Printer Friendly

Syntactic direction and obviation as empathy-based phenomena: a typological approach.


In previous studies, various syntactic/semantic factors (person hierarchy, animacy, topicality, etc.) have been discussed as relevant to linguistic phenomena known as syntactic direction and nominal obviation. This article develops and motivates a uniform analysis of the direct / inverse and obviation marking (DIO-marking), based on the (extended) theory of linguistic empathy. Drawing on data from four languages that belong to different Jamilies (Cree, Navajo, Jinghpaw, and Japanese), I discuss that the empathy-based approach ( i) provides a uniform analysis of DIO-systems in different languages, as well as the yaru/kureru opposition in Japanese, which have been believed to be controlled by different sets of syntactic/ semantic factors, and (ii) dispenses with construction-specific rules/ constraints such as the person constraint, the possessive constraint, and the ban on multiple proximates within a clause. I also demonstrate that the empathy-based account allows us to model similarities/contrasts among DIO-systems in a comprehensive way, reducing crosslinguistic differences into two planes: (i) the plane of E-marking: how and to what extent empathy relations are encoded, and (ii) the plane of E-ranking: what factors affect (more) empathy relations.

1. Introduction

In the literature, various syntactic/semantic factors (e.g., person hierarchy, animacy, topicality, discourse prominence, control force) have been discussed as relevant to linguistic phenomena known as syntactic direction and nominal obviation. This article explores the hypothesis that the opposition of direct/inverse and obviation are most directly controlled by the notion of linguistic empathy, drawing on data from different groups of languages.

Direct/inverse systems are attested in various families of languages, such as Algonquian, Athabaskan, and Tibeto-Burman; an analog of such systems is found in Japanese too. Nominal obviation has a more limited distribution, and is best known from the Algonquian languages. The basic function of these syntactic devices is to rank participant NPs along a certain dimension or hierarchy, which is known as the animacy hierarchy, obviation tier, etc. The subject of a direct construction must outrank (or be of the same rank as) the object, while the opposite holds for an inverse construction; similarly, an NP marked as proximate outranks NPs marked as obviative in the relevant discourse stretch. Various factors, such as person, animacy, definiteness, topicality, discourse prominence, control force, etc., have been discussed as determinant of the choice of direct/inverse and obviation. In the present work I argue that the key factor controlling the direct/inverse opposition and obviation marking (DIO-marking) is linguistic empathy. Under this hypothesis, the effects of various semantic/pragmatic factors, which have been observed and discussed in previous studies, neatly follow from the general theory of linguistic empathy; also, the empathy-based approach allows us to model crosslinguistic similarities/contrasts among DIO-systems in an elegant way.

The organization of this article is as follows. In Section 2, I review the theory of linguistic empathy and illustrate some empathy-related phenomena, drawing on Japanese data. In Section 3, I illustrate the DIO-marking systems in Cree (Algonquian), Navajo (Athabaskan), and Jingh-paw (Tibeto-Burman), and argue for a uniform, empathy-based analysis of them all. In Section 4, I discuss similarities and differences among the discussed DIO-systems, and draw out certain typological generalizations.

2. Preliminaries

This section provides an overview of the theory of linguistic empathy, which plays a central role in the arguments to be developed in the following sections. In Section 2.1, I illustrate various constraints on empathy relations drawing on Japanese data, which will be repeatedly referred back to in the later sections. In Section 2.2, I make conceptual and terminological clarifications about the notion of empathy.

2.1. Empathy hierarchies

The basic idea of the theory of empathy is that linguistic expressions may reflect the speaker's point of view, from which he or she describes a state of affairs. The notion, which was first discussed by Kuno and Kaburaki (1977), has been characterized in metaphorical terms such as "speaker's identification with a participant", "camera angle", and "point of view". Linguistic phenomena on which linguistic empathy has direct or indirect effects include, but are not limited to, anaphora and reference tracking, split case marking, and various types of syntactic alternation such as passivization, dative alternation, and syntactic direction (Kuno 1987; DeLancey 1981a; Nariyama 2003; Kozai 2000). As Kuno and Kaburaki (1977) and Kuno (1978, 1987) show at length, Japanese has several lexical devices which represent the "point of view" from which an event is described, including the two kinds of giving verbs yaru and kureru: (1)

(1) a. Taro-ga Hanako-ni okane-o yar-u.

Taro-Nom Hanako-Dat money-Acc give-Pres

'Taro gives Hanako money.'

b. Taro-ga Hanako-ni okane-o kure-ru.

Taro-Nom Hanako-Dat money-Acc give-Pres

'Taro gives Hanako money.'

(Kuno 1987: 246)

Both (1a) and (1b) describe a situation in which Taro gives money to Hanako, but (1a) describes it from Taro's or the neutral perspective and (1b) from Hanako's. More generally, yaru is a verb that is used when the action is looked at from the point of view of the referent of the subject or the neutral (objective) point of view, whereas kureru is a verb used when the event is described from the point of view of the referent of the dative object. Similar observations hold for compound verbs with auxiliary giving verbs, which convey the benefactive meaning:

(2) a. Taro-ga Hanako-o tasukete-yat-ta.

Taro-Nom Hanako-Acc help-Ben-Past

'Taro helped Hanako (for her/my/ ... sake).'

b. Taro-ga Hanako-o tasukete-kure-ta.

Taro-Nom Hanako-Acc help-Ben-Past

'Taro helped Hanako (for her/my/ ... sake).'

Kuno and Kaburaki introduce the term "empathy" in reference to the point of view or "camera angle" that a speaker takes when he describes an event. In Kuno (1987), the notion of empathy is defined as follows:

(3) Empathy: Empathy is the speaker's identification, which may vary in degree, with a person/thing that participates in the event or state that he describes in a sentence.

Degree of Empathy: The degree of the speaker's empathy with x, E(x), ranges from 0 to 1, with E(x) = 1 signifying his total identification with x, and E(x) = 0 a total lack of identification. (Kuno 1987: 206)

The empathy relationships (i.e., the relative degree of the speaker's identification with participants) encoded by yaru and kureru can be summarized as follows (cf. Oshima 2004):
(4) yaru (main verb): E(Agent) [greater than or equal to]

 kureru (main verb): E(Recipient) > E(Agent)

 yaru (auxiliary verb): E(Benefactor) [greater than or equal to]

 kureru (auxiliary verb): E(Beneficiary) > E(Benefactor)

Kuno and Kaburaki (1977) also observe that there are several constraints on possible or favored choices of point of view (empathy locus). For example,

(5) Speech Act Empathy Hierarchy: The speaker cannot empathize with someone else more than with himself.

E(speaker) greater than or equal to] E(other)

(Kuno 1987: 212)

(6) Topic Empathy Hierarchy: Given an event or state that involves A and B such that A is coreferential with the topic of the present discourse and B is not, it is easier for the speaker to empathize with A than with B.

E(discourse topic) [greater than or equal to] E(nontopic)

(Kuno 1987: 210)

(7) Surface Structure Empathy Hierarchy: It is easier for the speaker to empathize with the referent of the subject than with the referents of other NPs in the sentence.

E(subject) > E(other NPs)

(Kuno 1987: 211)

(8) Descriptor Empathy Hierarchy: Given descriptor x (e.g., John) and another descriptor f(x) (e.g., John's brother), the speaker's empathy with x is greater than with f(x).

E(x) > E(f(x))

(Kuno 1987: 207)

The examples below illustrate that violation of these constraints make sentences unacceptable or marginal. (2)

(9) (violation of the Speech Act Empathy Hierarchy)

a. Boku-ga Max-ni hon-o yat-ta.

I-Nom Max-Dat book-Acc give-Past 'I gave Max a book.'

b. *Boku-ga Max-ni hon-o kure-ta.

I-Nom Max-Dat book-Acc give-Past

cf. Kimi-ga Max-ni hon-o kure-ta.

you-Nom Max-Dat book-Acc give-Past

'You gave Max a book.'

(10) (violation of the Topic Empathy Hierarchy)

a. Taro-wa saikin keiki-ga i-i.

Taro-Top recently business-Nom good-Pres

Dareka-ga (kare-ni) okene-o

someone-Nom he-Dat money-Acc



'Taro is prosperous these days. Someone must have given him money.'

b. Taro-wa saikin okane-ni komatte-i-ru.

Taro-Top recently money-Dat have.trouble-Asp-Pres (Kare-wa) dareaka-ni okene-o

he-Top someone-Dat money-Acc



'Taro is short of money these days. He must have given his money to somebody.'

(11) (conflict of the Surface Structure Empathy Hierarchy and the Descriptor Empathy Hierarchy)

a. [Max.sub.i]-wa [kare.sub.i]-no musuko-ni tasuke-rare-ta.

[Max.sub.i]-Top [he.sub.i]-Gen son-Dat help-Pass-Past

'Max was helped by his son.'

b. * [Max.sub.i]-no musuko-wa [kare.sub.i]-ni tasuke-rare-ta.

[Max.sub.i]-Gen son-Top [he.sub.i]-by help-Pass-Past

'Max's son was helped by him.'

In (9), the agent participant refers to the speaker and thus the use of kureru causes violation of the Speech-Act Empathy Hierarchy. (10) illustrates that a discourse topic must receive at least as much empathy as a non-discourse topic. (3) In (11), the two arguments are expressed as 'Max' and 'his son' ([f.sub.son], (Max)); the Descriptor Empathy Hierarchy dictates that the speaker's empathy is with Max. (1lb) is marginal because the NP referring to Max is demoted from the subject, which is the natural position for the empathy locus, by passivization.

Finally, let me point out the (somewhat trivial) effect of humanhood/ animacy on empathy relations, which has not been, to my knowledge, explicitly discussed in the literature. Observe the following examples: (4)

(12) a. Herumetto-ga tentoo-no syokku-o

helmet-Nom fall-Gen shock-Acc



'The helmet lessened the shock when he fell down (for his sake).'

b. Kono saku-ga hituzi-tati-o ookami-kara

this fence-Nom sheep-Pl-Acc wolf-from



'This fence protects sheep against wolves.'

The data above suggest that animate objects generally receive more empathy than inanimate objects; to capture this I stipulate the following: (5)

(13) Animacy Empathy Hierarchy: It is easier for the speaker to empathize with animate objects than with inanimate objects.

E(animate) > E(inanimate)

2.2. Empathy and related notions

In previous studies, the notion of linguistic empathy has been metaphorically characterized as "point of view". The term "point of view", however, has been used loosely and ambiguously in the literature, and thus does not reveal much about the exact psychological/ontological nature of linguistic empathy. Therefore, I believe that it is important to make some clarifications about the notion, before proceeding to data-oriented analyses of syntactic direction/obviation.

First of all, the notion of empathy should be distinguished from other types of "point of view", in particular deictic center (see Levinson 2003; Fillmore 1982; Iida 1996; among others) and logophoricity (or narrative perspective; Schlenker 2003; Banfield 1992; Sells 1987). Although the latter two notions and linguistic empathy have strong correlations, they cannot be reduced to one another; while the locus of empathy of a clause tends to match the center of deixis and the logophoric individual, dissociation of them is not impossible (Culy 1997; Oshima 2004, 2007). Empathy must be distinguished from the notion of topicality as well, as a topic does not always match the empathy locus (see Note 3). Topicality is, whether it is given an aboutness--or givenness-based definition, conceptually orthogonal to factors like person and animacy; that is, most standard definitions of topicality do not preclude the possibility that 3rd person entities/inanimates outrank SAPs/animates in the hierarchy of topicality (Lambrecht 1994, Ariel 1990, among others). (6)

If empathy cannot be identified with any of other correlating notions like topicality, can it be decomposed to more primitive notions? A possible way to decompose linguistic empathy is to derive empathy relations from a harmonic alignment of various relevant hierarchies (scales), such as the hierarchies of grammatical functions (subj > obj > obl), topicality (more topical > less topical), person (local person > nonlocal person), and animacy (animate > inanimate) (Aissen 1999, 2003; Prince and Smolensky 1993; see also Giv6n 1994). (7) The partial ordering derived from a harmonic alignment would (conveniently) allow a certain range of cross-linguistic variety (see Section 3.2).

From the material at hand, it seems difficult to choose between the harmonic alignment-based, reductionist approach and the nonreductionist approach. A conceptual advantage of the former would be that it allows us to understand the elusive concept of linguistic empathy in more concrete terms; at the same time, the nonreductionist approach seems to provide a somewhat simpler picture, where formal oppositions (e.g., yaru and kureru) are associated with a certain, functional/conceptual primitive. In the present work, I will not pursue the reductionist approach in further detail, and adopt the working hypothesis that linguistic empathy is a primitive notion that reflects a certain psychological construct. At any rate, I believe that the choice between the nonreductionist and (harmonic alignment-based) reductionist hypotheses does not have a crucial bearing on the discussion to follow; should it turn out to be the case that the latter is more appropriate, the term empathy would remain as a convenient cover term to refer to relative rankings in harmonic scales.

3. Syntactic direction and obviation

In certain languages (e.g., Algonquian: Cree [Dahlstrom 1986; Wolfart 1973], Blackfoot [Pustet 1995], Ojibwa [Rhodes 1990; Jelinek 1990], Athabaskan: Navajo [Young and Morgan 1980; Hale 1973], TibetoBurman: Jinghpaw [DeLancey 1981b], Nootkan [Whistler 1985]; Otomanguean: Sochiapan Chinantec [Foris 1993]; see Klaiman 1991 and Givon 1994 for more references) (a subset of) transitive/ditransitive verbs have two forms called DIRECT and INVERSE. Following the view of typologists like Dixon and Aikhenvald (1997), I will assume that the direct-inverse alternation does not affect the mapping between semantic roles and grammatical functions (parallel to the Japanese yaru/kureru alternation), and that this is a defining characteristic of syntactic direction; it contrasts with the opposition of active/passive/antipassive in this respect (cf. Givon 1994). Algonquian languages also have systems of nominal OBVIATION, which are closely related to the choice of direct/inverse (Aissen 1997; Klaiman 1991; Dahlstrom 1986; among others). The basic function of these syntactic devices is to rank participant NPs along a certain dimension or hierarchy, which is known as the animacy hierarchy, obviation tier, etc. The subject of a direct construction must outrank (or be of the same rank as) the object, while the opposite holds for an inverse construction; similarly, an NP marked as proximate outranks NPs marked as obviative in the relevant discourse stretch.

In the literature, the role that empathy plays in systems of direct/ inverse and obviation (DIO-marking) has been largely overlooked, although a few authors count it as one of the relevant factors (Dahlstrom 1986; Navarro 2001; DeLancey 1981a, 1981b). In this section, I take up the DIO-systems in Cree, Navajo, and Jinghpaw, and develop a uniform, empathy-based analysis of them. (8) Namely, I propose that the syntactic direction in these languages is primarily controlled by linguistic empathy in a way parallel to the Japanese yaru/kureru opposition, and that the obviation is a morphological device to designate an NP as the empathy locus of a certain discourse stretch (minimally a clause); these ideas can be schematically represented as follows:

(14) Direct: E(Agent) [greater than or equal to] E(Recipient), or E(Agent) [greater than or equal to] E(Patient) in the absence of the recipient role

Inverse: E(Recipient) > E(Agent), or E(Patient) > E(Agent) in the absence of the recipient role

(15) [s ... NP[[Proximate].sub.i] ... NP[[Obviative].sub.j] ... NP[[Obviative].sub.k] ...] [??] E([NP.sub.i]) > {E([NP.sub.j]), E([NP.sub.k])}

I will also argue that the proposed analysis is superior to previous accounts, in that it provides straightforward accounts of various constraints on DIO-marking without stipulating construction-specific rules, and allows us to capture some aspects of crosslinguistic similarities/differences in a simple way, i.e., in terms of different "weights" on factors that affect empathy relations.

3.1. Cree

3.1.1. Basic facts. In Algonquian languages, verb stems are split into four classes according to the valence and semantic class (gender) of arguments: (9) Intransitive Inanimate (II; intransitive verbs with an inanimate subject), Intransitive Animate (IA; intransitive verbs with an animate subject), Transitive Inanimate (TI; transitive verbs with an inanimate object), Transitive Animate (TA; transitive verbs with an animate object). When a transitive verb selects for a recipient or a beneficiary in addition to a patient, the former is treated as a core argument to the exclusion of the latter (Klaiman 1991: 289). (10)

A subset of TA verbs have alternative forms (called THEME SIGNS) which indicate the syntactic direction. When the two arguments of a TA verb differ in person, the choice of direct/inverse forms is controlled by the person hierarchy: 2 > 1 > 3. That is, the direct is used when (i) the subject is first or second person (a SAP; speech-act participant) and the object is third person (a non-SAP), or (ii) the subject is second person and the object is first person; the inverse is used elsewhere. The following examples are from Cree, taken from Dahlstrom (1986):

(16) a. ni-wa.pam-a.-w

1. see.dir.3

'I see him.'

b. /ni-wa-pam-ekw.-w/ [right arrow] niwa.pamik

1. see. inv. 3

'He sees me.'

(17) a. ki-wa.pam-i-n


'You (sg.) see me.'

b. ki-wa.pam-iti-n


'I see you (sg.).'

As is manifest in the data above, agreement affixes in Cree generally do not specify grammatical functions of their target nominals. (11) The sentences in (16) and (17) would thus be ambiguous without the direct/ inverse suffixes. It should be noted that the use of full NPs would not resolve the ambiguity, since Cree (and Algonquian languages in general) does not have a fixed word order or a system of case marking for core arguments.

When the subject of a TA verb is inanimate, only the inverse form is possible (Dahlstrom 1986: 56-59):

(18) ni-se.kih-iko-n


'It scares me.'

Another factor that restricts the choice of direct/inverse is obviation. Obviation is a grammatical opposition which distinguishes one non-SAP NP from all others in a certain discourse stretch, minimally a clause (see below); the one singled out is called PROXIMATE, and the other non-SAPs are OBVIATIVE. Proximate nominals are morphologically unmarked. Obviative animate nominals are marked by the ending -a, and sometimes their obviative status is reflected in verbal inflection too. The obviative status of inanimate nominals is not marked by an ending, but it is reflected in verbal inflection when the inanimate NP is the subject of an intransitive verb (Dahlstrom 1986: 13).

When both core arguments of a TA clause are animate non-SAPs (and thus are on a par with one another in terms of person/animacy), the choice between direct/inverse is constrained by their obviation statuses. That is, when a proximate subject acts on an obviative object, the direct form is used, and when an obviative subject acts on a proximate object, the inverse is used:

(19) aya.hciyiniwah nisto e.h-nipaha.t awa na-pe.sis.

Blackfoot (obv) three kill.3.obv.dir this boy

'This boy (prox.) had killed three Blackfoot (obv.).' (Bloomfield 1934: 98)

(20) osa.m e.-sa-kihikot ohta.wiyah aw o-skini.kiw.

too.much love.obv.3.inv his.father.obv this

'For his father (obv.) too much cherished this young man (prox.).' (Bloomfield 1934: 53)

Both the subject and object of a transitive clause may be obviative, while they cannot be both proximate. When both of the core arguments are obviative, either a direct or inverse form is possible, although an inverse with two obviatives is very rare (Dahlstrom 1986: 53-54). The sole non-SAP participant in an intransitive clause is usually proximate, but it may be obviative. From the data at hand, it is not clear whether the configuration where one core argument is a SAP and the other is obviative is possible.

The obviative status of a nominal is said to be determined by discourse factors. Dahlstrom (1986: 108) notes: "The proximate third person may be the topic of discourse [... ]. The proximate third person is also usually the focus of the speaker's empathy (Kuno and Kaburaki 1977); in narratives, proximate often corresponds to the character whose point of view is being represented". (12)

Besides the topicality/discourse factor, obviation is restricted by two alleged "syntactic" constraints. First, there can be at most one proximate NP within a clause (or more precisely, when there are two or more proximates within a clause, they must be coreferential). Thus, as mentioned above, both the subject and object of a transitive clause cannot be proximate (Dahlstrom 1986:116):

(21) a. [s ... NP[Proximate] ... NP[Obviative] ...]

b. *[s ... NP[[Proximate].sub.i] ... NP[[Proximate].sub.j] ...]

c. [s ... NP[[Proximate].sub.i] ... NP[[Proximate].sub.i] ... ]

(22) awa na-pe.sis o.hih ka.-kaskatahomiht niya.nan

this boy this.obv be.wounded.obv five

miye.w misatimwah.

give.3.obv horse.obv

'The boy (prox.) gave five horses (obv.) to the man who had been wounded (obv.).'

A proximate participant NP and a proximate possessor NP within a clause must be coreferential (Dahlstrom 1986:119; Wolfart and Carroll 1981: 26-27).

(23) wa.pam-e.-w o-kosis-a.

see.obv.dir.3 3.son.obv

'[He.sub.i] (prox.) saw [his.sub.i] (prox.) son (obv.).'

Aissen (1997) proposes that there can be at most one proximate within a domain that she terms an OBVIATION SPAN. An obviation span can be indefinitely long, covering many sentences; as to the lower bound, she proposes the following as a tentative constraint:

(24) MINIMAL SPAN: Let A be a set consisting of a head and its arguments. Then, for each pair [of third person nominals] [alpha], [beta] in A, if [alpha] bears a relation [(of obviation)] to B, B an obviation span, and [beta] bears a relation [(of obviation)] to C, C an obviation span, then B = C.

(Aissen 1997: 714)

This roughly amounts to saying that an obviation span for a proximate is the minimal clause containing it, excluding adjuncts. This makes a correct prediction as to the facts mentioned above, as well as data like the following, as (24) allows a single sentence to contain more than one obviation span:

(25) e.h-takohte.cik e-kotah, a.say ka.-pa-skiswa.t there already this.inan shoot.3.obv



'When they (prox.) arrived there, he (prox.) had already shot the buffalo.'

(Dahlstrom 1986: 138)

(26) aya.hciyiniwah nisto awa

also Blackfoot.obv three this boy

miywe.yihtamwak o.k a.yisiniyiwak. these

'Also the people (prox.) were glad that the boy (prox.) killed three Blackfoot (obv.).'

(Dahlstrom 1986: 138)

Second, in a possessive construction, the possessum must be obviative:

(27) o.hta.wiya e.h-okima.wiyit.

his.father.obv be.thief.obv

'His (prox.) father (obv.) was a thief.'

Both possessor and possessum NPs may be obviative, but it is impossible for both to be proximate, or for only the possessum noun to be proximate; below I will refer to this phenomenon as the "possessive constraint". When the possessor is already obviative, it is possible to explicitly mark the possessum as "further" obviative (or subobviative; Dahlstrom 1986: 55-56; Wolfart 1978; see also Aissen 1997: 718-719; Pustet 1995):

(28) wa.pam-im-e.-w o-kosis-iyiw-a.

see.obv.dir.3 3.son.obv.obv

'[He.sub.i] (prox.) saw [his.sub.j] (obv.) son (obv.).'

3.1.2. The direct/inverse opposition and obviation as empathy-based phenomena. In the last section we surveyed three syntatico-semantic factors that affect the direct-inverse opposition in Cree: the person hierarchy, gender (animacy), and obviation. We have also seen that obviation is constrained by discourse factors as well as alleged syntactic rules (i.e., the minimal obviation span and possessive constraint).

As the reader may have noticed, the Cree direct/inverse opposition is strikingly similar to the Japanese yaru/kureru opposition. It seems reasonable, thus, to hypothesize that the empathy-based account of the latter, discussed in Section 2, can be extended to the former. I propose that the choice between direct and inverse in Cree is based on the empathy relations among the participants of a clause, analogous to the case of the Japanese donatory/benefactive constructions, and conversely, that the Japanese yaru/kureru can be construed as lexicalized direct/inverse verbs (cf. Nariyama 2003; Shibatani 2003).

(29) (= [14])

Direct: E(Agent) [greater than or equal to] E( Recipient), or E(Agent) [greater than or equal to] E(Patient) in the absence of the recipient role

Inverse: E(Recipient) > E(Agent), or E(Patient) > E(Agent) in the absence of the recipient role

In the following, I will closely examine the similarities of the two paradigms of constructions, and argue that the empathy-based analysis is more plausible than previous analyses in that it dispenses with various construction-specific constraints.

First, the person hierarchy-based restriction is reminiscent of the Speech-Act Empathy Hierarchy. Recall that in Japanese, a first person participant is generally required to be the empathy locus of a clause, to the effect that when it is the agent participant of a donatory event yaru must be chosen, while when it is the recipient kureru is the only option. If yaru and kureru are construed as lexicalized direct/inverse forms, the effect of person on the yaru/kureru alternation is analogous to that on the Cree direct/inverse alternation, except for the relevant person hierarchies (1 > {2, 3} in Japanese and 2 > 1 > 3 in Cree). Second, the effect of animacy can be attributed to the Animacy Empathy Hierarchy, which I proposed in Section 2.l based on Japanese data. Third, various constraints on obviation too are given straightforward accounts by the theory of empathy. While obviation has no counterpart in Japanese or English (which have been the major sources of data for studies of linguistic empathy), the correlation between empathy and obviation has been pointed out by at least two authors (Dahlstrom 1986; Navarro 2001). Following their remarks, I assume that the obviation status of an NP directly reflects the speaker's empathy with its referent, i.e., a proximate non-SAP empathically outranks other non-SAPs in the relevant discourse stretch. As we saw in the last section, the correlation between the empathy and topicality is captured by the Topic Empathy Hierarchy in the theory of empathy. The effect of topicality on obviation is thus a welcome consequence of this assumption.

Let us now address the two alleged syntactic constraints on nominal obviation: the minimal obviation span and possessive constraint. As we saw above, there can be at most one proximate within a certain discourse stretch (minimal obviation span; see [24]). Under the hypothesis that the syntactic direction and obviation are empathy-based phenomena, the minimal obviation span can be understood as the minimal domain within which there can be at most one empathy locus.

Within Japanese syntax, it has been argued that such a domain ("empathy domain") exists, and there has been debate as to its proper definition, especially in connection with the binding properties of (the perspectival use of) zibun (Kuno 1978; Kuroda 1973; Kameyama 1984; Oshima 2004, 2007, cf. Iida 1996; Katagiri 1991). Kuno (1978) stipulates that the empathy domain for an expression is the minimal clause or NP that contains it. The binding condition for zibun in its perspectival use (13) can be roughly stated as: (i) zibun in its perspectival use must be the empathy locus of its empathy domain (i.e., must empathically outrank all other co-participants in its empathy domain) and (ii) it must be bound to the subject of a higher clause. As such, a sentence like (30a) is precluded, because in the relative clause the use of zibun indicates that its referent, Max, empathically outranks all other coparticipants, while the use of yaru indicates that the subject of the relative clause, Pat, receives at least as much empathy as Max:

(30) a. *[Max.sub.i]-wa Pat-ga [zibun.sub.i]-ni kasite-yat-ta hon-o [Max.sub.i]-Top Pat-Nom [self.sub.i]-Dat lend-Ben-Past book-Acc nakusite-simat-ta. lose-end.up-Past 'Max lost the book which Pat lent to him.' cf. [Max.sub.i]-wa Pat-ga [kare.sub.i]-ni kasite-yat-ta hon-o he nakusite-simat-ta.

b. Max-wa Pat-ga zibun-ni kasite-kure-ta hon-o [Max.sub.i]-Top Pat-Nom [self.sub.j]-Dat lend-Ben-Past book-Acc nakusite-simat-ta. lose-end.up-Past 'Max lost the book which Pat lent to him.'

The ban on multiple proximates within a clause can be stated in a way very similar to binding condition (i) of perspectival zibun: i.e., it must empathically outrank coparticipants within its empathy domain. This formulation is more plausible than previous ones (Dahlstrom 1986; Aissen 1997), making it possible to draw a generalization that an expression designating an empathy locus is associated with a certain domain (within which it must outrank all other participants).

It should be noted, however, that a proximate NP differs from perspectival zibun in two respects (apart from the fact that the former is not necessarily anaphoric). First, its empathy domain ("minimal obviation span") is not equivalent to the empathy domain for zibun: in particular, the empathy domain for a possessive NP in Cree is the minimal clause that contains it, while the empathy domain for a possessive zibun is the minimal NP that contains it. Thus, a possessive proximate NP must be coreferential with the empathy locus of the clause that minimally contains it (i.e., the proximate argument of the clause; see [23], repeated below as [31]), while perspectival zibun may not be, as shown in (32):

(31) (= [23]) wa.pam-e.-w o-kosis-a. see.obv.dir.3 3.son.obv '[He.sub.i] (prox.) saw [his.sub.i] (prox.) son (obv.).'

(32) [Max.sub.i]-wa [zibun.sub.i]-no hon-o Pat-ni {yat/kure}-ta. [Max.sub.i]-Top [self.sub.i]-Gen book-Acc Pat-Dat give-Past '[Max.sub.i] gave [his.sub.i] book to Pat.'

I do not have an answer as to why this is the case. It might be worth pointing out, however, that the illustrated contrast is somewhat reminiscent of the cross/intralinguistic parametrization of "binding domains" for anaphoric expressions (see Kiparsky 2002; Huang 2000; among others).

Second, unlike zibun in its perspectival use, a proximate NP can co-occur with a first or second person argument; in other words, a proximate NP is required to outrank other non-SAPs in the minimal obviation span, but not SAPs.

(33) ni-wa.pam-88.-w atim. 1.see.dir.3 dog 'I see the dog (prox.).'

(34) [Max.sub.i]-wa boku-ga [zibun.sub.i]-ni kasite-{*yat/?*kure}-ta hon-o [Max.sub.i]-Top I-Nom [self.sub.i]--Dat lend-Ben-Past book-Acc nakusite-simat-ta. lose-end.up-Past '[Max.sub.i] lost the book I lent to [him.sub.i].'

The possessive constraint, finally, directly follows from the Descriptor Empathy Hierarchy and the ban on multiple proximates. Recall that in Cree when a possessor NP modifies a noun, either the possessor is proximate and the possessum is obviative, or both are obviative. The other two combinations are impossible:

(i) Proximate Proximate impossible
(ii) Proximate Obviative possible
(iii) Obviative Proximate impossible
(iv) Obviative Obviative possible

Configuration (i) is excluded by the ban on multiple proximates, as the possessor and the possessum necessarily belong to a single obviation span. Configuration (iii), on the other hand, is blocked by the Descriptor Empathy Hierarchy. Consider the following example:

(36) Ca.n ki.-ma-kwamik. John(prox.) his(prox.).dog(obv.) bite.inv 'John's dog bit him.' (Wolfart 1973: 25)

If the obviation statuses of the two nominals are switched, a conflict of empathy relations arises as the Descriptor Empathy Hierarchy dictates that: E(John)> E(John's dog) while obviation indicates the opposite.

3.2. Navajo

3.2.1. Basic facts. Navajo is another well-studied language with a DIO-system. Below, I briefly illustrate syntactic/morphological properties of the Navajo syntactic direction and semantic/pragmatic factors that affect it.

Navajo is a head-marking language with subject/object agreement affixes. In a transitive clause, a third person object is marked by one of the three alternative forms: (i) [empty set] (null), (ii) yi, and (iii) bi. The [empty set] form is used when the subject is a SAP. The yi form is used when the subject outranks the object in the so-called animacy hierarchy (see below), while bi is used when the object outranks the subject. The choice between yi/bi is also mirrored in the linear order of the subject and the object: with yi, the subject precedes the object, and with bi, the order is reversed: (14)

(37) tii' setat. horse horse 3Acc(empty])-1sgNom-kicked 'I kicked the horse.' (Jelinek 1990: 228)

(38) at'eed to yodlaa'. girl water 3Acc(yi)-3Nom-drank 'The girl drank the water.' (Hale 1973: 303)

(39) dibe to 'abiteel. sheep water 3Acc(bi) 'The water swept the sheep off.' (Hale 1973: 302)

Several authors regard this alternation as indicative of syntactic direction and/or obviation, rather than passive, etc. (Jelinek 1997, 1990; Klaiman 1991; Aissen 2000; Thompson 1996, 1989a, 1989b). (15) Like in Algonquian, the direct/inverse alternation with ditransitives involves only the agent and recipient arguments (Jelinek 1990:231).

As to the morphological properties of yi and bi, there remains some disagreement. Some authors consider subject/object prefixes in general to be pronominal (Jelinek 1997, 1990; Aissen 2000; Thompson 1996, 1989a, 1989b), whereas Speas (1990) and Uyechi (1991, 1996) argue that these prefixes are agreement markers, except for bi, which is an incorporated pronoun (see also Bresnan 2001: 161-168). In the present work I do not go into the details of the morphological facts and refer to yi and bi simply as "object prefixes", which indicate the direct and inverse construction respectively.

Navajo does not have a morphological obviation marker corresponding to -a in Cree; (16) the linear order of NPs can, however, be viewed as a functional analog of obviation, as it indicates that the first NP outranks, or is of the same rank as, the second on the hierarchy that determines the direction.

(40) a. Cree: {[[NP.sub.i] NP-[a.sub.j] V], [NP-[a.sub.j] [NP.sub.i] V], .. } [??] [NP.sub.i] > (? [greater than or equal to]) [NP.sub.j]

b. Navajo: [[NP.sub.i] [NP.sub.j] V] [??] [NP.sub.i] [greater than or equal to] > [NP.sub.j]

Now, let us examine factors that affect the direct/inverse alternation (and word order) in Navajo. When one core participant is animate and the other is inanimate, the choice between yi and bi is restricted, parallel to the case of the Cree syntactic direction:

(41) a. at'eed to yodlaa. girl water 3Acc(yi)-3Nom-drank 'The girl drank the water.'

b. *to at'eed bodlaa. water girl 3Acc(bi)-3Nom-drank (Hale 1973: 303)

(42) a. dibe to 'abiiteel. sheep water 3Acc(bi) 'The water swept the sheep off.'

b. *to dibe 'ayiiteel. water water 3Acc(yi) (Hale 1973: 302)

Creamer (1974) proposes the following as the "animacy hierarchy" that controls the yi/bi alternation: yi is used when the subject outranks the object in this hierarchy, whereas bi is chosen when the object outranks the subject.

(43) Human > Animals (Lg > Med > Sm) > Insects > Natural forces > Plants & inanimate objects > Abstract notions

The effect of the animacy hierarchy is not absolute, but relative (see Hale 1973 among others). For example, when the two core participants of a transitive clause are ranked close (e.g., human and animal), violation of the selection principles makes the sentence only awkward, rather than unacceptable.

(44) a. dine bii yiyisxii. man deer killed-Dir 'The man killed the deer.'

b. ?bii dine biyisxii. deer man killed-Inv (Hale 1973: 302)

(45) a. ?teechaa'i shiye yishxash. dog my.son bit-Dir 'The dog bit my son.'

b. shiye teechaa'i bishxash. my.son dog bit-Inv (Hale 1973: 302)

Such reversal phenomena are observed even between animate and inanimate nouns, as in: (17)

(46) a. tii' 'ii'ni' biisxi. horse lightning killed-Inv 'Lightning killed the horse.'

b. 'ii'ni' tii' yiisxi. lightning horse killed-Dir (Hale 1973: 305)

The following examples are from San Carlos Apache, a language closely related to Navajo:

(47) a. izee ncho'i gidi yi-yeshi. medicine bad cat killed-Dir 'Poison killed the cat.'

b. gidi izee ncho'i bi-yeshi. cat medicine bad killed-Inv (Shayne 1982: 389)

Based on such data, several authors conclude that the yi/bi alternation is controlled by relative potency or control force, rather than animacy (Frishberg 1972; Hale 1973; Shayne 1982; Witherspoon 1977; Klaiman 1991).

When the two core participants of a transitive clause are equally ranked in the animacy hierarchy, it is said that the direction is determined by the topicality/discourse prominence: i.e., yi indicates that the subject (agent) is a topic (or discourse-prominent), and bi indicates that the object (patient) is a topic (or discourse-prominent). In a neutral description with two equally animate arguments, the yi form is used (italics in the gloss indicate foci; cf. Aissen 2000; Willie and Jelinek 2000; Jelinek 1997; Hale 1973):

(48) a. 'ashkii 'at'eed yizts'os. boy girl 3Acc(yi)-3Nom-kissed 'The boy kissed the girl. / The boy kissed the girl.'

b. 'ashkii 'at'eed bizts'os. boy girl 3Acc(bi)-3Nom-kissed 'The girl kissed the boy. / The boy was kissed by the girl.' (Willie and Jelinek 2000: 283)

When a transitive clause has two human participants and only one of them is realized as an overt nominal, the interpretation of this nominal is constrained by the choice of yi/bi (Willie 1991; cf. Frishberg 1972). (18)

(49) a. [dii 'ashkii] yi-ztat. this boy kicked-Dir 'He kicked this boy.' NOT: 'This boy kicked him.' (Willie 1991: 74)

b. ['ashkii dii] bi-ztat. this boy kicked-Inv 'This boy kicked him.' NOT: 'He kicked this boy.' (Willie 1991: 75)

As Aissen (2000) points out, this phenomenon follows from the assumption that yi (hi) indicates that the subject (object) is a topic, as generally a null pronominal is more discourse-prominent than a lexically realized nominal (cf. Givon 1983; Ariel 1990).

3.2.2. Empathy and the Navajo direction. The DIO-system of Navajo is similar to that of Cree in that it is affected by both animacy and topicality. I hypothesize, as I did with the Cree DIO-system, that the Navajo syntactic direction and word order (which is a functional analog of the obviation in Cree) are directly controlled by the notion of empathy. The DIO-system in Navajo, however, contrasts with that in Cree in two respects: (i) not only animacy, but other semantic features of nouns such as humanhood, size, and potency are reported to affect the possibility/ likelihood of the alternation; also, the effect of these semantic features is not absolute or categorical, (ii) the opposition is present (or explicit) only if both core arguments are non-SAPs. Below, I examine these two points in some detail, and discuss that the empathy-based analysis nicely explains language-internal facts of Navajo, as well as the reasons behind crosslinguistic differences among DIO-systems.

As we saw above, the Navajo DIO-system is affected by at least two factors: the animacy (or alternatively, the scale of attributed control force, etc.) and topicality. This means that any accounts of the Navajo direct/inverse opposition based solely on the animacy hierarchy (Creamer 1974), relative potency (Shayne 1982), control force (Hale 1973; Witherspoon 1977), or topicality (Willie and Jelinek 2000; Jelinek 1997) fall short (cf. Thompson 1996: 96). A tentative approximation of the rules to determine the Navajo direction would be the following:

(50) i. When the agent of a transitive clause outranks the recipient/ patient in the animacy hierarchy, the direct form is used.

ii. When the recipient/patient of a transitive clause outranks the agent in the animacy hierarchy, the inverse form is used. iii. When the agent and the patient are equally ranked in the animacy hierarchy:

a. if the agent is more topical, the direct form is used.

b. if the recipient/patient is more topical, the inverse form is used.

This formulation too, however, fails to capture the phenomenon of animacy reversal and wrongly excludes sentences like (44b), (45a), (46b), and (47a). What brings about animacy reversal, I presume, is the conflict between the effects of animacy hierarchy and topicality. That is, in the Navajo syntactic direction, unlike those in Japanese and Cree, the effect of topicality may override that of animacy, although the latter takes precedence over the former in principle.

The empathy-based account of the syntactic direction captures the effects of both animacy and topicality, by the Animacy Empathy Hierarchy and the Topic Empathy Hierarchy. Furthermore, it allows us to model conflicts among the relevant hierarchies in a rather simple way. I submit that languages may vary as to how much each of the relevant factors (e.g., person, animacy, topicality, etc.) "matters" to the determination of linguistic empathy. The relation between the relevant factors and linguistic empathy can be approximated by the following scheme:

(51) E(x) = [F.sub.E] (Person (x), Topicality (x), Animacy (x), Potency (x), ... ) = a x Person (x) + b x Topicality (x) + c x Animacy (x) + d x Potency (x) + ..., where Person, Topicality, Animacy, Potency, ... are functions whose ranges are positive real numbers and a, b, c, d, ... are constants ("weights").

Crosslinguistic differences, such as the possibility of occasional animate-inanimate reversals, can be attributed to the weights of each argument, i.e., the size of the constants a, b, c, d,... in the formula above: that is, in Navajo, the weight on the topicality factor relative to that on the animacy factor is greater than in Japanese and Cree.

Now, let us address the second characteristic of the Navajo inverse system, which concerns the person hierarchy. As mentioned above, the object prefix position of a verb is empty when the subject is a SAP and the object is a non-SAP. When the object is a SAP, the alternation does not occur since there is no third person object prefix. In other words, the yi/bi alternation is present only when both of the core arguments are non-SAPs.

(52) a. tii' setat. horse 3Acc([empty])-1sgNom-kicked 'I kicked the horse.'

b. tii' shiztat. horse 1sgAcc-3Nom-kicked 'The horse kicked me.' (Jelinek 1990: 228)

Does this mean that person in Navajo has no bearings on the empathy relations among the participants? No. If this were the case, the yi/bi alternation would be present in sentences like (52a) just as in those in which both core arguments are non-SAPs.

The absence of opposition in such cases rather suggests that a variant of the Speech-Act Empathy Hierarchy, i.e., E(SAP) > E(non-SAP), is operative in Navajo; that is, if it is always the case that first and second persons outrank third person, the direction marking (yi in [52a]) would be redundant and thus its absence is motivated by economy, at the expense of explicitness. This surmise is supported by the fact that, when a lexical first person pronoun and a non-SAP NP cooccur within a clause, it is preferred that the former precedes the latter (Willie and Jelinek 2000:264-265; recall that Navajo word order reflects the empathy relations between the core participants). What cannot be inferred from the data presented so far is the ranking between first person and second person. Either of the three possibilities: (i) 1 > 2 (as in Japanese), (ii) 2 > 1 (as in Cree), and (iii) 1 = 2 (as in Jinghpaw; see below) is viable, and thus investigation with other diagnostics is required to identify the exact person hierarchy in Navajo. Note that, in any event, underspecification of the empathy relations between the core participants arises only if (iii) is the case.

3.3. Jinghpaw

Jinghpaw is a Tibeto-Burman language with SOV order. DeLancey (1981b) illustrates two sets of morphological empathy markers in the Hprang-Hkadung dialect of this language. The first is verbal affixes d-/ m-, which he calls VIEWPOINT MORPHEMES: d- indicates that the agent is the empathy locus of a transitive or ditransitive clause, while m- designates the patient in a transitive clause or the recipient in a ditransitive clause as the empathy locus. 19 The opposition of d- and m- is thus entirely analogous to that of direct/inverse markers in Algonquian. In the following I refer to d- and m- as direct/inverse markers.

The second is the person agreement. In Jinghpaw, the target of person agreement is the empathy locus, rather than a certain grammatical function. Consider the following examples (all examples in this subsection are taken from DeLancey [1981b]):

(53) ngai MaNaw hpe gumhpraw jaw n-i? ai. I MaNaw OBJ money give Asp-1st 'I gave money to Manaw.'

(54) ?MaNaw ngai hpe gumhpraw jaw n-u? ai. MaNaw I OBJ money give Asp-3rd 'Manaw gave me money.'

(55) MaNaw ngai hpe gumhpraw jaw n-i? ai. MaNaw I OBJ money give Asp-1st 'Manaw gave me money.'

In sentences like the above where d-/m- do not occur, (20) agreement with third person in preference to first or second person is disfavored. When both the subject and the object are SAPs, the agreement can go with either of them, depending on which is "spotlighted". Based on such data, DeLancey proposes that the person hierarchy in Jinghpaw is: (1 = 2) > 3. (21)

Interestingly, in Jinghpaw, unlike in Japanese and Cree, the effect of person can be overruled by that of topicality. This can be already seen from the status of (54), which is not entirely unacceptable. In sentences with the viewpoint morphemes d- and m-, this is further borne out:

(56) nang shihpe ndai jaw d-it ai. you he OBJ this give d-2nd 'You gave him this.'

(57) nang shi hpe ndai jaw m-u? ai. you he OBJ this give m-3rd 'To him, you gave this.'

(58) *nang shi hpe ndai jaw d-u? ai you he OBJ this give d-3rd ( 59) *nang shih pe ndai jaw m-it ai you he OBJ this give m-2nd

In (57), both the inverse marker m- and the agreement with third person indicate that the recipient receives more empathy than the agent, despite the fact the former is outranked by the latter in the person hierarchy.

Under the hypothesis that languages may differ as to the weights on semantic/pragmatic factors that affect linguistic empathy, this phenomenon can be accounted for by assuming that in Jinghpaw the weight on the person factor relative to that on the topicality factor is smaller than in Japanese, etc. (see Note 3). On the other hand, from the premise that d-/m- and person agreement both indicate the empathy locus, it is correctly predicted that (58) and (59) are unacceptable, where they designate two different participants as empathy loci of a single clause.

DeLancey, however, challenges Kuno's "unitary" theory of empathy, arguing that in Jinghpaw it is possible to specify two empathy loci within a clause. The crucial set of data is the following, in particular (62):

(60) shi nang hpe ndai jaw d-u? ai. he you OBJ this give d-3rd 'He gave you this.'

(61) shinang hpe ndai jaw m-it ai. he you OBJ this give m-2nd 'To you, he gave this.'

(62) shi nang hpe ndai jaw d-it ai. he you OBJ this give d-2nd 'You were given this by him.'

(63) *shi nang hpe ndai jaw m-u? ai he you OBJ this give m-3rd

Under the assumption that the person agreement and d- both designate the empathy locus, (62) should involve a conflict. DeLancey takes this as supporting evidence for his claim that the notion of empathy consists of two primitive components, which he terms viewpoint and attention flow (Delancey 1981a). He proposes that d- and m- identify what the attention focus (i.e., the starting point of the hearer's attention flow) of a clause is, while the person agreement marks the viewpoint from which an event is described. With this move, however, it becomes unclear why sentences like (58) and (59) are excluded; why are these sentences unacceptable, if the split of attention focus and viewpoint is allowed, the attention focus can go with the object, and the viewpoint can go with a third person participant?

A simpler and more plausible analysis would be that in (62) the event is described from the neutral perspective. If we assume that the opposition of d- and m- is an analog of that of yaru and kureru, it would not be surprising that d- allows the neutral perspective. To explain the acceptability of (62), thus, we only need to stipulate that agreement goes with a SAP in preference to a non-SAP when the two core arguments of a clause are equally empathized with. To conclude, the data shown above does not force us to abandon the "unitary notion of empathy" (DeLancey 1981b: 46).

3.4. Summary

In this section, I argued that the notion of empathy plays the central role in the DIO-systems in Cree, Navajo, and Jinghpaw. Syntactic directions in these languages are analogs of the yaru/kureru opposition in Japanese, and obviation can be understood as a morphological device to indicate the empathy locus of the relevant discourse stretch (obviation span).

The advantages of the uniform, empathy-based analysis of the DIO-systems over past analyses are summarized below:

1. To account for the conditions on the direct/inverse alternation and obviation in Cree, it has been believed that construction-specific constraints, such as the person/animacy-related restrictions, topicality effect, and possessive constraint must be postulated (Dahlstrom 1986; Aissen 1997). The empathy-based account allows us to eliminate them; under the empathy-based hypothesis, the effects of person and animacy follow from (a variant of) the Speech-Act Empathy Hierarchy and the Animacy Empathy Hierarchy, which have been motivated on independent grounds. In the same vein, the discourse effect and possessive constraint on obviation can be accounted for by the Topic Empathy Hierarchy and the Descriptor Empathy Hierarchy respectively.

2. While various notions, such as the animacy hierarchy (Creamer 1974), relative potency (Hale 1973), control force (Witherspoon 1977; Shayne 1982; Klaiman 1991), and topicality (Willie and Jelinek 2000; Jelinek 1997), have been proposed as the determinant factor of the yi/bi alternation in Navajo, none of them has fully adequate empirical coverage. The empathy-based analysis is more appropriate, capturing the effects of both information structure-based and semantic conditions.

3. DeLancey (1981b) proposes to split the notion of empathy into two components (viewpoint and attention flow), in order to account for data in Jinghpaw that involve apparent empathy conflicts. Not only is this move implausible on grounds of parsimony, it makes wrong predictions regarding certain data. The unitary notion of empathy must be maintained to provide a consistent account of the relevant phenomena.

4. Typology

So far we have overviewed the DIO-systems in four languages which belong to different families: Japanese (Japanese), Cree (Algonquian), Navajo (Athabaskan), and Jinghpaw (Tibeto-Burman). These systems exhibit certain striking similarities, based on which I developed a uniform, empathy-based analysis. Namely: (i) the alternation is controlled or affected by the person hierarchy, animacy, and topicality; (ii) for a ditransitive clause, it is the recipient, rather than the patient/theme, whose relative ranking with the agent determines the syntactic direction; (22) and (iii) the direct form allows the neutral perspective, while the inverse does not.

On the other hand, there are interesting differences among them; namely they differ as to (i) how and to what extent empathy relations are encoded (E-marking) and (ii) what ranks higher than what in the empathy hierarchy (E-ranking). In this section, I will summarize and discuss how these systems differ from one another, to establish a preliminary typology of DIO-systems.

4.1. E-marking

4.1.1. Domain. In Japanese, only verbs of giving and derived benefactive verbs exhibit the opposition. In Cree, a wider range of transitive/ ditransitive verbs, which belong to the Transitive Animate class, have direct/inverse forms. In Navajo and Jinghpaw, as a rule all transitive/ ditransitive verbs potentially alternate.

4.1.2. Marking. The opposition of yaru and kureru in Japanese is lexical. In Navajo, the direction is indicated by third person pronoun object prefixes. Cree and Jinghpaw have suffixes specialized to indicate the direction, which are called theme signs and viewpoint morphemes respectively. Jinghpaw m-/d- may be blocked by other suffixes such as the aspect marker n-, while in Cree theme signs are always present when appropriate.

4.1.3. Linking strategy: empathy locus and grammatical functions. In the four languages, only Cree has the morphological obviation of nominals. As mentioned in Section 3.2, however, the word order in Navajo reflects empathy relations among the core participants of a clause and thus conveys more or less equivalent information as morphological obviation. An important generalization is that there is a negative correlation between the presence of such markers of the empathy locus (including word order) on the one hand, and grammatical function (GF) marking devices such as GF-based word order and case marking on the other. GF marking and empathy locus marking can be understood as alternative strategies to carry out the surface form/semantic role linking, the presence of empathy locus marking (with direction marking) compensating the lack of GF marking (cf. Klaiman 1991:164-170). The system in Navajo is a "hybrid", in that it also has person agreement with GF encoding; a possible linking problem arises only when both arguments are third person, and only in this case does the empathy-based word order play a significant role in the linking resolution.

4.1.4. Active pairs. In Navajo, the direction is indicated only when both arguments are third person. In the other three languages, the direction-marking is present in all person combinations. (23) As remarked in 3.2, from the functional perspective this contrast can be understood as a trade-off between economy and explicitness. When the two core arguments of a clause differ in person, the person hierarchy restricts possible empathy relations, to the effect that the direction marking may be redundant; for example, in the following Japanese example, the choice of yaru over kureru does not have any semantic or pragmatic import, as the empathy relation designated by yaru (E(the speaker) [greater than or equal to] E(Max)) is predictable from the person hierarchy (the Speech Act Empathy Hierarchy).

(64) Boku-wa Max-ni hon-o yat-ta.

I-Top Max-Dat book-Acc give-Past

'I gave Max a book.'

In Navajo, this type of redundancy is not present, the direct/inverse opposition being neutralized. In languages like Cree which do not have GF-based person agreement or case-marking, on the other hand, the direction marking is functionally motivated in all person combinations to resolve the surface form/semantic role linking.

4.2. E-ranking

4.2.1. Person hierarchy. In Japanese, first person outranks second and third (1 > {2, 3}). In Cree, second person outranks first person, and first person outranks third person (2 > 1 > 3). In Jinghpaw, first person and second person are equally ranked, and local persons outrank third person ({1,2} > 3). In Navajo, local persons outrank third person, but the ranking between first person and second person is unknown (SAP > 3). From these hierarchies we can draw out the generalization that third person never outranks a local person (i.e., SAP [greater than or equal to] 3).

Such representations of hierarchies are, however, only approximations. In Jinghpaw, for example, the effect of person hierarchy can be overruled by the topicality/discourse prominence, while this never happens in Cree, etc. Also, in Japanese, although second person can be outranked by third person (see [9]), there seems to be a tendency for the direct form to be used in the [Agent:2nd, Recipient:3rd] configuration, and the inverse in the [Agent:3rd, Recipient:2nd] configuration. These points suggest that the "distances" between ranked members within person hierarchies too may vary from language to language, so that the ranking is perceived as a mere tendency in certain languages but as a categorical effect in others.

An interesting question concerning this point is whether such rankings are determined culturally or linguistically. In other words: Do varieties of E-ranking strategies reflect speakers' ontology or social/cultural backgrounds? Or are they rather arbitrarily encoded in individual grammars? I am not prepared to discuss this issue at this time, and leave it for future study.

4.2.2. Animacy hierarchy. In Navajo, it is possible that certain "potent" inanimate nouns (e.g., 'lightning', 'poison') empathically outrank animate nouns (presumably only when the former is more discourse-prominent than the latter). This does not happen in Cree, in which animacy is by and large a grammaticized property of nouns. In Japanese too, an animate/inanimate reversal seems not to take place even when an inanimate NP is both potent and topical. As to Jinghpaw, I do not have enough data to assess this possibility. Again, I leave it as an open question to what extent such contrasts reflect the world view/ontology of speakers.

4.3. Summary

The points discussed above are summarized in Table 1.

The observations from the four languages reveal that there can be considerable variety among DIO-systems in diverse aspects. At the same time, the empathy-based account allows us to draw out certain typological generalizations/predictions such as:

(65) i. THERE IS NO "REVERSED ALIGNMENT": There are no DIO-systems where, for example, the direct (or inverse) construction indicates that the subject is more topical than the object or less animate than the object, or that the subject outranks the object in the person hierarchy or the subject is less topical than the object.

ii. THERE IS A TRADE-OFF BETWEEN EMPATHY LOCUS MARKING AND GF MARKING: There are no or few languages that have both empathy locus marking on nominals (such as nominal obviation) and GF marking (on top of the syntactic direction), because having both causes redundancy. Conversely, there are no or few languages that lack both nominal obviation and GF marking, because at least one of them is required to resolve the linking between surface forms and semantic roles.

Table 2 illustrates the point (65ii) in more detail.

In a language like Cree (row [i] in Table 2), which has both direction and obviation, the combination of these two devices convey the information about both semantic linking and empathy relations. In a Cree transitive sentence, thus, it is always encoded which of the core arguments is the agent and which is the recipient/patient, plus which argument is more empathized with than the other. Navajo has direction marking, GF marking and a functional analog of obviation, but the direction marking and GF marking are both deficient, in the sense that they are not always present (row [ii]). The combination of these three devices again conveys the information about semantic linking and empathy relations. A language like Jinghpaw has direction and GF marking, instead of obviation (row [iii]). The information specified in a clause, however, ends up being the same: both semantic linking and empathy relations are encoded. In Japanese, GF marking is always present, but direction is lexically restricted (row [iv]). As a result, empathy relations are not always specified in a clause. A language like English, next, has only GF marking (row [v]). In such a language empathy relations within a clause cannot be explicitly encoded. Rows (i)-(v) are possible and attested patterns. What we expect not to exist is languages that have patterns in the bottom three rows (and to my knowledge, such languages do not exist). If a language has only direction or obviation and not GF marking, semantic linking cannot be resolved. The intuition behind this is that the resolution of semantic linking is more crucial for linguistic communication than the specification of empathy relations (as the former concerns the truth-conditional meaning of the sentence), and thus a language may lack the specification of empathy relations but not the specification of semantic linking. The row at the bottom corresponds to a language that has all of (complete) direction, GF marking, and obviation. It seems likely that such a language does not exist either, because it yields redundancy; as shown in rows (i) and (iii) of this table, obviation or GF marking on top of direction is sufficient to specify both semantic linking and empathy relations, so for a language to have both leads to redundancy.

5. Conclusion

In this article, I argued that the direct/inverse alternation and obviation are most directly controlled by the notion of linguistic empathy, drawing on data from Japanese, Cree, Navajo, and Jinghpaw, and developed a preliminary typology of DIO-systems. The empathy-based approach neatly integrates observations and insights in past studies of syntactic direction and obviation. Its advantages over previous analyses are threefold. First, it provides a uniform analysis of the DIO-systems in different languages, as well as the yaru/kureru opposition in Japanese, which have been believed to be controlled by different sets of syntactic/semantic factors. Second, it dispenses with construction-specific rules such as the person constraint and the possessive constraint. Third, it allows us to capture contrasts among DIO-systems in a simple way, reducing crosslinguistic differences into two planes: (i) the plane of E-marking: how and to what extent empathy relations are encoded, and (ii) the plane of E-ranking: what factors affect (more) empathy relations.

Let me conclude by noting that the extended theory of empathy might be applicable to a wider variety of syntactic phenomena besides DIO-marking. A good example is the person constraint on the passive; as argued in Kuno's work, within the theory of empathy the awkwardness of a sentence like the following is explained as a result of the conflict of the Surface Structure Empathy Hierarchy and the Speech-Act Empathy Hierarchy (Kuno 1987: 230-231; Kato 1979):

(66) He was hit by me.

In certain languages like Lummi the effects of the person hierarchy are categorical and a sentence like (66) is judged as ungrammatical/ unacceptable, rather than just awkward (Jelinek and Demers 1983, 1994). This contrast between English and Lummi can be given a purely empathy-based account (e.g., without constructing harmonic alignments; cf. Bresnan et al. 2001), if we assume that the weight on topicality relative to the weights on subjecthood and person is bigger in English than in Lummi, so that only in English the effect of topicality can overrule the joint effect of grammatical function and person.


Aissen, Judith (1997). On the syntax of obviation. Language 73, 705-750.

--(1999). Markedness and subject choice in Optimality Theory. Natural Language and Linguistic Theory 17, 673-711. [Reprinted in (2001) Optimality-theoretic Syntax, Geraldine Legendre, Jane Grimshaw, and Sten Vikner (eds.), 61-96. Cambridge, MA: The MIT Press.]

--(2000). Yi and bi: proximate and obviative in Navajo. In Papers in Honor of Ken Hale, MIT Working Papers on Endangered and Less Familiar Languages 1, Andrew Carnie, Eloise Jelinek, and M. Willie (eds.), 129-150. Cambridge, MA: Department of Linguistics, MIT.

--(2003). Differential object marking: iconicity vs. economy. Natural Language and Linguistic Theory 21, 435-483.

Akmajian, Adrian and Anderson, Steven (1970). On the use of the fourth person in Navajo, or Navajo made harder. International Journal of American Linguistics 36, 1-8.

Ariel, Mira (1990). Accessing NP Antecedents. London: Routledge.

Banfield, Ann (1992). Unspeakable Sentences. London: Routledge.

Bloomfield, Leonard (1934). Plains Cree Texts, American Ethnological Society Publications 16. New York: G. E. Stechert and Co.

Bresnan, Joan (2001). Lexical Functional Syntax. Malden, MA: Blackwell.--; Manning, Chris D.; and Dingare Shipra (2001). Soft constraints mirror hard constraints: voice and person in English and Lummi. In The Proceedings of the LFG 01 Conference, Muriel Butt and Tracey Holloway King (eds.), 13-32. Stanford, CA: CSLI.

Creamer, Mary H. (1974). Ranking in Navajo nouns. Dine Bizaad N'anil'iih/Navajo Language Review 1(1), 29-38.

Culy, Christopher (1997). Logophoric pronouns and point of view. Linguistics 35, 845-859.

Dahlstrom, Amy L. (1986). Plains Cree morphosyntax. Unpublished doctoral dissertation, University of California at Berkely.

DeLancey, Scott (1981 a). An interpretation of split ergativity and related patterns. Language 57(3), 626-657.

--(1981b). Parameters of empathy. Journal of Linguistic Research 1(3), 40-49.

Dik, Simon C. (1989). The Theory of Functional Grammar Part I: The Structure of the Clause. Dordrecht: Foils.

Dixon, Robert M. and Aikhenvald Alexandra Y. (1997). A typology of argument-determined constructions. In Essays on Language Function and Language Type, John Bybee, John Haiman, and Sandra A. Thompson (eds.), 71-113. Philadelphia: John Benjamins.

Fillmore, Charles J. (1982). Towards a descriptive framework for spatial deixis. In Speech, Place, and Action. Studies in Deixis and Related Topic's, Robert J. Jarvella and Wolfgang Klein (eds.), 31-72. Chichester: John Wiley and Sons.

Foris, David (1993). A grammar of Sochian Chinantec. Unpublished doctoral dissertation, University of Auckland.

--(1997). Sochiapan Chinantec GIVE: a window into clause structure. In The Linguistics of Giving, Typological Studies in Language 36, John Newman (ed.), 209-248. Amsterdam: John Benjamins.

Frishberg, Nancy (1972). Navajo object markers and the great chain of being. In Syntax and Semantics 1, Charles Li (ed.), 259-266. New York: Academic Press.

Givon, Talmy (1983). Topic continuity in discourse: an introduction. In Topic Continuity in Discourse: A Quantitative Cross-language Study, Talmy Givon (ed.), 1-42. Amsterdam: John Benjamins.

--(1994). The pragmatics of de-transitive voice. In Voice and Inversion, Typological Studies in Language 28, Talmy Givon (ed.), 3-44. Amsterdam: John Benjamins.

Gundel, Jeanette K. (1974) The role of topic and comment in linguistic theory. Unpublished doctoral dissertation, University of Texas.

Hajicova, Eva; Barbara, Partee P.; and Sgall, Peter (1998). Topic-focus Articulation, Triparate Structure, and Semantic Content. Dorderecht: Kluwer.

Hale, Kenneth (1973). A note on subject-object inversion in Navajo. In Issues in Linguistics: Papers in Honor of Henry and Renee Kahane, Braj B. Kachru, Robert B. Lees, Yacov Malkiel, and Angelina Pietrangeli (eds.), 300-309. Chicago: University of Illinois Press.

Hara, Takaaki (2002). Anaphoric Dependencies in Japanese. Utrecht: LOT.

Huang, Yan (2000). Anaphora: A Crosslinguistic Study. New York: Oxford University Press.

Iida, Masayo (1996). Context and Binding in Japanese. Stanford, CA: CSLI.

Jelinek, Eloise (1990). Grammatical relations and coindexing in inverse systems. In Grammatical Relations. A Cross-Theoretical Perspective, Katarzyna Dziwirek, Patrick Farrell, and Errapel Mejias-Bikandi (eds.), 227 246. Stanford, CA: CSLI.

--(1997). Topic and focus in Navajo inverse. West Coast Conference of Formal Linguistics 15, 241-255.

--and Demers, Richard (1983). The agent hierarchy and voice in some Coast Salish languages. International Journal of American Linguistics 49, 167-185.

--and Demers, Richard (1994). Predicates and pronominal arguments in Straits Salish. Language 70, 697-736.

Kameyama, Megumi (1984). Subjective/logophoric bound anaphor zibun. The Proceedings of the 20th Annual Meeting of Chicago Linguistic Society 20, 228-238.

Katada, Fusa (1991). The LF representation of anaphors. Linguistic Inquiry 22(2), 287-313.

Katagiri, Yasuhiro (1991). Perspective and the Japanese reflexive zibun. In Situation Theory and Its Applications, Vol. II, Jon Barwise, Jean Mark Gawron, Gordon Plotkin, and Syun Tutiya (eds.), 425-447. Stanford, CA: CSLI.

Kato, Kazuo (1979). Empathy and passive resistance. Linguistic Inquiry 10, 149-152.

Kiparsky, Paul (2002). Disjoint reference and the typology of pronouns. In More Than Words, Studia Grammatica 53, Ingrid Kaufmann and Barbara Stiebels (eds.), 179-226. Berlin: Akademie Verlag.

Klaiman, M.-H. (1991). Grammatical Voice. Cambridge: Cambridge University Press.

Kozai, Soichi (2000). Viewpoint distribution and transitivity in Japanese and English. Unpublished doctoral dissertation, University of Hawaii.

Kruijff-Korbayova Ivana and Steedman, Mark (2003). Discourse and information structure. Journal of Logic, Language and Information 12, 249-259.

Kuno, Susumu (1978). Danwa no bunpoo [Grammar of discourse]. Tokyo: Taishukan.

--(1987). Functional Syntax: Anaphora, Discourse, and Empathy. Chicago: University of Chicago Press.

--and Etsuko Kaburaki (1977). Empathy and syntax. Linguistic Inquiry 8, 625-672.

Kuroda, Shige-Yuki (1973). On Kuno's direct discourse analysis of the Japanese reflexive zibun. Papers. in Japanese Linguistics 2, 136-147.

Lambrecht, Knud (1994). Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press.

Levinson, Stephen C. (2003). Space in Language and Cognition. Cambridge: Cambridge University Press.

Machida, Nanako (1998). On the null beneficiary in benefactive constructions in Japanese. In Japanese/Korean Linguistics 7, Noriko Akatsuka, Hajime Hoji, Shoichi Iwasaki, Sung-Ock Sohn, and Susan Strauss (eds.), 409-425. Stanford, CA: CSLI.

Nariyama, Shigeko (2003). Ellipsis. and Reference Tracking in Japanese, Studies in Language Companion Series 66. Amsterdam: John Benjamins.

Navarro, Samuel (2001). Obviation in Cree and the theory of empathy. Paper presented at Workshop on the Structure and Constituency of Languages of the Americas 6.

Oshima, David Y. (2004). Zibun revisited: empathy, logophoricity and binding. In The Proceedings of the 20th NWLC, University of Washington Working Papers 23, Amy McNamara and Sylwia Tur (eds.), 175-190. Seattle: Department of linguistics, University of Washington.

--(2007). On empathic and logophoric binding. Research on Language and Computation 5, 19-35.

Prince, Alan and Paul Smolensky (1993). Optimality Theory: constraint interaction in generative grammar. Technical Report TR-2. New Brunswick, NJ: Center for Cognitive Science, Rutgers University.

Prince, Ellen (1981). Toward a taxonomy of given-new information. In Radical Pragmatics, Peter Cole (ed.), 223-255. New York: Academic Press.

Pustet, Regina (1995). Obviation and subjectivization: the same basic phenomenon? A study of participant marking in Blackfoot. Studies in Language 19, 37-72.

Rhodes, Richard A. (1990). Obviation, inversion and topic rank in Ojibwa. In The Proceedings of the Annual Meeting of Berkeley Linguistic Society 16, 101-115.

Schlenker, Philippe (2003). A plea for monsters. Linguistics and Philosophy 26, 29-120.

Sells, Peter (1987). Aspects of logophoricity. Linguistic Inquiry 18, 445-479.

Shayne, Joanne (1982). Some semantic aspects of yi- and bi-. In Studies in Transitivity, Syntax and Semantics 15, Paul Hopper and Sandra Thompson (eds.), 379-407. New York: Academic Press.

Shibatani, Masayoshi (2003). Directional verbs in Japanese. In Motion, Direction and Location in Languages: In Honor of Zygmunt Frajzyngier, Typological Studies in Language 56, Erin Shay and Uwe Seibert (eds.), 259-286. Amsterdam: John Benjamins.

Speas, Margaret (1990). Phrase Structure in Natural Language. Dordrecht: Khiwer.

Thompson, Chad (1989a). Voice and obviation in Navajo. In The Proceedings of the Fourth Meeting of the Pacific Linguistics Conference, 466-488.

--(1989b). Pronouns and voice in Koyukan Athapaskan. International Journal of American Linguistics 55, 1-24.

--(1996). The history and function of the yi-/bi- alternation in Athabaskan. In A thabaskan Language Studies: Essays in Honor of Robert W. Young, Eloise Jelinek, Sally Midgette, Keren Rice, and Leslie Saxon (eds.) 81-100. Albuquerque, NM: University of New Mexico Press.

Uyechi, Linda (1991). The functional structure of the Navajo third person alternation. The Proceedings of the Annual Meeting of Chicago Linguistic Society 27, 434 446.

--(1996). The Navajo third person alternation and the pronoun incorporation analysis. In Athabaskan Language Studies: Essays in Honor of Robert PK Young, Eloise Jelinek, Sally Midgette, Keren Rice, and Leslie Saxon (eds.), 123-136. Albuquerque, NM: University of New Mexico Press.

Vallduvi, Enric (1992). The Informational Component. New York: Garland.

Whistler, Kenneth W. (1985). Focus, perspective, and inverse person marking in Nootkan. Grammar Inside and Outside the Clause, Johanna Nichols and Anthony C. Woodbury (eds.), 227-265. Cambridge: Cambridge University Press.

Willie, MaryAnn (1991). Navajo pronouns and obviation. Unpublished doctoral dissertation, University of Arizona.

--and Jelinek, Eloise (2000). Navajo as a discourse configurational language. In The Athabaskan Languages: Perspectives on a Native American Language Family, Theodore B. Fernald and Paul R. Platero (eds.), 252-287. Oxford: Oxford University Press.

Witherspoon, Gary J. (1977). Language and Art in the Navajo Universe. Ann Arbor, MI: University of Michigan Press.

Wolfart, H. Christoph (1973). Plains Cree: A Grammatical Study, American Philosophical Society Transactions 63-65. Philadelphia: American Philosophical Society.

--(1978). How many obviatives: sense and reference in a Cree verb paradigm. In Linguistic Studies of Native Canada, Eung-Do Cook and Jonathan Kaye (eds.), 255-299. Vancouver: University of British Columbia Press.

--and Carroll, Janet F. (1981). Meet Cree: A Guide to the Language. Lincoln: University of Nebraska Press.

Young, Robert and Morgan, William (1980). The Navajo Language, 2nd ed. Albuquerque, NM: University of New Mexico Press.

Ibaraki University

Received 16 March 2004

Revised version received

30 March 2005


(1.) Note that yaru and kureru share the same argument structure; the relation between them thus cannot be analogized with that of "give" and "receive"; Japanese also has a verb corresponding to "receive", morau. In many contexts, however, morau (and English receive, etc.) implicates that the speaker's empathy is with the recipient participant, because of its noncanonical argument linking pattern and the effect of the Surface Structure Empathy Hierarchy ([7] below; see Kuno and Kaburaki: 643-645; Kuno 1987: 216-219; Kozai 2000:69-76). Correspondence address: International Student Center, Ibaraki University, Mito-shi Bunkyo 2-1-1, Ibaraki 310-8512, Japan. E-mail:

(2.) Following Kuno and Kaburaki's convention, *, ?, etc. in this section indicate that sentences marked with them "are syntactically grammatical, but are unacceptable (in varying degrees) due to violation of various constraints on empathy foci" (Kuno and Kaburaki 1977: fn. 1).

(3.) When the Topic Empathy Hierarchy and the Speech-Act Empathy Hierarchy conflict, the latter always takes precedence (in Japanese):

(i) a. Taro-wa saikin keiki-ga i-i. Boku-ga (kare-ni)

Taro-Top recently business-Nom good-Pres I-Nom he-Dat

okene-o. {yat/*kure}-ta-noda.

money-Ace give-Past-Emph

'Taro is prosperous these days. I gave him money.'

b. Taro-wa saikin okane-ni komatte-i-ru. (Kare-wa)

Taro-Top recently money-Dat have.trouble-Asp-Pres he-Top

boku-ni okene-o {*yat/kure}-ta-noda.

I-Dat money-Acc give-Past-Emph

'Taro is short of money these days. He gave me his money.'

Such interactions among empathy constraints will be discussed in more detail in the followingsections.

(4.) A sentence like (12a) becomes awkward with the beneficiary argument overtly expressed with a phrase like kare-no tame-ni 'for him'; this is presumably because such a phrase in Japanese implies volitionality of the agent, and thus is incompatible with its inanimacy. See Machida (1998) for relevant discussion.

(5.) Humanhood seems to have some effect on empathy relations too; the contrast between human and nonhuman animates is, however, less clear than the one between animates and inanimates, and can be overruled by pragmatic factors.

(i) a. Kono syoonen-ga sono inu-ni hone-o {yat/??kure}-ta.

this boy-Nom that dog-Dat bone-Acc give-Past

'This boy gave that dog a bone.'

b. Kono syoonen-ga {uti-no inu/Fido} -ni hone-o {yat/kure}-ta.

this boy-Nom our dog Fido -Dat bone-Acc give-Past

'This boy gave our dog/Fido [referring to a dog] a bone.'

(ii) a. Kono inu-ga sono syoonen-o tasukete-{??yat/kure}-ta.

this dog-Nom that boy-Acc help-give-Past

'This dog saved the boy (for his sake).'

b. {Uti-no inu/ Fido} -ga sono syoonen-o tasukete-{yat/kure}-ta.

our dog Fido -Nom that boy-Acc help-give-Past

'Our dog/Fido [referring to a dog] saved the boy (for his sake).'

(6.) Kuno's formulation of the Topic Empathy Hierarchy might be, however, too simplistic (or not sufficiently specific) in that it takes the primitive notion of topicality as granted. In the literature, the notion of topicality has had varied definitions; some authors even claim that topicality is a complex notion that consists of various components or subtypes (see Lambrecht 1994; Prince 1981; Gundel 1974; Kruijff-Korbayova and Steedman 2003; among others). It seems thus possible that linguistic empathy is more sensitive to particular components or subtypes of topicality than others. Such a possibility is in fact supported by data like the following (as contrasted with (10)):

(i) a. Max-wa Pat-ni hon-o kasite-{yat/(?)kure}-ta.

Max-Top Pat-Dat book-Acc lend-Ben-Past

'Max lent the book to Pat.'

b. Pat-ni-wa Max-ga hon-o kasite- {(?)yat/kure}-ta.

Pat-Dat-Top Max-Nom book-Acc lend-Ben-Past

'Max lent the book to Pat.'

The sentences in (i) illustrate that the empathy locus of a clause does not necessarily match the participant marked by wa, which is known as a "topic marker". This can be explained if we assume, say, that (i) wa marks a topic in the sense of "what the sentence is about" (Vallduvi 1992; Hajicova et al. 1998; Dik 1989), while (ii) empathy correlates more directly with topicality in the sense of "discourse prominent" or "given" (Givon 1983; Ariel 1990; Prince 1981). I am indebted to the anonymous reviewer who drew my attention to this point.

(7.) The formal definition of harmonic alignment adopted by Prince and Smolensky (1993) is as follows:

(i) Alignment. Suppose given a binary dimension [D.sub.1] with a scale X > Y on its elements {X, Y}, and another dimension D2 with a scale a > b ... > z on its elements. The harmonic alignment of [D.sub.1] and [D.sub.2] is the pair of Harmony scales.

[H.sub.x]: X/a > X/b > ... > X/z

[H.sub.y]: Y/z > ... > Y/b > Y/a

where the connective > is read as "more harmonic than". To construct a harmonic alignment of more than two scales, the definition must be modified so that each harmonic scale may be partially (rather than totally) ordered.

(8.) The three languages are chosen because (i) the DIO-systems in these languages are relatively well studied in the literature, and (ii) they exhibit theoretically interesting contrasts, on the basis of which I will draw (in the next section) a preliminary typology of DIO-systems.

(9.) Cree nouns belong to one of the two genders, animate and inanimate. The animate class includes people, animals, most plants, and also some objects (such as ospwa.kan 'car' and se.hke.payi.s 'pipe'). The inanimate class includes most objects, most body parts, and some parts of plants (Dahlstrom 1986:11-12; Aissen 1997: 714).

(10.) Dahlstrom (1986: 17) remarks that three-place predicates like mi.y- 'give' assigns the recipient role to the first object and the theme to the second object.

(11.) First and second person agreement prefixes appear in the same affix slot, and thus cannot co-occur. When both first and second persons appear as core arguments, the second person prefix is chosen, as in (17).

(12.) The observation that the direct form is generally preferred when both of the core arguments are obviative, thus, seems to suggest that the direct is used when the speaker's perspective is neutral.

(13.) Zibun has (at least) three distinct uses, which are referred to as: reflexive, logophoric, and perspectival (empathic). Among the three uses, only the perspectival use is inherently sensitive to linguistic empathy (Oshima 2004; see Katada 1991; Hara 2002 and references therein for detailed syntactic discussion of Japanese anaphora).

(14.) Thompson (1996: 82-83) remarks that the link between the word order and yi-/bimarking is not as strict as commonly believed, showing data like the following:

(i) a. John gat yikfi' nagu'.

John cedar on fell

'John fell on the cedar.'

b. John gat bika' nagu'.

John cedar on fell

'The cedar fell on John.'

c. gat John bika' nagu'.

cedar John on fell

'The cedar fell on John.'

d. gat John yika' nagu'.

cedar John on fell

'John fell on the cedar.'

A sentence like (ic) is problematic for Speas' (1990) claim that bi is a pronoun anaphoric to the NP in the topic position.

(15.) See Uyechi (1991) for comparative discussion of alternative analyses.

(16.) The so-called fourth person in Navajo (and other Apachean languages) is sometimes referred to as "obviative" (Klaiman 1991: 180). l find this terminology misleading and thus do not adopt it; the functions of the Apachean fourth person are different in kind from those of Algonquian obviation (see Akmajian and Anderson 1970).

(17.) Recall that sentences like (46b) and (47b) are impossible in Cree, where an inanimate object acts on an animate and yet the inverse is used. This is probably not unrelated to the fact that in Algonquian animacy is by large a grammaticized property of nouns.

The soft effect of humanhood, size, etc., on the other hand, might well be attested in the Algonquian syntactic direction too, although, to my knowledge, it has not been explicitly discussed in the literature.

(18.) Frishberg (1972: 263) reports that a sentence like (49b) is ambiguous (while a sentence like (49a) is not).

(19.) According to DeLancey's analysis, what d- and m- indicate is an attention focus, which is a subtype of empathy locus; see below.

(20.) The occurrence of d-/m- is blocked by other prefix morphemes, such as n- in (53)-(55). They also do not occur in certain constructions types, such as gnomic sentences.

(21.) This, incidentally, is reminiscent of person agreement in Algonquian. As mentioned in Note 11, in Cree, when one core argument is first person and the other is second person, agreement goes with second person which outranks first person.

(22.) The syntactic direction in Sochiapan Chinantec (Otomanguean) slightly diverges from this pattern (Foris 1997). The verb for 'give' in Sochiapan Chinantec selects for the patient (theme) as a direct object, and the recipient as an indirect object or oblique. When the patient is animate (as in a sentence like 'He will give me Peter [in marriage]', etc.), the direction is determined by the relative ranking (which is subject to principles very similar to those for Algonquian) between the agent and patient; when the patient is inanimate, it is determined by the ranking between the agent and recipient.

(23.) It is reported that in some languages (e.g., Ojibwa, another Algonquian language) the direction is present only when at least one of the core arguments is a SAP (Jelinek 1990; Rhodes 1990).
Table 1. Comparison of the DIO-systems in
Japanese, Cree, Navajo, and Jinghpaw

 Japanese Cree

Direction lexical by specialized
marking affixes (theme

Domain only giving verbs TA (transitive
 and benefactive animate) verbs

Obviation no yes

Active pairs all combinations all combinations

Linking strategy case-particles by the
 combination of
 direction and

Person hierarchy 1 > {2, 3} 2 > 1 > 3

Other remarks

 Navajo Jinghpaw

Direction by agreement by specialized
marking (pronominal?) affixes
 prefixes (viewpoint

Domain transitive transitive verbs
 verbs in in general

Obviation no; but the no
 word order

Active pairs 3-3 only all combinations

Linking strategy by the GF-based word
 combination order
 of direction,
 based word
 order, and

Person hierarchy SAP > 3 {1, 2} > 3

Other remarks animate/ person reversal
 inanimate possible

Table 2. Strategies for the specification of
grammatical functions and empathy relations

 Marking information Languages

(i) direction, empathy semantic linking, Cree
 locus marking (by empathy relations

(ii) direction semantic linking, Navajo
 (deficient), empathy relations
 GF marking
 (deficient), empathy
 locus marking (by
 word order)

(iii) direction, GF semantic linking, Jinghpaw
 marking empathy relations

(iv) direction (lexically semantic linking, Japanese
 restricted), GF empathy relations
 marking (limited)

(v) only GF marking semantic linking English, etc.

(vi) only direction empathy relations likely to be

(vii) only empathy locus empathy relations likely to be

(viii) direction, GF semantic linking, likely to be
 marking, empathy empathy relations nonexistent
 locus marking
COPYRIGHT 2007 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Oshima, David Y.
Publication:Linguistics: an interdisciplinary journal of the language sciences
Date:Jul 1, 2007
Previous Article:Phonological development: toward a "radical" templatic phonology *.
Next Article:Aspects of Vietnamese clausal structure: separating tense from assertion *.

Related Articles
Voice and Grammatical Relations: In Honor of Masayoshi Shibatani.
English mediopassive constructions; a cognitive, corpus-based study of their origin, spread, and current status.
Unique focus.
Reciprocals and reflexives; theoretical and typological explorations.
New perspectives on historical Latin syntax; v.1: Syntax of the sentence.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters