語言，空間，與情緒(2/3)

行政院國家科學委員會專題研究計畫 期中進度報告

語言，空間，與情緒(2/3)

中 華 民 國 93 年 5 月 24 日

Table of Contents

1.Seediq: Spatial Representation in a language without prepositions ………..1 2.Spatial representation in Tsou………...20 3.Spatial representation in Squliq……….35 4.Emotion words in Tsou: a classification by five modes and seven categories…….58 5.Morphology of emotion verbs in Tsou……….79 6.Tsou is different: a cognitive perspective on language, emotion and body……….99 7.A preliminary study of the structure of emotion expressions in Squliq…………..122 8.Spatial Representation in Saisiyat………..……….146 9.Emotion in Saisiyat……….………173 10. Reference to motion events in six western Austronesian languages—toward a

1

Seediq: Spatial Representation in a Language without Prepositions

Abstract

This paper presents the linguistic mechanisms used to represent spatial relations in Seediq, focusing on the syntax and semantics of locative nouns and path verbs. It is demonstrated that most of the path categories are never specified and that topological notions are encoded only sparingly in Seediq. Where path relations are encoded, they are done in a way distinct from what is found in a "satellite-framed" language or in a "verb-framed" language. Instead, the path relations are distributed over two or more different spatial form classes in a sentence. Finally, the paper examines Jackendoff's theory of "universalist" conceptual structure in the context of the Seediq data. It is clear that the kind of conceptual structure constructed by Jackendoff has been biased by the particular languages used in building that structure, and the universality claim of Jackendoff's conceptual structure remains at best controversial. It is suggested that research into the role of linguistic form and its interaction with pragmatics to yield understandings of spatial relations would be a far more productive line of investigation.

1. Introduction

The purpose of this paper is to provide a description of the way in which spatial information is linguistically encoded in the Paran dialect of Seediq, a moderately endangered Formosan language spoken in central Taiwan, which belongs to the Ataylalic branch of the Austronesian family. The genetic classification of Seediq in the Austronesian family is shown below (Blust 1977, 1985; Li 1985):

Proto-Austronesian Malayo-Polynesian Paiwanic Tsouic Atayalic Seediq Atayal Paran Toda Truku Squliq Cu'uli'

The aim of the paper is to present the Seediq solution to the problem of how to specify the relations of the focal object and the reference object in spatial expressions. What is particularly interesting about the Seediq solution is that Seediq is a language that lacks the syntactic category of prepositions entirely and that spatial information is coded by a combination of spatial predicates and locative nouns or path verbs. This coding strategy found in Seediq and possibly other Formosan languages is different from the coding strategy found in a language like English where spatial locations are canonically encoded by prepositional phrases headed by spatial prepositions. While there is a vast amount of literature on the English system, the Seediq system has to my knowledge never been investigated.

The organization of the paper proceeds as follows. First, the basic spatial constructions in Seediq are introduced in Section 2. Section 3 deals with two types of genitive constructions involving spatial morphemes and argues that Seediq is a language that lacks the class of prepositions entirely. This is followed by a discussion of spatial representations of path notions in the language. Section 4 touches on briefly the concepts of daya and rahuc. Section 5 evaluates Jackendoff's theory of conceptual structure in the context of Seediq spatial expressions. Section 6 is the conclusion.

2. Basic Seediq Spatial Constructions

Seediq, like other Western Austronesian languages, possesses a highly complex system of focus morphology. The term 'focus' refers to the attachment of verbal focus markers in order to express a special relation between the verb and a particular noun participant. In other words, the focus system can be regarded as the manifestation of agreement in a sentence between the verb and the noun in focus. This sort of agreement is one of the characteristics of Austronesian languages.

There are four verbal focuses in Seediq: agent focus (AF), indicated by the verbal affixes -m-/mV-/ ; patient focus (PF), indicated by the verbal suffix  -un or -an; locative focus (LF), indicated by the verbal suffix -an, and referential focus (RF), indicated by the verbal prefix s-. PF, LF, and RF are collectively known as non-agent focus (NAF) systems (See Huang et al. for details).

The basic word order in Seediq is, in an agent focus construction, VOA, except where A is a pronoun, in which case, it becomes VAO. In a non-agent focus sentence, the word order is VAO, except where both argument pronouns occur in a portmanteau form, in which case, ordering is based on person rather than case: 2>1>3 (The second person pronoun precedes the first person pronoun, which in turn precedes the third person pronoun).

In a Seediq spatial construction, a focal object -- an object that is located -- is coded as an NP, and a reference object is coded as an NP which is sometimes but not always dependent on the presence of a locative noun that specifies a search domain with respect to the reference object. The relation between the focal object and the reference object is expressed by a spatial predicate waga 'be located' as in (1), (2) and (3):

Relation Reference Object Focal Object (1) Waga Pungerah Ga nasi

be located Bowl DEF pear 'The pear is in the bowl.'

Relation Reference Object Search Domain Focal Object (2) Waga Pungerah Turuma ga nasi

be located Bowl Inside DEF pear 'The pear is (deep) inside the bowl.'

Relation Reference Object Focal Object

(3) Waga Sapah Ka Takun

be located House SM PN 'Takun is at home.'

In (2), as the glosses show, there is a locative noun turuma 'the inside (of some object)' that specifies a search domain; in (1) and (3) there is no such a locative noun and the reference object is represented by an inherently locative noun (a toponym) sapah 'home, house' or by a nominal pungerah 'bowl' which can be and must be interpreted as an object defined region 'in the bowl' in the present context. In effect sentences like (1) and (3) mean that the coding for the search domain (a containment relation) relative to the reference object and the coding for the reference object itself are collapsed into a single noun. This behavior is comparable to situations in languages with case systems where nouns referring to places are unmarked in the locative case. Whereas all the spatial information in sentences (1)-(3) is carried in English by the prepositions in/inside/at, the main verb is being nearly vacuous, the semantic load in the Seediq sentences is distributed over the locative predicate waga 'be located', an NP that denotes the reference object, and an optional locative noun that specifies where the focal object is in relation to the reference object. To encode a containment relation, though not a contact and support relation in Seediq, sentences like (1) represent the canonical strategy and there is no need to linguistically specify the search domain, as in (2). The hearer must infer the exact topological nature of the spatial relationship on the basis of their common sense knowledge of the objects and events in the world. The speaker would be compelled to use (2) only when the focal object is, for example, too small in size relative to the reference object as to be inaccessible to visual perception, or when the speaker knows with certainty that the focal objet is contained in the reference object. This brings us to the important distinction in Seediq between ga, the definiteness marker and ka, the subject marker. The former is instantiated in (1) and (2) and the latter in (3).

Unlike English, Seediq provides explicit information on the accessibility to visual perception of the focal object. If the focal object is visible, or if the speaker is sure of the presence of a referentially definite focal object in the discourse context, the definite marker ga would be used; otherwise the subject marker ka would normally be the default choice. (4) and (5) are further illustrations:

(4) Waga tibu Ka babuy

be located sty SM pig

'The pig is in the pigsty.'

(5) Waga tibu Ga babuy

be located sty DEF pig

'The pig is (visibly) in the pigsty.'

Since the function of a spatial description is to tell us where to find the focal object, it is important to be fairly explicit about the geometrical nature of the search domain projected from the referent object, or the trajectories of focal objects in motion. This, however, is what Seediq usually does not or can't do. While Seediq can and sometimes does provide for functional and topological notions of containment (in), support and contact (on), occlusion (under) and proximity (beside), and relations involving projective order (in front of, in back of, behind), it lacks the grammatical machinery for specifying trajectories of objects in motion (across, over, up, down, off). Based on naturally occurring narrative data from retellings of the well-known Pear film for a total of 594 clauses, we tabulate below the distribution of spatial morphemes in Seediq which would be functionally equivalent to prepositions in English.

Table 1 Distribution of spatial morphemes in Seediq functionally equivalent to prepositions in English (Based on six Pear Stories for a total of 594 clauses)

Seediq spatial morpheme English equivalent Frequency baro ‘up above’ over, above 16 turuma ‘inside of some object’ in/inside 6 bobo 'surface of some object ' on/on top of 10 siyo 'side of some object ' by/beside 15 berah 'front of some object ' in front of 6 bukuy 'back of some object ' in back of \behind 4

muquri 'face, toward' Toward 2

kundalax 'from' From 2

toma 'underside of some object ' Under 6

? To 0 ? Up 0 ? Down 0 ? across 0 ? over 0 ? with 0 ? off 0

Table 1 shows that Seediq uses locative nouns to encode topological relations, but the language lacks the class of path prepositions entirely. As indicated above, Seediq uses locative nouns to express topological notions of containment and support and contact, though it does so only sparingly, constrained in part by discourse pragmatics (see further below for detail). Surely one does find locative nouns that gloss rather similarly to prepositions in English; still there is no underlying conceptual parallelism between, say, English on and Seediq bobo 'surface (of)'. On, like other English prepositions, is highly abstract and schematic and has a topological flexibility of applications; bobo, on the other hand, can only be applied to objects that have an upper surface. Consider the following English spatial expressions involving on and their Seediq equivalents:

(6)

a. Takun is sitting on the bed.

a'. waga tileung bobo qulapo ka Takun be located sit surface bed SM PN 'Takun is sitting on the bed.'

b. There is a poster on the wall.

b'. miqan kingan kulabuy punqupahan ka qunabin to have one poster be hung SM wall 'The wall has a poster hanging up.'

c. On the ceiling was a bug.

c'. miqan kingan kuyi ka toma na tezo baro. to have one bug SM under GEN ceiling up 'Underneath the ceiling up there is a bug.'

d. Takun put the pears on the ground.

d'. Wada pusaun na dheran nasi ka Takun ASP be put GEN ground pear SM PN 'Takun put the pears on the ground.'

e. A shirt was hanging on the hook.

e'. Waga quyaanan so daquc ka kingan syacu be located be hung vicinity hook SM one shirt 'A shirt was hung on the hook' (lit. 'in the vicinity of hook')

f. The buttons on his shirt are missing.

f'. Waga tuting botang na syacu na

ASP fall button GEN shirt his

'The buttons on his shirt are missing.' (lit., "His shirt's buttons have fallen.")

In (a) on encodes the notion of contact and support and corresponds to bobo in Seediq. In (b) since on means attachment to the side of a wall, but not to its uppermost surface, bobo can't be used in Seediq. Similarly, in (c), since on means attachment to the underneath of some object (in the present case a ceiling), bobo would be inappropriate in (c'). In (d), on indicates a surface which some object moves toward (equivalent to onto). Now, as shown above in Table 1, path categories are rarely specified in Seediq; in particular, the endpoint of a trajectory (the goal) of an object in motion is never encoded, whether that relation is indicated in English by to, onto or into. This is why bobo cannot be applied to (d'). In (e), since on indicates attachment to the end of something rather than to its surface, it is clear that bobo would not be appropriate in Seediq. Finally, in (f), on means attachment of a focal object to something and the focal object is considered part of it. In Seediq bobo applies only to the surface of a reference object on which a focal object can move. This accounts for why the attachment sense of on finds no functional equivalents in bobo.

The exercise above strongly suggests that the spatial concepts underlying the English preposition on and the Seediq locative noun bobo are different and so, consequently, are the overall sets of spatial scenes they pick out. Similar arguments can be constructed for the conceptual differences between in in English and turuma 'inside (of an object)' in Seediq. Thus the Seediq data pose a challenge to the universality claim of the basic nature of some prepositions that has been made in the literature. Thus Herskovits (1986:127) states that

"At, on and in constitute in English a fundamental set of prepositions, with the large number of distinct types of uses derived from three ideal meanings which are cognitively basic, essentially topological relations. ...

The most basic human perception of space is surely universal, and claims of universality of basic conceptual building blocks to refer to spatial relationships would seem well motivated (see, however Levinson 1996, Haviland 1993 for dissenting views). When the claim is couched in such language-specific terms, however, it has little to recommend itself.

3. Spatial Morphemes in Genitive Construction

In this section I present an overview of the spatial morphemes in Seediq describing their semantic and syntactic properties. In Seediq, locative nouns are typically construed as parts or part defined regions inherently related to reference objects. This is evident from the way in which they are coded in the grammar with respect to reference objects, i.e., in genitive constructions. Two types of genitive construction can be distinguished in Seediq (cf. also Ameka 1995). The first type has the form X na Y, where X is the possessee, na the genitive marker and Y the possessor. The second type has the form XY in which the possessee and the possessor are juxtaposed to each other without any intervening genitive marker. Although some of the detail remains to be worked out, it is safe to say that the first type, involving a genitive marker, is used to code relations between the entities perceived as less inherent or more accidental, and the second construction, involving no genitive marker, is used to code relations between the intrinsic parts of an object and the object itself, as in body-part relations and part-whole relations, including spatial relations. The narrative data from retellings of the Pear film also bear this out. In (1) below are genitive relations that make use of the first type of construction and in (2) the second type, with examples drawn from the Pear data.

(1) Relations coded as X na Y, where X is prossessee, y possessor: laqi 'child' na Takun PN "Takun's child" lunengo 'thinking' na Takun PN "Takun's thinking" rawa 'basket' na nasi 'pear' "basket of pears" Papah 'foot' na laqi 'child' "child's feet" Hunyegan 'stature, looks' na mqedin 'girl'

"looks of the girl"

Guguwan

'stealing; object stolen

na laqi 'child'

(2) Relations coded as XY, where X is possessee and Y is possessor:

siyo elu "road side" (lit. side road)

rawa turuma "inside of basket" (lit. basket inside) Quhuni baro "tree top"

Tunux laqi "head of the child" (lit.: head child) elu bbtunux "gravel road" (lit. road gravel) Papah laqi "feet of child" (lit.: foot child) bobo rulu "top of bike" (lit. top bike) heyi quhuni "tree fruit" (lit. fruit tree) toma baki "underside of old man"

bobo dheran " ground surface" (lit.: surface ground)

All of the expressions in (2) have to do with either body part relations or spatial relations containing such locative nouns as siyo 'side', turuma "inside", bobo "top (of something)", toma "underside (of something)", and baro "region over something". Table 2 gives a list of the class of spatial morphemes in Seediq (including both locative nouns and path verbs).

Table 2 Seediq spatial morphemes

Spatial Morpheme Gloss Source

baro 'above, over' baro 'region above some object berah 'in front of' berah 'bosom'

bukuy 'behind bukuy 'buttocks'

bobo 'on' bobo 'upper surface (of object)' ngerac 'outside' ngerac 'outside (of object)' so 'about, around' so 'vicinity (of something)' siyo 'beside' siyo 'side (of an object)' toma 'under' toma 'underside, armpit' turuma 'inside' turuma 'inside (of an object)' irin 'left' irin 'left hand'

narac 'right' narac 'right hand' (mu)quri 'face; toward' (mu)quri 'face; toward'

kundalax 'from' kundalax 'from'

Some of the locative nouns have evolved from body-part terms, others from object-part terms and none has evolved from environmental landmarks (cf. Bowden 1991, Svorou 1994). Left hand (baga irin) and right hand (baga narac) are still in use,

but left and right as spatial morphemes are disappearing from among the current generations of Seediq speakers, who often have to resort to some such locations as the weaker side/ the stronger side or this side/that side accompanied by gesture. Cardinal concepts (north, south, east, west) are not part of the vocabulary of the language.1 This leaves the present day Seediq language with only the intrinsic system, which is based on the inherent features of objects or topological descriptions, and the relative system (minus the concepts of right and left), which is based on anthropocentric concepts like front and back, left and right, in the linguistic representation of spatial orientation (cf. Fillmore 1971, Levinson 1994, Heine 1997).

It is well known that stative spatial grams (adpositions) often owe their genesis to the grammaticization of head nouns in genitive constructions, arriving ultimately at a stage at which they are bound to another form as an affix. Similarly, dynamic motion verbs pass through a stage at which they are frequently used in a serial verb construction and a stage at which they lose some of the verbal components in their semantics, subsequently arriving at a stage at which they become directional path grams (Svorou 1994). Neither locative nouns nor motion (path) verbs in Seediq can be justifiably argued to be spatial grams. A locative noun in conjunction with a nominal denoting the reference object routinely takes the subject marker ka and a predicate phrase to form a sentence, just as other types of NPs do, as in (7) and (8), or function as object of a transitive verb, as in (9), again as other types of NPs do:

(7) Niqan kingan kuyi ka toma na tezo baro exist one bug SM under GEN ceiling up "There is a bug up there on the ceiling."

(lit.: "the underside of the ceiling up there has a bug")

(8) Niqan patis ka bobo cukuwe exist book SM surface desk "There is a book on the desk."

(lit.: "the surface of the desk has a book")

(9) Muda so2 susiyo alang ka Takun pass vicinity side village SM PN

"Takun passes by the village."

(lit.: "Takun passes by the side of the village.")

Path verbs like quri 'face; move toward' and kundalax 'from' often occur in a

serial verb construction, as in (10) and give the impression, from the perspective of their translations into English, that they are syntactically prepositions. However, just as locative nouns are not prepositions, the path morphemes are morphologically verbs. First, in Seediq, as in other Formosan languages, constructions that string together two or three motion verbs are fairly common. (10) is an illustration.

(10) Wada mukukesa muquri gakko ka Takun leave walk move toward school SM PN "Takun has left walking toward the school."

Second, path verbs (or path spatial morphemes, in order not to prejudge the case), like other true verbs, are inflected for the verbal categories of focus, tense or aspect.

(11) muquri ku sapah yaku, quri yayung isu

face I home I face river you

"I face home, you face the river."

(12) Kundalaxi hiya bulebing isu

from there pull you

"You pull from there."

In (11) mu- is the agent focus prefix attached to the verb stem quri 'face, toward', and it is the stem form that appears in an imperative sentence, as in the second clause in (11). In (12), kundalaxi 'from' is suffixed with the imperative marker for patient focus -i.

The strongest piece of evidence for claiming that path spatial morphemes are indeed verbs comes from the fact that in a sentence like (13) kundalax 'form' takes the genitive marker na to mark an agent. Only verbs in patient or locative focus constructions, in Formosan languages and in many of the Austronesian languages are privileged to do so:

(13) Kundalax na rawa turuma mangan heyi quhuni ka Takun from GEN basket inside take fruit tree SM PN "Takun took the tree fruit from inside the basket."

Talmy (1985, 1991) has distinguished a 'satellite-framed' language from a 'verb-framed' language. 'Satellite-framed' languages, which include most

Indo-European languages, characteristically express path notions (movement into, out of, up, down, off...) in a constituent that is a satellite to the main verb, such as a prefix or a particle/preposition. On the other hand, languages such as Hebrew, Turkish, Spanish and Korean, which express path in the verb itself, are verb-framed languages (See Bowerman 1996 for facts about Korean). Talmy's scheme of classification is an oversimplification, however. We have shown above that Seediq, a language which lacks the grammatical category of prepositions, is generally not interested in, or lacks the grammatical machinery for, encoding certain aspects of the path notions. When it is capable of doing so, its strategy is typically to use combinations of motion verbs, path verbs and locative nouns in complex serial verb construction-like configurations. (14) and (15) are further illustrations.

(14) Wada tuting yayung ka Takun Leave fall river SM PN "Takun fell into the river."

(lit.: "Takun fell river.")

(15) Mutugiya kususiyo sapah tumalang ka Takun

Circle along house run SM PN

"Takun ran around the house."

(lit.: "Takun circled by running and following the sides of the house.")

Since the relevant path notions are distributed across the spatial morphemes (including motion verbs, path verbs and locative nouns) in a Seediq sentence, it would be more appropriate to type it and other similarly behaved languages as "sentence-framed" languages, in contradistinction to satellite-framed and 'verb-framed' languages.

Consider what a Seediq speaker must know in order to function effectively using this spatial system? It is clear that Seediq speakers must be able to locate the relevant topological relations and path concepts so they can describe object locations and trajectories. They must constantly monitor the relevant relations and directions. In some sense, of course, understanding sentences where topological relations are not fully specified seems trivial enough. Thus a sentence like Many people the street could only mean, given some knowledge of the basic facts of Seediq syntax, that there are many people on the streets; Be located hiding the vicinity of bushes Takun could only mean Takun is hiking in the bushes. Even path categories seem in general easily inferable. Thus Be put by someone the bike must mean was put on the bike by someone; Is writing blackboard Takun must mean Takun is writing

on the blackboard; Has fallen river Takun must mean Takun has fallen into the river. It is as if we were asked to take a cloze test in English, with all of the appropriate path prepositions left out, and we would still fare reasonably well, since we can rely on a combination of multiple semantic cues and discourse pragmatics to solve the problems.

4. Daya/rahuc

The part of the Seediq region where I did my field work consists of a string of villages located in a valley surrounded on the north and south sides by steep rolling hills, part of the rugged, powerful Central Mountain Range that dominates that landscape. From this area, known as Gluban, the nearest town, Puli, is ten miles to the east, which the villagers can get to by car or motorcycle or bus. Villages that live uphill are termed alang daya (lit. village uphill) and villages that live downhill are alang turahuc (or alang hunac) by villagers. The words daya (or tugudaya) means ‘(land which is) uphill’ and turahuc (or tugurahuc) means ‘(land which is) downhill’. Daya has evolved from PAN * daya ‘upriver, towards the interior’ and tu(gu)-rahuc from PAN * lahud ‘downriver, towards the sea’. They are reflexes in a huge number of daughter languages in the Austronesian area. Blust (1997) points out that Proto-Malayo-Polynesian (the ancestor of all non-Formosan Austronesian languages) makes reference to two fundamental orienting features: a land-sea axis, which is more localized, and the south-east Asian monsoons that represent an axis with a broader geographical basis. No modern Formosan Austronesian languages utilize a land-sea axis for spatial reference, suggesting that the Formosan system is one adapted to a life on land-locked hills and mountains in which access to the sea has been entirely absent following their split from Proto-Austronesian several millennia ago.

5. Seediq and the Status of "Conceptual Structure"

This section will evaluate Jackendoff (1983, 1990)'s work on conceptual semantics in the context of our understanding of spatial representation in Seediq discussed in the preceding sections. The central issue to be addressed is whether a universal conceptual structure of spatial relationships can be meaningfully constructed independently of different syntactic or semantic structures in different languages (Cf. Tai 1993). Jackendoff's proposal is a universality claim of basic conceptual building blocks to refer to spatial relationships. It involves a number of technical innovations and is conceived within a modular theory of generative grammar. Jackendoff proposes a level of conceptual structure defined as "a single level of mental representation ... at which linguistic, visual, sensory, and motor information are

compatible" (1983:17) and concludes that "the terms semantic structure and conceptual structure denote the same level of representation" (1983:95). Conceptual structures are generated by conceptual formation rules, just as syntactic structures are generated by syntactic formation rules. Conceptual structures are linked to syntactic structures by correspondence rules.

Jackendoff's conceptual formation rules operate on a finite set of ontological categories such as THING, PLACE, PATH, EVENT and STATE. These conceptual categories can be further elaborated by a series of conceptual functions characterized by a system of context-free rewrite rules. Three of the most important conceptual formation rules formulated in Jackendoff (1990:43) for the spatial domain are (16) , (17) and (18):

(16) [PLACE] [ place PLACE - FUNCTION([THING])]

(17)

[

]

(18) [EVENT] [event GO ( [THING] , [PATH] )]

Rule (16) states that the conceptual category PLACE is construed as a PLACE-function with an argument of the category THING. The reference object serves as an argument for the PLACE-FUNCTION to define a region. In a spatial expression in the bowl, the bowl denotes the reference object and preposition in maps the reference object into the region inside it. Rule (17) states that the conceptual category PATH is construed as a function that maps a THING or PLACE into a path. In an expression from the park, the preposition to maps the reference object to a path. Rule (18) states that the conceptual category EVENT is a function that maps a THING to a PATH function. In these rules, spatial prepositions are construed as functions which map reference objects or places into regions or paths. The features and functions are chosen as primitives to express linguistically and conceptually significant generalizations in the data. The result is a conceptual structure which encodes meaning as a complex algebraic expression.

The approach reminds us of the lexical decomposition of the kind first advanced by Katz and Fodor (1963), despite the difficulties spelt out by a number of distinguished researchers in semantics over the years (Kempson 1977; Fillmore 1985;

Fodor et al. 1975, among others). Functions like TO, FROM, TOWARD, VIA etc. are 'primitives' where similarity to English prepositions is neither addressed nor explained. Indeed, Jackendoff's conceptual formation rules like (16) (17) and (18) are stated entirely in terms of a 'satellite-framed' language. As in much of the linguistic theorizing where one's meta-language is often, perhaps inescapably, biased by one's object language, Jackendoff's theory of conceptual semantics is no exception. Jackendoff's conceptual formation rules depend crucially on the role of prepositions as function mappers. Now prepositions in English encode highly schematic information about topological regions, main axes and trajectories, but not Euclidean information about angles and distances or about the exact shape or nature of the ground and especially the focal objects. Spatial expressions in other languages do not necessarily work this way, however. Levinson (1996) argues that Tzeltal is a language that utilizes absolute coordinates, together with a rich system of intrinsic distinctions that pick out a number of Euclidean properties of the focal objects and that the Guugu Yimithirr speakers of N. Queensland use a system of absolute orientation (similar to cardinal directions) which fixes absolute angles regardless of the orientation of the reference object. There is simply no analogue of the Indo-European prepositional concepts in these languages.

Similarly we have shown that the path categories in Seediq (excepting the concepts of toward and from) are never encoded for lack of the necessary grammatical machinery and that topological notions are expressed only sparingly. This is no doubt a consequence of the fact that Seediq lacks a class of spatial prepositions entirely, as demonstrated above. Spatial relation is something which must be expressed in English by prepositions, but it is something which is only optionally encoded in Seediq.

Even as a lexical category in English, Jackendoff's treatment of prepositions is seriously inadequate (Deane 1996). Prepositional polysemy is pervasive and systematic in the language (and perhaps in other languages as well), casting serious doubt on their status as 'primitives", and consequently on the whole project of constructing universal conceptual structure, motivated largely by the spatial semantic structures of English.

6. Conclusion

This overview is preliminary and incomplete. Within the scope of this paper it would be an impossible task to treat exhaustively the whole range of the system of spatial representation in Seediq. I have therefore limited myself to areas with which

I am more familiar, focusing on the syntax and semantics of locative nouns and path verbs. For some spatial expressions in Seediq it is the combinations of the spatial predicate waga and locative nouns in construction with a nominal reference object that are crucial for their understanding. For others, it is the combinations of the spatial predicate and a nominal reference object (a toponym or an activity noun) that are relevant. The search domain is generally not specified for the containment relation. For other relations, given sufficient context, combinations of the spatial predicate and a nominal reference object are also all that is necessary. The interpretation of the constructions is the result of the interaction between the meaning of the spatial predicate and the semantics of the nominal that represents the reference object plus discourse pragmatics.

The paper has also shown that most of the path categories are never specified in Seediq. Where path relations are encoded, they are done in a way distinct from what is found in “satellite-framed” languages and “verb-framed” languages. Instead, the path relations are distributed over two or more different spatial form classes in a sentence. Finally the paper examines Jackendoff’s theory of “universalist” conceptual structure in the context of the Seediq data. It is clear that the kind of conceptual structure constructed by Jackendoff has been biased by the particular languages used in building that structure, the universality claim of Jackendoff’s conceptual structure remains at best controversial. It is suggested that research into the role of linguistic form and its interaction with pragmatics. to yield understandings of spatial relations would be a far more productive line of investigation.

Linguistic meaning is always underspecified and unique reference is never purely a linguistic matter. What secures uniqueness is the user of the expression and the context in which it is used together with the expression. So is spatial reference. It is appropriate as a final note to paraphrase Quine (1971:144), “spatial reference is impossible apart from the network of terms, predicates and auxiliary devices that speakers of a language share. Like semantic content more generally, spatial reference arises only through combining a linguistic expression with an interactive context.”

Footnotes

1. This is consistent with Blust’s (1997) statement that in reconstructing earlier stages of Austronesian no basis is found for positing terms for cardinal directions, even though cardinal direction terms have entered a number of modern languages either through borrowing or through internal semantic change.

2. So ‘vicinity’ has a variety of functions. It can be attached to spatial morphemes or expressions to mean ‘around, about’, as in

so hiya ‘around there’

so tutingan rulu ‘around where one gets off the bus’ so nigan boru hiya ‘around where the ball is’

when it occurs with a noun denoting an object, it means ‘something like’, as in kulaan so nasi ka kiya ‘that could be a pear or something like that’ when it occurs with a predicate, it means ‘somehow or other’, as in so dumayo ‘help a little; help here and there’

References

Ameka, Felix K. 1995. The linguistic construction of space in Ewe. Cognitive Linguistics 6.2:139-181.

Blust, Robert. 1977. The Proto-Austronesian pronouns and Austronesian subgroupings. Working Papers in Linguistics 9.2:1-15.

---. 1985. The Austronesian homeland: a linguistic perspective. Asian Perspectives 26.1:45-67.

---. 1997. Semantic Change and the Conceptualization of Spatial Relationships in Austronesian Languages. In Senft, Bunter (ed.).

Bowden, John. 1991. Behind the preposition: grammaticalization of Locatives in Oceanic Languages. MA thesis, University of Auckland.

Bowerman, Melissa. 1997. Learning how to structure space for language: a cross-linguistic perspective. In P. Bloom, M. Peterson et al. (eds), Language and Space, 385-435. Cambridge University Press.

Deane, Paul D. 1996. On Jackendoff's conceptual semantics. Cognitive Linguistics 7.1:35-91.

Fillmore, C. J. 1971. Santa Cruz lectures on deixis. Bloomington: Indiana University Linguistics Club.

---. 1985. Frames and the semantics of understanding. Quaderni di Semantica 6.2:222-253.

Fodor, J. D. , Fodor, J. A. and Garrett, M. 1975. The unreality of semantic representation. Linguistic Inquiry 4.515-531.

Haviland, J. B. 1993. Anchoring, iconicity and orientation in Guugu Yimithirr pointing gestures. Journal of Linguistic Anthropology 3.1:3-45.

Heine, Bernd. 1997. Cognitive foundations of grammar. Oxford University Press. Herskovits, A. 1986. Language and spatial cognition: an interdisciplinary study of

prepositions in English. Cambridge University Press.

Huang, Shuanfan, I-wen Su and Hsiu-hsu Lin. 1998. A functional reference grammar of Seediq. Unpublished manuscript, National Taiwan University.

Jackendoff, Ray. 1983. Semantics and cognition. Cambridge: MIT Press. ---. 1990. Semantic Structures. Cambridge: MIT Press.

Katz, J. and J. Fodor. 1963. The structure of a semantic theory. Language 39.2-170-210.

Kempson, R. 1977. Semantic Theory. Cambridge University Press.

Levinson, S. 1994. Vision, shape and linguistic description. Linguistics 32: 791-855. ---. 1996. Relativity in spatial conception and description. In J. Gumperz and S.

University Press.

Li, Paul Jen-kuei. 1985. The position of Atayal in the Austronesian family. In Andrew Pawley and Lois Carrington (eds), Austronesian Linguistics at the 15th Pacific Science Congress. Pacific Linguistics, C-88, pp.257-280.

Quine, W. 1971. The inscrutability of reference. In d. Steinberg and L. Jacobovits (eds.), Semantics: an interdisciplinary reader in philosophy, linguistics and psychology. 142-154. Cambridge University Press.

Senft, Gunter (ed.). 1997. Referring to Space: studies in Austronesian and Papuan Languages. Oxford University Press.

Svorou, Soteria. 1994. The grammar of space. John Benjamins.

Tai, James H-Y. 1993. Conceptual structure of Chinese spatial expressions. CLS 29: 347-362.

Talmy, Len. 1983. How language structures space. In H. Pick and L. Acredolo (eds). Spatial orientation, 225-281. NY: Plenum.

2

Spatial Representation in Tsou

This chapter is a preliminary study of the basic linguistic mechanisms used to represent spatial relation in Tsou, an endangered Formosan language which belongs to the Tsouic branch of Austronesian family. We would like to argue that Tsou is a language without prepositions and its spatial information is encoded into spatial predicates, the semantic interpretation of “focal object”, or the interaction between two of them. It is instructive to observe that this language does not distinguish at/on/in which are treated as the fundamental set of prepositions in English by Herskovits (1986). However, it provides topological notions of containment “inside”, relations involving projective order (in front of, in back of, behind), trajectories of objects in motion (by, across, over, around), and so on. The mechanisms it uses to perceive spatial information include the interpretation of focal objects and reference objects, the predicates with rich spatial information, the predicates that contain identifiable spatial morphemes, the combinations of both of focal objects and spatial predicates, locative nouns, and reduplications. There are five sections in this chapter. Section 1 is an introduction to the languages, especially its case marking system; section 2 presents the syntactic evidence for its lack of prepositions. Section 3 presents the linguistic mechanisms used to represent spatial information in this language. Section 4 is a summary.

1. Introduction

Tsou is a VOS language syntactically with a Philippine-style focus system including Agent Focus and Non-Agent Focus (PF: Patient Focus; LF: Locative Focus; BF: Benefactive Focus). It has a rich system of case markers (or articles) involving intrinsic semantic features of visibility, definiteness, distance, and so on. Table 1 is tsou’s case marking system in isolated utterances:

Table 1. Tsou’s case markers in isolated utterances

Visibility Nominative case markers

Oblique case markers Visible proximate ‘e

(specific/definite) intermediate si ta

distal ta specific ‘o to

Invisible not experienced na no

locative ne

sensation co ninca/nca The case marking system in Tsou is divided into two categories: nominative vs.

oblique case markers. /’e/, /si/, /ta/ are nominative case markers used to mark NPs which are [+visible] and [+specific]/[+definite]. Because of certain semantic extension, the nominative case marker /’e/ can also mark NPs which are [–visible] but [+specific]/[+definite]. The selection of ‘e/si/ta depends on the speaker’s recognition of the distance between other objects and himself: near/middle/far. The nominative case marker /’o/ marks NPs which are [–visible] but [+specific]. /na/ is used when its following NP is [–experienced] (and [–visible], of course). For oblique case markers, /ta/ is the corresponding one of ‘e/si/ta; /to/ of /’o/, and /no/ of /na/. The nominative marker /co/ and its counterpart /ninca/ are special for they deal with sensation only. /ne/ is an oblique case marker and must relate to [+location]. The purpose of introducing Tsou’s case marking system is to show that this language has requirement to describe certain spatial information. For example, case markers

can reveal information like visibility, distance (between speakers and objects as a spatial relation), and even how the information of objects is gained (i.e., through visual perception or audio perception). Similar distinctions on spatial relation can also be found in its demonstratives as /eni/ “here”, /sico/ “there”, /tonoi/ “there (further)”. Rukai, a Formosan language of Paiwanic branch of Austronesian family, also has detailed distinctions of “distance” between speakers and objects. The case marker plus its following NP in Tsou may reveal a lot of semantic information more than we imagine, and their interaction with spatial predicates represent perfectly any kind of spatial information even though this language lacks prepositions.

2. The Lack of Prepositions

The basic spatial construction in Tsou is based on the combination of spatial predicate (V) relation and the interpretation of Reference Object and Focal Object. Unlike English, there is no conceptual parallelism of in/on/at in this language. Consider first some basic examples:

Relation Focal Object Reference Object (1) mo eon ta takubingi ‘e nasi.

AF locate Obl bowl Nom pear

“This pear is in the bowl.” (Fieldnotes) (2) mo eon ta emoo ‘e pasuya.

AF locate Obl house/home Nom Pasuya

“Pasuya is at home.” (Fieldnotes) (3) mo eusungu ta hopo ‘e pasuya.

AF sit Obl bed Nom Pasuya

“Pasuya is sitting on the bed.” (Fieldnotes)

The three sentences above are all in Agent Focus, and in the surface structure the “reference object” (or trajectory in Langackerian terminology) must be in subject

position and the “focal object” (or landmark in Langackerian terminology) in object position. (1)-(3) in which the focal objects represent the ideas of in/at/on respectively in English do not make any distinctions in Tsou. If tested in existential sentences (the pan-to construction), we still get the same result. Here are some

examples of “on” in English under existential construction:

(4) pan-to conci eoi ta sofU. Exist-Obl. one worm Gen ceiling

“There’s a worm on ceiling.” (Fieldnotes) (5) pan-to [i-si ngov’eni to tonghifza ci] poyave.

Exist-Obl NAF-3rd hang-NAF Obl wall Rl sword

“There’s a sword which is hung on the wall.” (Fieldnotes) (6) pan-to [mo cUnghU to f’uhu-su ci] eoi.

Exist-Obl AF stick Obl back -2nd Rl worm

“There’s a worm which sticks on your back.” (Fieldnotes) Although Tsou does not distinguish the spatial information inherent in English in/at/on, it doesn’t mean that this language lacks the equivalent conception embedded in the English prepositions. Aside from case markers which are known to be pregnant with rich spatial information, in many cases the spatial predicates themselves reveal enough spatial information, for example, nooeo “inside” in example (7), f ’uhu

“(in) back (of)” in (8) (note that f ’uhu can also be a noun; see (6)), and ~caefi

“over/cross” in (9):

(7) mo nooeo ta takubingi ‘o eoi.

AF inside Obl bowl Nom worm

“The (invisible) worm is inside the bowl.” (Fieldnotes) (8) mo coni ‘o mo f’uhu ta ca’nU ci tposU.

AF one Nom AF in back of Obl chair Rl book

“There’s a book which is in back of the chair.” (Fieldnotes) (9) mo su-caefi to taipahu nehucma ‘o pasuya.

AF _over/cross Obl Taipei yesterday Nom Pasuya

3. Mechanisms for Spatial Representation

From the previous section we know that Tsou is a language without overt prepositions and it doesn’t distinguish in/on/at which were treated as the most fundamental prepositions in English. We also know that though there is no preposition, it can still represent certain spatial information by using spatial predicates as shown in (7)-(9). In this section we’ll argue that there are more than one mechanism used in this language: by the interpretation of focal objects, the predicates with spatial information incorporated, the spatial predicates which represent the conception of prepositions themselves, the combination of both spatial predicates and focal objects, and the use of locative nouns. The following sub-sections will take up each of these possible mechanisms, offering both syntactic and morphological evidence to support the thesis.

3.1 The Interpretation of Focal Object & Reference Object

Recall that in examples (1) and (2) we have eon ta takubingi to be “in the bowl”

and eon ta emoo to be “at home”. And in (18) we have eon ta ca’nU to be “on the

chair”. Since in all of the three examples the same spatial predicate is used: eon,

why does each predicate expression have a different interpretation? The obvious explanation must be that we get the spatial information from semantic interpretations of “focal objects” in relation to the “reference object”, for example, the “bowl” in (1), “home” in (2), and “chair” in (18). Now consider the example (10). The spatial predicate “hang” either expresses an act ngov’eni or a state ngov’o. A check through

the entry “hang” in English shows that it can go with a wide array of prepositions such as hang on, hang down, hang of, hang about, hang back, hang out, and hang over. This suggests that the verb “hang” does not really contain any relevant spatial information until it is combined with a spatial preposition in English. The predicates

ngov’eni “hang” in (10a) and ngov’o “hang” in (10b) do not entail any prepositional

information, but why does the final translation contain the preposition “on”? Such spatial recognition is obtained through the semantic interpretation of the focal object

tonghifza “wall” in relation to the reference object poyave “sword”.

(10) a. i-si ngov’eni ta tonghifza ‘e poyave.

NAF-3rd hang-NAF Obl wall Nom sword

“The sword was hung on the wall.” (Fieldnotes) b. mo ngov’o ta tonghifza ‘e poyave.

AF hang-AF Obl wall Nom sword

“The sword is hanging on the wall.” (Fieldnotes)

3.2 Predicates Containing Rich Spatial Information

In this section a number of spatial predicates containing rich spatial information are introduced. This category is different from the one in section 3.3 for the verbs here contain no identifiable spatial morphemes but themselves embody inherent spatial information given by their intrinsic lexical semantics. (11)-(14) are the examples:

(11) mo sunghucu ‘o pasuya.

AF fall_into_water Nom Pasuya

“Pasuya fell into the water.” (Fieldnotes) (12) p’etpUti ta pasuya ‘o mali.

step_on Obl Pasuya Nom ball

“The ball was stepped on by Pasuya.” (Fieldnotes) (13) ‘e paicU, mita bichipi ta saungU.

Top PaicU AF-3rd stand beside Gen SaungU

“PaicU stands beside SaungU.” (Fieldnotes) (14) nahocu teongasi si emucu-su.

allow put_down Nom hand-2nd

3.3 Predicates Containing Identifiable Spatial Morphemes

The use of spatial predicates is the most common linguistic mechanism for representing spatial information in this language. These spatial predicates are of two types. To the first type belong spatial predicates that contain no identifiable spatial morphemes but are themselves rich in spatial information; the second type refers to those spatial predicates that contain an identifiable spatial morpheme. These two types are taken up separately below:

/nooeo/: inside

(15) mo nooeo ta feongo ‘o fkoi.

AF inside Obl cave Nom snake

“The (invisible) snake is inside the cave.” (Fieldnotes) /nooeo/ also occurs in (7) nooeo ta takubingi “inside the bowl”. /nooeo/ is

used, according to our informants, only when the “reference object” is invisible and the “focal object” is a container. Thus the reference objects in both (7) and (15) must be invisible and in a relation of “containment” with its focal object. This usage distinguishes eon “to be located; (in/on/at)” from nooeo “inside”.

/i’mi/: from

(16) te-ko i’mi ta’e ho e’tUi; te’o i’mi tan’e ho etUi.

Fut-2nd from there Conj pull Fut-1st from here Conj.pull “You’ll pull from there and I’ll pull from here.” (Fieldnotes) (17) aUlU cono hie ceoconU ho mo i’mi to emoo-taini ho

just one day walk if AF from Obl house-his Conj uh to emoo ta pasuya.

go Obl house Gen Pasuya

“Going to Pasuya’s house from his house takes a whole day’s walk.”(Fieldnotes) (18) mio i’mi ne khagi.

AF-1st from Obl Chia-Yi

“I am from Chia-Yi.” (Shoes: 1)

rather than a preposition by its syntactic position, since a verb in Ttsou is located between an auxiliary and an NP.

/~’unu/: to

(19) mi-cu mongoi ho emo’unu ne ‘oeona-tmopsU

AF-Perf leave Conj walk-to Obl location-study

“(He) had left and walked to the school.” (Fieldnotes) (20) mi‘o mi’unu ta emoo’u, mi-ko mi’unu ta va’nU.

AF-1st face-to Obl house-1st AF-2nd face-to Obl river

“I face my house, and you face that river.” (Fieldnotes)

The only difference in Tsou between /i’mi/ “to be from” and /~’unu/ “to” is that /i’mi/ must be in a free form while /~’unu/ must be in a bound form as a suffix. For example, we have toe’unu “run to~”, mi’unu to be “face to~”, (e)mo’unu “walk to~”.

The spatial morpheme /~caefi/ must also be a suffix rather than a word. It is interesting to observe that /~caefi/ covers a range of meanings including “through”, “by”, and “across”. Recall that in (9) we have su-caefi to taipahu “pass through

Taipei” which is very similar to to-caefi su “walk by you” in (22). However, if we

take the idea “by” to interpret mei-caefi in (21), it would be a little inappropriate (but

still work, in certain sense). /~caefi/ can also be treated as “cross/across”, as is suggested by example (28) in section 3.4.

/~caefi/: through/by/across

(21) o’a mocu i’si tiv’v’oha hu-sansana, mi-cu meicaefi

Neg NAF-3rd in time for see-clearly AF-Perf fly-over ta fnguu-si ‘o zomU.

Obl head-3rd Nom bird

“Before he realized it, the bird had flown over his head.” (Fieldnotes) (22) mon’a aUlU tocaefi su ‘o pasuya,

then just walk-by 2nd Nom. Pasuya o’a i-ko maka teolUi? Neg NAF-2nd to one’s surprise see

/~upu/ is a spatial suffix which means “with”: /~’upu/: with

(23) su’upu si mo maica ci hia yonghu ci mamespingi,

go_with Nom AF like_this Rl how beautiful Rl female mi-ta cus’a noana’o tUe’Ue’UhU?

AF-3rd must for a long time laugh in secret

“Such a beautiful lady sat beside him; he must laughed in secret for a long time.” (24) mo no’upu su na hamo.

AF be-with 2nd Nom God

“God is being with you.” (Fieldnotes) /~eafo/:out

(25) mo euso ci polo ‘e mo eueafo ta ceoa.

AF two Rl earthworm Nom AF crawl_out Obl earth “The crawling (things) out of the earth are two earthworms.” (Fieldnotes) (26) isi taeafa ta motoevi ‘e chumu ta kopu.

NAF-3rd shake-out Obl earthquake Nom water Gen cup “The earthquake shook the water in the cup out.” (Fieldnotes) /~ovei/: back

(27) mei-ovei ‘fly back’; mei~: fly e-ovei ‘come/go back to the topic’; e~: words eu-ovei ‘go back’; eu~: move eo-ovei ‘return’ eo~: hand /~avovei/: back and forth

(28) mei-avovei ‘fly back and forth’

3.4 Combinations of Space Predicates and Focal Nouns

So far we have seen that the spatial information in Tsou is expressed either by an interpretation of focal objects in relation to reference objects, or predicates with inherent spatial information, or spatial predicates that contain identifiable spatial morphemes. However, sometimes we have to take both the verbs and the objects into consideration so that we can fully understand what the sentence is about. Take (29) for example, su’ means “go”, and only when it goes with the oblique to + NP can

sucaefi ta va’hu in (30) means “cross the river” & sucaefi to mo eusno emoo in (31)

means “pass by two houses”.

(29) mo smoteUsU nehucma ho su’ to eo’hunge.

AF slip yesterday Conj go Obl valley

“(He) slipped yesterday and fell into the valley.” (Fieldnotes) (30) ake’ i n’a buveici; ho tac’u afu’u sucaefi ta va’hU

a little be patient if/when only go_across Obl river, ‘a tec’u sUc’UhU. (Fieldnotes)

arrive

“Be a little patient; we’ll arrive (there) only if we have crossed the river. (31) ta-hoza tan’e, sucaefi to mo eusno emoo, ta-ko cu

begin here go_by Obl AF two house AF-2nd Perf sUc’UhU to emoo to pasuya. (Fieldnotes) arrive Obl house Gen Pasuya

“From here you just pass by two houses, and you will arrive at Pasuya’s house.”

3.4 The Use of Locative Nouns

The characteristic of this category is that the spatial information resides in locative nouns. The locative nouns in Tsou include the notions of ‘left & right’, ‘nearness & side’, ‘this side & that side’, ‘below & under’:

/veina/ vs. /vhona/: left vs. right

(32) cuma na ongko ta hcuyu ta veina/vhona-su.

what Nom. name Gen hill Gen left/right-2nd.Poss

“What is the name of the hill in your left/right side? (Fieldnotes) There is no distinction of ‘west/east’ or ‘north/south’ in Tsou but ‘left and right’ (veina vs. vhona) and ‘high/low place’ (omza vs. oii). Though Tsou does not have

the ideas of ‘west’ and ‘east’, it has expressions like esmomha hie, which means

‘the place in which the sun rises’. The use of /cum’u/ ‘nearness’ and /feona/ ‘side’ is similar to ‘left & right’:

/cum’u/: nearness

(33) mo eon ta cum’u ta hcuyu ‘e pasuya.

AF exist Obl near Gen hill Nom Pasuya “Pasuya lives near the hill.”

/feona/: side

(34) mi’o eon to feona to hia-peoza.

AF-1st exist Obl side Gen bridge “I live beside the bridge.”

/tanesi/ vs. /taesi/: this side vs. that side

The semantic interpretation of two deictic locative nouns tanesi “this side”and taesi “that side” are jointly determined by the relations among the trajector (=subject),

the reference point, and the speaker’s location at the speech act time. The uses of

tanesi and taesi are thus completely speaker-oriented. Consider (35) & (36). For

example, if the speaker is facing the main entrance of the store veiyo and his house is

closer to him than the store, (35) will be used. On the other hand, if the speaker is facing the main entrance of the store but his house is farther away from him than the store is, (36) will be used.

(35) mi’o eon to tanesi ne veiyo.

AF-1st exist Obl this side Gen Veiyo (name of a store)

“I live in the place which is in Veiyo’s side.” (Fieldnotes) (36) mi’o eon to taesi ne veiyo.

AF-1st exist Obl that side Gen Veiyo (name of a store)

“I live in the place which is cross Veiyo.” (Fieldnotes) Similar interpretations apply to (37) & (38). In (37), PaicU is the focal object

and SaungU is the reference object. Assume the speaker is as shown in the picture

below at the time of speech act, then (37) means PaicU is closer to the speaker (i.e.

she is at this side of SaungU). In (38), while SaungU is also a reference object, PaicU is farther away from the speaker (i.e. she is at that side of SaungU).

(37) ‘e paicU, mita eon ta tanesi ta saungU.

Top PaicU AF-3rd exist Obl this side Gen SaungU

(38) ‘e paicU, mita eon ta taesi ta saungU.

Top PaicU AF-3rd exist Obl that side Gen SaungU

“PaicU is in back of SaungU.” (Fieldnotes)

reference

point

Î

PaicU front reference SaungU front

point Î

Speaker Speaker

tanesi ta saungU ‘e paicU taesi ta saungU ‘e paicU

/f’uf’u/ vs. /~peohna/: underside vs. downside

(39) & (40) are examples of f ’uf ’u ‘the underside of something ’; (41) is of ~peohnU ‘the downside’.

(39) pan-to mo aemo’U ta f’uf’u ne evi ci beahci.

Exist-Obl AF spilt Obl underside Gen tree Rl fruit

“There are fruits which are spilled under the tree.” (Fieldnotes) (40) mo coni ‘o mo nof’uf’u ta ca’nU ci tposU.

AF one Nom AF be-underside Obl chair Rl book

“There’s a book which is under the chair.” (Fieldnotes) (41) mo supeohU ‘o potingta ta eUsU.

AF go-downside = fall Nom button Gen cloth

“The button on the clothes fell/was missing.” (Fieldnotes) Similar to the distinction of “under” and “below” in English, f ’uf ’u and peohU/peohna differ in whether there is a cover above them. f ’uf ’u is the underside

of some reference object, so that it must be an NP that specifies the reference object, as shown in (39) and (40). The antonym of f ’uf ’u is skopu. We have m’eoskopu

‘tread on the upside’ & m’ef ’uf ’u ‘tread on the underside’. Different from f ’uf ’u, peohU/peohna only refer to downside and need not to specify a reference object. In

f’uf’u vs. peohna:

peohna f’uf’u

3.6 Reduplication

Reduplication in Tsou, generally speaking, is realized by reduplicating the first syllable of a stem morpheme. In (42), /ma’~/ is a morpheme which means “walk”, and /~kikiegni/ is formed by reduplicating the first syllable of /~kiegni/. (43)-(45) are the examples of the verb stem /~kukuyunvu/ ‘around’: fly around, be around the house, and walk around. According to our informant, there is no free morpheme functioning as a syntactically independent word kuyunvu. Besides, the difference

between /~kikiengi/ and /~kukuyunvu/ is not as yet clear.

(42) mon’a aUlU ma’ki-kiengi ta emoo ‘o oko.

then just Red-walk around Obl house Nom child “The child kept on walking around the house from that moment.” (Fieldnotes)

/kukuyunvu/: around

(43) mei-kukuyunvu ‘fly around’

(44) au-kukuyunvu ta emoo

_around Gen house “around the house”

(45) emo-kukuyunvu ho pasunaeno

walk-around Conj sing “walk in a circle and sing”

5. Summary

We have shown that Tsou is a VOS language with a complex focus system and a system of case marking which distinguishes a set of case markers along such parameters as [visibility] [proximate] [distal] as part of their intrinsic semantic features. Such detailed distinctions in this language imply the importance of spatial relation in the semantics of case markers. We have also shown that prepositions are not a syntactic category of the language. Despite of the absence of preposition in Tsou as a syntactic category, it has other linguistic mechanisms to represent spatial information. As has already been mentioned, there are several ways to understand the spatial representation in a sentence, including the interpretation of focal objects, the predicates with rich incorporated spatial information, the spatial predicates which directly represent certain spatial notions, the combinations of both focal objects and predicates, locative nouns, and the reduplication. Based on these observations, suggest that Tsou is a “verb-framed” language whose spatial information is obtained through the understanding of the verbs/predicates in most of the case. Locative nouns play an important role, too. Of course, the interactions among verbs, focal nouns, and even the location of the speaker are also key points to have a full understanding of spatial information in a sentence. To summarize, there are a lot of ways to get spatial information in Tsou, but only four syntactic mechanisms are used:

Table2. The syntactic mechanisms of space

Category Morphemes Examples Gloss in English

Identifiable nooyo ‘inside’

spatial predications i’mi ‘from’

~’unu/~’uni emo’unu mi’unu/mi’uni ‘to’ ~caefi meicaefi tocaefi sucaefi ‘by/cross’ ~’upu su’upu no’upu ‘with’ ~eafo/~eafa eueafo taeafo/taeafa ‘out’ ~ovei meiovei eovei euovei eoovei ‘back’

~avovei meiavovei ‘back & forth’ Predications with sunghucu ‘fall into water’ rich spatial information p’etpUti ‘step on’

bichipi ‘stand beside’

teongasi ‘put down’

Locative nouns veina/vhona ‘left/right’

omza/oii ‘high/low’

cum’u ‘nearness’

feona ‘side’

tanesi/taesi ‘this/that side’

f’uf’u nof’uf’u

m’ef’uf’u ‘underside’

skopu m’eoskopu ‘upside’

peohU/peohna supeohU ‘downside’

f’uhu ‘back’

Reduplication ~kikiengi ma’kikiengi ‘around’

~kukuyunvu meikukuyunvu aukukuyunvu emokukuyunvu ‘around’

3

Spatial Representation in Squliq

Abstract

Space is a fundamental concept in human cognitive and linguistic system, since every human being is necessarily aware of his or her spatial location. As claimed by Svorou (1994), “it is in our nature to locate objects with respect to other objects, in a relativistic way” (Svorou, 1994: 8). Therefore, focusing on the semantics of the spatial experiences, this paper discusses how Squliq, the major dialect of Atayal, structures space by means of the focal object and the reference object.

As in Seediq (Huang, 1998), Squliq can also be shown to be a language that codes spatial information by a combination of spatial predicates and locative nouns or path verbs rather than prepositions. This paper comprises two parts. In the first part, we discuss locative nouns in detail, and categorize locative nouns into four basic systems of spatial orientation (Heine, 1997). We find that either the reference object’s characteristics such as animacy or the distance between the focal object and the reference object shape the way the native speaker uses the locative nouns. In the second part, we examine the semantics of spatial verbs, especially deictic motion verbs. We conclude that Squliq is a verb-framed language.

1. Introduction

To structure the space, we need a lot of entities such as the object, the location, the observer, etc. Langacker’s suggestion (1986) is adopted more-spread. His major two entities structuring the space are the trajector and the landmark. The trajector is the entity to be located; the entity with respect to which the trajector is located is the

landmark. In this paper, I adopt the two entities, the focal object and the reference object, to frame the spatial representation in Squliq; the focal object refers to the same object as Langacker’s trajector and the reference object is equal to his landmark.

The purpose of this paper is to discuss how Squliq structures spaces in terms of the notions of focal object and reference object. In section 2, we introduce basic Squliq spatial construction, and divide all spatial terms into two types: locative nouns and spatial verbs. In section 3, we discuss each locative noun in detail and examine the locative nouns in terms of Heine’s four systems of spatial orientation. In section 4, we discuss spatial verbs.

2. Basic Squliq Spatial Construction

As in other Formosan languages, Squliq makes use of the focus system to indicate the complex agreement relationship between a focus VP and its corresponding focus NP. There are two main types of focus systems in Squliq: one is the agent focus (AF) construction with the basic word order in Squliq, VOS; the other is the non-agent (NAF) construction manifested by patient focus (PF), benefactive/instrumental focus (B/IF), and locative focus (LF). In a NAF sentence, its word order is VSO.

In this section, I introduce the basic spatial construction in Squliq. To frame the Squliq spatial construction, the combination of spatial predicate relation bases on the interpretation of the reference object and the focal object is the main dimension I intend to explain. Consider first the basic construction as following:

Relation Reference Object Focal Object

(3). cyux maki ska yawa cuygan buwe_gitu. [Asp. be:located middle basket three pear]

Relation Reference Object Focal Object (4). cyux mlniq ska qsiya quleh. [Asp swim middle water fish] “Fish swims in the water. ”

The two sentences above are AF constructions; the focal object is in subject position and the reference object is in object position. The spatial predicates, maki



‘be located’ and mlniq ‘swim’ in sentences (3) and (4) respectively code the spatial

relation framed by the focal object and the reference object. In (3) and (4), cyux is

treated as a remote aspectual marker in opposition to nyux used to denote the

immediate relation of the speaker’s interaction with the event he is describing (Lou, 1994).

Besides functioning as an aspectual marker, cyux (or nyux) can also function as

the main verb in a sentence, meaning ‘to be located’ or ‘to be in existence’. First, as in (5), cyux as a locative verb constructs the spatial representation framed by the

reference object, zik niqan and the focal object, qutux

Ν

Relation Reference Object Focal Object (5). cyux zik niqan qutux Νiaw. [be:located below desk one cat]

“A cat is under the desk.”

Second, in existential sentences, cyux or nyux denotes the existence of the focal object

as in (6).

Reference Object Relation Focal Object

(6). qsahuy na biru? nyux qu lalu na bnrwan_sota?. [inside Gen book Ext Nom name Gen President’s sign]

“There is a sign of President inside the book.”

So far I have introduced briefly the basic spatial construction in Squliq. In a basic construction, the spatial predicate relation is introduced first, then the reference object, and finally the focal object. The position of the focal object can shift from the sentence final to sentence initial position as in (6) or (5’) below.