4. Headedness
4.5 An Alternative of Determining Headedness
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(68) Zhangsan da-si-le Lisi
John hit-die-ASP Lee
‘John hit Lee to the extent of making Lee die.’
(69) shoupa ku-shi-le
handkerchief cry-wet-ASP
‘The handkerchief was wet with tears.’
(70) ta pao-lei-le
he run-tired-ASP
‘He ran and he got tired.’
Since the sentences in (68), (69), and (70) belong to different categories of
headedness, according to the head feature percolation condition, it is impossible to
generalize that all resultative compounds are headless. Even though some compounds
do feature headlessness, it cannot be plausibly argued that all do.
4.5 An Alternative of Determining Headedness
As noticed from previous sections, researchers have delved into the headedness
issue of Mandarin resultative compounds, yielding various conclusions that are
‧
nevertheless proven to be inconclusive. It is discovered that Mandarin resultative
compounds do not fall into only one category of headedness. A semantic approach to
determining headedness seems fraught with difficulty, since it relies too much on the
context. The syntactic aspect, which is less context-dependent, by contrast, could be a
more reliable criterion to judge where the head should be.
To provide a relatively comprehensive explanation for the headedness of
Mandarin resultative compounds, an alternative analysis with respect to syntax is
therefore given. This analysis assumes that the head of a Mandarin resultative
compound is only determined when arguments are overt (not suppressed). Given this
assumption, the present analysis adopts some concepts from the constraints or
definitions of head proposed by Zwicky (1984), Y. Li (1990, 1995), Chung (2006),
and C. Li (2008, 2009).
First, regarding the definition of a head in previous research, Zwicky (1984)
defines headedness from a distributional point of view, assuming that “the head of X
is a part with the same distribution as X.” When applied to resultative compounds,
this definition indicates that the argument structure of the head ought to be identical
with that of the entire compound. The same holds for the head feature percolation
condition (Y. Li, 1990, 1995; C. Li, 2008, 2009), which states that the arguments and
properties of the head are reflected on the compound level. Again, Chung (2006),
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
whose stance of headedness is based on The Concise Oxford Dictionary of Linguistics
(1997) and Jackendoff (1970) (as cited in Chung, 2006), concludes that “a head is the
constituent that determines the properties of the complex word or phrase as a whole.”
In brief, these definitions or conditions all specify that a head is able to determine or
characterize the properties of the whole compound; that is, the properties of the head
are realized as those that the compound bears. Therefore, for our analysis, since
argument structures are related to certain syntactic and distributional properties, it is
admissible to determine the head of a resultative compound if its argument structure
remains the same as that of the whole compound.
Second, the idea of headedness being dependent on the sentence structure arises
from the full range of possible patterns of argument structures in Mandarin resultative
compounds. One resultative compound can emerge from various constructions, as in
(71).
(71) a. ta ku-shi-le shoupa
he cry-wet-ASP handkerchief
‘He cried and the handkerchief got wet.’
b. shoupa ku-shi-le
handkerchief cry-wet-ASP
‧
‘The handkerchief was wet with tears.’
If the compound ku-shi ‘cry-wet’ alone is examined, as discussed previously, the
pair in (71) poses a problem. It seems contradictory to regard the compound ku-shi
‘cry-wet’ as a headless compound when it also fits a structure like (71b) or as a
headed compound while it fits a structure like (71a). However, if the headedness of a
compound is decided on the basis of the overt arguments in the argument structure,
the controversy can be resolved. Since the argument structure is a crucial property for
judging headedness, as the head feature percolation condition and other definitions
suggest, it is reasonable to determine the head of one compound when arguments are
present. Moreover, with overt arguments in the argument structure, the head of a
compound can be revealed if the head alone behaves syntactically the same as the
compound. Thus, in the case of the compound ku-shi ‘cry-wet,’ its argument structure
is detailed in (72).
(72) a. ta ku-shi-le shoupa
he cry-wet-ASP handkerchief
‘He cried and the handkerchief got wet.’
Vcaus <a> ＋ Vres <1> <a[caus] 1[af]>
b. shoupa ku-shi-le
handkerchief cry-wet-ASP
‧
‘The handkerchief was wet with tears.’
Vcaus <a> ＋ Vres <1> <a[caus] 1[af]>4
The compound is justifiably headless when appearing along with arguments like (72a)
because both V1 ku ‘cry’ and V2 shi ‘wet’ do not require two arguments. None of the
composing verbs behaves exactly the same as the compound does. On the contrary,
the V2-headness of the compound is indicated by the argument structure shown in
(72b), where only V2 behaves like the compound.
Furthermore, another case for this intransitive-intransitive compound justifies
this analysis. This analysis is able to distinguish the head of xiao-feng ‘laugh-crazy’ in
(73), simply based on a glimpse of its argument structure.
(73) ta xiao-feng-le
he laugh-crazy-ASP
‘He laughed so much it seemed he went crazy.’
Vcaus <a> ＋ Vres <1> i. <a-1>
ii. <a-1>
Two argument structures are valid for this compound. In terms of (73i), the overt
argument is <a>, which happens to be the only argument subcategorized by V1 xiao
‘laugh.’ The argument arrangement of V1 remains exactly the same as that of the
4 The agent role <a[caus]> in (72b) is apparently suppressed, causing <1[af]> to surface as SUBJ.
‧
entire compound. As a result, with the argument structure (73i), the compound is
headed by V1. On the other hand, regarding (73ii), as the only argument required by
V2 feng ‘crazy,’ <1> is overt in the argument structure. Since the argument order of
the whole compound is identical to that of V2, the compound is V2-headed with an
argument structure like (73ii).
The examples presented above show that the present analysis suffices to predict
the head of each resultative compound. According to this analysis, in each argument
structure, only the unsuppressed arguments are considered to participate in
determining headedness. Given conditions such as the head feature percolation
condition, if the order which a composing verb’s overt arguments are aligned is
reflected in the way arguments of the entire compound are arranged, this composing
verb is the head. By this criterion, headedness of each type of compound successfully
falls out as follows.
(74i) is headed by V1 because the overt <a, b> of V1 behaves exactly the same
‧
as the compound; so is the case for (74ii). However, for (74ii) and (74iii), no
argument structure of any composing verb matches that of the compound because one
argument of each composing verb is suppressed. As a result, (74ii) and (74iii) are
headless.
The left-headedness of both (75i) and (75iii) is attributed to the fact that the overt
argument arrangement which is maintained on the compound level belongs to V1. On
the other hand, (75ii) and (75iv) are headless for two reasons. First, V1 does not
behave like the whole compound because one argument required by V1 is suppressed.
Second, the way V2’s single argument <1> is realized differs from the final
realization of the compound since the compound subcategorizes for two arguments.
(76) intransitive-transitive verb compounds
Vcaus <a> ＋ Vres <1, 2> i. <a[caus]-1 2[af]> -- Headless
ii. <a-1 2> -- V2-headed
In (76i), with an argument suppressed, V2 fails to serve as the head. Similarly,
‧
V1, as a one-place verb, also falls short of matching the two-argument-needed
argument structure of the entire compound. Nevertheless, in (76ii), the two overt
arguments entirely come from V2 and their arrangement is maintained on the
compound level. (76ii) is hence headed by V2.
(77) intransitive-intransitive verb compounds
Vcaus <a> ＋ Vres <1> i. <a-1> -- V1-headed
ii. <a-1> -- V2-headed
iii. <a[caus] 1[af]> -- Headless
In cases like (77i) and (77ii), each composing verb simply selects one argument;
so does the whole compound. They are headed by V1 and V2 respectively due to
different overt arguments. In the case of (77iii), while V1 and V2 are both intransitive,
the compound itself subcategorizes for two arguments. Unfortunately, neither V1 nor
V2 syntactically behaves the same as the compound. The compound in (77iii) is thus
headless.
The present analysis not only settles the controversy of headedness that
resultative compounds generate but at the same time eases the comprehension or
interpretation of resultative compounds. Since the expressiveness of a resultative
compound may complicate one’s comprehension, determining the head and specifying
where the head is would facilitate comprehension. In addition, with the arguments
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
being overt, headedness is readily revealed by applying principles such as the head
feature percolation condition. The problem of controversial headedness is hence
solved. Based on our standpoint, headedness of a compound is not fixed but
dependent on its overt arguments appearing alongside with the compound. Thus how
the overt arguments are arranged, as a vital factor in determining headedness, may
cause the headedness of one compound to vary.
‧ 國
立 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Chapter 5 Argument Realization and Headedness in Lexical Mapping Theory
From the analysis shown in the preceding sections, both argument realization and
headedness should be interpreted on a sentence level where arguments are present.
With arguments being overt, the argument-function mapping and the reading are thus
understood and the head is thus determined.
This analysis concerning argument realization and headedness would be more
complete and strictly governed if it is applied within a theoretical framework. The
theoretical framework adopted in the present study is Lexical Mapping Theory, which
is constructed under Lexical-Functional Grammar.
5.1 Lexical Mapping Theory
As one of the theories under the framework of Lexical-Functional Grammar
(LFG), Lexical Mapping Theory (LMT) is related to the argument-function realization
of a predicate. It emphasizes the existence of an argument-structure (a-structure), an
interface located between the semantic or conceptual dimension and the syntactic
dimension (Bresnan, 1995; Falk, 2001). As (78) illustrates, a-structure bridges the
lexical semantic structure and the syntactic structure (Bresnan, 1995).
71
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(78) lexical semantics
a-structure
syntactic structure (f-structure)
In terms of thematic roles in LMT, (79) presents a thematic hierarchy adopted
from Bresnan and Kanerva (1989, 1992) (cf. Jackendoff, 1990; Grimshaw, 1990). In
this hierarchy, roles from left to right are in an order with descending prominence.
(79) agent > beneficiary > goal/experiencer > instrument > patient/theme
> proposition > location
Matched with thematic roles through the a-structure, grammatical functions are
determined by the property of restrictedness that certain grammatical functions are
restricted to a few specific thematic roles: OBLθ and OBJθ, SUBJ, and OBJ.
Categorized by this distinctive feature [±r], along with objective feature [±o], the
properties of these four grammatical functions SUBJ, OBJ, OBLθ, and OBJθ can be
summarized as the following chart (80) shows (Dalrymple, 2001; Falk, 2001; Her,
2004, 2007):
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(80)
[－o] [＋o]
[－r] SUBJ OBJ
[＋r] OBLθ OBJθ
From the binary distinctive features provided above, a markedness hierarchy is
obtained. It is assumed that minus features are considered unmarked while plus
features are marked. As a result, the markedness hierarchy is demonstrated in (81)
(Her, 2004, 2007).
(81) SUBJ > OBJ/OBLθ > OBJθ
The following step by step demonstrates mapping principles that regulate the
mapping from θ-structure to a-structure, from a-structure to f-structure and other
mapping conditions (c.f. Dalrymple, 2001; Falk, 2001; Her 2004, 2007). First, a
principle, known as the “intrinsic classification of argument roles for functions” (IC)
specifies the mapping from θ-structure to a-structure, as in (82).
(82) Intrinsic Classification of Argument Roles for Functions (IC) (Dalrymple,
2001; Falk, 2001; Her 2004, 2007):
a. patient/theme [－r]
b. secondary patient/theme [＋o]
‧
Another principle that governs the mapping process is described in (83). Except
the most prominent role (theta-hat), all other roles are assigned a restricted feature
[＋r]. Note that feature information can only be added rather than be changed. Theme
and patient are not suitable for this feature assignment since they are already
registered with an unrestricted feature [－r] previously. Only non- roles which are
not yet given the [－r] feature can be assigned a [＋r] feature.
ˆ
ˆ
(83) Default Morphosyntactic Classification of Argument Roles (DC)
(Dalrymple, 2001; Her 2004, 2007):
θ≠ˆ, θ [＋r]
Next, the a-structure to f-structure mapping, demonstrated in (84), describes how
roles on the a-structure map to grammatical functions (Falk, 2001).
(84) A-structure to F-structure Mapping (Falk, 2001):
a. a [－o] argument which is theta-hat maps to SUBJ.
b. [－r] may map to SUBJ.
c. add to non-SUBJ arguments positive feature values [＋r]/ [＋o] where
possible.
From a more simplified aspect, Her (2004, 2007) revised the Lexical Mapping
Theory. In addition to adopting the general thematic and markedness hierarchies as
‧
well as the IC and DC, a unified mapping principle (UMP) is further suggested to
ensure a strict one-to-one argument-function linking (Her, 2004, 2007).
(85) The Unified Mapping Principle (UMP) (Her, 2004, 2007):
Map each role in a-structure to the highest compatible* AF
available+.5
(*An AF is compatible if and only if it contains no
conflicting features. +An AF is available if and only if it is
not fully specified by a role and not linked to a higher role.)
According to the hierarchies and principles demonstrated above, the mapping
between thematic roles and grammatical functions can thus be organized. The next
section illustrates the application of Lexical Mapping Theory (mainly the revised
Lexical Mapping Theory) on Mandarin resultative compounds, presenting the
realization of their argument-function linking.
5.2 The Application
The revised Lexical Mapping Theory, based on Her (2004, 2007), correctly
predicts that the consequence of the argument-function linking for four different types
of resultative compounds (transitive-transitive, transitive-intransitive,
intransitive-transitive, and intransitive-intransitive verb compounds) automatically
5 Her modified the UMP after a class discussion in 2009. The UMP in (85) is the latest version. AF:
argument function. A-structure: argument structure.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
falls out. We will present how the argument-function linking of these compounds is
derived under this framework, along with their headedness below.
The first type of resultatitve compounds consist of two transitive verbs. Example
sentences and their application of LMT are presented in (86), (87), and (88).
(86) a. Zhangsan da-shu-le Lisi
John hit-lose-ASP Lee
‘John hit Lee and John lost to Lee.’
b. Zhangsan du-dong-le zhe-shou shi.
John read-understand-ASP this poem
‘John read the poem and he understood it.’
i. <a-1, b-2> (38i) (a = agent, b = patient)
IC [－r]
CF S/O/… S/O
UMP SUBJ OBJ V1 as head
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
ii. <a-1, b-2> (38iv) (1 = theme, 2 = patient)
IC [－r] [＋o] applied to (86a)
CF S/O OBJ/ OBJθ
UMP SUBJ OBJ V2 as head
<a-1, b-2> (38iv) (1 = experiencer, 2 = patient)
IC [－r] applied to (86b)
CF S/O/… S/O
UMP SUBJ OBJ V2 as head
(86i) and (86ii) ideally explain the reading and grammatical structure of example
sentences such as (86a) and (86b). Given that <a-1>/<a-1> receives the
agent/experiencer role and <b-2>/<b-2> the patient role, an unrestricted feature [－r]
is assigned to <b-2>/<b-2> according to the IC. Compatible grammatical functions
are thus revealed. Even though <a-1>/<a-1> can be assigned to all grammatical
functions, its thematic role, as the more prominent agent/experiencer naturally
occupies the subject position as the UMP conditions. Since the SUBJ is taken, the
UMP next maps OBJ to the patient role <b-2>/<b-2>, which receives the [－r] feature
and only has SUBJ and OBJ to choose from.
For the argument structures with theme and patient, it is specified previously that
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in cases like da-shu ‘hit-lose,’ theme is primary to patient. Thus, the IC requires that
the theme <a-1> receives [－r] and the patient <b-2> [＋o]. Compatible grammatical
functions for theme remain SUBJ and OBJ, while patient is permitted to choose
between OBJ and OBJθ. The UMP finally relates <a-1> to the more prominent SUBJ
and <b-2> to OBJ.
As we can see, by the LMT, these two argument structures lead to the same
outcome as discussed in 3.4.1; that is, one syntactic realization and one reading.
(87) Zhangsan zuo-pei-le maimai.
John do-lose-ASP business
‘John did business and he lost it.’
<a[caus]-1, b-2[af]> (38ii) (a = agent, 2 = theme)
IC [－r]
CF S/O/… S/O
UMP SUBJ OBJ Headlessness
Concerning a sentence that corresponds to an argument structure like (87), the IC
offers a [－r] to theme <b-2[af]>. The agent role <a[caus]-1> is permitted to select all
possible grammatical functions, while the theme role <b-2[af]> can only link to SUBJ
and OBJ. The UMP then requests that the most prominent agent <a[caus]-1> occupies
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the most prominent subject position and that the least prominent theme <b-2[af]> is
left with OBJ for mapping. Moreover, note that causitivity is valid in this argument
structure, also predicting that <a[caus]-1> and <b-2[af]> map to SUBJ and OBJ
respectively due to causative prominence difference between [caus] and [af].
(88) Zhangsan da-shu-le Lisi
John hit-lose-ASP Lee
‘Lee hit John and Lee lost to John.’
<a-1[af], b[caus]-2> (38iii) (1 = theme, b = patient)
IC [－r] [－r]
CF S/O S/O
UMP OBJ SUBJ Headlessness
(88) lends evidence to show that the reading of Zhangsan da-shu-le Lisi ‘Lee hit
John and Lee lost to John’ exists. First, the IC distributes [－r] to theme <a-1[af]> and
patient <b[caus]-2> separately. Second, by the CF, SUBJ and OBJ are selected for
<b[caus]-2>; same holds for <a-1[af]>. Even though theme and patient bear an even
status on the thematic hierarchy, a prominency hierarchy is built between the two due
to causativity assignment. Given that the causativity distribution requests that [caus] is
more prominent than [af], the UMP finally links the more prominent <b[caus]-2> to
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the least marked SUBJ and maps the less prominent <a-1[af]> to the second available
grammatical function in line.
The application of the second type of resultative compounds whose first verb is
transitive and whose second verb is intransitive is detailed in (89), (90), and (91)
below.
(89) a. Zhangsan zhui-lei-le Lisi.
John chase-tired-ASP Lee
‘John chased Lee to the extent of making Lee tired.’
b. ta shuai-po-le huaping.
he throw-break-ASP vase
‘He threw the vase and the vase was broken.’
c. ta ma-ku-le xiaohai.
he scold-cry-ASP children
‘He scolded the children and the children cried.’
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
i. <a, b-1> (41iii) (a = agent, b = patient)
IC [－r]
CF S/O/… S/O
UMP SUBJ OBJ V1 as head
ii. <a[caus], b-1[af]> (41iv) (a = agent, 1 = theme)
IC [－r]
CF S/O/… S/O
UMP SUBJ OBJ Headlessness
Resembling the analysis in transitive-transitive compounds, resulting in the same
result, (89i) and (89ii) under the LMT framework could account for the reading of
(89a) and (89b). Either <a> or <a[caus]> refers to agent; <b-1> or <b-1[af]>
represents patient/theme. As patient and theme, both <b-1> and <b-1[af]> receive [－r]
like the IC specifies. With compatible grammatical functions, the UMP predicts that
<a>/<a[caus]> automatically occupies SUBJ and <b-1>/<b-1[af]> surfaces as OBJ,
showing that the consequences of (89i) and (89ii) remain the same.
(89c) also holds identical argument structures with (89a) and (89b), containing a
causative reading. The process of generating the argument-function linking remains
the same. Nevertheless, one point is worth noticing in the following. Even though it is
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
observed that the single argument of ku ‘cry’ seems to be agent, it is claimed that such
agent arguments behave like theme when subcategorized by intransitive verbs (Huang
et al., 2009). Therefore, the unsuppressed theme-like <1> from Vres can be added a [af]
feature after it connects with <b> when the unsuppressed <a> from Vcaus receives a
[caus] feature. The argument structure (89ii) can be applied to (89c), explaining its
causative reading.
(90) a. Zhangsan zhui-lei-le Lisi.
John chase-tired-ASP Lee
‘John chased Lee and John got tired’
b. ta ting-ni-le zhe-shou ge.
he listen-bored-ASP this-CL song
‘He listened to this song (too often) and he got bored.’
<a-1, b> (41i) (a = agent, b = patient)
IC [－r]
CF S/O/… S/O
UMP SUBJ OBJ V1 as head
The argument structure in (90) matches sentences like (90a) and (90b). Since
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
<a-1> preserves the role of agent and <b> entails patient, <a-1> and <b> naturally
map to SUBJ and OBJ respectively, through the process from the IC to the UMP.
(91) a. Zhangsan zhui-lei-le Lisi.
John chase-tired-ASP Lee
‘Lee chased John and Lee got tired.’
b. zhe-zhong yao chi-si-le ren
this-CL medicine eat-dead-ASP people
‘People took this medicine and people were dead (by taking it).’
<a-1[af], b[caus]> (41ii) (1 = theme, b = patient/theme)
IC [－r] [－r]
CF S/O S/O
UMP OBJ SUBJ Headlessness
For a reversed argument-function mapping in (91), both <a-1[af]> and <b[caus]>
accept a patient/theme role, thus receiving a [－r] due to the IC. The CF also assigns
accept a patient/theme role, thus receiving a [－r] due to the IC. The CF also assigns