According to Haegman (1991, p. 384), “A precise definition of the conception of heaviness has not been formulated but the intuitive idea is clear.” The following are some explanations about what heaviness is from the perception of Chomsky (1975, p.
477).
It is interesting to note that it is apparently not the length in words of the object determines the naturalness of the transformation, but, rather, in some senses, its complexity. Thus “they brought all the leaders of the riot in” seems more natural than “they brought the man I saw in.” The later, though shorter, is more complex.
Hence, though in the literature the general idea of heaviness is recognized, the precise definition has not been decided yet. Some people claimed heaviness is determined by the complexity of NP. Ross (1967) claimed that an NP is heavy if it dominates “a sentence.”, such as “The book which I borrowed from the library.”
Emonds (1976, p. 112) argued that the constraint on the complex NP shift Is that the NP dominates an S or PP, for example, John took into account only the people he knew. On the other hand, Hawkins (1990) proposed the idea that the heaviness is determined by the number of words dominated by a NP; Hawkins (1994) elaborated on his proposal and claimed that heaviness is decided by the number of nodes dominated by the NP. Due to the two different accounts to explain how heaviness is named, different researchers had done investigations on its definition.
In Wasow’s (1997a) investigation of the corpus in consideration of the complexity of constituent influenced the heavy NP shift, dative alternation, and verb particle constructions, the complexity effect and length effect were used to analyze the data. For the complexity effect, three criteria are identified: an NP dominating a sentence, dominating a sentence or an NP, and a coordinated or postmodified NP. As for the length effect, he identified three criteria: words, nodes, and phrasal nodes.
Wasow found that in complexity effect, the result is not able to account for the heavy NP shift since the percentage of NP which get shifted is more than the complexity account indicated. That is, some NPs which are not identified by the complexity effect
still get shifted. In addition, Wasow further indicated that only the account “dominates a sentence” shows that most NP identified as heavy are shifted , but there are still NPs which are identified as heavy do not get shifted. In other words, the number of NPs identified as heavy by the account “dominates a sentence” is far more than those get shifted. That is, there are still many NPs which are identified as heavy do not get shifted. By contrast, the length effect perfectly predicted the heavy NP shift and the dative alternation; in dative sentence pattern, the mean weight of direct object and indirect object is two words to sic words, and in double object construction, the mean weight of indirect object and direct object is one word to six words. However, we should note here that Wasow did not use the complexity effect to test the dative alternation since he considered that the complexity effect had failed to predict the heavy NP shift, thus it is unnecessary to continue to test the dative alternation.
Therefore, we may doubt here that if the result meets the end-weight rule, putting a heavy constituent at the final position of a sentence, when we examine the heaviness of constituents in dative alternation by the complexity effect.
Furthermore, Wasow (2002) conducted an experiment on heavy NP shift, dative alternation and verb particle concerning the complexity of NP by distributing questionnaires to native speakers in America. Wasow conducted pairs of NPs, which are identical in length and similar in meaning but differ in that one contains a sentence, but the other contains a simple NP, and made them into constructions in postverbal position and in position later in the clause. Hence each sentence has four variants:
sentential vs. non sentential NP, crossed with immediately postverbal vs. later position.
Subjects were asked to judge the acceptability of the sentences. The finding was that the participants judge the sentences containing a sentential NP at the final position of a sentence more acceptable. This result indicated that complexity of constituents might play a role in the definition of grammatical weight. This result was indeed
puzzling when comparing to Wasow (1997a). In Wasow’s study (1997a), he claimed that the complexity of constituents was not able to predict the grammatical weight.
We may wonder if it is the difference between the corpus analysis and the grammatical judgment to cause the different results. Hence, this thesis includes corpus analysis and grammatical judgment.
Rickford et al. (1995) did an investigation on the absence of verb in as far as construction to examine the influence of NP complexity. They did the research by giving questionnaires and collecting the actual uses in daily conversations. In the questionnaire study, they designed questionnaire containing thirteen as far as construction sentences--- four with concerned, one with go, and eight verbless, with four filler sentences and scrambled order.
Respondents were asked to rate all the sentences on a four-point scale, ranging from totally acceptable (scored as 4) to totally unacceptable (scored as 1). The elicited responses were seventy-nine responses. Besides, the collected tokens of the as far as construction on actual uses in daily life conversation are twelve hundred over an eight-year period. Through their experiment, among the seven factors which contribute to the verb absence in as far as construction including syntactic complexity of the NP, number of words in the NP, prosodic structure of the NP, position in sentence, age of speaker, sex of speaker, and mode, syntactic complexity is the most significant factor. That is, when the NP is a simple NP in as far as construction, the verb was usually kept, for example, as far as Tom is concerned; whereas when the NP is a complex one, the verb is usually omitted, for example, as far as playing baseball.
The following examples are more discussions to the NPs discussed.
Simple NPs: as far as sociolinguistics is concerned
Conjoined NPs: as far as better house and better cars
NPs containing PPs: the whole situation upset me as far as the outcome of the verdict
Sentential NPs: people think that I’m constantly in motion, as far as making films
On the other hand, number of words in the NP follows as far as proved to be nonsignificant in the VARBRUL statistics. Based on the study of Rickford et al., the complexity effect is considered to be the most important in deciding the word order variation; whereas the length effect is regarded as to be even nonsignificant at all.
Besides, Rickford et al. (1995) argued that ninety-nine percent of the second object (either direct object or indirect object) in a dative alternation is heavier than the previous one, at least as the same weight as the first object. The heaviness was determined by the complexity of the constituents. Consider the example from Rickford et al. (1995):
(32) They immediately give [a natural chemical] to [patients who have suffered heart attacks].
Feng (2005) proposed that the dative alteration in Chinese can be solved by the Nuclear Stress Pattern. The heavier one should be put at the end, and the lighter one should show earlier. Consider the example from Feng:
(33) a. * Please pick THE BOOK I BOUGHT YESTERDAY up.
請 拿 這本 書 我 買的 昨天 起來 請把我昨天買的那本書拿起來!
b. Please pick up THE BOOK I BOUGHT YESTERDAY.
請 拿 起來 這本 書 我 買的 昨天 請把我昨天買的那本書拿起來
Based on Feng, (33a) is ungrammatical because the noun phrase is heavier than
“up”. It is unacceptable that the noun phrase be put in front of “up.” Therefore, that’s why when a pronoun occurs in the verb-particle pattern construction, the pronoun can not be placed after “up” for a pronoun is considered to be the lightest. In other words, when the object in the dative construction is very heavy, it is placed at the final position due to the Normal Stress Pattern. However, in Feng’s work, he fails to identify what “heaviness” exactly means. Does heaviness refer to the word number of a phrase or the complexity in phrase’s structure or in pronunciation?
From the above mentioned studies, the different results proposed by different researchers lead us to ask the question why such difference appears when accounting for the word order variations?
Concerning the discourse effect, Niv (1992, p. 3) claimed the concept of heaviness is formalized by giveness in the discourse, which means the given information is usually followed by new information. Thus, the dative word order variation is determined by the giveness of the constituents. In addition, Gundel (1988, p. 229) also proposed given before new principle, which stated that what is given before what is new in relation to it. However, the precise definition of what can termed as given is still in debate in the literature. A more straightforward definition is proposed by Prince (1992).
In Prince’s term, the information status of a constituent can be divided as discourse-given, inferable, and discourse-new. In Prince’s term, when the referent of an NP is completely new to the discourse, it is called discourse-new. Prince labelled an NP as discourse-given when it is mentioned earlier in the discourse. As for inferable, Prince defined it as a NP which is not mentioned earlier in the discourse but the hearer can recognize it by logical reasoning. For example, a landlord must have a land, so a land can be inferred from landlord.
In the study of the interaction of heaviness and newness based on the production experiment and the corpus of elicited sixty-nine examples of give for dative alternation, Arnold et al. (2000) claimed that both heaviness and newness are significantly correlated with constituent ordering, and that neither correlation can be reduced to the other. Among the two types of word order variations: heavy NP shift and dative alternation, length effect is considered by the number of words of the constituent. Therefore, it can be classified into three categories. The first one is the theme is shorter than the goal, theme NP length - goal NP length = -2 or less. The second one is the theme is equal to the goal, theme NP length - goal NP length = between - 1 and 1. The third one is the theme is longer than the goal, theme NP length - goal NP length = 2 or more. Newness is coded as the following two categories:
given and new. These are determined according to whether the referent had previously appeared within each context. In their production task, the participants worked in pairs and were presented with three sets of three animals each and over a hundred objects. The participants worked together, one give instructions and the other carry out the instructions. There are twenty-four pairs of participants, each of which performed the experiment twice, switching roles the second time. That is, there are fourty-eight sessions in total. Giveness is determined when the referent had been referred to. Length is determined by the number of words of the constituents, theme length shorter than goal length, theme length equal to goal length, and theme length longer than goal length. The result showed that newness and heaviness correlated with the subjects’ choice of word order. That is, the participants tend to use more double object construction when the indirect object is given and is shorter than the direct object.
Although many researches have been done concerning the grammatical weight on the dative alternation, few have considered two possible factors, namely the
newness and the heaviness, at the same time in one study. The study conducted by Arnold et al. is one which considered the interplay of two factors.
As for the information order in Chinese sentences, the position of the NP earlier or later in a sentence distinguishes definite and indefinite status (Chao, 1968; Li, 1971;
Teng, 1975; Li & Thompson 1975). In other words, new information is indefinite thus occupies the final position of a sentence; on the other hand, given information is usually definite and occupies the sentence initial position. According to Chang (2001), the devices which are commonly used in Chinese to differentiate the given and new information are topicalization, BA-construction, and pre-verbal or post-verbal position. According to Chang (2001), the function of BA-construction is to distinguish the given information from the new information. The given information is placed after BA, so that the infotmation can be arranged in the given-new order.
Consider the examples from Chang (2001):
(34)a ?Ta zhong shu zai yuanzi li he plant tree at garden in “He is planting the tree/s in the yard.”
“他種樹在院子裡”
(34)b Ta ba shu zhong zai yuanzi li he tree plant at garden in “He is planting tree/s in the yard.”
“他把樹種在院子裡。”
Chang claimed that (34a) is less acceptable due to the position where the information is placed. If shu “the tree” is the given information, it is usually followed by BA to separate it from the new information.
As for the topicalization, Chang (2001) proposed that a topic means what a sentence is about, or what is known to both the interlocutors. Besides, the topic is
usually placed at the sentence initial position. And the rest of the sentence functions as a comment to interpret the topic. Consider the example from Chang (2001):
(35) Zhangsan wo yijing jian-guo le Zhangsan I already meet-past Asp “Zhangsan, I have already met (him).”
“張三，我已經見過了。”
Since Zhangsan is the person that the speaker talks about, it is placed in the sentence initial position for the hearer to identify. Zhangsan may be mentioned in the previous discourse and is treated as a topic rather than the object of “meet”.