2. Literature Review
2.3. Overall Remarks
tai-shang zuo zhe Yuehan.
stage-top sit ASP John
‘On the stage was sitting John.’
zuo/sit < x y > x=th, y=loc
In the demonstration of (27a), the SC rule applies first, assigning the patient/theme role [－r] feature, and then the DM rule assigns a [＋r] feature to all the non-θˆroles.
Likewise, in (27b), the SC also applies first, but next comes the application of the Locative Inversion rule (LI) which assigns [＋o] and [－r] to the two roles,
respectively. Then, the DM is blocked as its [＋r] feature conflicts with the existing features, [－r], that SC and LI have assigned.
2.3. Overall Remarks
Comparing the two different theories above, traditional transformational
grammar uses only two levels to form constituent structure: one level (the underlying level) to account for the lexical role structure of verbs, and the one (the surface level) to arrange the syntactic functions of these roles.9 The grammatical functions/relations
9 The traditional transformational grammar here refers to the version adopted by Coopmans (1989), Li (1990), and Radford (2004), but not the modern version of transformational grammar.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
are not independent but rely on the constituent structure. For example, the subject is the position higher than the verb in hierarchy structure. The two levels are related by syntactic movement operations which are restricted by syntactic principles (such as the Projection Principle, Case theory and the θ-Criterion) so as to preserve the
information of the applied movement. In addition, transformational grammar is barely concerned with lexicon in grammar but only with the derivation or movement rules.
This thus induces the need for many arbitrary rules to generate locative inversion.
They sort of develop them an ad hoc manner—case by case—so there is no coherence to the analyses.
However, LFG has no distinction between D-structure and S-structure, unlike transformational grammar, Instead, there is only a one-level constituent structure. And grammatical functions are abstracted from the constituent structure and presented as values or attributes in the functional structure. Functional structure appears as an underlying structure in transformational grammar; however, it involves no movement but only mapping theory, which means that it is non-derivational grammar. Moreover, LFG takes great account of lexicon. In argument structure, the thematic role is
mapped to its correspondent syntactic argument function in accordance with the universal hierarchy of prominence. The mapping principle of LFG seems rational in taking universal hierarchy into consideration, and produces results which are not as
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
arbitrary as derivational rules.
In view of this, the present paper will account for locative inversion under the framework of LFG, and the discussion will be with regard to the three level structures LFG takes: a(rgument)-structure, c(onstituent)-structure, and f(unctional)-structure, respectively. In addition to syntactic LFG grammar, the discussion will also include the discourse grammar—the information structure. Thus, in the next chapter, locative inversion will be discussed from these four structures (argument, constituent,
functional, and information structures), mainly focusing on examples from the two languages, English and Mandarin.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
Locative Inversion in English and Mandarin
3.1. Argument Structure
While it is well-established that locative inversion verbs bear unaccusativity, the definition of unaccusativity varies from one theory to another. Take traditional
movement theory and LFG, for example. Under the framework of syntactic movement theories, unaccusativity exists only in unaccusative verbs and passive verbs, which take an internal argument (logical object) but no external one (logical subject) in D-Structure. However, this definition cannot explain why some classified unergative verbs, taking an external argument in D-Structure, undergo locative inversion as well, as shown in (28): (28a) presenting an English example while (28b) a Mandarin one.
(28) a. Into the room walked an old man.
b. 洞裡 跳進 一隻 兔子。
dong-li tiao jin yi-zhi tu-zi.
cave-inside jump-in one-CL Rabbit
‘Into the cave jumped in a rabbit.’
To explain this exception, we resort to another theory—LFG to re-define unaccusativity. Instead of defining the unaccusativity within the framework of constituent hierarchy or grammatical relation, LFG defines unaccusative verbs by means of argument structure; locative inversion verbs are those unaccusative verbs taking the a-structure of <theme locative> (Bresnan and Kanerva 1989). Under this
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
assumption, such verbs as called unergative verbs in GB theory actually perform unaccusativity in possessing an a-structure of <theme locative>. Return to example (28a). Since the verb ‘walk’ can be interpreted as requiring a theme role, ‘an old man’, and a locative role, ‘into the room’, simultaneously, it can be a locative inversion verb without violating the unaccusativity principle.
Although the account <theme locative> seems workable for (28), it is not applicable in example (29).
(29) a. John (intentionally) sat on the stage.
b. On the stage (*intentionally) sat John.
In (29), the verb ‘sat’ is a locative inversion verb in that it has the inverted form in (29a). However, being modified by an agent-oriented adverb ‘intentionally’, it is predicted that the verb ‘sat’ takes an agent role, which is incompatible with Bresnan and Kanerva’s definition for a locative inversion verb: that it should take an
a-structure of <theme locative>. In this case, the strict a-structure <theme locative>
for a locative inversion seems to be unsatisfactory and inappropriate.
3.1.1. Suppressionist Approach
To solve the problem in example (29), Her (2004, 2007, 2009b) suggests suppressionist approach. Given the approach of suppressionist theory, all semantic interpretations of an argument will be taken into consideration, while, at the same
‧
time, the strict one-to-one argument-function linking (or θ-Criterion) is still
maintained when the thematic role is mapped to a syntactic argument. According to Her (2004, 2007, 2009b), in the intrinsic thematic structure, a verb may take two composing roles, agent and theme, in a composite role, agent-theme. When the composite role is mapped to its correspondent grammatical function, triggered by the one-to-one θ-Criterion that one role can only be assigned to one argument, one of the composing roles is suppressed, which allows the other to be linked to the grammatical function. See the example in (30).
(30) a. 約翰 坐 在 台上。 (Her 2009b: 11) Yuehan zuo zai tai-shang.
John sit at stage-top
‘John is sitting on the stage.’
i. zuo/sit < x-y z > x=ag, y=th, z=loc
In the canonical form of (30a), the verb zuo ‘sit’ is assumed to take a composite
‧
role of <agent-theme> and a locative role <locative>, which leads to an a-structure of
<agent-theme locative>. If the suppression rule is then applied, the a-structure will be either < agent-theme locative> or < agent-theme locative>, which leads to a separate meaning but the same grammatical relation is maintained. In the case of applied suppression, in (i), the SC rule assigns no features while DM assigns a [＋r] feature to the locative role. Based on UMP, the theme is linked to the more prominent SUBJ, and similarly, the locative to the less marked OBLθ. On the other hand, in (ii), SC assigns a [－r] feature to the theme role <z>, and DM assigns a [＋r] feature to the locative role <y>. And again, the UMP guarantees the most prominent OBLθ and SUBJ to the locative and the theme role, respectively. The two mapping processes imply two readings in the sentence (30a), Yuehan ‘John’ can be either an agent or a theme role.
(30) b. 台上 坐 著 約翰。 (Her 2009b: 11)
tai-shang zuo zhe Yuehan.
stage-top sit ASP John
‘On the stage was sitting John.’
zuo/sit < x-y z > x=ag, y=th, z=loc
As for the inverted form (30b), only the suppression of the agent is allowed: the
‧
a-structure of <agent-theme locative>. The theme role is assigned a [－r] feature from SC, and then the LI rule assigns feature [＋o] and [＋r] to <y> and <z>, respectively.
After the UMP, the <z> role is mapped to the SUBJ while the <y> role is the OBJ.
The inverted form in (30b), contrary to (30a), only induces one reading that Yuehan
‘John’ is a theme role.
This new account challenges the conventional concept that locative inversion verbs are restricted to a strict a-structure of <theme locative>, but suggests that some locative inversion verbs take also an a-structure <agent-theme locative>, which is evidenced in the contrast between (31) and (32).
(31) a. * 一疊 鈔票 故意 躺 在 地上。
yi-die chao-piao guyi tang zai di-shang.
one-CL bill intentionally lie at ground-top
‘A pile of bills is lying on the ground.’
b. * 地上 故意 躺 著 一疊 鈔票。
di-shang Guyi tang Zhe yi-die chao-piao.
ground-top Intentionally lie ASP one-CL bill
‘On the ground is lying a pile of bills.’
(32) a. 一個 醉漢 故意 躺 在地上。
yi-ge zui-han Guyi tang Zai di-shang.
one-CL drunk intentionally lie at ground-top
‘A drunk is lying on the ground.’
b. *地上 故意 躺 著一個 醉漢。
*di-shang guyi tang zhe yi-ge zui-han.
ground-top intentionally lie ASP one-CL Drunk
*‘On the ground is intentionally lying a drunk.’
Given that the volitional adverb guyi ‘intentionally’ can only modify an agent role, the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
sentences that bear this adverb must have an agent role. In this case, it can be inferred that the canonical form (31a) and its inverted form (31b) have the same a-structure
<theme locative>. However, in (32), the canonical form (32a) and the inverted form (32b) seem to differ in a-structure: the former taking <agent locative> or < theme locative > while the latter < theme locative >. This is also true in English, as in (33)
and (34).
(33) a. *A bullet intentionally shot through the window.
b. *Through the window intentionally shot a bullet.
(34) a. A man intentionally shot through the window.
b. *Through the window intentionally shot a man.
The examples from (31) to (34) suggest the possibility that locative inversion and its canonical form may not necessarily have the same truth value; sometimes the
canonical form may have two distinctive readings.
3.1.2. Resultative Compounds in Locative Inversion
Another supporting argument for the suppressionist theory concerns the
resultative compound in Mandarin locative inversion, as exemplified in (35), provided by Li (1995).
(35) 這場 饑荒 餓死了 很多 人。
zhe-chang jihuang e-si-le henduo ren.
this-CL famine hungry-dead-ASP many people
‘Many people starved to death in this famine.’
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In Li (1995), this sentence is interpreted as a resultative compound inversion. The compound verb e-si in (35) is composed of two verbs: e ‘hungry’, indicating the cause, and si ‘death’, the result; the former is termed as Vcaus and the latter as Vres. Vcaus and Vres both assign theta roles to the object position, but none to the subject, leaving the subject position unoccupied by roles from either Vcaus or Vres. On the other hand, although acting like the cause of the starvation of the people, the argument zhe-chang jihuang ‘this famine ‘ cannot be directly linked to either Vcaus or Vres. However, the argument receiving no theta roles from the compound verb turns to fill the subject position, which violates the principle of θ-Criterion.
To solve the problem, we can alternatively interpret this sentence as a locative inversion: attaching yu ‘in’ to the subject zhe-chang jihuang ‘this famine’ and rephrase the sentence as in (36). Yu, in Mandarin, can refer to the time, place, or the event of the cause. In this case, ‘this famine’ can be viewed as a location, and (36) is a canonical form of a locative inversion while (35) is the inverted form.
(36) 很多 人 餓死 於 這場 饑荒。
henduo ren e-si yu zhe-chang jihuang.
many people hungry-dead in this-CL Famine
‘Many people starved to death in this famine.’
Applying the suppressionist theory, the mapping process of locative inversion in (35) and its canonical form in (36) will be illustrated in (37a) and (37b).
‧
henduo ren e-si yu zhe-chang jihuang.
many people hungry-dead in this-CL Famine
‘Many people starved to death in this famine.’
In the canonical form (37a), the causing verb e takes a theme and a locative role
<x z>, while the resulting verb si takes a theme role <y>; when forming a compound, the compound e-si then takes an a-structure <x-y z> with a composite role. After applying the suppression rule, the a-structure that the compound verb takes turns to either <x-y z> or <x-y z>, and either of them will lead to the same reading, <theme locative>. In the process of operation, the theme recieves a [－r] feature from SC, and the locative receives a [＋r] from DM. Based on UMP, the theme is linked to the more prominent SUBJ, and similarly, the locative is mapped to the less marked OBLθ.
‧
zhe-chang jihuang e-si-le henduo ren.
this-CL famine hungry-dead-ASP many people
‘Many people starved to death in that famine.’
e ‘hungry’ <x > si ‘dead <y z>
The inverted form in (37b) has the same a-structure after suppression. However, after receiving a [－r] feature from SC, the locative inversion rule (LI) is applied, assigning [＋o] and [－r] features to the theme and the locative role, respectively. DM cannot assign the [＋r] feature since it is contradictory to the [－r] feature which has been assigned to the < x-y > and <z> roles. Again, by UMP, the locative chooses the more prominent SUBJ while the theme has only one choice, the OBJ. Since the interpretation of locative inversion allows the phrase ‘this famine’ to fill the subject position, it is suggested that the sentence in (35) is acually a locative inversion instead
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
of a resultative inversion.
3.2. Constituent Structure
3.2.1. C-structure in English and Mandarin
After examining the argument structure of locative inversion, this section will move on to investigate its constituent structure. Under the framework of LFG, the discussion of c-structure here focuses only on the lexical category of arguments and no grammatical functions are involved.
As noted, locative inversion verbs require two arguments, a postposed theme role and a preposed locative one. As to the former, much of the research proposes a
consistent definition for it—that it is a noun phrase, which is true both in English and in Mandarin. At the same time, however, the lexical category of the preposed locative phrase varies across languages. Take English and Mandarin, for example; the former treats the locative as a prepositional phrase, while the latter treats it as a noun phrase.
The following section will discuss the different categories of locative phrase in the two languages.
3.2.2. The Preposed Locative Phrase
The preposed locative phrase of locative inversion in English is always
accompanied by a preposition like ‘in’, ‘at’, ‘on’, a directional preposition like ‘into’,
‧
‘onto’, or an adverb like ‘away’, ‘above’, etc., as exemplified in (38).
(38) a. On the stage sat a man.
b. Into the pool jumped a man.
c. Away flew the birds.
Due to the fact that a locative phrase in English is obliged to be accompanied by a preposition, it is defined as a preposition phrase (PP).
Contrary to the PP category in English, the preposed locative phrase in Mandarin seems more like a noun phrase than a preposition phrase, in that it is composed of no preposition but a localizer. The literature has indicated the phenomenon that in Mandarin, the preposed locative phrase requires no preposition (Huang 1982, Chao 1961, Wang 2005); and few, if any, are qualified apart from zai ‘at’ and cong ‘from’
(Teng 1975)10, as seen in (39) and (40).
(39) (?在) 空中 飄 著 一片 落葉。
(?zai) kong-zhong piao zhe yi-pian luo-ye.
in air-inside float ASP one-CL falling-leaf
‘In the sky is floating a falling-leaf.’
(40) (從) 洞裡 跳出 一隻 老虎。
cong dong-li tiao-chu yi- zhi lao-hu.
from cave-inside jump-out one-CL Tiger
‘From the cave jumped out a tiger.’
However, the existence of these prepositions in Mandarin locative inversion is still under question for two reasons. First, these prepositions tend to be omitted, and even
10 Teng (1975) claims that locative prepositions include 1) zai ‘to be located’ (to indicate Base), 2) cong ‘from’ (to indicate Source), and 3) dao ‘to’ (to indicate Goal).
‧
need to be obligatorily deleted in some cases. Huang (1982) attributes the omission to their highly-verb-like aspect11, which is in conflict with the stative property of
locative inversion. What is more, (40) is not a typical locative inversion on closer examination, for there is no corresponding canonical form, as shown in the ungrammaticality in (41).
(41) * 一隻 老虎 跳出 從 洞裡。
yi- zhi lao-hu tiao-chu cong dong-li.
one-CL tiger jump-out from cave-inside
‘A tiger jumped out from a cave.’
In view of this, it seems that the preposition in locative inversion in Mandarin tends to be omitted rather than be present, which provides evidence for the nominalization of the preposed locative phrase in Mandarin (Hou 1977). Another argument for the nominal property of the preposed locative is the obligatory requirement of localizer nouns, like li ‘inside’, pang ‘beside’, hou ‘back’, qian ‘front’, wai ‘outside’, shan ‘up’, and xia ‘down’, etc (Huang 1982).12 Considering the absence of a preposition and the appearance of a localizer—that the locative phrase is headed by a noun but not by a preposition—it is suggested that the preposed locative in Mandarin is categorized as a noun phrase.
In summary, the preposed locative phrase belongs to different categories in
11 Huang (1982) indicates that some prepositions in Mandarin are evolved from intransitive verbs, and thus inherit verb-like features, such as zai ‘in/at’, cong ‘from’, xiang ‘to’, cao ‘to’, and dao ‘to’, etc.
12 Chao (1961) regards them as post-positional particles while Tang (1988) claims they are localizers.
See Huang (1982).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
English and Mandarin; in the former it is a preposition phrase (PP), while in the latter it is a noun phrase (NP).
3.3. Functional Structure
3.3.1. Literature Review
While there is a generally-accepted understanding of the position of c-structure in locative inversion, the literature on grammatical relation lacks consistent and complete arguments. The preposed locative is regarded as a subject (Li 1990, Huang and Her 1998, Her 2003, Levin and Rappaport 1995) or a topic (Coopmans 1989), and the postposed theme is suggested as a subject (Rochemont and Culicover 1990) or an object (Coopmans 1989, Bresnan and Kanerva 1989, Li 1990, Huang and Her 1998, Her 2003). Some of these proposals are based on the word order SVO (the preposed locative as a subject), some are based on the lexical category of the preposed locative (where the preposed locative is a topic but not a subject), and still others are based on the verb-agreement rule (the postposed theme as the subject).
Recent literature is nearly all consistent in claiming that the postposed theme is an unaccusative object, which is true both in English and Mandarin. In light of this, this section will not focus on the postposed theme; instead, we will focus on the preposed locative phrase, the other argument in locative inversion, in respect to its
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
grammatical relation.
3.3.2. Bresnan’s Analysis (1994)
Bresnan’s analysis of the preposed locative phrase in English will be adopted to resolve the inconsistent interpretations of its grammatical function.
In c-structure, the sentence-initial position of a preposed locative phrase implies its subject grammatical status. Nevertheless, as a preposition phrase in English, the preposed locative, is unqualified to fill the nominal subject position in c-structure, as described in (42), which illustrates the mapping restriction between c-structure and f-structure.
(42) If C is a c-structure position restricted to subjects or objects, then C is nominal. (Bresnan 1994:104)
In accordance with the mapping constraint in (42), only nominals can fill the subject and object position of phrase structure. However, in English, the preposed locative is not nominal and thus is unable to fill the subject position.
To tackle this problem, we examine the c-structure and f-structure of the preposed locative phrase in English, respectively, as presented in (43)13.
To tackle this problem, we examine the c-structure and f-structure of the preposed locative phrase in English, respectively, as presented in (43)13.