IDEA GROUP PUBLISHING ITJ3045osh/publication/A21-OT-LMT.pdf · Lexical Mapping Theory: A Case Study of Locative Inversion One-Soon Her, National Chengchi University, Taiwan ABSTRACT

Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.

International Journal of Technology and Human Interaction, 2(1), 67-94, January-March 2006 67

Optimality-TheoreticLexical Mapping Theory:A Case Study of Locative Inversion

One-Soon Her, National Chengchi University, Taiwan

ABSTRACT

Locative inversion verbs seem to share the same argument structure and grammatical functionassignment (i.e., <th-OBJ loc-SUBJ>) cross-linguistically. This article discusses the nature ofargument-function linking in LFG and demonstrates how the Lexical Mapping Theory (LMT)rendered in Optimality-Theoretic (OT) terms, where argument-function linking is governed byuniversal violable constraints that consistently favor the unmarked function, accounts forlocative inversion straightforwardly. Within this OT-LMT, locative inversion is due to a universalmorphosyntactic constraint, and language variation in locative inversion is due to the differencein its relative ranking. This account also offers a potential explanation for the markedness ofthe locative inversion construction.

Keywords: algorithms; case study; semantic matching; very high-level languages

INTRODUCTIONThe locative inversion construction,

as shown in Figure 1, cross-linguisticallyhas similar characteristics in discourse in-formation packaging, which allows themore familiar information to precede theless familiar information (Ackerman &Moore, 2001b; Birner, 1994; Cheng, 1983;Tan, 1991). Between the canonical con-struction in Figure 1a and the inverted form

of 1b, along with the switch of focus fromthe locative to the theme, is the change insyntactic function assignment. An examplefrom Chinese is given in the figure. Thetheme role in Figure 1a is assigned the sub-ject function and locative an oblique func-tion; the canonical linking is, thus, <th-SUBJ loc-OBL>. In the inverted Figure1b, however, the locative is the subject,while the theme now occupies the object

IDEA GROUP PUBLISHING

This chapter appears in the publication, International Journal of Technology and Human Interaction Volume 2, Issue 1edited by Bernd Carsten Stahl © 2006, Idea Group Inc.

701 E. Chocolate Avenue, Suite 200, Hershey PA 17033-1240, USATel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com

ITJ3045

http://www.pdfcomplete.com/1002/2001/upgrade.htm

68 International Journal of Technology and Human Interaction, 2(1), 67-94, January-March 2006


position (Her, 1990; Huang, 1993; Huang& Her, 1998; Tan, 1991).

This <th-OBJ loc-SUBJ> argument-function “mismatch” was first identified andconvincingly argued for in locative inver-sion verbs in Chichewa (Bresnan, 1994;Bresnan & Kanerva, 1989) and in English(Bresnan, 1989; Tan, 1991). Examples inFigure 2 are from Bresnan and Kanerva(1989).

The subjecthood of the inverted loca-tive phrase tai-shang “stage-top” in Figure1b is evidenced by the fact that it is a bareNP and occupies the usual position for sub-jects. This is further confirmed by the usualraising test. As shown in Figure 3, tai-shang“stage-top” is, indeed, the raised subject,while the “demoted” theme in the post-ver-

bal position, also a bare NP, must be recog-nized as the object (see Figure 3).

Chinese data thus further confirmBresnan’s (1994) observation that cross-linguistically, locative inversion verbs sharean identical argument structure <th loc>and the function assignment of the canoni-cal <th-SUBJ loc-OBL¸> in Figure 1a andthe inverted <th-OBJ loc-SUBJ> in Fig-ure 1b. This article aims mainly to accountfor the syntactic assignment of the argu-ment roles in locative inversion verbs.

In any syntactic theory that aims atcharacterizing UG, it would be a consider-able compromise to simply leave the syn-tactic assignment of argument roles to lexi-cal idiosyncrasies (Pesetsky, 1995). Thisarticle focuses on how the syntactic as-

Figure 1.

(1) a. Amei zuo zai tai-shang.Amei sit at stage-top

‘Amei is sitting on the stage.’

b. Tai-shang zuo-zhe Amei. stage-top sit-ASP Amei ‘On the stage is sitting Amei.’

a. A-lendo-wo ku-ba-bwer-a ku-mu-dzi. ( p.3 (2b) ) 2-visitor-2 those 17 SB-REC-PST-come-IND 17-3-villiage a’ ‘Those visitors came to the village.’

b. Ku-mu-dzi ku-ba-bwer-a a-lendo-wo. ( p.3 (1b) ) 17-3-villiage 17 SB-REC-PST-come-IND 2-visitor-2 those b’ ‘To the village came those visitors.’

Figure 2.

(1) a. Tai-shang kanqilai zuo-le henduo ren.stage-top appear sit-ASP many person‘On the stage appears to be sitting many people.’

b. Tai-shang you zuo-zhe henduo ren ma?stage-top YOU sit-ASP many person Q‘Is it the case that on the stage was sitting many people?’

Figure 3.




signment of argument roles is accountedfor universally in the syntactic theory ofLexical Functional Grammar (LFG). Thisarticle is organized as follows. The firstsection discusses how argument-functionlinking is accounted for by the lexical map-ping theory (LMT) in LFG. We will alsodemonstrate how certain versions of thistheory do not account for the locative in-version data from Chinese and Englishstraightforwardly. Furthermore, we willdemonstrate how the theory can be im-proved upon for more consistency and com-putational efficiency. In the second sectionwe propose a revised LMT formulated asdeclarative constraints in Optimality-Theo-retic (OT) terms. The locative inversiondata from Chinese is then accounted for inthe third section. The fourth section con-sists of a discussion on the implications ofthis study, and the fifth section concludesthe article.

The goal of the article is, thus, two-fold: (1) to come up with a universal lexicalmapping theory based on violable declara-tive constraints in OT terms; and (2) toaccount for Mandarin locative inversionwithin this comprehensive OT-LMT.

LEXICAL MAPPING THEORYLFG mainly posits three distinct, par-

allel planes of grammatical description: theargument structure, the functional structure,and the constituent structure (Bresnan,2001; Dalrymple, 2001; Falk, 2001). Theargument structure, or a-structure, consistsof the predicate’s thematic and non-the-matic argument roles, while the constitu-ent structure, or c-structure, represents theconfigurational structure, which is the sur-face structure and allows no syntactic deri-vation. The functional structure, or f-struc-ture, is the locus of grammatical informa-

tion, such as grammatical functions (e.g.,SUBJ and OBJ), case, person, number,gender, and so forth. The linking of thesestructures, each with a distinct formal na-ture, is constrained by correspondence prin-ciples. The lexical mapping theory (LMT)is the UG component that constrains thelinking between a-structure roles and f-structure functions. The f-structure thuscan be viewed as the interface level thatlinks the a-structure and the c-structure.An argument role thus is linked to a gram-matical function in the f-structure, which,in turn, is linked to a certain c-structure con-figuration. The lexical mapping theory(LMT) is the subtheory within LFG, whichconstrains the syntactic assignment of a-structure roles.

The pioneering work by Levin (1987)started the exploration of more principledaccounts to replace the earlier stipulatedfunction-changing rules in LFG. The firstcomprehensive formulation of LMT wasproposed in Bresnan and Kanerva (1989).Since then, even though the essential un-derpinning assumptions have remainedlargely stable, the issue of argument-func-tion linking, especially its precise formula-tion, has yet to be resolved (Butt & King,2000). A number of different versions ofthe theory have been proposed (Ackerman,1992; Ackerman & Moore, 2001a; Alsina,1996; Bresnan, 1989, 2001; Butt, Dalrymple& Frank, 1997; Her, 1998; Huang, 1993;Zaenen, 1987), among others. A review ofthese existing versions is clearly outside thescope of this article1. Instead, we will out-line mainly the version that seems to be themost widely circulated, found in Chapter14 of Bresnan (2001), which, in turn, isbased largely upon Bresnan and Zaenen(1990)2.




The Theory of A-StructureConceptually, LMT consists of two

components: the theory of a-structure andthe mapping constraints. LFG assumes auniversal hierarchy among a-structure rolesin terms of their relative prominence in theevent denoted by the predicate. This scaledescends from the most prominent agent roleto the least prominent locative role (Bresnan& Kanerva, 1989, 1992)3 (see Figure 4).

The most prominent role in an a-struc-ture is called the “logical subject” and isdesignated Ô (pronounced “theta-hat”). InFigure 5, the two-place predicate breakrequires two argument roles in a-structure,agent (also Ô) and theme; the three-placepredicate put requires agent (again the Ô),goal, and theme. Roles in a-structure, byconvention, descend in prominence accord-ing to the thematic hierarchy.

Grammatical functions (GFs) that canbe linked to argument roles are called argu-ment functions. In Figure 6, LFG distin-guishes argument functions (shown in bold)from non-argument functions (in italics).

It is important to note that in struc-ture-oriented theories, such as Transforma-tional Grammar (TG) and all its later incar-nations, notions such as subject and objectare secondary and are derived from struc-tural configurations. In contrast, in relation-oriented theories, such as Relational Gram-mar (RG) and LFG, these are primary no-tions in syntax. However, in LFG, argumentfunctions are further decomposed by twobinary features: [r] (whether the functionis restricted to having an argument role)and [o] (whether the function is objective)(see Figure 7).

In this system, each argument func-tion is composed of exactly two featuresand natural classes can be identified, asshown in Figure 8. Furthermore, assumingthe minus feature to be the unmarkedvalue, a markedness hierarchy also canbe obtained.

Similar to the intrinsic classificationof argument roles in Bresnan and Kanerva(1989), Bresnan (2001) assumes that theunderlying lexical semantics partially de-

Figure 4. Thematic hierarchy

(5) a. break < x y > (x = ag, y = th)b. give < x y z > (x = ag, y = go, z = th)

Figure 5.

TOP FOC SUBJ OBJ OBJ OBL 4 ADJUNCTS

Figure 6.

-r +r-o SUBJ OBL+o OBJ OBJ

[+r] = (un)restricted [+o] = (un)objective

Figure 7. Feature decomposition of argument functions

agent > beneficiary > experiencer/goal > instrument >patient/theme > locative




termine the syntactic assignment of differ-ent event participants. The universal clas-sification shown in Figure 9 is proposed tocapture these predetermined choices ofgrammatical function assignment.

Cross-language variation in the syn-tactic assignment of a-structure roles is thussubject to the above universal constraints.The agent role, for example, as a non-patientlike role, is classified [-o] by Figure9c and is thus not associated with OBJ ca-nonically. Patient and theme roles, with the[-r] classification, are associated canoni-cally with either SUBJ or OBJ. Under theassumptions in Figure 9, each role in the a-structure is assigned one and only one fea-ture for syntactic function assignment, asmorpholexical processes are not allowedto add syntactic features. Language-spe-cific morpholexical operations are allowed,however, to alter the “lexical stock” of ana-structure by adding, suppressing, or bind-

ing thematic roles (Bresnan, 2001).Passivization, for example, suppresses Ô,the most prominent role, from syntacticassignment (see Figure 10).

In summary, the theory of a-structurerenders the argument roles a given predi-cator requires into an a-structure represen-tation, where roles are listed in a descend-ing order in prominence, and each role isassigned exactly one feature specificationfor function assignment. The second com-ponent in LMT (i.e., the universal set ofmapping constraints) then determines ex-actly which GF each role is assigned to.

Mapping PrinciplesArgument-function linking is subject

to certain universal constraints; otherwise,each argument role is freely mapped ontoany and all GFs with compatible features.Bresnan (2001) proposes the principlesshown in Figure 11.

Figure 8. Markedness hierarchy of argument functions

SUBJ[-r –o] > OBJ[-r +o]/OBL [+r –o] > OBJ [+r +o]

Figure 9. Semantic classification of a-structure roles for function(5) Semantic Classification of A-Structure Roles for Function

a. patientlike roles: [-r] b. secondary patientlike roles: [+o] c. other semantic roles: [-o]

Figure 10. Passivization

< … >↓∅

Figure 11. Mapping principles (MPs)Mapping Principles (MPs)a. Subject roles: (i) Ô[-o] is mapped onto SUBJ when initial in the a-structure; Otherwise,

(ii) [-r] is mapped onto SUBJ.b. Other roles are mapped onto the lowest compatible function in the

Markedness Hierarchy.




Two more well-formedness conditions(WFs) are needed in addition to the map-ping principles in order to further constrainthe non-deterministic argument-functionlinking: the function-argument biuniquenessconstraint and the subject condition:

• Function-Argument Biuniqueness. Eacha-structure role must be associated witha unique function, and conversely.

• The Subject Condition. Every predica-tor must have a subject.

The function-argument biuniquenessconstraint ensures a strict one-to-one map-ping relation between roles and functions.Computationally, it forces a deterministicassignment to an “unattached” GF betweenthe two GFs with which a role is compat-ible. The subject condition serves the obvi-ous purpose to ensure that one role in a-

structure must be mapped to SUBJ. Thiscondition also forces a deterministic choicewhen a role’s syntactic assignment is com-patible with SUBJ and some other func-tion and when all other roles in the a-struc-ture, if any, are incompatible with SUBJ.

We demonstrate how three differenttypes of verbs receive correct argument-function linking in the LMT just described.An unaccusative verb is given in Figure 12,while an unergative verb is illustrated inFigure 13. A typical transitive verb is givenin Figure 14.

Improvement to theConventional LMT

There are several areas at the theo-retical level upon which the conventionalLMT may be improved. First, the uniformunderspecification of each role with exactlyone syntactic feature can be relaxed to al-

Figure 12.

Figure 13.

Bing hua le.ice melt ASP‘The ice has melted’

‘melt < x >’ (x = th) SC: [-r]

---------------S/O

MPs: S WFs: S

Mama xiao le.Mama laugh ASP

‘Mama laughed.’

‘laugh < x >’ (x = ag) SC: [-o]

--------------- S/OBL

MPs: S WFs: S




low the formalism to be more expressive,yet without compromising its formal power.This classification scheme also may be toorigid in that it does not allow the possibilityof agentive objects, which have been ob-served in several languages (Bresnan,2001; Dalrymple, 2001). A desirable im-provement to the theory is to allow suchlinking possibilities and, at the same time,be able to express the marked nature ofsuch a linking as agent-OBJ.

As for the mapping principles, twodisjunctions are observed. The first one isin the mapping principles of subject roles: adisjunction exists between Ô[-o] and θ[-r],each a stipulation for linking to SUBJ. In amore general theory of UG, it would be de-sirable not to include such function-specificlinking conditions. Notice also the specifica-tion that Ô[-o] be the initial role in the a-struc-ture. This principle thus must refer explicitlyto the ordering in the a-structure5. The sec-ond disjunction is found between subject rolesand non-subject roles. For the former, a quali-fied role is mapped to SUBJ (i.e., the mostprominent GF). However, on the contrary,non-subject roles must be linked to the leastprominent compatible GF. A consistent prin-ciple for all roles would make a simpler andmore general theory.

Finally, note that the Subject Condi-tion in LFG states explicitly that every

clause must have a subject. Similar con-straints are also necessary in other syntac-tic frameworks; for example, the same isaccomplished by the Extended ProjectionPrinciple (EPP) in Transformational Gram-mar and the Final-1 Law in RelationalGrammar. However, as it often has beennoted, such an inflexible stipulation may notbe empirically accurate6. As cited inAckerman and Moore (2001a), clausesmay truly be without a subject (Babby,1989; McCloskey, 2001). Bresnan (2001)thus hinted that this condition should per-haps be stipulated as a parameter. Again,ideally, a UG theory should be able to ac-count for such subjectless clauses and theirmarked nature at the same time.

Assuming that the conventional LMTtakes the same position advocated first inAlsina and Mchombo (1993) and does notallow morphological operations to add fea-tures, this version of LMT described pre-viously also does not seem to account forlocative inversion straightforwardly.

As shown in Figure 16, the argument-function linking of <th-OBJ loc-SUBJ> inthe locative inversion construction cannotbe obtained, even though the canonical link-ing of <th-SUBJ loc-OBL¸> is accountedfor in Figure 16. Therefore, it would makesense empirically to allow morphologicalprocesses in the theory to alter syntactic

Figure 14.

Amei mai changpian.Amei sell record’Amei sells records.’

‘sell < x y >’ (x = ag, y = th) SC: [-o] [-r]

------------------ S/OBL S/O

MPs: S O WFs: S O




assignments by adding features, as pro-posed in Zaenen (1987), Ackerman (1992),Markantonatou (1995), and Her (1998,2003). The default locative classificationemployed by Bresnan and Kanerva (1989)and Bresnan (1989), which assigns loc[-r]when th is focused, likewise can be viewedas such a feature-adding morphologicaloperation. In the fourth section, we alsowill discuss the advantages of feature-add-ing morphological operations from thestandpoint of expressivity and formalpower.

In the second section, we will pro-pose an LMT in OT terms, thus an OT-LMT, that attempts to incorporate the de-sirable improvements suggested here.

AN OPTIMALITY-THEORETIC LMT

Optimality Theory has exerted greatinfluence over the field of phonology; how-ever, its application in syntactic theory is

still in its infancy. Recently, there have beensome explorations within the OT-LFGframework (aka Optimal Syntax [Bresnan,2000]). From the OT point of view, OT-LFG can be seen as OT with a universalLFG as GEN. From the point of view ofLFG, a constraint-based grammaticalframework, generalizations are interpretedin OT terms with (violable) constraintsranked in relation to one another (Sells,2001). A number of studies have been car-ried out within this general framework(Mikkelsen, 2003; Sells, 2001). There alsohave been efforts to render argument-func-tion linking in OT terms (Butt et al., 1997;Lødrup, 1999).

An OT-LFG OverviewBresnan (2000) depicts the basic

structure of OT-LFG, or Optimal Syntax,where LFG’s correspondence theory ofparallel structures serves as a model forGEN. The standard OT-LFG assumes in-

Figure 15.

Figure 16.

Amei zuo zai tai-shang.Amei sit at stage-top‘Amei is sitting on the stage.’

‘sit < x y >’ (x = th, y = loc) SC: [-r] [-o]

------------------ S/O S/OBL

MPs: S OBL WFs: S OBL

(5) Tai-shang zuo-zhe Amei.stage-top sit-ASP Amei‘On the stage is sitting Amei.’

‘sit < x y >’ (x = th, y = loc SC: [-r] [-o]

------------------*O *S




put to be “a (possibly underspecified) fea-ture structure representing some givenmorphosyntactic content independent of itsform of expressions” (Bresnan, 2000). Anexample is given in Figure 17, which as-sumes I saw her as its optimal form of ex-pression. Note that in the input structure,<x, y> is the a-structure of see and GF1and GF2 are unspecified grammatical func-tions that argument roles with which x andy are associated.

The candidate set comprises pairs off-structure and corresponding c-structure(and perhaps other corresponding planesof information) generated by the LFGgrammar (Bresnan, 2000; Kuhn, 2001). Forease of presentation, I am simplifying thematter by taking the input to be an a-struc-ture <x y>, and a set of <x-GF1 y-GF2>pairs as candidates in OT-LMT, which is amodule within OT-LFG that constrains ar-gument-function linking specifically. Thecandidates are evaluated by a universal setof lexical mapping constraints. The outputis taken to be the most harmonic, or opti-mal, candidate pair; namely, the one withthe least (serious) violations (Kuhn, 2001).

A Comprehensive OT-LMTThe OT-LMT proposed here modi-

fies and expands the LMT component inBresnan (2001) and is based specificallyon the particular formulation of LMT in Her(1997), Huang and Her (1998), and Her(2003), where syntactic feature assign-ments are simplified, and the multiple map-ping principles and well-formedness con-ditions in the conventional LMT are all uni-fied into a single consistent mapping prin-ciple. Here, I will take this further and re-interpret the entire simplified LMT as a setof Optimality-Theoretic constraints and thusoffer a comprehensive OT-LMT.

Crucial to the theory are two promi-nence scales discussed earlier: a universalthematic hierarchy and a markedness hi-erarchy of grammatical functions (GFs)(see Figures 18 and 19).

Mapping constraints are classified intothree categories: well-formedness con-straints on argument roles, well-formednessconstraints on argument functions, and con-straints on linking. Note that we are ignor-ing athematic arguments in this article7. “R”is thus a thematic role in a-structure, and“F” is a corresponding grammatical func-

Figure 17.

Input f-structure: I saw her.

PRED ‘see <x, y>’

PRED ‘PRO’GF1 PERS 1 NUM SG x

PRED ‘PRO’GF2 PERS 3 NUM SG GEND FEM y

TNS PAST




tion. We first examine the well-formednessconstraints on the representation of argu-ment roles (see Figure 20).

UniqRol ensures the uniqueness ofeach and every role in the a-structure andthus rules out a-structures like <ag ag th>and <th loc loc>. DescendRol further for-malizes the a-structure representation,where argument roles descend in promi-nence. For example, given the locative verbsit and its two roles, theme and locative, ina-structure, <th loc> is the only well-formedrepresentation; <loc th> is ill-formed. Twocorresponding constraints are proposed forargument functions (see Figure 21).

UniqFun ensures the uniqueness ofeach and every function in the a-to-f map-ping; thus, both of the following are ill-formed:<θa-SUBJ θb-SUBJ>, and <θa-OBJ θb-OBJ>. DescendFun penalizes a candidatewith a violation of the descending order inprominence. For example, because SUBJoutranks OBJ, <θa-SUBJ θb-OBJ> has 0violation and is favored over the inverted<θa-OBJ θb-SUBJ>, which incurs one vio-lation. Thus, given n GFs in a candidate form,there are at most n – 1 violations as there

are n – 1 consecutive pairs (Kuhn, 2001).Inversion is still possible, given that all OTconstraints are violable in order to satisfyhigher-ranked constraints, including lan-guage-specific morphosyntactic operations.(We will discuss the possibility of a language-specific component in the next section.) Next,we move on to the general constraints onthe linking between roles and functions.

The two constraints in Figure 22,LinkRol and LinkFun, ensure that eachexpressed role is linked to a GF and thateach GF is linked to a role. A role that isnot linked to an argument function causesincompleteness, while an argument func-tion that is not linked to an argument role ina-structure causes incoherence. Notice thatthere is no need to specify a constraint justto ensure that a role is linked to a GF withcompatible features. This is accomplishedautomatically by the universal constraintson the morphosyntactic properties of argu-ment roles shown in Figure 23.

LinkPtTh reflects the unaccusativehypothesis that cross-linguistically the pri-mary patient/theme is encoded as an unre-stricted [-r] GF (i.e., SUBJ or OBJ)

(5) ag > ben > go/exp > inst > pt/th > loc

Figure 18.

Figure 19.

SUBJ[-r –o] > OBJ[-r +o]/OBL [+r –o] > OBJ [+r +o]

Figure 20. Well-formedness constraints on argument roles

Figure 21. Well-formedness constraints on grammatical functions

Well-formedness Constraints on Argument Rolesa. UniqRol(Ra, Rb): Given <..Ra-Fa..Rb-Fb..>, Ra Rb

b. DescendRol(Ra, Rb): Given <.. Ra-Fa Rb-Fb..>, Ra > Rb in prominence

Well-formedness Constraints on Grammatical Functionsa. UniqFun(Fa, Fb): Given <.. Ra-Fa..Ra-Fb..>, Fa Fb

b. DescendFun(Fa, Fb): Given <.. Ra-Fa R-Fb..>, Fa > Fb in prominence




(Bresnan & Kanerva, 1989; Bresnan &Zaenen, 1990; Zaenen, 1993)8. LinkRolRescaptures the generalization that a non-pa-tient/theme internal argument prefers thesyntactic assignment of a thematically re-stricted function. Finally, LinkUnobj andLinkUnres consistently favor the assign-ment of a role to the most unmarked func-tion, SUBJ, [-r –o]9. Each function thusmay have zero to two violations. These twoconstraints together are more general andinsightful than the previous Subject Condi-tion, which simply stipulates that everyclause should have a subject.

Note that LinkRolRes does not applyto agent, the external argument. Being thehighest-ranked role, it is linked to SUBJ dueto LinkUnobj and LinkUnres. This thusaccounts for the fact that, for the majorityof the world’s languages, agent cannot berealized as an object. However, given theviolable nature of these constraints and their

variable ranking, the possibility of agent-OBJdoes exist as a marked morphosyntacticoption. This reflects the insight of Falk (1989)cited in Lødrup (2000) that in Norwegian,“what has been called external theta rolesare in fact structurally unspecified thetaroles” (p. 173).

I will follow the standard view inOT and assume that these constraints areuniversal, but their ranking may be lan-guage-specific. For Chinese, I propose theranking shown in Figure 24.

An Illustration of OT-LMTWe will now look at the lexical map-

ping of three different verbs in their ca-nonical active construction as examples:“melt<th>,” “laugh<ag>,” and “sell<agth>.” To save time and space, the (many)candidates that violate any of the five high-est-ranked well-formedness constraints willbe excluded and we only will be concerned

Figure 22. General constraints on argument-function linking

a. LinkRol(R, F): Given <..R..>, R is linked to an F such that <..R-F..>.b. LinkFun(F, R): Given <..F..>, F is linked to an R such that <..R-F..>.

Figure 23. Specific constraints on argument-function linkingSpecific Constraints on Argument-Function Linking

a. LinkPtTh(R, F): Given <..R-F..>, where R = pt/th, F is [-r]b. LinkRolRes(R, F): Given <..R-F..>, where R Ô, F is [+r]c. LinkUnobj(R, F): Given <..R-F..>, F is [-o]d. LinkUnres(R, F): Given <..R-F..>, F is [-r]

Figure 24. OT ranking of lexical mapping constraints (Chinese)OT Ranking of Lexical Mapping Constraints (Chinese)

UniqRol/DescendRol/UniqFun/LinkRol/LinkFun>>

LinkPtTh>>

LinkRolRes>>

DescendFun>>

LinkUnobj/LinkUnres




with the lower five. Following standard OTnotation, in Figure 25, a violation is markedwith “*”; a “fatal” violation causing a can-didate to lose in evaluation is highlighted with“!”. The shaded area covers the constraintsthat are no longer relevant in the evaluationof a particular candidate, and finally, thesign indicates the optimal selection.

The candidate C1, <th-SUBJ>, withno violation, is clearly the optimal selec-tion, where SUBJ is an unrestricted [-r]function allowed by LinkPtTh, and also theunmarked [-r –o] function preferred byLinkUnobj and LinkUnres.

Next, we turn to the a-structure ofan unergative verb “laugh<ag>” (see Fig-ure 26). Here, the only relevant constraintsare LinkUnobj and LinkUnres, which againselect SUBJ, the unmarked function.

Again, the candidate with no viola-tion (C1) is the optimal selection.

The final example, sell, is a transitiveverb with an agent role and a theme role.

Again, Figure 27 excludes candidates thatviolate any of the five highest-ranked con-straints.

Among the candidates, C1, <ag-SUBJ th-OBJ>, is the optimal selection,even though it does violate one of the twolowest-ranked constraints (i.e., LinkUnobj)due to the linking of theme to OBJ, a func-tion with the marked feature [+o]. All othercandidates, however, violate at least onehigher-ranked constraint. Note that a can-didate a-structure where both roles arelinked to the unmarked function, thus <ag-SUBJ th-SUBJ>, violates the highestranked UniqFun and, therefore, is not in-cluded in the figure.

AN OT-LMT ACCOUNTOF LOCATIVE INVERSIONIN CHINESE

We first apply the OT-LMT to thecanonical a-structure of the locative verb.

Figure 25. Input a-structure: ‘melt <th>’

Candidate LinkPtTh LinkRolRes DescendFun LinkUnobj LinkUnres

C1 <th-SUBJ>

C2 <th-OBJ> *!

C3 <th-OBL > *! * *

C4 <th-OBJ > *! * *

Figure 26. Input a-structure: ‘laugh <ag>’


C1 <ag-SUBJ>

C2 <ag-OBJ> *!

C3 <ag-OBL > *!

C4 <ag-OBJ > *! *




The theory correctly predicts the followingoptimal argument-function linking: <th-SUBJ loc-OBLθ>, shown in Figure 28.

With C2, <th-SUBJ loc-OBLθ>, asthe optimal selection, this constraint rank-ing obviously does not account for locativeinversion, which is represented by candi-date C4, <th-OBJ loc-SUBJ>. An addi-tional constraint is needed.

The constraint shown in Figure 29draws on the insight found in the defaultrule for focused theme posited by Bresnanand Kanverva (1989) and also faithfully re-flects Bresnan’s (1994) observation on theuniversals of locative inversion verbs. Giventhe fact that the complement of the predi-cator usually carries the discourse func-tion of marking the less familiar informa-tion and that the subject is the default gram-matical function for topic or more familiarinformation, the locative inversion opera-tion forces the locative to map onto SUBJsuch that the focused theme can surface

as a complement of the locative verb. Withthis constraint in place, we now have animportant decision to make; that is, whetherto posit LinkLocInv as a language-specificconstraint. Recall that locative inversion isfound in many languages, and locative in-version verbs share an identical a-struc-ture and function assignment. However,locative inversion certainly does not occurin all languages. In non-configurational lan-guages with extensive case-marking forgrammatical relations (e.g., Korean andJapanese), locative inversion may not befound (Huang & Her, 1998). Figure 30 de-picts a Japanese example.

Notice that even though the locativephrase indeed may invert positions with thesubject and thus affect the focus in Figure30b, its grammatical functions remain thesame. In other words, locative inversiondoes not affect argument-function linking.Recall the standard OT view that con-straints are universal and that only their

Figure 27. Input a-structure: ‘sell <ag th>’


C1 <ag-SUBJ th-OBJ> *

C2 <ag-SUBJ th-OBL > *! *

C3 <ag-SUBJ th-OBJ > *! * *

C4 <ag-OBJ th-SUBJ> *! *

C5 <ag-OBJ th-OBL > *! * * *

C6 <ag-OBJ th-OBJ > *! * * *

C7 <ag-OBL th-SUBJ> *! *

C8 <ag-OBL th-OBJ> *! *

C9 <ag-OBL th-OBJ > *! * *

C10 <ag-OBJ th-SUBJ> *! * *

C11 <ag-OBJ th-OBJ> *! * * *

C12 <ag-OBJ th-OBL > *! * * * *




ranking is subject to variation. Therefore,if we follow the standard OT view and positLinkLocInv as a universal constraint, lan-guages such as Japanese also must be ac-counted for, but only with a different rank-ing of the same constraints. This is the pathwe will explore. Figures 31 and 32 show

the revised ranking we propose for Chi-nese. Notice that LinkLocInv outranksLinkRolRes and is outranked by LinkPtTh.Again, we continue to ignore the five high-est-ranked well-formedness constraints.

We also need to point out thatLinkLocInv is irrelevant in the selection of

Figure 28. Input a-structure: ‘sit <th loc>’


C1 <th-SUBJ loc-OBJ> *! *

C2 <th-SUBJ loc-OBL > *

C3 <th-SUBJ loc-OBJ > *! *

C4 <th-OBJ loc-SUBJ> *! * *

C5 <th-OBJ loc-OBL > *! *

C6 <th-OBJ loc-OBJ > *! *! *

C7 <th-OBL loc-SUBJ> *! * * *

C8 <th-OBL loc-OBJ> *! * * *

C9 <th-OBL loc-OBJ > *! * * *

C10 <th-OBJ loc-SUBJ> *! * * * *

C11 <th-OBJ loc-OBJ> *! * * * * *

C12 <th-OBJ loc-OBL > *! * * * *

Figure 29.

LinkLocInv(R, F): Given a-structure <Ra–Fa Rb–Fb>, where Ra = th[foc]and Rb = loc, Fb is [-r -o].

Figure 30.

a. Herikoputa ga yama no ue ni orimashita.helicopter NOM mountain POSS top LOC land‘A helicopter landed on top of the mountain.’

b. Yama no ue ni herikoputa ga orimashita.mountain POSS top LOC helicopter NOM land‘On top of the mountain landed a helicopter.

c.*Yama no ue ga herikoputa o orimashita.mountain POSS top NOM helicopter ACC land

‘On top of the mountain landed a helicopter.’




the canonical in Figure 28, because there,the theme is not focused. Now, to accountfor the data from languages like Japanese,where the focused theme does not resultin mismatches of the function assignmentof argument roles, we posit the ranking inFigures 33 and 34. Notice here thatLinkLocInv is outranked by all other con-straints.

DISCUSSIONThis section discusses three issues in

further detail. The first issue relates to the

nature and the scope of the OT-LMT pro-posed in the article. The second issue con-cerns the potential advantages that the OT-LMT may have over the conventionalLMT. Finally, we explore some of the di-rections for further research concerning theOT-LMT.

Morphosyntactic vs.Morpholexical Processes

Given the often idiosyncratic natureof language-specific lexical information, itis not yet clear how the technical integra-

Figure 31. OT ranking of lexica mapping constraints (Chinese, revised)

LinkPtTh>>

LinkLocInv>>

LinkRolRes>>

DescendFun>>

LinkUnobj/LinkUnres

Figure 32. Input a-structure: ‘sit <th[foc] loc>’ (Chinese)

Candidate LinkPtTh LinkLocInv LinkRolRes DescendFun LinkUnobj LinkUnres

C1 <th-SUBJ loc-OBJ> *! * *

C2 <th-SUBJ loc-OBL > *! *

C3 <th-SUBJ loc-OBJ > *!* * *

C4 <th-OBJ loc-SUBJ> * * *

C5 <th-OBJ loc-OBL > *! * *

C6 <th-OBJ loc-OBJ > *!* * * *


C8 <th-OBL loc-OBJ> *! * * * *

C9 <th-OBL loc-OBJ > *! ** * * *


C11 <th-OBJ loc-OBJ> *! * * * * * *

C12 <th-OBJ loc-OBL > *! * * * * *




tion of the lexicon should be envisaged inOT syntax, in general (Kuhn, 2001). Thisarticle clearly does not address this largerissue. In order to have an insightful lexicalmapping theory in OT syntax, we first mustbe explicit about its nature and scope. TheOT-LMT envisioned here is part of a uni-versal OT-LFG theory that constrains ar-gument-function linking. In other words, itconstrains the syntactic function assignmentof argument roles required by a predicator.Thus, this OT-LMT, as it is currently for-mulated, has nothing to say about

morpholexical processes that alter the“lexical stock” in a-structure (Bresnan,2001; Bresnan & Kanerva, 1989). There-fore, it is purely morphosyntactic in natureand scope. Crucially, Ackerman (1992) dif-ferentiates and characterizes morpholexicaland morphosyntactic operations as follows:

Morpholexical (Operations), affect thelexical semantics of predicates by alteringthe semantic properties associated withp r e d i c a t e s … M o r p h o s y n t a c t i c(Operations), assign features supplemental

Figure 33. OT ranking of lexical mapping constraints (Japanese, Korean, etc.)

LinkPtTh>>

LinkRolRes>>

DescendFun>>

LinkUnobj/LinkUnres>>

LinkLocInv

Figure 34. Input a-structure: ‘land <th[foc] loc>’ (Japanese)

Candidate LinkPtTh LinkRolRes DescendFun LinkUnobj LinkUnres LinkLocInv

C1 <th-SUBJ loc-OBJ> *! * *

C2 <th-SUBJ loc-OBL > * *

C3 <th-SUBJ loc-OBJ > *! * **

C4 <th-OBJ loc-SUBJ> *! * *

C5 <th-OBJ loc-OBL > *! * *

C6 <th-OBJ loc-OBJ > *! *! * **


C8 <th-OBL loc-OBJ> *! * * * *

C9 <th-OBL loc-OBJ > *! * * * **


C11 <th-OBJ loc-OBJ> *! * * * * * *

C12 <th-OBJ loc-OBL > *! * * * * *




to those supplied by IC assignment10: theseoperations can affect the final GFassignments to arguments but cannot affectthe lexical semantics. (p. 56)

Morpholexical operations thus areword-formation processes that producepredicates with an altered inventory of ar-gument roles, or a-structures, which serveas input to OT-LMT. Morphosyntactic op-erations, however, are within the proper do-main of LMT. Assuming that onlymorpholexical operations may be language-specific, the OT-LMT proposed thus uni-versally governs how argument roles aremapped to GFs, with constraints that mayvary from language to language only in termsof ranking. Thus, as Huang and Her (1998)have argued, given the nature of syntacticassignment of argument roles in the theory,it, in fact, makes the theory more coherentby allowing syntactic feature assignment inmorphosyntcatic operations. This is pre-cisely how we treated locative inversion.Similar proposals that allow morphologicalprocesses to affect syntactic assignmentsby adding features are found as early asZaenen (1987) and Her (1990) and later inAckerman (1992), Markantonatou (1995),Her (2003), among others.

Allowing feature-adding morphosyntacticoperations, in fact, also offers a computationaladvantage. Morpholexical operations consti-tute a much more powerful formal devicecomputationally in that they are not subject tothe general monotonicity condition that infor-mation only can be added but cannot bedeleted or changed (Bresnan, 1990; Falk,2001)11. Monotonic morphosyntactic op-erations with the feature-adding capacityenable a formalism that is more consistentand also more expressive, without any in-crease in its formal power. Empirically, such

operations also have been adopted to accountfor syntactic variations in several languages;for example, Greek (Markantonatou, 1995),Chinese (Huang, 1995; Her, 1999), and En-glish (Zaenen, 1987).

We will now illustrate this view of theOT-LMT with two more constructions fromChinese that are related to locative verbs.The first one is a passivized locative con-struction. Three-place transitive predicateslike xie “write,” with the argument struc-ture <ag th loc>, do not allow inversion inspite of the locative role it requires. How-ever, there may be locative inversion if theagent role is suppressed. This is observedin Chinese (Huang & Her, 1998) and otherlanguages (Bresnan, 1989; Bresnan &Kanerva, 1989). The examples in Figures35 and 36 are from Chinese and English.

Recall that passivization, repeated inFigure 36, suppresses the logical subject.In effect, it gives rise to an argument struc-ture <ag th loc>, precisely that of a loca-tive inversion verb. Locative inversion,therefore, is allowed, as in Figure 35c.Passivization thus falls outside of the realmof LMT and is regarded as a language-specific operation. One indication of its lan-guage-dependence is in the indirect expres-sion of the suppressed agent role as an ad-junct; for example, the English by-expres-sion (Bresnan, 1994). Chinese, however,does not allow such indirect expressions12.

The second construction we will ex-amine is the transitivized locative verb. Ithas been noted that, in Chinese, a two-placelocative verb with an argument structure<th loc> in fact allows its locative phraseto be a PP or an NP. The locative phrasethus may be alternatively mapped ontoOBL or OBJ¸ (Huang & Her, 1998). This,however, is not allowed in English, as shownin Figures 37 and 38.




Huang and Her (1998) treat this func-tion change as a morphosyntactic variationof the same argument structure; thus, <th-SUBJ loc-OBL¸> in Figure 37 and <th-SUBJ loc-OBJ> in Figure 38. However,further evidence indicates that this viewmay be incorrect and that locativetransitivization involves a morpholexicalchange, instead. In other words, without

the preposition zai, the argument structureis, in fact, no longer <th loc>. Note thatthe presence of the locative preposition zairequires a place noun as its complement inFigure 39. In Chinese, certain nouns areplace nouns inherently, such as xuexiao“school,” zheli “here,” and gongyuan“park,” and thus can be the complement ofpreposition zai directly. Non-place nouns,

Figure 37.

a. Amei zuo zai yizi-shang.Amei sit at chair-top‘Amei sits on the chair.’

b. Amei shui zai diban-shang.Amei sleep at floor-top‘Amei sleeps on the floor.’

Figure 38.

a. Amei zuo yizi-shang.Amei sit chair-top‘Amei sits *(on) the chair.’

b. Amei shui diban-shang.Amei sleep floor-top‘Amei sleeps *(on) the floor.’

Figure 35.

a. Amei xie le yi ge zi zai qiang-shang.Amei write ASP a CL character at wall-top‘Amei wrote a Chinese character on the wall.’

b*Qiang-shang xie le yi ge zi Amei.wall-top write ASP a CL character Amei

*’On the wall was written a Chinese character (by) Amei.’

c. Qiang-shang xie le yi ge zi.wall-top write ASP a CL character

‘On the wall was written a Chinese character.’

Figure 36. Passivization

< … >↓∅




however, must form a constituent with alocative affix such as –shang and -xia, ora locative noun such as shangmian andxiamian; or there must be a place noun asthe complement of locative preposition zai.Notice in Figure 40 that the object requiredby the transitivized verb zuo “sit” and shui“sleep” is free of this restriction.

Therefore, it is clear that the objectsin Figure 40 do not denote the locationwhere the theme that undergoes the move-ment ends up; rather, they are the entitiesthat receive the action denoted by the verbs.To account for this construction, I proposea morpholexical operation (see Figure 41).

Two more syntactic tests, shown inFigures 42 and 43, confirm that this argu-ment structure is now <ag th>:passivization and resultative compounding.

In the passive construction, the sup-pression of the agent results in the themerole’s “promotion” to SUBJ, as seen in Fig-ure 42. In Figure 43, the single composite

role, formed by the binding of the themerole of the action verb and the theme ofthe result state verb, maps to SUBJ (Her,2004)13. Based on the prevailing evidence,locative transitivization should be treatedas a morpholexical operation that alters thelexical stock of an argument structure, and,as such, it is again outside of the realm ofthe OT-LMT proposed here.

According to the previous discussion,it is now possible to indicate exactly howthe OT-LMT system is envisaged as themodule in LFG that links the lexical semanticstructure and the syntactic structure of apredicator (Bresnan & Kanerva, 1989;Bresnan & Zaenen, 1990). The particularconceptualization of the a-structure as-sumed here, as shown in Figure 44, is basedon Bresnan (1996, 2001), which, in turn,follows Baker (1983).

The a-structure is a lexical syntacticrepresentation with the minimally neces-sary information on the syntactic arguments

Figure 39.

a. Amei zuo zai yizi-*(shang).Amei sit at chair-top‘Amei sits on the chair.’

b. Amei shui diban-*(shang).Amei sleep floor-top‘Amei sleeps on the floor.’

Figure 40.

a. Amei zuo yizi.Amei sit chair‘Amei sits *(on) the chair.’

b. Amei shui diban.Amei sleep floor‘Amei sleeps *(on) the floor.’

Figure 41. Locative transtivization

<th loc> <ag th>




of a predicator, such as the number of ar-guments, their thematic and syntactic types,and their hierarchical organization. Asshown in Figure 44, the a-structure “sink<ag pt>” states that the verb sink requirestwo arguments, one of the type agent andthe other theme, and also that agent is the-matically more prominent than theme. Thea-structure thus contains information nec-essary for the final syntactic manifestationor more precisely, the mapping of agentand theme to SUBJ and OBJ, respectively.

Morpholexical operations interact spe-cifically with lexical semantics and, as such,are outside the proper domain of the LMT,while morphosyntactic operations are partof the LMT, which constrain the syntacticassignment of a-structure roles. All OT-LMT constraints thus are conceived to bemorphosyntactic and universal in nature,while morpholexical operations may be lan-guage-specific.

a. Yizi bei zuo le.Chair BEI sit ASP‘The chair has been sat *(on).’

b. Diban bei shui le.Floor BEI sleep ASP‘The floor has been slept *(on).’

Figure 42.

a. Yizi zuo-lan le.Chair sit-threadbare ASP‘The chair is threadbare from (over)sitting.’

b. Diban shui-kua le.Floor sleep-collapse ASP‘The floor was slept on and collapsed.’

Figure 43.

Figure 44.

(5) lexical semantics (e.g., sink <sinker sunk>)

a-structure (e.g., sink <agent theme>)

syntactic structure (e.g., [ PRED ‘sink <( SUBJ) ( OBJ)> ])

(5) lexical semantics morpholexical operations

a-structure OT-LMT & morphosyntactic operations

syntactic structure

Figure 45.




Potential Advantages of OT-LMTThe OT-LMT proposed here targets

specifically at the universal constraints onargument-function linking. We leave theOT formulation of morpholexical opera-tions to further research. In this section,we discuss some of the advantages thatthe OT-LMT may afford.

We start from the fact that all OTconstraints are declarative. In the conven-tional LMT, the two subject conditions mustapply sequentially, not simultaneously, toprevent θ[-r] from mapping to SUBJ, whenÔ[-o] is present. Likewise, only after themapping of the subject role can other rolesbe mapped. In contrast, all constraints inOT-LMT apply declaratively and, thus, si-multaneously. Furthermore, in the conven-tional LMT, all roles are uniformly assignedexactly one feature for function assignment,while the OT-LMT allows a more expres-sive system with only the patient/theme rolepre-assigned to unrestricted functions. ThisOT-LMT thus allows the possibility ofagent-OBJ as a marked morphosyntacticselection, which is ruled out in the conven-tional LMT.

Recall also that two disjunctions areobserved in the conventional LMT: the dis-junction between the two principles of sub-ject role mapping and the disjunction be-tween subject roles and non-subject roles.The OT-LMT, however, consistently favorsthe unmarked values for all roles. This char-

acteristic ultimately may lead to the replace-ment of the stipulation in the Subject Con-dition while preserving its insight. Thus, ingeneral, this revised LMT formulated in OTformalism offers a potentially more con-sistent and simpler computational system14.

As noted earlier, the LinkLocInv con-straint proposed in the OT-LMT accountessentially reflects the insight of Bresnanand Kanerva’s (1989) locative default,which assigns loc[-r] when th is focused.Thus, both accounts are descriptivelyequivalent in explicating locative inversionin the various languages observed, Chineseincluded. However, Bresnan and Kanerva’s(1989) account would need to state thatlanguages like Japanese and Korean lackthe mechanism of linking loc to [-r]. TheOT account, on the other hand, has theadvantage of a more general solution inattributing the presence or absence of loca-tive inversion in a language to the relativeranking of LinkLocInv, which, like all OTconstraints, is universal15.

Finally, we will indicate exactly howthe OT-LMT better reflects the intuitionthat the locative inversion construction of<th-OBJ loc-SUBJ> is marked in compari-son to the canonical locative constructionof <th-SUBJ loc-OBL¸>. In Figure 46, weexamine the constraints that each of thetwo violates.

Notice that the canonical form con-stitutes only one violation of one of the two

Figure 46.


C2 <th-SUBJ loc-OBL > *


C4 <th-OBJ loc-SUBJ> * * *




lowest-ranked constraints. The invertedform, on the other hand, violates two of thehigher-ranked constraints, LinkRolRes andDescendFun, in addition to one of the twolowest-ranked constraints. Nonetheless,even with such violations, the inverted formstill outranks all other candidates. It is, there-fore, still the optimal choice, in spite of itsmarkedness. The OT-LMT is, therefore,more expressive and flexible, accounts fora wider range of data, and reveals the(un)markedness of different linking rela-tions. It is, in short, a simpler, more consis-tent, and more general theory.

Directions of Further ResearchConsidering its limited number of prin-

ciples, LMT is a relatively small theory butwith ambitious goals. In the previous sec-tions, an OT version of the theory has beenlaid out and tested against cross-linguisticdata of locative inversion, as generalizedby Bresnan (1994). However, there arelocative constructions closely related tolocative inversion that have not been cov-ered. For instance, locative inversion mightbear some relation to sentences with anexpletive subject and a locative argument(Bresnan, 1994). Figure 47 is an examplefrom French16.

The expletive subject is an athematicargument, and, as such, it must receive an

intrinsic [-r] classification by the very na-ture of thematic restrictedness [r],(Bresnan, 2001). Given its initial position, itinvariably links to SUBJ17. Similar to thelocative inversion discussed earlier, thisconstruction also indicates that the objectfunction of the theme role renders it morefocal than the oblique locative role. Anathematic argument in the a-structure isindicated by an underscore outside of theangled brackets, while thematic argumentsare within the angled brackets. Thus, thea-structure of Figure 47 is “arrive _<th[foc]loc>.” An expletive subject also may beassociated with the linking of agentive ob-jects. Figure 48 is an example from French.

Similar phenomena also are observedin Bantu languages (Demuth, 1990; Demuth& Mmusi, 1997; Harford, 1990; Machobane,1995). All these issues are important andinteresting but cannot be adequately ad-dressed in the current article. Further re-search is needed on how the a-structure“arrive _<th[foc] loc>” and “work_<ag[foc]>” come about, whether they aremorpholexically or morphosyntactically re-lated to “arrive <th loc>” and “work_<ag[foc]>,” respectively, and how best toincorporate such relations within the OT-LMT proposed here.

Further development of this OT-LMTalso will need to address the issue of sec-

Figure 47.

Il est arrivé beaucoup de gens à la plageit is arrived many of people at the beach

‘There were many people arriving at the beach.’

Figure 48.

Il travaille deux mille ouvriers dans cette usineit works two thousand workers in this factory

‘There are two thousand workers working in the factory.’




ondary patient-like roles as a parameter ofvariation in double object constructions,known as the Asymmetrical Object Param-eter (AOP) (Alsina & Mchombo, 1993;Bresnan, 2001). In a non-AOP language,all patient-like roles are linked to an unre-stricted function, while AOP languages mustlink the secondary patient/theme to an ob-ject function. An additional constraint maybe necessary, and constraint ranking thenmay reflect this variation. This asymmetri-cal object parameter is stated in Figure 49.

Finally, the OT-LMT developed hereneeds to be applied to a much wider rangeof data cross-linguistically; for example,complex predicates in various languages(Abaitua, 1988; Ackerman, 1992; Alisina,Bresnan & Sells, 1997; Her, to appear;Ishikawa, 1985), the valence-changing mor-phemes and inversion constructions inGeorgian (Blevins, 2005; Harris, 1981;Holisky, 1981), among others. A solid analy-sis of some of these facts would be a sig-nificant test of the linking theory proposedhere.

CONCLUSIONIn this article, we set out to accom-

plish two goals. The more ambitious one isto come up with a simpler and more gen-eral lexical mapping theory in OT terms, orOT-LMT. The second one is to test thistheory and account for locative inversionin Chinese, English, and Chichewa on theone hand and Japanese and Korean on theother hand. Following the standard view inOptimality Theory, the mapping constraintswe proposed are all universal, and language

variation in locative inversion is accountedfor by different constraint rankings. TheOT-LMT we proposed is the UG compo-nent that constrains the argument-functionlinking, or morphosyntactic processes. Itthus does not govern language-specificmorpholexical processes, such aspassivization, locative transitivization, andresultative compounding. Locative inver-sion, on the other hand, involves onlymorphosyntactic operations and, therefore,is accounted for within the OT-LMT.

In summary, the OT-LMT we pro-posed not only covers a wider range ofempirical data, but it also affords a simpler,more consistent, and more general theory.

ACKNOWLEDGMENTSI am genuinely grateful to the three

anonymous reviewers for their insightfulcomments and exceedingly generous sug-gestions, and also to Adams Bodomo for hisencouragement. I also thank Nissa Hui-shanLin for discussions on OT. However, I amsolely responsible for the content of the ar-ticle. Research reported in this article hasbeen funded partly by NSC grants 92-2411-H-004-024 and 93-2411-H-004-006. Part ofthe research for the article was done whileI visited the School of Information Technol-ogy, Bond University in 2003 and 2004. Ithank the dean, Professor Ron Davison, forhis kind and continuous support.

REFERENCESAbaitua, J. (1988). Complex predicates in

Basque: From lexical forms to func-tional structures. Unpublished doctoral

Figure 49. Asymmetrical object parameter (AOP)Asymmetrical Object Parameter (AOP)

* => [-r] [-r] [-r] [+o]




dissertation, Department of Languagesand Linguistics, University of Manches-ter.

Ackerman, F. (1992). Complex predicatesand morpholexical relatedness: Locativealternation in Hungarian. In I. Sag, &A. Szabolcsi (Eds.), Lexical matters(pp. 55-83). Stanford, CA: CSLI Publi-cations.

Ackerman, F., & Moore, J. (2001a).Dowtyian proto-properties and lexicalmapping theory. In Proceedings of the6th International Lexical-FunctionalGrammar Conference, Hong Kong.

Ackerman, F., & Moore, J. (2001b). Proto-properties and grammatical encod-ing: A correspondence theory of ar-gument selection. Stanford, CA: CSLIPublications.

Aissen, J. (1999). Markedness and subjectchoices in optimality theory. NaturalLanguage and Linguistic Theory, 17,673-711.

Alsina, A. (1996). The role of argumentstructure in grammar. Stanford, CA:CSLI Publications.

Alsina, A., Bresnan, J., & Sells, P. (Eds.).(1997). Complex predicates. Stanford,CA: CSLI Publications.

Alsina, A., & Mchombo, S. (1993). Objectasymmetries and the Chichewaapplicative construction. In S. Mchombo(Ed.), Theoretical aspects of Bantugrammar (pp. 17-45). Stanford, CA:CSLI Publications.

Asudeh, A. (2001). Linking, optionality, andambiguity in Marathi. In Sells (Ed.), pp.257-312.

Babby, L. (l989). Subjectlessness, externalsubcategorization, and the projectionprinciple. Zbornik Matice Srpske ZaFilologiju i Lingvistiku, 32, 7-40.

Baker, M. (1983). Objects, themes, andlexical rules in Italian. In L. Levin, M.

Rappaport, & A. Zaenen (Eds.), Papersin lexical-functional grammar. Indi-ana: Indiana University Linguistics Club.

Birner, B. (l994). Information status andword order: An analysis of English in-version. Linguistic Inquiry, l7, 347-354.

Blevins, J. (2005). Thematic inversion inGeorgian. University of Cambridge.Retrieved June 15, 2005, http://www.cus.cam.ac.uk/~jpb39/drafts/tig.pdf

Bresnan, J. (l989). The syntactic projec-tion problem and comparative syntax oflocative inversion. Journal of Informa-tion Science and Engineering,5, 287-303.

Bresnan, J. (1990). Monotonicity and thetheory of relation-changes in LFG. Lan-guage Research, 26(4), 637-652.

Bresnan, J. (l994). Locative inversion andthe architecture of universal grammar.Language, 70, 72-l3l.

Bresnan, J. (l996). Lexicality and argu-ment structure. Paper presented atCollogue de Syntax et Semantique, Paris.Retrieved June 15, 2005, from http://csli-www.stanford.edu/users/bresnan/

Bresnan, J. (2000). Optimal syntax. In J.Dekkers, F. van der Leeuw, & J. van deWeijer (Eds.), Optimality theory: Pho-nology, syntax and acquisition (pp.334-385). Oxford: Oxford UniversityPress.

Bresnan, J. (2001). Lexical-functionalsyntax. Oxford: Blackwell Publishers.

Bresnan, J., & Kanerva, J. (1989). Loca-tive inversion in Chichewa: A case studyof factorization in grammar. LinguisticInquiry, 20, 1-50.

Bresnan, J., & Kanerva, J. (l992). The the-matic hierarchy and locative inversionin UG: A reply to Schachter’s comments.Syntax and Semantics: Syntax and theLexicon, 111-l25.




Bresnan, J., & Zaenen, A. (l990). Deepunaccusativity in LFG. In K. Dziwirek,P. Farrell, & E. Mejias (Eds.), Gram-matical relations: A cross-theoreticalperspective (pp. 45-57). Stanford, CA:CSLI Publications.

Butt, M., Dalrymple, M., & Frank, A.(1997). An architecture for linking theoryin LFG. In Proceedings of the LFG97Conference. Stanford, CA: CSLI On-line Publications. Retrieved June 15, 2005from, http://csli-publications.stanford.edu/LFG/2/lfg97-toc.html

Butt, M., & King, T. H. (Eds.). (2000).Argument realization. Stanford, CA:CSLI Publications.

Cheng, R. (1983). Focus devices in Man-darin Chinese. In T. Tang, R. Cheng, &Y. Li (Eds.), Hanyu jufa yuyixue lunji(pp. 50-102). Taipei: Student Book Co.

Dalrymple, M. (2001). Lexical functionalgrammar. New York: Academic Press.

Demuth, K. (1990). Locatives, impersonalsand expletives in Sesotho. The Linguis-tic Review, 7, 233-249.

Demuth, K., & Mmusi, S. (1997). Presen-tational focus and thematic structure incomparative Bantu. Journal of AfricanLanguages and Linguistics, 18, 1-19.

Dowty, D. (1991). Thematic proto-rolesand argument selection. Language,67(3), 547-619.

Falk, C. (1989). On the existential construc-tions in Germanic languages. WorkingPapers in Scandinavian Syntax, 44,45-59.

Falk, Y. (2001). Lexical-functional gram-mar: An introduction to constraint-based syntax. Stanford, CA: CSLIPublications.

Grimshaw, J. (1990). Argument structure.Cambridge, MA: MIT Press.

Harford, C. (1990). Locative inversion in

Chisona. In J. Hutchison, & V. Manfredi(Eds.), Current approaches to Africanlinguistics 7 (pp. 137-144). Foris:Dordrecht.

Harris, A. (1981). Georgian syntax: Astudy in relational grammar. Cam-bridge: Cambridge University Press.

Her, O.-S. (1989). An LFG account forChinese BEI sentences. Journal of theChinese Language Teachers Associa-tion, 24(3), 67-89.

Her, O.-S. (1990). Grammatical functionsand verb subcategorization in Man-darin Chinese. Doctoral dissertation,University of Hawaii, 1991.

Her, O.-S. (1997). Interaction and varia-tion in the Chinese VO construction.Taipei: Crane Publishing.

Her, O.-S. (1998). Lexical mapping in Chi-nese inversion sonstructions. In Pro-ceedings of the LFG98 Conference.Stanford, CA: CSLI On-line Publications.Retrieved June 15, 2005, from http://csli-publications.stanford.edu/LFG/3/lfg98.html

Her, O.-S. (1999). Interaction of thematicstructure and syntactic structures: OnMandarin dative alternations. ChineseLanguages and Linguistics Vol. V:Interaction, 373-412.

Her, O.-S. (2003). Chinese inversion con-structions within a simplified LMT. InA. Bodomo, & K. K. Luke (Eds.), Jour-nal of Chinese Linguistics Mono-graph No. 19: Lexical-FunctionalGrammar Analysis of Chinese.

Her, O.-S. 2004. Argument-function link-ing in resultatives. Concentric: Studiesin Linguistics, 30(2), 1-34.

Her, O.-S. (to appear). Argument-functionmismatches in Mandarin Chinese: A lexi-cal mapping account. Natural Lan-guage and Linguistic Theory.

Holisky, D. A. (1981). Aspect and Geor-




gian medial verbs. New York: Caravan.Huang, C.-r. (1993). Mandarin Chinese and

the lexical mapping theory: A study ofthe interaction of morphology and argu-ment changing. Bulletin of the Instituteof History and Philology, 62(2), 337-388.

Huang, H.-t. (l995). Lexical mappingtheory: Mandarin dative shift andlocative inversion. Unpublishedmaster’s thesis, National Chengchi Uni-versity, Taiwan.

Huang, H.-t., & Her, O.-S. (1998). Man-darin locative inversion and relation-changing rules. In S. Huang (Ed.), Se-lected papers from the second inter-national symposium on languages inTaiwan (pp. 287-304). Taipei: CranePublishing.

Ishikawa, A. (1985). Complex predicatesand lexical operations in Japanese.Unpublished doctoral dissertation, De-partment of Linguistics, Stanford Uni-versity.

Kaplan, R., Riezler, T. H., King, T. H., Max-well, J. T., III, Vasserman, A., & Crouch,R. (2004). Speed and accuracy in shal-low and deep stochastic parsing. In Pro-ceedings of the Human LanguageTechnology Conference and the 4thAnnual Meeting of the North Ameri-can Chapter of the Association forComputational Linguistics, (HLT-NAACL’04). Boston, MA.

Kuhn, J. (2001). Generation and parsing inoptimality theoretic syntax: Issues in theformalization of OT-LFG. In Sells (Ed.),pp. 313-366.

Kuhn, J. (2003). Optimality-theoretic syn-tax: A declarative approach. Stanford,CA: CSLI Publications.

Lee, H. (2001). Markedness and word or-der freezing. In Sells (Ed.), pp. 63-128.

Levin, L. (1987). Toward a linking theory

of relation changing rules in LFG(Report No. CSLI-87-115). Stanford,CA: CSLI Publications.

Li, Y. (1995). The thematic hierarchy andcausativity. Natural Language andLinguistic Theory, 13, 255-282.

Lødrup, H. (1999). Linking and optimalityin the Norwegian presentational focusconstruction. Nordic Journal of Lin-guistics, 22(2), 205-229.

Lødrup, H. (2000). Underspecification inlexical mapping theory: The case ofNorwegian existentials and resultatives.In M. Butt, & T. H. King (Eds.), Argu-ment realization (pp. 55-83). Stanford,CA: CSLI Publications.

Machobane, M. M. (1995). The Sesotholocative constructions. Journal of Af-rican Languages and Linguistics, 16,115-36.

Markantonatou, S. (1995). Modern Greekdeverbal nominals: An LMT approach.Journal of Linguistics, 31, 267-299.

Maxwell, J. T., III, & Kaplan, R. M. (1991).A method for disjunctive constraint sat-isfaction. In M. Tomita (Ed.), Currentissues in parsing technology (pp. 173-190). Dordrecht: Kluwer Academic Pub-lishers.

Maxwell, J. T., III, & Kaplan, R. M. (1993).The interface between phrasal and func-tional constraints. Computational Lin-guistics, 19(4), 571-590.

Maxwell, J. T., III, & Kaplan, R. M. (1996).An efficient parser for LFG. In M. Butt,& T. H. King (Eds.), Proceedings ofthe LFG96 Conference [online].

McCloskey, J. (2001). On the distributionof subject properties in Irish. In W. D.Davies, & S. Dubinsky (Eds.), Objectsand other subjects: Grammaticalfunctions, functional categories, andconfigurationality. Dordrecht: KluwerAcademic Publishers.




Mikkelsen, L. H. (2003). Reanalyzing thedefiniteness effect: Evidence from Dan-ish. Working Papers in ScandinavianSyntax, 69, 1-75.

Pesetsky, D. (1995). Zero syntax:Experiencers and cascades. Cam-bridge, MA: MIT Press.

Prince, A., & Smolensky. (2004).Optimality theory: Constraint interac-tion in generative grammar. Oxford:Blackwell Publishers.

Sells, P. (Ed.) (2001). Formal and empiri-cal issues in optimality theoretic syn-tax. Stanford, CA: CSLI Publications.

Sharma, D. (2001). Kashimiri case cliticsand person hierarchy effects. In Sells(Ed.), pp. 225-256.

Simpson, J. (1991). Warlpiri morpho-syn-tax: A lexicalist approach. Dordrecht:Kluwer Academic Publishers.

Tan, F. (1991). Notion of subject in Chi-nese. Unpublished doctoral thesis,Stanford University.

Ting, J. (1998). Deriving the Bei-construc-tion in Mandarin Chinese. Journal ofEast Asian Linguistics, 7(4), 319-354.

Zaenen, A. (1987). Lexical informationin LFG: An overview. Palo Alto, CA:Ms. Xerox-PARC.

Zaenen, A. (1993). Unaccusativity in Dutch:Integrating syntax and lexical semantics.In J. Pustejovsky (Ed.), Semantics andthe lexicon (pp. 129-161). Dordrecht:Kluwer Academic Publishers.

Zaenen, A., & Engdahl, E. (1994). Descrip-tive and theoretical syntax in the lexi-con. In B. T. S. Atkins, & A. Zampolli(Eds.), Conceptual approaches to thelexicon (pp. 181-212). Oxford:Clarendon Press.

ENDNOTES1 See Bresnan (2001, chapter 14) for a

brief exposition of other formulations.2 Falk (2001) also presents a concise in-

troduction to LMT and a more preciselydefined theory of argument roles.Dalrymple (2001) offers more examplesin her introduction to the theory.

3 The concept of thematic hierarchy iswell-established (Grimshaw, 1990; Li1995). The hierarchy in “The SubjectCondition” also might be derived fromthe proto-role properties proposed byDowty (1991) (Ackerman & Moore,2001b; Bresnan, 2001).

4 Note that following Zaenen and Engdahl(1994), the two propositional argumentfunctions COMP and XCOMP aretreated as instances of OBL¸.

5 Note that Ô refers to the most promi-nent role in the a-structure. The fact thatit is also the left-most role within theangled brackets is inconsequential. Ô isusually also the initial role, unless thereis an initial athematic argument.

6 Alsina (1996) also argues that the func-tion-argument bi-uniqueness condition,which is fully integrated in the OT-LMTproposed later, is too strong.

7 A constraint for athematic roles must re-strict such roles to [-r]. Alternatively, aconstraint may be proposed to outrankLinkFun in Figure 22b and thus allow aGF in a-structure to be unmatched.

8 An additional constraint is needed forthe secondary patient/theme, which re-stricts the secondary pt/th to [+o].Again, we will ignore this issue in thisarticle.

9 DescendFun and DescendRol effect aparallel alignment between the thematichierarchy and the markedness hierarchy.This parallel alignment is similar to a




harmonic alignment, but with an impor-tant difference. In a harmonic alignment,according to Prince and Smolensky(2004), the correspondence between amore prominent element on one scaleand a less prominent element on the otheris worse than the correspondence be-tween two elements that are equal inprominence (Aissen, 1999; Asudeh,2001; Lee, 2001; Sharma, 2001). Thus,it is better for agent, the most prominentrole, to link to SUBJ, the most promi-nent function, and likewise for locative,the least prominent role, to link to OBJ¸.However, in our scheme here, due toLinkUnobj and LinkUnres, a moreprominent GF is favored, regardless ofthe prominence of the role.

1 0 IC refers to the intrinsic classificationof argument roles. See Figure 7.

1 1 Therefore, as I have proposed elsewhere(Her, 2003), morpholexical operationslikewise can add features and, thus, al-ter syntactic assignments of argumentroles, besides changing the lexical se-mantics of a predicator.

1 2 See Her (1989) and Ting (1998), amongothers, for compelling arguments againstviewing the bei-NP phrase as a PP by-phrase.

1 3 In fact, the same concept of suppres-sion in passives is used here, as well.See Her (2004) for details of mappingthe composite role, formed by two roles,to a single GF.

1 4 Note that I am only referring to compu-tational efficiency in formulation andformalism, not in practical terms of anactual computational implementation.See Kuhn (2003) for extensive discus-sions on the computational aspects ofOT. However, there is little practicalevidence for the computational effi-ciency of a large-scale OT implementa-tion of a grammar, as there seems to beno such practical systems yet. For thecomputational efficiency of LFG in gen-eral, see Maxwell and Kaplan (1996,1993, 1991); for LFG in practice, referto Kaplan, et al. (2004). I thank theanonymous reviewer who made thispoint and provided the references.

1 5 I thank another anonymous reviewer forpointing this out to me.

1 6 I thank the anonymous reviewer whosuggested this direction of further re-search and provided this French exampleand its discourse analysis.

1 7 Refer to Bresnan (2001, section 14.1) fora more in-depth discussion on athematicarguments in raising constructions.

One-Soon Her received his PhD in linguistics from the University of Hawaii in1990. He worked on a LFG-based multilingual machine translation system from1986 to 1992 in a R&D firm in Provo, Utah. He has since been teaching at theGraduate Institute of Linguistics, National Chengchi University in Taiwan, andhas been a full professor since 1998. He has held visiting positions with the MiddleEast Technical University (Turkey), University of Hawaii, and three universities inAustralia: University of Queensland, Griffith University, and Bond University. Heis also a senior consultant for Application Technologies, a Washington DC-basedlanguage software developer. His research interests include syntactic theories,machine translation, grammatical interaction, Chinese grammar, language learningand teaching.


Documents

IDEA GROUP PUBLISHING ITJ3045osh/publication/A21-OT-LMT.pdf · Lexical Mapping Theory: A Case Study of Locative Inversion One-Soon Her, National Chengchi University, Taiwan ABSTRACT