not yet acquired the feature [+/-Polar] for differentiating the pair.
1.2.5 Vendler’s (1967) Classification of Verbs
Vendler (1967) classifies verbs into four types, that is, Stative, Activities, Accomplishments, and Achievements, as stated in Table 1.1:
Table 1.1 Vendler’s (1967) Verb Classification
Type Example
1. Statives3 Static, durative, atelic know, love 2. Activities Dynamic, durative, atelic run, walk
3. Accomplishments Dynamic, durative, telic build a house, walk to school 4. Achievements Dynamic, instantaneous recognize, find
Based on the discussion, States are verbs lacking continuous senses, for instance, know and love belong to this category. They can not have a progressive aspect since they do not possess continuous senses. Activities such as run and walk are verbs denoting processes going on in time with no terminal point. Accomplishments are verbs denoting processes going on in time with a terminal point, and after achieving the goal, the action can no longer continue. Verbs like build a house and walk to school are members of this type. Achievements are verbs that have a terminal point, but the time for reaching this point is instantaneous, for example, recognize and find are verbs of this type.
In the present study it is assumed that the classification of verbs here is also applicable to Taiwanese. Also, later in Chapter Four, we will develop the feature specifications of Taiwanese verbs based on Vendler’s (1967) classification of verb types.