Aspect in Bonggi

MISSING IMAGE

Material Information

Title:
Aspect in Bonggi
Physical Description:
x, 240 leaves : ill. ; 29 cm.
Language:
English
Creator:
Boutin, Michael E., 1953-
Publication Date:

Subjects

Subjects / Keywords:
Bonggi language -- Aspect   ( lcsh )
Grammar, Comparative and general -- Aspect   ( lcsh )
Austronesian languages -- Malaysia -- Sabah   ( lcsh )
Linguistics thesis Ph.D
Dissertations, Academic -- Linguistics -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1994.
Bibliography:
Includes bibliographical references (leaves 230-238).
Statement of Responsibility:
by Michael E. Boutin.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002027300
oclc - 33002993
notis - AKL4898
System ID:
AA00002059:00001

Full Text










ASPECT IN BONGGI


By

MICHAEL E. BOUTIN



























A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1994


















Copyright 1994


by


Michael E. Boutin























This work is dedicated to my father who taught me to work hard, be honest, and have integrity.

He also loaned me the money for my first linguistics course, and

recognized that I enjoyed "book learning" long before I did.

If dad were still alive,

he would love to hear me say that he was right about the "book learning"

and find that his investment is bringing forth dividends.














ACKNOWLEDGEMENTS


I would like to thank my Bonggi friends for all their help and patience in teaching my wife and

I about their language and culture. Living and working with the Bonggi has been a rich experience

for us. The number of Bonggi who have befriended and helped us is too large to mention. Those

who stand out as being especially close to us for different reasons include: Mual bin Tugulan, Tagi

binti Limbatan, Sayad bin Liad, Aplan bin Liad, Darah binti Kundasan, Nunga binti Latip, and

Tipah binti Timisup.

I am very grateful to the Sabah State Government for permission to undertake this research

and the cooperation I have enjoyed from them. I especially thank Patricia Regis and Rita

Lasimbang of the Sabah State Museum for their interest in the Bonggi.

My colleagues in Sabah, who have also been involved in language research, have provided a

good sounding board for some of my ideas. I especially appreciate the healthy dose of skepticism I

have received from Richard Brewis, Hope Hurlbut, Julie King, Paul Kroeger, and Dave Moody. I

am also grateful for the help I received from linguistic workshops conducted in Sabah by Steven H.

Levinsohn and Ivan Lowe and for the encouragement I have received from different administrators

in Sabah including: Ken Smith, Gene Fuller, Paul Setter, and Wayne King. I also thank all those

who have helped me to attend various linguistic conferences and the Linguistic Institute at the Ohio

State University in 1993.

My SIB friends in Sabah have blessed me with their encouragement, especially Kenneth Ng,

Philip Lynn, Paul Wang, Simon Goh, and Michael Chen. Beyond Sabah, I appreciate both the

friendship and the encouragement of Malaysian linguists: Mashudi Kader, Boon Seong Teoh, Bibi

Aminah binti Abdul Ghani, and Eileen Yen Ee Lee.









I wish to thank Professor Haig Der-Houssikian who has served as chairman of my doctoral

committee. Working one-on-one with him has been a real privilege. I have benefited greatly from

both his excellent linguistic help and his counsel and encouragement.

I also would like to express my gratitude to the other members of my committee: Professor

Chauncey Chu, Dr. William J. Sullivan, and Professor Donald E. Williams. Each of them

provided valuable suggestions and encouragement to me.

My time at the University of Florida would have been very different had I been discouraged

from working on a minority language of Southeast Asia. I thank the following members of the

faculty for their encouragement and interest in my work: Michel Achard, Robert de Beaugrande,

Diana Boxer, Jean Casagrande, Gary Miller, David Pharies, Roger Thompson, Andrea Tyler, and

Ann Wyatt-Brown. I also appreciate the encouragement I have received from fellow graduate

students, especially Timothy Ajani, David Hankins, Wayne King, and Suzanne Norris.

Outside of the University of Florida, I am indebted to Robert Van Valin for his

encouragement and correspondence. His suggestions and critique of my earlier work have been

extremely helpful.

I thank all my family and the many friends in the U.S. who have encouraged and supported

Alanna and I these years. My sons Seth and Micah have been very understanding and loving when

I have not had as much time to play as they would have liked.

Finally, my heartfelt thanks go to my wife Alanna who is a real Proverbs 31 woman. She has

been a help in every sense of the word. She has enjoyed learning Bonggi with me and living in

remote villages where most westerners could not cope for a month, much less for years. Alanna

has been my chief encourager and has edited all my work always with a cheerful attitude and an

uncomplaining spirit.














TABLE OF CONTENTS



ACKNOWLEDGEMENTS................. ........... .............. iv

KEYS TO SYMBOLS AND ABBREVIATIONS ............................. viii

ABSTRACT ............................. ................. ix

CHAPTERS

1. INTRODUCTION ............ .............................. 1

1.1. The Bonggi Language and People . ........... ................ .. 1
1.2. Aspect in Bonggi .................. ............ ........... 6
1.3. Basic Descriptive Sketch of Bonggi ................................ 15
1.3.1. Simple Clauses ........................................ 15
1.3.2. Simple Nominal Phrases ................................... 22
1.3.3. W ord Classes ............ ...... ..................... 26

2. SITUATION ASPECT IN BONGGI................................. 31

2.1. States ....................... ............. ......... .. 34
2.1.1. Equational States .............. .... .... .... ........... 36
2.1.2. Locative States ............. ......................... 42
2.1.3. Condition States ............ ...... .................. 50
2.1.4. Possession States ....................................... 68
2.1.5. Cognition States ........... ........ .. ................. 72
2.2. Achievements . ........... .............................. .. 79
2.3. Activities ...................... ...................... 99
2.3.1. M otion Activity Verbs . ......... ..................... .. 100
2.3.2. Nonmotion Activity Verbs . .......... .................. .. 113
2.4. Accomplishments ........... .... ... ...................... 122








3. VIEWPOINT ASPECT IN BONGGI............................... 148

3.1. Viewpoint Aspect in Verb Morphology .................... ... 151
3.1.1. Morphophonemic Contrasts within the Peg- Group ................ 152
3.1.2. Semantic Structure of the Peg- Group ......................... 158
3.1.3. Iterativity: Specific and Distributive Viewpoints ................... 187
3.2. Viewpoint Aspect inside Verb Phrases ..... ..................... 199
3.3. Viewpoint Aspect outside Verb Phrases ..... .................... 218

4. CONCLUSION ............................................ 223

4.1 Summary ................. .. ........... .. ..... .......... 223
4.2 Implications ................. .... ............................ 225

REFERENCES ....................... ........................... 230

BIOGRAPHICAL SKETCH......................................... 239















KEYS TO SYMBOLS AND ABBREVIATIONS


ACC
ACH
ACT
ACY
APL
Arg
AS
CAU
d
DEF
DIR
DIS
EMPH
exc
EXP
FPFT
GEN
IMP
INSTPL
IT
LINK
LOC
NEG
NOM
NP
NPIV
NPST
NV
0
p
PFT
PIV
pred
POS
PRF
PST
REC
s
S
ST
UND
UPL
US
V
VA


Explanation or Gloss
1 first person
2 second person
3 third person
@ consonant in root/stem which is
separatedfrom root/stem by
an infix
[-mk]UPL unmarked undergoer pivot
accomplishment
[+mk]UPL marked undergoer pivot
accomplishment


Explanation or Gloss
accusative case
achievement
actor
activity
actor pivot accomplishment
argument
actor subject
causative
dual
definite
directional verb
distributive
emphatic
exclusive
experience
future perfect
genitive case
imperative mood
instrument pivot accomplishment
iterative
linkage particle
locative
negative
nominative case
noun phrase
nonpivot
nonpast tense
nonvolitional
object
plural
perfect
pivot
predicate
positive
perfect
past tense
reciprocal
singular
subject
state
undergoer
undergoer pivot accomplishment
undergoer subject
verb
viewpoint aspect














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


ASPECT IN BONGGI

By
Michael E. Boutin

August, 1994

Chairman: Haig Der-Houssikian
Major Department: Linguistics

This dissertation presents an analysis of aspect in Bonggi within the framework of Role and

Reference Grammar (RRG). Bonggi is a Western Austronesian language of Sabah, Malaysia.

The morphology of Bonggi distinguishes situation types: states, achievements, activities, and

accomplishments. Because these four situation types are the starting point for a RRG grammatical

analysis, there is a reciprocal harmony between the RRG model and Bonggi.

Bonggi verbs are classified semantically according to the relationships which exist between

predicates and their arguments. These relationships are described in terms of logical structures

which are linked to the verb morphology by a series of rules including the assignment of thematic

relations, semantic macroroles, syntactic functions, case, and verbal cross-referencing.

Each situation type has a unique set of inherent aspectual properties (Aktionsart) which are

reflected in the logical structures by predicates and a small set of operators such as BECOME and

CAUSE.

The model highlights the distinction between Aktionsart and viewpoint aspect by treating

viewpoint aspect as an operator. Whereas Aktionsart properties are determined from the logical

structures in a constituent projection that accounts for argument structure, the assignment of

viewpoint aspect belongs to an operator projection which includes viewpoint aspect, tense,









modality, negation, and illocutionary force. Unlike Aktionsart which is determined from the logical

structure, viewpoint aspect is independent of the logical structure. Although each situation type

has a unique logical structure and a unique set of Aktionsart properties, the same situation can be

presented from different viewpoints. That is, the inherent Aktionsart properties do not change with

a change in viewpoint aspect.

Viewpoint aspect in Bonggi is formally expressed in: 1) the verb morphology, 2) a system of

free form auxiliaries, 3) a system of enclitic particles, and 4) a system of temporal adverbs.

Although aspect, tense, and modality all belong to the operator projection, they modify

different layers of the clause. For example, viewpoint aspect is an operator modifying the clause

nucleus. This model not only provides a framework for treating aspect independently of modality

and tense but also for treating the interrelationship of aspect with modality, tense, and other verbal

categories.














CHAPTER 1
INTRODUCTION


This chapter briefly answers two questions which arise from the title of this dissertation. The

first question, 'Who are the Bonggi?' is answered in 1.1, while the second question, 'Why study

aspect in Bonggi?' is addressed in 1.2. Finally, a condensed sketch of Bonggi is provided in 1.3.


1.1. The Bonggi Language and People

Bonggi is a Western Austronesian language spoken by approximately 1,400 people on Banggi

and Balambangan islands in the Kudat District of Sabah, Malaysia. (See Figure 1 for the location

of Banggi and Balambangan in Southeast Asia.) Outsiders usually use the exonym 'Banggi' or

'Banggi Dusun' to refer to the people, while the Bonggi refer to themselves by the autonym Bonggi.

The term 'Banggi' is somewhat derogatory in nature as banggi means 'corpse' and is used in the

curse, Banggi nu! which is equivalent to the English phrase 'Drop dead!'

The term 'Dusun' is a misnomer from a linguistic point of view. Linguistically, the term

Dusun refers to a group of languages in Sabah. The majority of the indigenous languages of

Sabah belong to one of three groups: Dusunic, Murutic, or Paitanic.1 Bonggi is a linguistic isolate

which does not fit into any of these three groups. In fact, the Murutic languages are more closely

related to the Dusunic languages than Bonggi is to either group. Thus, calling the Bonggi Dusunic

as opposed to Murutic or Paitanic is a linguistic error. However, 'Dusun' has become a highly

politicalized term in Sabah, and it is in this sense that the label 'Banggi Dusun' is often used. I use






1These groups are based solely on lexicostatistics from Smith (1984). Previous groupings based
on lexicostatistics were proposed by Dyen (1965) and Prentice (1970). The most well-known
subgrouping argument dealing with any of the languages of Sabah is Blust (1974).









the term 'Bonggi' to refer to both the language and the people, and the term 'Banggi' to refer to the

island.2


Figure 1: Location of Banggi and Balambangan in Southeast Asia


On the basis of lexicostatistics, the closest linguistic relative of Bonggi is the Molbog

language spoken on Balabak and Ramos islands in the Philippines south of Palawan Island. A

comparison of basic wordlists indicates that Bonggi and Molbog are only about 50% cognate.

Furthermore, the two languages are far from being mutually intelligible. There are some Molbog

speakers on Banggi Island who are referred to as Belobog by the Bonggi. Many Bonggi speakers

living near the two Molbog villages on Banggi island have learned to speak Belobog. On the other




2For a discussion of the origin of the term 'Dusun' see Lebar (1972:148). For a recent discussion
of problems associated with the term 'Dusun' see Lasimbang and Miller (in press). For a general
discussion of nomenclature problems in Borneo see Appell (1968), Blust (1974), and Lebar
(1972).









hand, only a couple of Molbog speakers who have married Bonggi women have learned Bonggi.

This is typical of the sociolinguistic situation on Banggi and Balambangan islands.

Besides Bonggi and Molbog, there are four other language communities on these two islands:

Suluk (known in the Philippines as Tausug) and three languages from the Sama language family:

Bajau, Ubian, and Kagayan. All of the languages, with the exception of Bonggi, have their

geographical center in the Philippines. That is, Molbog, Suluk, Bajau, Ubian, and Kagayan

speakers are relatively recent arrivals in Sabah compared to the Bonggi.3 On the other hand, there

are no Bonggi speakers in the Philippines.

The general sociolinguistic picture pertaining to the Bonggi is as follows: Bonggi villages are

somewhat scattered over the two islands, but group into village clusters consisting of at least two

geographically close villages belonging to the same ethnic group.4 For example, the Limbuak

cluster in the southwest portion of Banggi Island includes the following Bonggi villages: Limbuak

Darat (Pega'Diaa), Batu Layar Darat (Liag Diaa), Lumanis Darat (Lemeis Diaa), and Kuda-

Kuda (Kuda'-Kuda);5 the Palak cluster in the southeast portion of Banggi Island includes Palak

Darat (Giparak Diaa) and Memang (Milimbiaa).

In general, with the exception of a few people who marry into a Bonggi community, speakers

of other languages do not learn Bonggi. It is the Bonggi who have learned to speak the languages

of the other ethnic groups. The other ethnic groups are also scattered around the islands in villages

which tend to cluster in groups. For example, the primary ethnic group near the Limbuak cluster is





3However, Bajau are known to have lived on Banggi island since at least as early as 1769
(Dalrymple 1769:65). For a discussion of relationships among the languages in the Sama language
family see Pallesen (1985). For an overview of relationships among Philippine languages see
Walton (1979). For an overview of the language situation and distribution of languages on Banggi
and Balambangan see Boutin and Boutin (1985).

4The term bunua 'community' is used refer to a hamlet, village, or village cluster. For a discussion
of the importance of the term bunua in Bonggi society see Boutin (1990).

5Village names are given in Malay, the national language, with Bonggi names provided in
parentheses.









Bajau who live in the two Bajau villages of Limbuak Laut (Pega'Loud) and Batu Layar Laut

(Liag Loud), while the primary ethnic group near the Palak cluster is Kagayan who live in the two

Kagayan villages of Palak Laut (Giparak Loud) and Laksian (Leksiadn).

As a general rule, the Bonggi learn the language of the outside community which is

geographically closest to their cluster. Thus, most Bonggi speakers over thirty years of age who

have lived in the Limbuak cluster speak Bajau, while those who have lived in the Palak cluster

would speak Kagayan due to their respective exposure to these two languages. The older a Bonggi

and the longer he or she has lived in a village cluster, the more likely he can speak the language of

the nearby outside community. Among younger speakers from different ethnic groups there is a

greater tendency to use Malay because: 1) Malay is the language of the schools; 2) the young

people have been increasingly exposed to Malay via radio and television; and 3) relationships

between Bonggi children and children from other language groups often begin during their school

years where they are required to speak Malay to each other.

Very few Bonggi are monolingual; most Bonggi over thirty are trilingual (speaking Bonggi,

Malay, and the language of the geographically closest outside community). It is not uncommon to

find Bonggi who speak four or five languages. In terms of language use, Bonggi is always spoken

between Bonggi speakers, except in cases where a speaker may use a few phrases from another

language. When a Bonggi meets a non-Bonggi, the language of the non-Bonggi is used if the

Bonggi has sufficient knowledge of the language; otherwise, Malay is spoken. This

accommodation strategy fits a possible sociolinguistic universal suggested by Quakenbush

(1986:242):

In a multilingual setting where language groups are of markedly different social
status, the group on 'bottom' will accommodate to the group on 'top' by using that
group's first language in face-to-face interaction, regardless of other components
of the social situation such as role relationships, location, formality, etc.
Despite the fact that few Bonggi are monolingual, Bonggi is not being replaced by a different

language. Bonggi children are learning the language as their first language, except in some cases









involving mixed marriages.6 The Bonggi have borrowed a number of words from Malay and some

words from English. The majority of English borrowings are nouns, although a few verbs have

also been borrowed (Boutin 1994).

There are differences in the Bonggi language spoken in different areas. However, these

differences are not great enough to cause any difficulty in communication. According to Bonggi

folk linguistics, the most important dialect variation is between the Bonggi spoken on Banggi

Island and that spoken on Balambangan Island. However, I have found that 'alleged differences'

tend to be exaggerated. Bonggi speakers have a very ethnocentric perspective of their own dialect.

The home village cluster dialect is always best, and the "Bonggi over there" are always the ones

who kiara youk 'have an accent'. The majority of Bonggi on Banggi Island have never been to

Balambangan. They tend to stereotype the Balambangan Bonggi and attribute to them a number of

noncognate lexical items. I have found in talking to Balambangan Bonggi that many of the lexical

items attributed to them are either archaic or unknown.

The vast majority of the Bonggi data presented in this book is from either the Limbuak village

cluster or the Palak village cluster on Banggi Island. There are some regional differences on

Banggi Island as illustrated by variation in the use of [ei] and [oi] in a small set of words.

Limbuak Darat NW part of island Palak Darat Gloss
/soid/ solidd' ] ['seid'] ['soig'] 'inside'
/toin/ ['toidn] ['teidn] ['toidn] 'jungle'
/sin.doin/ [sm'doidn] [sm'deidn] [sm'doidn] 'fingernail'
/oid/ ['oid'] ['oid'] ['og] 'boat'

The language variation which does exist is primarily phonetic. There are a few lexical

differences such as the use of the Bajau term angat 'said' by Bonggi in the Limbuak cluster.






6Although few Bonggi are monolingual, this is not meant to suggest that there is a high level of
bilingualism. Without having measured for degrees of bilingualism, my assumption is that the
level of bilingualism is not sufficient to pose any threat to the viability of Bonggi at this point.









However, angat is never used by speakers in the Palak cluster except in ridiculing the people from

Limbuak. No morphosyntactic differences have been observed between dialects.

The Bonggi are primarily subsistence farmers and fishermen. Subsistence agriculture is based

on cassava (Manihot esculenta Crantz).7 The actual origin of the Bonggi and the time period in

which they arrived in their present location are unknown. Early Bonggi settlers to Banggi and

Balambangan probably settled along the coast at estuaries, initiating a pattern which has persisted

to this day. Later, with the arrival of further immigrants, the Bonggi, being the weaker community,

were displaced into the interior.8 The only published record of the Bonggi language before the

1980s was a wordlist taken in 1937 (Schneeberger 1937). Several short Bonggi folktales appeared

in an English newspaper in 1923 (Agama 1923a, 1923b, 1923c, 1923d). They were probably told

in Malay and translated into English. Previous linguistic work on Bonggi which is most relevant to

the topic of this dissertation is Boutin (1991a) and Boutin (1992).


1.2. Aspect in Bonggi


In beginning an examination of aspect in Bonggi, it is important to realize that one structural

feature in a language often predicts, implies, or constrains the distributional possibilities of another

feature. My goal is to demonstrate how certain structural features in Bonggi predict and constrain

the distributional possibilities of aspect. Specifically, this dissertation answers two general

questions with respect to aspect in Bonggi: 1) what are the aspectual possibilities in Bonggi? and 2)

how are they constrained?

The term 'aspect' has been used to refer to two related, but distinct, phenomena:





7For an overview of Bonggi social organization see Boutin (1990).

8Many of the village names are a reflection of this pattern, diaa 'inland' (Malay: dart) and loud
'sea' (Malay: laut) are a reflection of this pattern; e.g. Limbuak Darat (Pega'Diaa) an interior
Bonggi village, and Limbuak Laut (Pega'Loud) a coastal Bajau village. For a discussion of
Bonggi history see Boutin (in press).








(a) Aspect refers to ways in which languages mark notions such as duration,
frequency, and completion in verbs. Simply speaking, from this standpoint, aspect
is a matter of the speaker's perspective or viewpoint on a situation. For example,
the speaker may choose to depict a situation as beginning, ending, continuing,
repeating, or completed.

(b) Aspect refers to the inherent features of situations or verbs; that is, whether the
situation or verb is static or dynamic, punctual or durative, telic or atelic. This
second use of the term aspect has been called situation aspect or Aktionsart 'kind
of action' (from German).
The distinction between (a) viewpoint aspect and (b) situation aspect (Aktionsart) is often

blurred in studies on aspect. My analysis of aspect in Bonggi shows that this distinction is

fundamental. Briefly speaking, each situation type has a specific situation aspect (Aktionsart)

associated with it, and given any situation type, there are constraints on the viewpoint the speaker

can have. That is, situation types have an inherent aspect (referred to here as situation aspect)

which constrains viewpoint aspect.

I provide a brief overview of situation types which are described in detail in Chapter 2. The

semantic notions which are discussed in this paper are interrelated and it is very difficult to discuss

one notional category such as aspect without discussing all the other categories which interact with

it and thus, effect the form in which aspect is realized. For example, past tense is marked

differently depending upon the situation type and which argument is selected as subject.

Following Lyons (1977:483), Smith (1983:481), and others, I use situation as a cover term

for both events and states.9 Vendler (1967) devised a universal four-way semantic distinction in

situation types: 1) states; 2) achievements; 3) accomplishments; and 4) activities. These four types

of situations correspond to major verb classes which are encoded in the verbal morphology of

Bonggi. In the remainder of this section, I provide a brief example of the four basic situation types

using data from both English and Bonggi.


9Situations are comparable to States of affairs (e.g. Dik 1981:32-36).









States are static situations. Several sub-classes of states are elaborated in Chapter 2. Only

condition states are discussed in this overview. The English sentence in (1) illustrates a condition

state.

(1) It is dry.
English condition states such as (1) are primarily realized by a form of the verb be followed

by an adjectival complement. Bonggi does not have a verb comparable to English be. Instead,

condition states like (1) are realized as a stative verb derived from an adjective root.10 This is

illustrated in (2) by the stative verb ngkorikng which is derived from the adjective root korikng

'dry'.11

(2) Sia ngkorikng.


it dry
'It is dry.'
Achievements are nonvolitional changes of state which are inception-oriented. Sentence (3)

illustrates an achievement in English, while (4) illustrates the Bonggi achievement verb kimorikng

which is derived from the adjective root korikng 'dry'.12


(3) It became dry.

(4) Sia kimorikng.
it become.dry
'It became dry.'
Accomplishments refer to changes of state which are brought about by a volitional actor and

have a final endpoint. Sentence (5) illustrates an accomplishment in English, while (6) illustrates

the Bonggi accomplishment verb ngorikng which is derived from the root korikng 'dry'.

(5) She dries coconuts.






10States whose complement is an adjective as in (1) are often referred to as attributive clauses.

11The morphological processes resulting in the Bonggi surface forms are elaborated in Chapter 2.

12Chapter 2 describes subclasses of achievements, accomplishments, and activities.








(6) Sia ngorikng piasu.
she dry coconut
'She dries coconuts.'
Activities involve a volitional actor and are activity-oriented referring to events which often

have no clear endpoint. Sentence (7) illustrates an activity in English, while (8) illustrates the

Bonggi activity verb ngkapak which is derived from the root kapak 'sing'.

(7) She sings.

(8) Sia ngkapak.
she sing
'She sings.'
The Bonggi root korikng 'dry' was used to contrast states, achievements, and accomplishments

resulting in respective surface contrasts between: ngkorikng (2), kimorikng (4), and ngorikng (6).

This root can also occur unaffixed as illustrated in (9).

(9) Bakng kiou meiti na, korikng.
if wood dead dry13
'If wood is dead, it is dry.'
However, an activity verb cannot be derived from korikng 'dry'. Thus, the Bonggi data

provided so far do not sufficiently illustrate a surface contrast between activities and the other

situation types. 14 However, there are other roots such as uru 'medicine' which can be used to

derive both an activity (ngguru) and an accomplishment verb (nguru) as in (10).

(10) Ama' Numpang ngguru tina robi. Sia nguru n Daya'.
father Numpang medicate later night he medicate NPIV Daya
'Numpang's father is medicating people later tonight. He will medicate Daya.'15




13In this section '-' is used to gloss certain grammatical morphemes which are described in Chapter
2 and not relevant to the current discussion.

14The roots korikng 'dry' and kapak 'sing' were chosen since both are /k/ initial roots. The initial
phoneme of the root is important in determining surface contrasts in Bonggi (cf. Chapter 2).
Although the paucity of data described in this overview may not convince the skeptical reader that
the primary function of the verbal morphology described here is to indicate situation type, detailed
evidence for my analysis is provided in Chapter 2.

15The medicating described by the verbs ngguru and nguru in (10) refers to a treatment performed
by a local spiritualist. Other types of medical treatment involve either a herbalist or western
medicine from a clinic (cf. Boutin & Boutin 1987). Sentence (10) is from a recorded text.









Dowty (1979) and Cooper (1985) state that it is sentences as a whole which belong to the

various situation classes and the classes are not determined simply by the verb that the sentences

contain. Thus, sentences (1 la) and (12a) are activities, whereas sentences (1 ib) and (12b) are

accomplishments. 16

(11) a. She ate fish.
b. She ate the fish.

(12) a. She walked.
b. She walked to the store.
Dowty's and Cooper's claim that situation type is not determined by the verb holds for

English. For example, the activity verb walked in (12a) can become an accomplishment walked to

the store when a definite goal is added (cf. FVV 1984:39). In English, the verb morphology of ate

and walked does not change whether they occur in activity or accomplishment situations.

However, this is not always the case in Bonggi, as seen in (10), where the verb morphology

distinguishes the activity verb (ngguru 'medicate') from the accomplishment verb (nguru

'medicate').

Some Bonggi verbs, such as the verb 'to walk' (mpanu), are lexically an activity verb, and can

occur in either activity clauses similar to (12a) or accomplishment clauses similar to (12b).

However, in either case, the verb mpanu 'to walk' always appears with activity verb morphology,

which is described in Chapter 2. Other Bonggi verbs, such as the verb 'to cut down a tree'

(nobokng), are lexically an accomplishment, and only occur in accomplishment clauses similar to

(5) and (6). Verbs which are lexically an accomplishment only occur with accomplishment verb

morphology which is also described in Chapter 2. Still other Bonggi verbs, such as the verb 'to

move', can be lexically either an accomplishment verb, in which case it occurs with

accomplishment verb morphology (nguhad), or an activity verb, in which case it occurs with

activity verb morphology (muhad).



16As noted by Van Valin (1990:225, footnote 4), the contrast between (1 la) and (1 lb) cannot be
reduced to the presence or absence of articles.









In Bonggi, the verbal morphology distinguishes whether a verb is lexically an activity or an

accomplishment, regardless of the presence or absence of verbal arguments or adjuncts. In fact,

one of the primary functions of verbal morphology in Bonggi is to indicate the lexical situation

type. This means that situation types are much more transparent in Bonggi than in English, and

that contrary to Dowty's and Cooper's claim, verbs can determine the situation type, at least

lexically in some languages.

As stated earlier, the four-way distinction in situation type corresponds to major verb classes:

1) states; 2) achievements; 3) accomplishments; and 4) activities. A number of semantic features

can be used to delimit the four different situation types and the corresponding verb classes. Some

semantic features such as dynamicity are associated with aspect, while others such as the thematic

role of the arguments) are less directly relevant to aspect. In fact, certain aspectual features such

as dynamicity are essential in the classification of verbs. It is in this sense that Dowty uses aspect

to distinguish the inherent meaning of verbs (Dowty 1979:52).

Table 1.1 characterizes the salient aspectual features which distinguish the four situation

types.



Table 1.1: Salient aspectual features of the four situation types17

state achievement activity accomplishment

dynamic + + +

telic + +

causative +-


17The absence of'+' or '-' means that the feature is not relevant to the situation type.









Situation is a cover-term for states and events. In terms of the typology of situations

presented above, events include achievements, activities, and accomplishments. States are static

and do not involve change over time, while events are dynamic and involve change (cf. Chung &

Timberlake 1985:214).

A situation which is telic has an endpoint or definite limits. Only events, not states, can be

telic. Telicity is often described in terms of the boundedness of events. Achievements and

accomplishments are telic or endpoint-oriented whereas activities are atelic and activity-oriented.

Accomplishments have a causal predicate, whereas achievements and activities do not. The caual

predicate links an activity to a resulting achievement (cf. 2.4). Thus, the feature causative

distinguishes accomplishments from other events.

Duration is an aspectual feature which is sometimes used to delimit situation types. That is,

some situations are punctual (momentary), while others are durative. However, duration is not a

salient aspectual feature of situations in Bonggi. States by definition are durative and this is

reflected within the feature dynamism. Achievement and accomplishment verbs are not limited to

punctual situations or to situations of short duration (e.g. (4), (6)). Some verbs can be viewed as

inherently punctual (e.g. break), whereas others are inherently durative (e.g. soften), but this

distinction is not encoded in Bonggi. Chapter 3 provides evidence that there are durative and

punctual verbs of the same situation type.

Thus, for each of the four situation types, there are corresponding aspectual features which

are referred to as situation aspect (Aktionsart). The aspectual features in Table 1.1 are not

language particular. Although reference was made to duration as a possible situation aspect

feature, the three features provided in Table 1.1 are sufficient to delimit the four situation types

and they are considered universal situation aspect features.

Van Valin (1993:34) refers to the four situation types (states, achievements, activities, and

accomplishments) as Aktionsart classes. His use of the term Aktionsart varies from the traditional

use of the term where Aktionsart refers to the inherent aspectual meaning of a verb (Dahl

1985:26). On the other hand, the term aspect is traditionally a strictly grammatical category. I use









the terms situation aspect and Aktionsart synonymously to refer to the salient aspectual features

found in Table 1.1 which distinguish the different situation types in Bonggi.

I- One property of aspect is that the same concepts are relevant at different levels (Chung &

Timberlake 1985:214). For example, telecity is not only important for differentiating situation

types, but also for constraining the way in which situations can be viewed.

Situations can be viewed from different points in order to highlight different phases of the

situation. For instance, the situation described in (5) can be viewed from an iterative perspective

as shown in (13), or an inceptive perspective as shown in (14).

(13) He always breaks the seal.

(14) He started to break the seal.
The viewpoints) available for referring to a particular type of situation depends upon the

properties of that situation (Smith 1983:491). For example, the stative situation described in (1)

cannot be viewed from an inceptive perspective as in (15).

(15) *He started to be tall.
Viewpoint possibilities reflect aspectual characteristics at the sentence level and are described

in terms of viewpoint aspect. In Bonggi viewpoint aspect is often encoded in aspectual words

which are part of the verb phrase. Viewpoint aspect is constrained by situation type. Thus, while

the accomplishment situation described in (6) can occur with continuous viewpoint aspect which is

marked by kahal 'still' as in (16), achievement situations such as that in (4) are not compatible with

continuous viewpoint aspect.

(16) Sia kahal ngorikng piasu.
she still dry coconut
'She still dries coconuts.'
One difference between situation aspect and viewpoint aspect is a distinction in formal

expression. Situation aspect correlates with situation type, which is formally expressed by the verb

morphology. On the other hand, viewpoint aspect is expressed by either verb morphology, as in

the use of meg- frequentativee' in (17), or periphrasis, as in the use of kahal 'still' in (16).








(17) Sia pandi meg-meikap.
she know DIS-make.up
'She uses make up all the time.'
Although situation aspect and viewpoint aspect can both be expressed by derivational verb

morphology, they are in complementary distribution in terms of the affixes employed. Other

correlations between formal expression and aspectual type are explored in Chapter 3. These

include the use of partial reduplication of verb roots to express continuous durativee) activity and

full reduplication of verb roots to express iterative activity. Chapter 3 also shows that some

aspectual notions must be studied outside the verb system in the adverbials; for example, habituals.

,- The conceptual framework for this study of aspect draws on the work of Carlota Smith (1991)

and Role and Reference Grammar. The major influence of Smith on my thinking is the distinction

between viewpoint and situation aspect (Aktionsart). The major points of difference between

Smith and me are: (1) her view of aspect is sentential, whereas my own view is that aspect is not

strictly sententially governed; and (2) whereas Smith claims that viewpoint aspect is generally

indicated morphologically and situation aspect (Aktionsart) is generally indicated syntactically, I

show that situation aspect in Bonggi is indicated morphologically and viewpoint aspect is indicated

morphosyntactically, suggesting that her claim is premature and based on insufficient data.

My approach to the study of aspect focuses on overt grammatical forms, as well as semantic

features of aspect. The data sources for this dissertation are a working dictionary (3,200+ roots

with multiple forms), recorded and transcribed texts (200+ pages, including glosses), and field

notes which my wife Alanna and I collected in Sabah between November 1982 and May 1992.18







18My wife and I had a house in Limbuak Darat on Banggi Island between November 1982 and
December 1985. During that period, about 60% of our time was spent in Limbuak Darat and 40%
in the city of Kota Kinabalu on the mainland. Afterwards, we resided in the United States from
April 1986 July 1987. Upon returning to Malaysia we had a house in Palak Darat from
November 1987 to April 1992. During that period, about 40% of our time was spent in the village
of Palak Darat and 60% in the city of Kota Kinabalu. Although Kota Kinabalu is outside of the
language area in which Bonggi is spoken, we almost always had Bonggi speakers living with us
whenever we were outside the village between August 1987 and May 1992.









Although much of the data presented is from texts, I have not excluded other types of data,

including elicited verb paradigms.

The theoretical model used for discussing aspect in Bonggi is Role and Reference Grammar

(RRG) as described by Robert Van Valin, Jr., editor (1993). From the perspective of RRG,

situation aspect results from inherent properties of situations which are defined in terms of lexical

semantics; on the other hand, viewpoint aspect is treated as an operator over the clause nucleus.

Before discussing Bonggi from the perspective of RRG (cf. Chapter 2), the remainder of this

chapter provides a brief overview of some of the linguistic features found in Bonggi.


1.3. Basic Descriptive Sketch of Bonggi


1.3.1. Simple Clauses

In terms of common word-order typology, Bonggi is a SVO language. Basic word-order in

simple transitive clauses with two arguments is SVO, as illustrated in (18), (19), (20), (21), and

(22).19

(18) S V 0
Ou ng-atad diha.
I AS-take you
'I take you there.'20

(19) S V 0
Sia ng-atad diaadn.
he AS-take me
'HE takes me there.'

(20) S V 0
Ou etad-adn nya.
I take-US him
'I am taken there by him.'





19This brief overview of clauses only includes simple, transitive, verbal clauses. Chapter 2
distinguishes transitive and intransitive clauses as well as verbal and nonverbal clauses, and
provides a more detailed analysis.

20The subject of the sentence is capitalized in the free translation.








(21) S V 0
Sia ng-atad siidn ku dii diaadn.
he AS-take money my to me
'HE takes my money to me.'

(22) S V 0
Siidn etad-adn ku dii diha.
money take-US me to you
'MONEY is taken to you by me.'
Actor and undergoer are the two primary arguments of a transitive predicate, either one of

which may be the subject. Actor refers to the entity which instigates or controls the action

expressed by the verb. Undergoer refers to the entity affected by the action or state expressed by

the verb. For example, the subject is an actor in (18), (19), and (21), whereas the subject is an

undergoer in (20) and (22). One NP in every verbal clause is indexed by the morphology of the

verb. Within Philippine linguistics, the indexed NP has been called various names including:

subject, topic, focus, pivot, and trigger. Although I use the term 'subject' in this overview, the term

pivot is used in the remaining chapters in keeping with the theoretical framework described in

Chapter 2. Also, within Philippine linguistics, the role of the verb morphology is usually

considered as encoding the semantic role of the indexed NP (or subject). Although I argue later

that encoding the semantic role of the subject is not the primary function of verbal morphology in

Bonggi, for ease of exposition in this overview, I have glossed the grammatical morphemes above

as 'AS' (actor subject) and 'US' (undergoer subject) to indicate the primary semantic function of the

subject.

In (22) the undergoer (siidn 'money') is subject and the actor (ku 'me') is labeled as '0'

(object). Note that the actor is not marked as an oblique argument in (22) as in English passives

where the agent is marked with the preposition by. This raises an interesting question regarding the

syntactic status of the actor in Bonggi.21 If we avoid the typological parameter 'O', word order in

simple Bonggi clauses is: Subject Verb Nonsubject core argument.


21This question is briefly addressed below.









There are three partially distinct sets of personal pronouns which are referred to as

nominative, accusative, and genitive case pronouns. Table 1.2 contrasts the three sets of pronouns

in Bonggi.



Table 1.2: Bonggi Pronouns


NOMINATIVE


GENITIVE


ACCUSATIVE


Singular ou ku diaadn

Idual kita ta dihita

Iplural-inclusive kiti ti dihiti

Iplural-exclusive ihi mi dihi

2singular aha nu diha

2plural uhu nyu dihu

singular sia nya nya22

3plural sigelama sigelama sigelama


Nominative case pronouns occur as subjects (e.g. ou T in (18) and (20), and sia 'he/she' in

(19) and (21)). Genitive case pronouns occur as the possessor in possessive NPs (e.g. ku 'my' in

(21)) and nonsubject actor (e.g. ku 'me' in (22)). Accusative case pronouns occur as nonsubject

undergoer (e.g. diha 'you' in (18), diaadn 'me' in (19), and nya 'him' in (20)) and following certain

oblique markers such as dii 'to' (e.g. diaadn 'me' in (21) and diha 'you' in (22)). Bonggi does not

have a pronominal dative case form, but expresses the notion of indirect object using prepostions

(e.g. dii to' in (21)).





22The contrast between 3s genitive and accusative case pronouns is neutralized as nya, and there is
no contrast in case in third person plural sigelama.









Bonggi is a nominative-accusative language with the grammatical subject being in the

nominative case. Every Bonggi verbal clause contains one and only one nominative argument.

The nominative argument can be either an actor (e.g. ou T in (18), and sia 'he/she' in (19) and

(21)) or an undergoer (e.g. ou T in (20)).23 Sentences such as (20) and (22) are different from

passives in English. Like English passives, the undergoer is the subject (e.g. siidn 'money in (22));

however, unlike in English, the actor (e.g. ku 'me' in (22)) is not marked as an oblique argument.

The syntactic status of arguments like ku 'me' in (22) has been widely debated for Philippine-type

languages with the primary question being the argument status of these agents.24 According to

Bell (1983), nonsubject agents are nonterms (nonarguments), while Kroeger (1992a) claims they

are always terms (arguments). RRG also takes the position that they are terms (cf. Chapter 4). At

any rate, in the active voice as in (18), (19), and (21), the verb root atad 'take' has two direct core

arguments; one, the actor, occurring in the nominative case, and the other, the undergoer, occurring

in the accusative case. Direct core arguments are defined as arguments which are not preceded

by a preposition (oblique case marker); for example, diaadn 'me' in (19) is a direct core argument.

Although every verbal clause has a nominative argument, it may not be realized due to

ellipsis, as in (23) (cf. (22)).

(23) Etad-adn ku gulu.
take-US me first
'IT is taken there by me first.'
Ellipsis of the subject as in (23) is common in Bonggi. In fact, in many instances, especially

imperative clauses, both core arguments are elided as in (24).25





23Common nouns such as siidn 'money' in (22) are not case marked (cf. below).

24Some people have argued that constructions like (22) are nominal, not verbal, since the agent is
treated in the same way as the owner in possessive NPs. See Dahl (1984) for an argument against
this nominal construction position. One argument against the nominal hypothesis is that the verbal
element in constructions like (22) can be inflected for tense.
25Cf. English, where the subject of transitive clauses, but not the object, is elided in imperatives.








(24) Etad-a' ba!
take-IMP.US EMPH
'Take IT there!
The semantic role of the nominative argument can be determined from the verbal affixation.

Thus, the verb affix functions like a placeholder for the subject nominal. For example, the suffix -

a' in (24) indicates imperative mood and that the subject is the undergoer, in this case 'the thing

being taken'.

Besides the morphological distinctions described for pronouns, Bonggi distinguishes personal

names and common nouns. Common nouns which are direct core arguments are not case marked.

Instead, word-order is used to encode the syntactic function subject. In (25) the common nouns

ama' '"father' and siidn 'money' are not marked because they are direct core arguments, but the

common noun indu "mother' is marked by dii 'to', indicating that it is an oblique argument.

(25) Ama' nya ng-atad siidn dii indu' nya.
father her AS-take money to mother her
'HER FATHER takes money to her mother.'
In (26) Butak is preceded by si, indicating that Butak is a personal name and the clause

subject; Tagi is preceded by ni (realized as [n]), indicating that Tagi is a personal name and

nonsubject.26

(26) Si Butak ng-atad siidn dii n Tagi.
PIV Butak AS-take money to NPIV Tagi
'BUTAK takes money to Tagi.'

Si marks nouns as both personal names and clause subject.27 Thus, one function of si is to

mark nominative case. However, ni is not a case marker because dii 'to' (an oblique marker)











26The grammatical markers si and ni have phonologically conditioned variants. Si and ni do not
occur with names beginning with /s/. Ni is reduced to [n] before names beginning with /t/ or Id/; it
is reduced to [n] before names beginning with a vowel; and it is reduced to [i] elsewhere.









governs the case of Tagi in (26). Furthermore, ni can occur with accusative case nouns (e.g. Tagi

in (27)) and genitive case nouns (e.g. Tagi in (28)).

(27) Si Butak ng-atad n Tagi.
PIV Butak AS-take NPIV Tagi
'BUTAK takes Tagi there.'

(28) Si Butak ng-atad siidn n Tagi.
PIV Butak AS-take money NPIV Tagi
'BUTAK takes Tagi's money there.'

The use of word-order instead of case-marking with common nouns to encode the syntactic

function subject is illustrated in (29) where the subject precedes the verb and the nonsubject core

argument follows the verb. This contrasts with most Philippine-type languages, including the

languages of Sabah which tend to be verb-initial and whose nominal forms are generally marked by

prepositional particles (Shibatani 1988:88). However, verb-medial languages like Bonggi do not

typically have case markings on NPs to establish their grammatical relations; instead, the

grammatical relations are signalled with word-order (cf. Bybee, et al. (1990:15)).

(29) Asu nya bas na ng-ohol asu ku.
dog his PST PFT AS-bite dog my
-HIS DOG bit my dog.'

To summarize, the subject is the prominent nominal in a clause. In Bonggi, subject NPs are

marked in one of three ways: (a) by the special particle si for personal names and nominals which

are treated like personal names (cf. footnote 27), (b) by nominative case pronouns, or (c) by word-

order. All three of these techniques for indicating the clause subject (i.e. the prominent nominal)

mask the semantic role of the subject NP. The subject depends on the verbal affix for its semantic






27Nicknames, death-names, teknonyms, the indefinite reference term anu 'so-and-so', and certain
kinship terms are treated morphosyntactically as personal names (Boutin 1991b). For example, the
kinship terms ama"father' and indu"mother' in (25) are marked as personal names when the
speaker is referring to his/her own parents, as in (i).

(i) Si ama' ngatad siidn dii ny indu'.
PIV father take money to NPIV mother
'FATHER takes money to mother.'









role. If word-order is included as a form of marking, subjects are marked twice: once on the verb

to indicate the semantic role of the subject, and once on the subject argument (by either si, or a

nominative case pronoun, or word-order). The different nominal categories show different degrees

of merger in form. Personal pronouns have three surface forms with accusative case and oblique

NPs distinguished on the basis of the presence or absence of an oblique marker. However,

personal names have two surface forms and common nouns have only one surface form with

oblique arguments in each instance being distinguished by an oblique marker. The different forms

of each nominal category are illustrated in Table 1.3.



Table 1.3: Nominal arguments


Direct core Noncore

Subject Nonsubject

Actor j NonActor


Nominative


Genitive


Accusative


Oblique NP


Possessive NP


Personal ou T ku 'me' diaadn 'me' dii diaadn koon ku 'my food'
pronouns __o me'

Personal si Tagi n Tagi n Tagi dii n Tagi 'to koon n Tagi
names 'Tagi' 'Tagi' 'Tagi' Tagi' 'Tagi's food'

Common asu 'dog' asu 'dog' asu 'dog' dii asu koon asu
nouns 'to (a) dog' '(a) dog's food'


Table 1.3 shows that: (a) oblique NPs are always marked by prepositions; (b) in possessive

NPs, the possessor always follows the possessed item; (c) the distinction between nonsubject actors

and nonsubject nonactors only exists for pronominal NPs; and (d) direct core arguments (subject

and nonsubject) are distinguished on the basis of case if they are pronominal, by the choice of si

versus ni if they are personal names, and by word-order otherwise.









In the transitive clauses illustrated by (18) (29), all of the nonsubject direct core arguments

are either actors or undergoers. However, this is not always the case; that is, nonsubject direct

core arguments are not necessarily actors or undergoers. For instance, clauses which are

traditionally described as dative-shift constructions have two nonsubject direct core arguments, but

only one is an undergoer. Sentence (30) illustrates the ditransitive predicate mori 'give' in a clause

without dative-shift, and (31) illustrates the same verb in a clause with dative-shift. There are two

nonsubject direct core arguments in (31), but only one, diaadn 'me', is an undergoer.

(30) Sia m-ori siidn dii diaadn.
PIV AS-give money to me
'SHE gives money to me.'

(31) Sia m-ori diaadn siidn.
PIV AS-give me money
'SHE gives me money.'

Chapter 2 describes certain intransitive clauses that have direct core syntactic arguments

which are neither actors nor undergoers. According to case-marking principles in Van Valin

(1990:241), the default case for direct core arguments which are neither actor or undergoer is

dative. However, as shown in Table 1.3, Bonggi does not have unique dative case forms; instead,

all nonsubject direct core arguments which are not actors occur in accusative case.


1.3.2. Simple Nominal Phrases

The head of nominal phrases is either a pronoun or a noun. Pronominal heads are either

personal pronouns (as shown in Table 1.2) or demonstrative pronouns. Noun heads can be simple

nouns such as apu "grandfather' in (32), compound nouns such as apu'-odu' '"ancestors' in (33), and

personal names such as Butak in (34) (cf. (28)).

(32) apu' nya
grandfather her
'her grandfather'

(33) apu'-odu' nya
ancestors her
'her ancestors'








(34) si Butak
PIV Butak
'Butak'
The primary word order within NPs whose head is a noun is: Determiner Noun Adjective -

Determiner, with some overlap between the determiners which precede the head-noun and those

which follow the head-noun. NPs with simply a noun and an adjective modifier are illustrated in

(35).

(35) anak kaha' toudn baru
child eldest year new
'eldest child' 'new year'

Unlike English, which can string a series of adjectives together within a single NP, Bonggi

NPs normally contain only one adjective.28 Even adjectives from different semantic domains such

as color, quality, and size rarely cooccur. Instead, when a speaker desires to ascribe two attributes

to a simple entity, he either uses two separate NPs (as in (36) where the first NP lama bulag 'blind

person' is qualified by the second NP indu'Mual 'Mual's mother') or stative verbs (as in (37) where

n-doot 'ST-bad' and m-pia 'ST-good' are stative verbs).

(36) Iketomu ou lama bulag, indu' Mual.
met I person blind mother Mual
'I met a blind person, Mual's mother.'

(37) Ndara n-doot sia, asal m-pia baar-baar sia.
not.exist ST-bad he surely ST-good true-true he
'HE is not bad, surely HE is very good.'

Some adjective roots rarely occur as adjectives in NPs; they are far more likely to be realized

as stative verbs. These include the stative verbs n-doot 'ST-bad' and m-pia 'ST-good' as in (37)

and mi-gia 'ST-big' as in (38). However, other adjective roots (such as toyuk 'small' in (38)) occur

more frequently as adjectives, even when stative verb forms are possible (e.g. n-toyuk 'ST-small').






28In English conversation there is a tendency for speakers to avoid complex noun groups
containing a series of adjectives and instead spread adjectives out over several clauses. This
contrasts with English written language which is more dense (cf. Bygate 1987:62-63; Richards
1990:73).






24

(38) Uubm nya, "Suhu ga toyuk tuni toyuk kei bagi,
said he all CONTRAST little body little also share

suhu mi-gia tuni mi-gia hei bagi."
all ST-big body ST-big also share
'He said, "All small bodies get a small share, all big bodies a big share."'

When the head of an NP is a noun, the noun can be followed by determiners which indicate

something about the definiteness and/or referentiality of the head noun. The set of determiners

which follow the head-noun include: possessors, as illustrated in (39) (cf. Table 1.3);

demonstratives, as illustrated in (40); and the definite article na, which is illustrated in (41).

(39) bali ku
house my
'my house'

(40) bali nti
house this
'this house'

(41) bali na
house DEF
'the house'

Demonstratives such as nti 'this' in (40) have a deictic function with the deictic center being

the speaker. Table 1.3 shows the demonstratives which can occur in Bonggi. The demonstratives

are distinguished on the basis of orientation with respect to the speaker/hearer. They can occur

alone as the head of a NP, in which case they are referred to as demonstrative pronouns. Each of

the three demonstratives has a shortened form which is also shown in Table 1.3.29








29The details regarding the conditioning factors for long versus short demonstrative forms have not
been worked out. Two factors which influence the occurrence of short versus long forms are the
phonological shape of the preceding head noun (shortened forms are more likely to occur when
contiguous segments are identical as shown below) and the speech style (shortened forms are more
likely to occur in fast/casual speech).

toudn nti 'this year' bali nti 'this house' bali ina 'that house' bali inoo 'that house over
yonder'
toudn ti 'this year' bali na 'that house' bali noo 'that house over
yonder'






25

Table 1.4: Demonstratives


Ina 'that' also has anaphoric function, as illustrated in (42) where the anaphoric reference is to

what has previously been said.30 In fact, ina na 'that', with the referent being an extended passage

of text, is one of the major cohesive devices in Bonggi (cf. Halliday & Hasan 1976:67).

(42) Ina na susuad nya.
that DEF say he
THAT is what he said.'

Nti this' has a cataphoric function as illustrated in (43) where the cataphoric reference

pertains to what is said immediately following.

(43) Moro' sikng nti, ...
tell like this ...
'Tell them something like this,...'
Nti 'this' also has an exophoric function whereby it is used to refer to current periods of time

such as odu nti today', minggu nti 'this week', buaidn nti this month', and toudn nti 'this year' (cf.

Halliday & Hasan 1976:61).

As stated previously, the three different types of determiners which follow the head noun

(possessors, demonstratives, and the definite article na) all indicate something about the

definiteness and/or referentiality of the head noun. Na 'definite' identifies a particular individual or

subclass within the class designated by the noun. Possessors and demonstratives are semantically

selective; they contain within themselves some referential element in terms of which the item in




30In Bonggi, as in most systems, it is the intermediate term of the deictic series which is used as a
general anaphoric element (Anderson & Keenan 1985:287).


Long Form Short Form Meaning

nti ti 'this' (near speaker)

ina na 'that' (near hearer)

inoo noo 'that' (not near speaker/hearer)









question is to be identified. With the possessives, it is person (e.g. bali ku 'my house'); with the

demonstratives, it is proximity (e.g. bali nti 'this house'). But na 'definite' has no content; it merely

indicates that the item in question is specific and identifiable (e.g. bali na 'the house'). Somewhere

the information for identifying it is recoverable (cf. Halliday & Hasan 1976:71).

Possessive pronouns can cooccur with demonstratives, as illustrated in (44).

(44) anak ku nti
child my this
'this child of mine'
Determiners which precede the head noun include: (a) personal name markers as illustrated in

(26)-(28) and Table 1.3; (b) indefinite quantifiers as illustrated by separu 'some', nentaadn na 'all',

and tiap-tiap 'every' in (45); and (c) numerals, classifiers, and a linkage particle which

phonologically links numerals and classifiers as illustrated in (46).

(45) separu lama nentaadn na lama tiap-tiap lama
some people all -- people every person
'some people' 'all people' 'every person'

(46) dua m batakng sikiou
two LINK long.thin.object cassava
'two cassava plant stems'


1.3.3. Word Classes

Since aspect is a verbal category, discussion of aspect, verb morphology, and verb phrases is

not included in this sketch, but is delayed until Chapter 2. However, besides the brief descriptions

of clauses and nominal phrases which were provided in 1.3.1 and 1.3.2, a short discussion of

word classes or lexical categories is needed to round out this sketch since much of the verbal

morphology of Bonggi is conditioned by the lexical category of the root. Verbs can only be derived

from certain lexical categories, and the category of the root or base form often predicts the verbal

situation types which can occur; therefore, some understanding of lexical categories is ultimately

very important to an understanding of aspect. For example, given any root and knowing its lexical

category, we can predict whether or not verbs can be derived from that root and which verbal

situations are possible.









As pointed out in 1.2, situation types have an inherent situation aspect which constrains

viewpoint aspect. Because the four types of situations (states, achievements, activities, and

accomplishments) are primarily derived from verb, adjective, and noun roots, only these three

words classes are considered here.31

Hopper and Thompson (1984:703f.) describe three criteria which have been used in

determining word classes or 'parts of speech': 1) morphological, 2) syntactic, and 3) semantic.

Following Schacter (1985), I assume that the primary criteria for determining word classes are

grammatical, not semantic. The grammatical properties which are relevant for classifying word

classes include distribution, syntactic function, and structural characteristics, especially inflectional

affixes.


Nouns

Bonggi nouns are divided into two main subclasses (personal names and common nouns) on

the basis of whether or not the nouns occur with the grammatical markers si or ni (where si marks

personal names in the nominative case (e.g. si Butak in (26)), and ni marks personal names which

are not in the nominative case (e.g. n Tagi in (26)) (cf. Table 1.3 and discussion in 1.3.1)).

Common nouns are further subdivided on the basis of whether they are count nouns or mass nouns.

Count nouns can cooccur with the quantifier barabm 'many', while mass nouns, such as beig

'water' and timus 'salt', are quantified with the stative verb migia 'large'. Verbs can only be derived

from common nouns, not personal names.









31Verbs are not derived from the other word classes mentioned in 1.3.2. However, evidence for a
class of directional verbs which are derived from spatial deictics is presented in 2.1.2. Evidence
is also provided in Chapters 2 and 3 for a class of aspectual auxiliary words and a class of
temporal adverbs.









Syntactically, nouns primarily function as arguments of verbs as in (47) where sigu hu 'my

teacher' is the subject and uubm Bonggi 'Bonggi language' is the object. Nouns can also function

as predicates in equational clauses as in (48) where sigu hu 'my teacher' is the predicate.32

(47) Sigu hu nda' pandi uubm Bonggi.
teacher my not know language Bonggi
'MY TEACHER does not know Bonggi.'

(48) Sia sigu hu.
he teacher my
'HE is my teacher.'

Morphological evidence for word classes is often established on the basis of inflectional

morphemes alone. For example, only verbs are inflected for tense. However, Bonggi word classes

cannot be determined from inflectional morphology alone because nouns and adjectives are not

inflected.33 Although case is not marked morphologically, noncore arguments are marked

syntactically by prepositions and are discussed in Chapter 2.


Verbs

I take the distinction between nouns and verbs to be a language universal (cf. Giv6n 1984;

Hopper & Thompson 1984; Schacter 1985; Thompson 1989:247). Morphological categories

associated with verbs include voice, tense, aspect, and mood. These categories are described in

Chapter 2. Although only verbs are inflected for tense, stative verbs in Bonggi are not inflected for

tense; thus, even tense inflection is not a definitive criteria for distinguishing verbs from nouns and

adjectives.









32Unlike English, there is no copula in Bonggi. Equational clauses are described in 2.1.1.

33Cf. McCawley (1983) who argues that word classes can be learned from inflectional morphology
regardless of any syntactic considerations.









Verbs usually function as predicates. However, both nouns (such as sigu hu 'my teacher' in

(48)) and adjectives (such as sega' 'red' in (49)) may also function as predicates.34

(49) Sega' mata nu.
red eye your
YOUR EYES are red.'
Some roots are very difficult to classify in terms of lexical category. One of the most

troublesome is kerai 'work' which can be a noun or a verb. This root is usually uninflected which

is characteristic of nouns, but it can be inflected like a verb. Furthermore, its syntactic distribution

can be either nominal (as in (50)) or verbal (as in (51)).

(50) Sia ndara kerai.
she not.have work
'SHE does not have work.'

(51) Sia kerai deirdn na ga.
she work self DEF CONTRAST
'SHE is working alone.'
Because the verb in (51) is unaffixed, the form of the word kerai 'work' is the same in both

(50) and (51). There is a functional shift from one part of speech to another. This is similar to

English where a verb like walk can occur as a noun by using it in a syntactic position reserved for

nouns, as in she took a long walk.35


Adjectives


Like nouns, adjectives are not inflected in Bonggi. Adjectives are distinguished from nouns

and verbs on morphological grounds only with respect to some derivational morphology.36


34The adjective root sega' 'red' can function as an adjective or a stative verb. As a stative verb, it
is either affixed (nsega) or unaffixed (sega). Compare the discussion of the root bulag 'blind' in
(52), (53), and (54).

35Unlike English, I am uncertain of the direction of the shift in Bonggi. In any case, it does not
matter for this study.

36Dixon (1977:62-63) claims that adjectives are a class of lexical items distinguished on
morphological and syntactic grounds from the universal classes noun and verb.









Chapter 2 shows how verbs derived from adjectives sometimes have a different form than those

derived from nouns.

Adjectives usually function as nominal modifiers which follow the noun they modify, as was

illustrated in (35). One of the biggest problems associated with adjectives in Bonggi is

distinguishing them from stative verbs. In some cases the distinction is clear-cut, while in others it

is not. For example, in (52) the adjective root bulag 'blind' follows its nominal head and is

unaffixed. Thus, both syntactically and morphologically bulag is an adjective in (52). In contrast,

the root bulag functions syntactically as a verb in (53) and is morphologically marked as a stative

verb by the prefix m-. On the other hand, in (54) bulag functions syntactically as a verb, but looks

like an adjective morphologically.

(52) Sia lama bulag.
she person blind
'SHE is a blind person.'

(53) Mbulag mata nya.
blind eye her
'HER EYES are blind.'

(54) Bulag mata nya.
blind eye her
THER EYES are blind.'
To summarize, this section has provided an overview of the distributional, syntactic, and

structural properties which are relevant for distinguishing noun, verb, and adjective roots in

Bonggi. None of the properties are definitive, working for all instances of a particular category.

Inflectional morphology is not very helpful, since nouns and adjectives are uninflected.

Derivational morphology is discussed in Chapter 2 where it is argued that some morphemes are

conditioned by the lexical category of the root to which the morpheme is affixed. Thus,

derivational morphology provides evidence for lexical categories. The distributional and syntactic

evidence discussed above provide the best evidence for the three categories: noun, verb, and

adjective.














CHAPTER 2
SITUATION ASPECT IN BONGGI

Whereas 1.3 provided a broad outline of Bonggi clauses and nominal phrases while skirting

issues related to verbs, this chapter focuses on the 'verbal group' and its involvement with aspect.

Furthermore, whereas 1.3 was descriptive and theoretical, this chapter approaches the verbal

system of Bonggi from the perspective of Role and Reference Grammar (RRG). Specifically, it

provides a RRG account of situation aspect (Aktionsart) in Bonggi.

.-- RRG has been chosen as a model because the morphology of Bonggi lends itself to this

framework. The primary descriptions of RRG are Foley and Van Valin (1984) and Van Valin

(1993) which are referenced here as FVV (1984) and VV (1993) respectively.

j, RRG is a structural-functionalist theory concerned with the interplay of syntax, semantics,

and pragmatics in grammatical systems. All of these factors are important in understanding

aspect, as will be shown in Chapter 3. This chapter does not assume a basic knowledge of RRG

on the part of the reader. Concepts from RRG are introduced as needed, and the reader is referred

to VV (1993) for elaboration of the theory and model. I begin the description of the verb system

with a discussion of verbal semantics which is fundamental for any discussion of aspect.

Verbal semantics is primarily concerned with the classification of verbs according to the

relationships which exist between predicates and their arguments. The relationship between a

predicate and its arguments is expressed by logical structures. It is important to keep in mind that

the verb classes described here are semantically defined. Some approaches to verbal semantics

attempt to account for the complete semantic content of the predicate, while others seek to identify

only the relevant components required for classification. Following Dowty (1979), FVV (1984),

and VV (1993), I pay attention to only those aspects of meaning relevant to the classification of









verbs (cf. FVV 1984:35; Jolly 1993:290). Verbs are classified on the basis of their inherent

aspectual properties (Aktionsart).

/, RRG starts with the classification of verbs into situation types: states, achievements,

activities, and accomplishments (VV 1993:34). As stated in Chapter 1, the notion of situation is

basic to understanding Bonggi verb morphology. Other notions like aspect, tense, mood, and

subject potential are restricted by the various situation types, and are secondary. The verbal

system which is described in this chapter is built around the notion of situation type. Other notions

are introduced as needed.

A fundamental distinction in situation types exists between events and nonevents (states).1

This distinction is captured in Figure 2.1 where events include achievements, activities, and

accomplishments. Table 1.1 accounts for this basic distinction in terms of the aspectual feature

dynamicity with states being [-dynamic] and events [+dynamic].



Situations
/ \
States Event
/ I \
/ I \
Achievements Activities Accomplishments

Figure 2.1: Relationship between states and events



RRG posits only a single level of syntactic representation. There is a direct linking between

the semantic and syntactic representations. Semantic representations are based on Dowty's (1979)

theory of verbal semantics in which verbs are classified into states, achievements, activities, and

accomplishments.2 Each verb class is given a formal representation called its Logical Structure




ISituation is neutral between event and state (Smith 1983:481).
2Cf. Van Valin (1990:222) for a brief overview. See 1.2 for a brief introduction to these four
classes in Bonggi.









(LS). The inherent aspectual properties associated with each verb class are accounted for in the

LSs by predicates and a small set of operators (cf. Dowty 1979:71; Van Valin 1990:223).

Dowty (1979:60) proposed a series of semantic and morphosyntactic tests to delineate each of

the four major situation types (cf. Van Valin 1990:223; Van Valin 1991:155). Some of these tests

are applicable to Bonggi and others are not.

Clause structure is viewed as hierarchially structured in RRG. The primary constituents of a

clause are the nucleus, which contains the predicate, the core, which contains the nucleus and the

arguments of the predicate, and the periphery, which is an adjunct to the core and includes

nonarguments of the predicate (cf. Van Valin 1993:5). These three clause constituents are

semantically motivated by two contrasts: a) between predicate and argument, and b) between

arguments and nonarguments (adjuncts).

Although the predicate is usually a verb, there are nonverbal stative clauses in Bonggi whose

predicate is either a nominal or a locative. The relationship between situation type and predicate

type is summarized in Table 2.1.



Table 2.1: Relationship between situation type and predicate type


SITUATION

State Achievement Activity Accomplishment
Nonverbal Verb Verb Verb Verb
type

nominal locative


The following sections describe the four situation types, their inherent aspectual properties,

and how these situation types and aspectual properties are reflected in Bonggi morphosyntax.






34

2.1. States

States are static situations with no activity. They last or endure through time, and are

homogenous throughout the period of their existence. They are the most noun-like verbal

situations in that they are time-stable. Stative situations are basic in the sense that the semantic

structure of achievements, most accomplishments, and some activities are derived from states.3

Bonggi stative situations can be divided between verbal and nonverbal situations. Nonverbal

stative situations are possible because of the time-stablity associated with states and the absence of

a copula verb. Nonverbal states include equational/identificational clauses (such as He is my

teacher (cf. (1)) and certain locative clauses (such as He is at his house (cf. (8)). Nonverbal states

differ from other states in that their predicate is not a verb. 2.1.1 deals with

equational/identificational statives which are always nonverbal. 2.1.2 deals with locative statives,

both nonverbal and verbal. 2.1.3 deals with condition statives, 2.1.4 with possession statives,

and 2.1.5 with cognition statives.4

Dowty (1979:60) presents a number of syntactic and semantic tests used to differentiate the

four situation types.5 Table 2.2 summarizes tests which have been found to be most useful in

differentiating situation types (cf. Van Valin 1991:155). Some of the tests in Table 2.2 are not

always useful cross-linguistically and can be very difficult to apply. In theory, these tests are very

important because they provide independent criteria for determining situation types and their

logical structures (LSs). However, in practice, any tests comparable to those in Table 2.2 which I

use for differentiating situation types have been developed in a post hoc fashion. This is because

Bonggi verb morphology is quite transparent with respect to situation type. If you know the

meaning of the verb root and the verb morphology, the situation type is usually straightforward.



3Dowty (1979) claimed that stative predicates were the only primitives in the LS, and that all other
predicates are derived from statives by means of logical operators.

4Cf. FVV (1984:47-53) and VV (1993) for a discussion of the various sub-classes of stative verbs.









However, from an analytical point of view, verb morphology cannot be used to explain situation

type because this leads to circular reasoning. That is, we cannot claim both (i) verb morphology

determines situation type, and (ii) situation type determines verb morphology. In terms of language

learning, Bonggi verb morphology is a very good indicator of situation type. In summary, I learned

the verb forms and their function in distinguishing situation types. Then, given the situation types

available in Bonggi, I asked what tests comparable to those in Table 2.2 could be established to

distinguish the different situations.



Table 2.2: Tests for differentiating situation types

Criterion States Achievements Accomplishments Activities
Durative Punctual
1. Occurs with progressive No Yes No Yes Yes
2. Occurs with adverbs like No No No Yes Yes
vigorously, carefully, etc.
3. Occurs with for an Yes Yes No Yes Yes
hour, spend an hour (ing
4. Occurs with 0 in an No Yes No Yes No
hour, take an hour to


In contrast to the tests used to distinguish situation types, there are no tests to distinguish the

various subclasses of situations (cf. Van Valin 1994). The question How do we know when we

have a stative situation? is answered by reference to the loose semantic notions mentioned at the

beginning of this section. That is, states are static situations with no activity. One criterion which


5These tests are summarized in FVV (1984:37), Van Valin (1990:223), Van Valin (1991:155), and
Van Valin (1993:35).









clearly distinguishes some states from events in Bonggi is whether or not the predicate is a verb.

Nonverbal situations are always states; however, states are not always nonverbal.


2.1.1. Equational States

Equational/identificational clauses contain two nominals, as illustrated in (1) by sia 'he' and

sigu hu 'my teacher'.6 The second nominal in equational clauses always identifies or attributes

some characteristic to the referent of the first nominal.

(1) Sia sigu hu.
3sNOM teacher IsGEN
'HE is my teacher.'

A general characteristic of states is that they attribute some property to an entity. The type of

property which is attributed to the entity influences the type of clause that occurs in syntax. For

example, when attributing a social role (such as sigu 'teacher' in (1)) or kinship relation (such as

kuman 'uncle' in (2)) to a person (that is, a human entity), the social role or kinship relation is

realized as a nominal in syntax and functions as the clause predicate. Thus, an

equational/identificational clause can be defined syntactically as a clause with a nominal predicate

and semantically as a clause which attributes a social role or kinship relation to an entity.7

(2) Sia kuman ku.
3sNOM uncle IsGEN
'HE is my uncle.'

In RRG, thematic relations are not primitives; instead, they are defined in terms of logical

structures (LSs) (cf. VV 1993:39, 43). Thus, the assignment of LSs is prior to the assignment of

thematic relations. The LS for equational clauses (e.g. (1) and (2)) is shown in (3) (cf. VV

1993:36).





6Cf. 1.3.1 and 1.3.2 for a discussion of the types of nominals that can occur in Bonggi.

71In contrast to equational clauses, locative clauses occur when the property attributed to an entity
is the location of that entity. The location is realized as either a deictic adverb, a locative phrase,
or a locative stative verb. See 2.1.2 for a discussion of locative stative situations.






37

(3) LS for equational states: be' (x,y)
The LS in (3) indicates that equational states are two-place stative verbs with 'x' and 'y' being

the two arguments.8 The next step in the analysis of a clause is to assign thematic relations to any

arguments in the LS (for instance, 'x' and 'y' in (3)). In RRG thematic relations are derived from

the LS. Therefore, the assignment of thematic relations is independently motivated.

The question then becomes what are the thematic relations of'x' and 'y' in (3) or, more

specifically, what are the thematic relations of sia 'he' and kuman ku 'my uncle' in (2)?9 Schwartz

(1993) takes (3) to be the LS for equational/identificational clauses, and claims the 6-roles which

are derived from (3) are locative and theme. She argues that in clauses such as (2), sia 'he' is the

locative argument and kuman ku 'my uncle' the theme (cf. VV 1993:40). Thus, the theme argument

is the attribute, and the locative argument is the bearer of the attribute (cf. Van Valin 1990:234).

In RRG, 6-roles are said to be language independent; the first argument 'x' of two-place equational

stative verbs being a locative, and the second argument 'y' a theme. I return to the question of 0-

roles after introducing the notion of macrorole.

RRG posits two tiers of semantic roles (VV 1993:39). Actor and undergoer are the two

primary arguments of a transitive predicate, either one of which may be the single argument of an

intransitive verb (VV 1993:43). Actor and undergoer correspond to 'logical subject' and 'logical

object' (VV 1993:43, 46).10 Actor and undergoer are called macroroles because each subsumes a

number of specific thematic relations. Macroroles are motivated by the fact that in grammatical

constructions, groups of thematic relations are treated alike (VV 1993:43). The number of

macroroles a verb takes is either 0, 1, or 2, and is largely predictable from the LS of the verb (VV

1993:46-47).






8FVV (1984) took the position that equational clauses are one-place stative verbs whose LS is
predicate' (x). Van Valin (1990; 1993) has since revised this position.

9Various solutions have been offered; see Schwartz (1993) for discussion.









The relationship between the macrorole tier and thematic relation tier is captured in the Actor-

Undergoer Hierarchy in (4) (cf. FVV 1984:59; VV 1993:44). This double hierarchy states that the

thematic relation that is leftmost on the cline and either agent, effector, experience, or source will

be the actor, and the thematic relation that is rightmost and either patient, theme, goal, locative, or

experience will be the undergoer. This is the unmarked situation; marked assignments to

undergoer are possible.

(4) Actor-Undergoer Hierarchy

Macrorole tier: ACTOR UNDERGOER


Thematic relation tier: Agent Effector Experiencer Locative Theme Patient
/ \
Source Goal
[-- = increasing markedness of realization of thematic relation as macrorole]

Recently, Van Valin (personal communication) has reformulated the Actor-Undergoer

Hierarchy as (5) where reference is not made to thematic roles at all, only to LS argument

positions.11

(5) Actor-Undergoer Hierarchy (Revised)

Macrorole tier: ACTOR UNDERGOER


ArgofDO Argofdo' Istargof 2ndargof Arg of state
pred'(x,y) pred'(x,y) pred'(x)
[-- = increasing markedness of realization of argument as macrorole]

The revised Actor-Undergoer Hierarchy in (5) is based on the RRG position that 0-roles are

not independently motivated. In fact, according to the view in (5), 0-roles are nothing more than

mnemonics for argument positions (cf. VV 1993; Wilkins & Van Valin 1993). This alleviates the

problem of 0-role assignment which has plagued linguists since Fillmore (1968).


10Actor and undergoer are similar to Dowty's protoroles (W 1993:154, footnote 25).

1iThe LS argument positions in (5) are described later in this chapter.









Given: a) the revised hierarchy in (5), and b) the LS for equational clauses in (3), the question

of thematic relations posed above does not even arise. The second argument of predicate' (x,y) is

the unmarked undergoer since it is rightmost on the cline in (5). However, given: a) the hierarchy

in (4), b) the LS in (3), and c) the RRG claim that the first argument 'x' of two-place equational

stative verbs is a locative, and the second argument 'y' a theme (Schwartz 1993; W 1993), then

the theme is the unmarked undergoer since it is rightmost on the cline in (4). Thus, in both

instances, it is expected that in (2) kuman ku 'my uncle' would be the undergoer, but this is not

what happens in equational clauses. In equational/identificational clauses, the second argument (or

theme) 'y' is incorporated into the predicate resulting in a nominal predicate (cf. Van Valin

1990:234; Van Valin 1993:40; Schwartz 1993:447). The second argument in the LS of equational

clauses is never realized syntactically as an argument. This is captured by the rule in (6) (cf. Van

Valin 1990:234; Schwartz 1993:447).12

(6) EQUATIONAL PREDICATE CREATION:

be' + theme -- predicate

The result of (6) is that there is only one argument which can can be the undergoer in both (1)

and (2), the first argument (or locative argument) 'x'. This results in a marked linking in terms of

the hierarchy in either (4) or (5). Stated another way, the undergoer is functionally marked.13

Because the theme argument is incorporated into the predicate by (6), it is not available as a

syntactic argument. Thus, sia 'he' is linked to undergoer in both (1) and (2) (cf. Van Valin

1990:234).

To summarize what has been said with respect to LSs, argument positions, thematic roles, and

the Actor-Undergoer Hierarchy, LSs are primitives, and semantic arguments such as 'x' and 'y' are



12The rule in (6) is stated in terms of thematic roles and the Actor-Undergoer Hierarchy in (4). It
could just as well be stated in terms of argument positions and the hierarchy in (5); i.e., be' + the
2nd argument of predicate' (x,y) -- predicate.
131 show later that some functionally marked undergoers are accompanied by corresponding
morphological marking which is absent in equational clauses such as (1) and (2).









included in LSs. On the other hand, thematic roles are not primitives. Within RRG, thematic roles

are traditionally defined in terms of LSs. This has been done by assigning semantic labels such as

'theme' to argument positions in the LS. Some thematic roles are then linked to macroroles (actor

or undergoer) following the hierarchy in (4). More recently, Van Valin suggests doing away with

thematic roles all together, and simply linking argument positions in the LS to the macroroles

following the hierarchy in (5).

Normally, arguments in the LS are linked to argument positions in syntax. However, the

analysis presented here requires that the second argument in the LS of equational states be

incorporated into the predicate, so that only one of the two LS arguments is linked to an argument

position in syntax. Restated in terms of thematic roles, the analysis presented here requires that the

theme argument be incorporated into the predicate, so that only one of the two thematic roles is

linked to an argument position in syntax (cf. Schwartz 1993:442). This analysis asserts that

although nominals normally function as arguments, they can also function as predicates. Thus,

nominals such as sigu hu 'my teacher' in (1) and kuman ku 'my uncle' in (2) are predicates which

are neither assigned a thematic role nor linked to an argument position in syntax.14 In Bonggi

equational clauses, the attribute functions as the predicate.15

Once thematic roles have been assigned to macroroles, the next step is to assign actor and

undergoer to specific morphosyntactic statuses (VV 1993:76). The most important syntactic

function is the subject or pivot. The pivot is the primary argument in a syntactic construction.

The highest ranking macrorole is assigned to the pivot. Bonggi is syntactically an accusative

language. The Pivot Choice Hierarchy for syntactically accusative languages is shown in (7) with

the leftmost item being the least marked pivot choice (cf. W 1993:59).





14Cf. Schachter's (1985:7) analysis of Tagalog equational clauses.

15Cf. Schwartz's (1993:443) discussion of Dakota. She uses the term 'identificational', whereas
VV (1993:39) uses the term 'equational.' I use these terms interchangeably.








(7) Pivot Choice Hierarchy

Actor > Undergoer > other
According to the hierarchy in (7), the undergoer sia 'he' in both (1) and (2) is assigned the

status of syntactic pivot since it is the only available macrorole.

Part of the process involved in assigning actor and undergoer to specific morphosyntactic

statuses is case and preposition assignment. The pivot is always assigned nominative case. As

pointed out in 1.3.1, pivots (subjects) are marked in one of three ways in Bonggi: (a) the special

particle si for personal names, (b) nominative case pronouns, or (c) word-order. For example,

because the pivot in (1) is a pronoun, it is assigned nominative case and the form is sia 'he' (cf.

Table 1.2). On the other hand, nominals such as sigu hu 'my teacher' in (1) cannot be replaced by

a pronoun and assigned case. The absence of pronominal forms and case strengthens the argument

stated above that attribute nominals in equational clauses (e.g. sigu hu 'my teacher' in (1)) are not

linked to an argument position in syntax.

The pivot choice hierarchy in (7) claims that the actor is the unmarked choice for syntactic

pivot in a clause. However, equational clauses have only one macrorole, an undergoer, which must

be the syntactic pivot. Choosing a pivot other than the actor is only a marked pivot choice when

there is also an actor macrorole.

In summary, equational states occur in Bonggi syntax as nonverbal clauses whose predicate is

a nominal which refers to the attribute ascribed to the clause pivot. Unlike verbal predicates which

index or cross-reference the clause pivot (subject) in the verb morphology,16 nonverbal predicates

are not cross-referenced.






16See the discussion of verbal indexing (or cross-referencing) in 1.3.1, following examples (18),
(19), (20), (21), and (22). The affixes involved in verbal cross-referencing are often referred to as
voice affixes (e.g. de Guzman 1986:349). De Guzman concludes that identificational clauses in
Tagalog are also nonverbal, and the predicate is nominal. However, her paper primarily deals with
cleft constructions which are derived from equational/identificational clauses like those described in
this section.









The analysis of equational states presented in this section is dependent upon the lexical rule of

argument incorporation found in (6). Although a claim has been made that incorporation takes

place, what remains to be accounted for is evidence for incorporation. Evidence for incorporation

includes: a) impossibility of case assignment to the incorporated argument; and b) impossibility of

prepositions occurring with the incorporated argument.


2.1.2. Locative States

This section introduces locative states, whereas later sections show how some verbs are

derived from these statives. There are two types of locative stative clauses in Bonggi, nonverbal

and verbal. I begin with a description of nonverbal locative clauses followed by verbal locative

clauses.

Nonverbal clauses are defined as clauses whose predicate is not a verb. The previous section

showed that Bonggi equational statives are always nonverbal, and the predicate is a nominal. Like

equational statives, many locative stative clauses in Bonggi are nonverbal.

As pointed out in the previous section, a general characteristic of states is that they attribute

some property to an entity. When the property attributed to an entity is the location of that entity,

the result is a locative clause. In nonverbal locative clauses, the location is realized in syntax as

either a deictic adverb or a locative prepositional phrase, and the deictic or preposition functions as

the clause predicate. Thus, a nonverbal locative clause can be defined as a clause with a deictic or

prepositional predicate. A nonverbal locative clause with a locative prepositional phrase is

illustrated in (8).

(8) Sia dii bali nya.
3sNOM at house 3sGEN
'HE is at his house.'
Locative stative predicates have the two-place abstract predicate, be-at' (x,y) in their LS

(Jolly 1993:277). The LS for locative statives (including (8)) is shown in (9).

(9) LS for locative statives: be-at' (x,y)









According to an RRG analysis in terms of thematic roles (cf. 2.1.1), the first argument 'x' of

two-place locative stative verbs is a locative, and the second argument 'y' a theme (VV 1993:40).

For example, in (8) bali nya 'his house' is a locative argument, sia 'he' a theme, and the predicate is

the preposition dii 'at'.

As stated in 2.1.1, the number of macroroles a verb takes is either 0, 1, or 2, and is normally

predictable from the LS of the verb. That is, if there are two or more arguments in the LS of a

verb, then the verb takes two macroroles in the default situation. The nature of the macroroles is

also derived from the verb's LS. Macrorole assignment principles are summarized in (10) (cf. Van

Valin 1990:227; Van Valin 1993:47).

(10) GENERAL MACROROLE ASSIGNMENT PRINCIPLES:
a. Number: the number of macroroles a verb takes is less than or equal to the number of
arguments in its LS.
1. If a verb has two or more arguments in its LS, it will take two macroroles.
2. If a verb has one argument in its LS, it will take one macrorole.
b. Nature: for verbs which take one macrorole,
1. If the verb has an activity predicate in its LS, the macrorole is actor.
2. If the verb has no activity predicate in its LS, the macrorole is undergoer.
According to the default situation in (10a. 1), (8) should have two macroroles since its LS in

(9) has two arguments 'x' and 'y'. However, (8) is a nonverbal locative clause, and such clauses are

an exception to (10a. 1) in that they have only a single macrorole. That is, these clauses are an

exception to the default situation in (10a. 1), but they do not contradict the general principle in

(10a). The nature of the single macrorole is predictable from (10b); that is, the single macrorole in

(8) is an undergoer since there is no activity predicate in its LS in (9).17

Following the Actor-Undergoer Hierarchy in (4), the theme argument is the unmarked choice

for undergoer since it is the rightmost of the two thematic relations locativee and theme). Thus, in

(8) the theme argument (sia 'he') is assigned the macrorole status undergoer. Furthermore, because

the undergoer is the only macrorole available, it is assigned to the pivot according to the Pivot





17Activities and subclasses of activity predicates are described in 2.3. Activity predicates are
defined as predicates with do' in their LS.









Choice Hierarchy in (7). As pointed out earlier, the pivot is always assigned nominative case.

Therefore, because the undergoer-pivot in (8) is a pronoun, it is assigned nominative case and the

form is sia 'he' (cf. Table 1.2).

It is important to keep in mind that the LSs described here provide only a partial semantic

representation of the verb and clause structure. Furthermore, the LS in (9) represents the class of

locative statives, and not simply the LS of (8). Thus, the presence of be-at' in (9) does not mean

that surface syntax must contain a locative phrase introduced by dii 'at' as in (8). Instead, any

member of certain sets of locative elements including dii 'at' can occur. Dii itself can be a

preposition as in (8) or a locative adverb.18 There are two sets of spatial deictic adverbs in

Bonggi. The first set, which includes dii 'there', refers to specific locations which are relative to the

speaker as shown in Table 2.3.



Table 2.3: Specific spatial deictics


The second set of spatial deictics, which are shown in Table 2.4, refer to nonspecific spatial

deictics; that is, they are more vague or have more approximate locations than their counterparts in

Table 2.3.





18Although locative adverbs and locative prepositional phrases belong to different word classes,
they have the same function. Adverbs can be seen as abbreviated prepositional phrases.


Form Meaning

diti 'here' (near speaker)

dia 'there' (near addressee)

dioo 'there' (not near speaker or addressee, but usually visible)

dii 'yonder' (not visible)






45

Table 2.4: Nonspecific spatial deictics


The unmarked or basic locative preposition is dii 'at' which can be combined with other

spatial prepositions to form more complex locative meanings as in (11) and (12).19

(11) Sia dii soid bali nya.
3sNOM at inside house 3sGEN
'HE is inside his house.'

(12) Sia dii sodi diaadn.
3sNOM at beside IsACC
'HE is beside me.'

Based on the analytical steps introduced above, (13) is a summary analysis of (11).

(13) a. LS : be-at' (x,y) x=locative, y=theme
b. Assign thematic relations: bali nya 'his house' <- locative20
sia 'he' <- theme
c. Assign semantic macroroles: undergoer +- theme 1 Macrorole
d. Assign syntactic functions: pivot <- undergoer-sia 'he'
oblique argument <- locative-bali nya 'his house'
e. Assign case and prepositions: pivot-sia 'he' <- nominative case
locative-bali nya 'his house' <- oblique case
dii soid 'inside'

(13a) indicates that the LS for (11) is the same as that shown in (9) for (8). Furthermore,

(13a) claims that, as always, the first argument 'x' of locative statives is a locative and the second

'y' a theme. (13b) assigns the thematic relations locative and theme to specific elements in (11).




19Cf. Jolly (1993:290) for references supporting the choice of'at' as the basic locative.

20The symbol <- is used to mean 'assigned', with the assignment being in the direction of the
arrow.


Form Meaning

kati' 'somewhere here' (near speaker)

kana'/kono' 'somewhere there' (near addressee)

kenoo 'somewhere there' (not near speaker or addressee, but usually visible)

kuii' 'somewhere yonder' (not visible)









Nonverbal locative clauses (e.g. (8), (11), and (12)) have two thematic roles, but only one core

syntactic argument position and one macrorole.21 This corresponds to the general principle in

(10a), whereby a predicate can have fewer macroroles than it has arguments in its LS. The nature

of the single macrorole is predictable from (10b); since there is no activity predicate in the LS in

(13a), the single macrorole has to be an undergoer. (13c) claims that theme is assigned to

undergoer which is the single macrorole. The assignment of theme to undergoer follows from the

Actor-Undergoer Hierarchy in (4), since theme is the rightmost of the two thematic relations

locativee and theme). (13d) assigns the undergoer (sia 'he') to the syntactic status of clause pivot

according to the Pivot Choice Hierarchy in (7), and it assigns the locative argument to the syntactic

status of oblique argument. Finally, (13e) shows that the pivot is assigned nominative case and the

locative argument oblique case.22

In locative stative situations the theme (the figure in gestalt terms) is located with respect to a

point of reference (the ground or location) (cf. Talmy 1985:61). The point of reference

(ground/location) does not have to be an inanimate object incapable of movement as shown by

diaadn 'me' in (12). A summary analysis of (12) would be the same as that provided for (11) in

(13), with the exception of diaadn 'me' being assigned the thematic relation locative in (13b) and

the assignment of accusative case to diaadn 'me'.

As stated above, dii 'at' is the basic locative preposition which combines with other spatial

prepositions as shown in Table 2.5.









21ln nonverbal locative clauses the locative argument is always an oblique argument which is
preceded by an oblique marker.

22The assignment of case and adpositions is discussed in more detail below. The locative elements
dii 'at', dii solid 'inside', and dii sodi 'beside' in (8), (11), and (12) respectively function as both
predicates and prepositions in these clauses.








Table 2.5: Locative preposition combinations


Because the LS in (9) represents the class of locative statives, any of the spatial deictics in

Tables 2.3 or 2.4, or any of the prepositions in Table 2.5 could function as the predicate in a

nonverbal locative clause. For example, the predicate in (14) is the spatial deictic kati' 'somewhere

here'.

(14) Sia kati'.
3sNOM here.somewhere
'HE is somewhere here.'
The locative states discussed thus far have all been nonverbal (e.g. (8), (11), (12), and (14)).

Nonverbal locative clauses are defined as clauses whose predicate is either a deictic adverb (e.g.

(14)) or a locative preposition (e.g. (8), (11), and (12)). Verbal locative clauses are defined as

clauses whose predicate is a stative verb. Such verbs are derived from adjective roots which have

a locative meaning. Sentence (15) provides an example of the locative stative verb mingad 'ST-

near'.

(15) Sia m-ingad bali nya.
3sNOM ST-near house 3sGEN
'HE is near his house.'




23Dii soid 'inside' is the correct pronunciation in the Limbuak Darat cluster, while dii soig 'inside'
is the correct pronunciation in the Palak Darat cluster (cf. 1.1).


Form Meaning Form Meaning

dii libuat on top of dii sidukng below; underneath

dii sirib below; underneath dii soid inside23

dii ruar outside dii sodi beside; in the vicinity of

dii tingguangan in the face of dii kimidiadn behind; after

dii puguluadn before; in front of dii simbela' on the side; on the other side

dii tenga'-tenga' in the middle of dii seborokng on the other side of a body of water









Thus far, I have discussed LSs as a general property of clauses; for example, be' (x,y) is the

LS for equational stative clauses (cf. (3)), and be-at' (x,y) is the LS for locative stative clauses (cf.

(9)).24 A distinction is made between general LSs and clause-specific LSs. For example, the

general LS for locative statives (e.g. (15) and (8), repeated here as (16)) is shown in (17a), whereas

the clause-specific LS for (15) is provided in (17b) and the clause-specific LS for (16) in (17c).

(16) Sia dii bali nya.
3sNOM at house 3sGEN
'HE is at his house.'

(17) a. General LS for locative statives: be-at' (x,y) x=locative, y=theme
b. Clause-specific LS for (15): ingad 'near' (bali nya 'his house', sia 'he')
c. Clause-specific LS for (16): dii 'at' (bali nya 'his house', sia 'he')
A summary analysis of (15) is provided in (18).25

(18) a. General LS for locative statives: be-at' (x,y) x=locative, y=theme
b. Clause-specific LS for (15): ingad 'near' (bali nya 'his house', sia 'he')
c. Assign thematic relations: bali nya 'his house' +- locative
sia 'he' +- theme
d. Assign semantic macroroles: undergoer -- theme 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sia 'he'
direct core syntactic argument <- locative-bali nya
'his house'
f. Assign case and prepositions: pivot-sia 'he' <- nominative case
direct core argument-bali nya 'his house' accusative
case (0)26
g. Cross-reference verb: ingad 'near' m- 'ST'
(18a) provides the general LS for all locative stative clauses, both verbal and nonverbal (cf.

(13a)). (18b) repeats the clause-specific LS for (15) (cf. (17b)). (18c) assigns the thematic

relations locative and theme to specific elements in (15). (18d) assigns the theme sia 'he' to

undergoer following the Actor-Undergoer Hierarchy in (4). (18e) assigns the undergoer (sia 'he') to

the syntactic status of clause pivot according to the Pivot Choice Hierarchy in (7), and it assigns




24Equational clauses are nonverbal stative clauses whose predicate is a nominal. Locative clauses
are stative clauses whose predicate is either verbal (e.g. (15)) or nonverbal in which case the
predicate is a deictic adverb (e.g. (14)) or a locative preposition (e.g. (8), (11), and (12)).

25Cf. the summary analysis of (11) found in (13).

26Common nouns such as bali 'house' are not case marked which is indicated by (0).









the locative argument to the syntactic status of direct core syntactic argument.27 (18f) assigns

nominative case to the pivot, and accusative case to the other direct core argument. (18g) cross-

references the verb with the prefix m-.

Previous discussion did not include cross-referencing the verb because the predicate of

nonverbal clauses is not cross-referenced. However, verbal clauses such as (15) usually require

verbal cross-referencing. The verbal cross-referencing system has traditionally been called a

"focus" system within Philippine linguistics, with one NP in a clause indexed by the morphology of

the verb as being "in focus." The role of the verb morphology is usually considered as encoding

voice distinctions or the semantic role of the NP. Only one NP is cross-referenced and it is the

pivot. The undergoer-theme in locative clauses is cross-referenced by the verbal prefix m- when

the predicate is a stative verb.

Besides the distinction between general LSs and clause-specific LSs is the distinction between

semantic arguments and syntactic arguments. Semantic arguments refer to arguments in the LS

such as 'x' and y' in (17a). The arguments in the LS of a VERB provide a strict definition of core

semantic arguments (VV 1993:40). In the simple case, there is a one-to-one correspondence

between the number of arguments in the LS of a verb and the number of arguments in syntax. This

is clearly the case in (18b) and (15), where there are two semantic arguments in the LS in (18b)

and two syntactic arguments in (15).

In some instances, however, there are arguments in the general LS which do not occur in

syntax. For example, the general and clause-specific LSs for (19) are shown in (20a) and (20b).








27Cf. the assignment of the locative argument to the syntactic status of direct core argument in
(18e), with the assignment of the locative argument to the syntactic status of oblique argument in
(13d). Like nonverbal locative clauses, the locative argument in verbal locative clauses can be
assigned the syntactic status of oblique argument. For example, in (15) the locative argument, bali
nya 'his house', can be preceded by the oblique marker dii 'at'.









The predicate in (19) is dii 'there',28 the theme is sia 'he', and there is no locative argument present.

The absence of a locative argument in syntax is accounted for in the clause-specific LS of (20b),

where the locative argument is unspecified being filled by "0".29

(19) Sia dii.
3sNOM there
tHE is there.'

(20) a. General LS for locative statives: be-at' (x,y) x=locative, y-=theme
b. Clause-specific LS for (19): dii there' (0, sia 'he')30
To summarize, locative states occur in Bonggi syntax as either verbal or nonverbal clauses.

The locative is realized morphosyntactically as a stative verb in verbal clauses, and functions

predicatively in nonverbal clauses.


2.1.3. Condition States

The previous two sections have provided an overview of two types of stative situations in

Bonggi. The former section showed that the function of equational/identificational clauses is to

attribute a social role (including kinship relation) to an entity, whereas the latter showed that the

function of locational clauses is to attribute a location to an entity. It was also pointed out that the

predicate of the former is always realized as a nominal, whereas the predicate of the latter is

realized as either a deictic adverb, a locative preposition, or a stative verb.

This section provides strong grounds, both semantically and morphosyntactically, for

distinguishing locative statives from condition statives. It begins with a description of the simplest

kind of verbal stative situations which are one-place condition stative clauses. The function of





28Recall that dii 'there/at' can be either a spatial deictic as in (19) (cf. Table 2.3) or a preposition
as in (16). In both of these instances, dii functions as the predicate of a locative clause. For
instances where dii functions as a simple preposition see 1.3.1, examples (25) and (26).
29Cf. WVV (1993:156; fi. 46).

30A similar clause-specific LS can be provided for (14).









condition stative clauses is to attribute a quality to an entity. The predicate of these clauses is

always a stative verb. Sentence (21) illustrates a condition stative clause.

(21) Sia ng-korikng (m-koring).31
3sNOM ST-dry
'IT is dry.'
The logical structure for condition statives is shown in (22).

(22) LS for condition statives: predicate' (x)
According to Van Valin (1993:39-40), the single argument of a one-place stative verb is a

patient. Based on the analytical steps introduced in the previous two sections, (23) is a summary

analysis of (21).

(23) a. General LS for condition statives: predicate' (x) x=patient
b. Clause-specific LS for (21): korikng 'dry' (sia 'it')
c. Assign thematic relations: sia 'it' <- patient
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot +- undergoer-sia 'it'
f. Assign case and prepositions: pivot-sia 'it' <- nominative case
g. Cross-reference verb: korikng 'dry' *- m- 'ST






31Morphophonemic alternations in Bonggi verbs are complex. Roots and accompanying verbal
affixes are enclosed in parentheses following the surface forms. Within roots, nasals are always
homorganic with contiguous consonants. The following nasal assimilation rule makes nasal
consonants homorganic with following nonsonorant consonants.
Nasal assimilation:
C C
[+nasal] -- Faanteriorl /__ -sonorant 1
Lacoronal J I anterior I
LBcoronal I
The nasal assimilation rule is ordered after vowel epenthesis which inserts vowels between prefixes
and certain consonants in order to avoid impermissible syllable structures. The vowel epenthesis
rule is:
Vowel epenthesis:
V C C V
0 [aF] / [+prefix] + F+sonorant 1 [aF]
L+continuant J
The vowel epenthesis rule is ordered before a prefix vowel harmony rule. The condition stative
verb mi-gia 'ST-big' (m-gia) is an exception to the rule of vowel epenthesis. As shown in (24),
roots whose initial consonant is /g/ such as garakng 'ferocious' are normally preceded by /3/ 'ng'.









According to (10a.2), the verb takes one macrorole since it has only one argument in its LS.

The nature of the single macrorole is predictable from (10b.2). Since there is no activity predicate

in the LS in (23a), the single macrorole has to be an undergoer.32 (23c) assigns the thematic

relation patient to the single argument in (21), and (23d) assigns the patient (sia 'it') to undergoer.

(23e) assigns the undergoer (sia 'it') to the syntactic status of clause pivot according to the Pivot

Choice Hierarchy in (7).33 (23f) assigns nominative case to the pivot, and (23g) cross-references

the verb with the prefix m-.

Since the attributes which are attributed to entities by condition stative clauses are qualities, it

is not surprising that many condition stative verbs are derived from adjective roots. For example,

the root koring 'dry' in (21) is an adjective root. Other condition stative verbs which are derived

from adjective roots are shown in (24). The roots are arranged according to how m- is realized.
(24) /m-/ In-/
m-ayad ST-pretty n-dalabm ST-deep
m-enta' ST-unripe n-doot ST-bad
m-ingi ST-crazy n-dupakng ST-foolish
m-iskidn ST-poor n-duruk ST-fast
m-omis ST-sweet n-sega' ST-red
m-udap ST-hungry n-tihukng ST-crooked
m-panas ST-hot n-togi' ST-pregnant
m-pia ST-good n-took ST-ripe
m-pagadn ST-difficult
m-pala ST-spicy hot
m-puti' ST-white
m-basa' ST-wet
m-basa ST-generous

lb/_- /mV-/
ng-garakng ST-ferocious me-lambat ST-slow
ng-gitiukng ST-dizzy me-langgu ST-long
ng-kapal ST-thick mi-libuat ST-high
ng-korikng ST-dry mi-liug ST-tall
ng-kotul ST-hard me-ramig ST-cold
mu-runggu' ST-slow




32Condition statives are like equational statives (cf. 2. 1.1) and nonverbal locative statives (cf.
2.1.2) in that they all have a single macrorole which is an undergoer according to (10b.2).

33In one-place condition stative constructions, the undergoer is always the pivot which is the noun
phrase that is crucially involved in a syntactic construction (Foley & Van Valin 1984:110).









Not all adjectives have the same status with respect to their potential for stative verb

affixation. The adjective roots in (25) can occur as condition stative predicates, but there is no

accompanying nasal prefix m-.34

(25) alus 'fine' arabm 'forbidden' baru 'new' biru 'blue' senang 'easy'
2.1.2 pointed out that in some instances there are arguments in the general LS which do not

occur in the syntax. Such arguments are unspecified in the clause-specific LS, being indicated by

"0" (e.g. (20b)). In other instances, there are adjuncts in syntax which are not part of the LS of the

verb, but are included in the clause-specific LS. For example, condition stative verbs such as

ngkorikng 'ST-dry' in (21) have a single syntactic argument which is semantically a patient.35

However, these clauses can have an optional adjunct such as ga' odu 'from the sun' in (26).36

(26) Sia ng-korikng (m-koring) ga' odu.
3sNOM ST-dry from sun
'IT is dry from the sun.'
The preposition ga' 'from' has an adjunct or predicative function in that it takes the entire LS

of the verb as one of its arguments. According to Jolly (1987) and (1993), prepositions are divided

into three classes within the RRG framework. Prepositions which have an adjunct function, such

as ga"from', are not case-marking (Jolly 1993:275, 281). The preposition ga"from' has a

causative meaning.37 Causality is expressed either directly or indirectly. Accomplishment verbs




34Perhaps the absence of a nasal prefix with alus 'fine' and arabm 'forbidden' arises from their
being borrowed from Malay halus 'fine' and haram 'unlawful; forbidden'. /h/ is phonemic in
Malay, but not in Bonggi. However, the presence of word-initial /h/ in the source language does
not block affixation in all cases. See the discussion of achievement verbs in 2.2, especially kem-
alus 'become-fine' and kem-arabm 'become-forbidden' in (76). The presence of baru 'new' and biru
'blue' in Malay may also be the reason for the absence of nasal prefixes in the Bonggi forms.

35See (22) for the general LS.

36Longacre (1976:64-65) refers to nouns like odu 'sun' in (26) as optional instruments. Bonggi
resorts to a quasi-passive in the surface syntax of condition stative clauses with optional adjuncts
(cf. Longacre 1976:65).

37Cf. Jolly (1987:104-06) and (1993:293-97) for a discussion of the English prepositionfrom in
its adjunct or predicative function.









express causality directly which is reflected in their LS by the logical operator CAUSE (cf. 2.4).

On the other hand, statives and achievements express causality indirectly, which is reflected in their

LS by ga"from' (cf. Jolly 1993:294).38

A summary analysis of (26) is provided in (27). Sentence (26) describes an antecedent cause

(ga' odu 'from the sun') and a resulting/consequent state (sia ngkorikng 'it is dry'). This is seen in

(27b) where 'x' refers to the antecedent cause and predicate' (y) to the resulting state.39 As seen in

(27a), the general LS for one-place condition statives does not contain an antecedent cause;

however, (27b) indicates that there is an indirect cause in the clause in question. Two clause-

specific LSs are provided in (27b). The latter is the same as the former with the exception that the

Bonggi arguments and predicate are provided in the latter. In (27b) the preposition ga"from' takes

the entire LS of the verb as one its arguments (cf. Jolly 1993:287). The first argument ofga'

'from' in (27b) is 'x' which is an antecedent cause, a subtype of effector.

(27) a. General LS for condition statives: predicate' (x) x=patient
b. Clause-specific LS for (26): from' (x, [predicate' (y)]) x=antecedent cause,
predicate' (y)=resulting
state,
y=patient
ga' 'from' (odu 'sun', [korikng 'dry' (sia 'it')])
c. Assign thematic relations: sia 'it' <- patient
odu 'sun' <- effector (antecedent cause)
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sia 'it'
oblique adjunct <- effector (antecedent cause)-odu
'sun'
f. Assign case and prepositions: pivot-sia 'it' <- nominative case
effector (antecedent cause)-odu 'sun' <- oblique ga'
'from'
g. Cross-reference verb: korikng 'dry' <- m- 'ST
According to the Actor-Undergoer Hierarchy in (4), the antecedent cause should be assigned

to actor since it is a subtype of effector, but it is not (cf. (27d)). The preposition ga' 'from' marks


38Cf. the discussion of cause satellites in Hokan languages in Talmy (1985:111-13).

39Jolly (1993:294) suggests that an analysis such as that provided in (27b) is provisional. She
provides a more detailed interval analysis for similar clauses in English.









effectors which would be expected to occur as actor in terms of the hierarchy in (4), but do not (cf.

FVV 1984:87; Jolly 1993:280). The antecedent cause cannot function as an actor because it is an

adjunct, not an argument of the verb. Recall that actor and undergoer are arguments of the

predicate. Since there is no actor, the undergoer is the pivot in (26) in accordance with the Pivot

Choice Hierarchy in (7) (cf. (27e)). In fact, condition stative clauses cannot have an actor; thus,

the pivot is always an undergoer. An explanation for the absence of an actor is found in the

general LS for condition statives (cf. (27a)) where the single argument is a patient which is linked

to undergoer. That is, condition stative verbs can only have an effector if it occurs as an optional

adjunct; under no circumstance can the effector be linked to actor or become the clause pivot. The

key point here is that the GENERAL MACROROLE ASSIGNMENT PRINCIPLES in (10) apply

to verbs and not clauses. Thus, for example, the principles in (10) apply to (27a) and not (27b).

According to (10a.2), the verb has one macrorole since there is only one argument 'x' in the LS in

(27a). If the principles in (10) were to apply to (27b), one might be misled to conclude that there

are two macroroles since there are two semantic arguments 'x' and 'y'. The application of principle

(10b.2) to (27a) results in the single macrorole being an undergoer (cf. (27d)). The absence of an

available actor explains why the antecedent cause cannot be linked to an actor (cf. above).

One function of logical structures is to show the relationship between a predicate and its

arguments.40 In RRG the semantic representation of verbs is accounted for in terms of LSs, which

are the main portion of the lexical entry for a verb (cf. W 1993:39, 43). Thus, we can talk about

the LS of clauses (which contain a predicate, its arguments, and adjuncts), or the LS of verbs

(which are the most common type of predicate). In many cases the LS of the clause is equivalent

to the LS of the verb of that clause. However, this is not necessarily so. For instance, optional

adjuncts (nonarguments) such as time, manner, and in some instances causality, are not associated

with the LS of verbs, but can occur as part of the LS of clauses. For example, antecedent cause is




40Or, stated another way, one function of LSs is to subcategorize verbs (cf. the introduction to this
chapter).









not associated with the LS of the stative verb ngkorikng 'ST-dry' (e.g. (23b)), but it can occur as

part of the LS of clauses (e.g. (27b)). The question arises as to how much of the semantic content

of clauses should be accounted for in descriptions of LS. This study primarily attends to those

aspects of meaning which are important for classifying verbs, since the hypothesis put forth in

Chapter 1 is that verb classification constrains aspectual possibilities.

The stative verbs described thus far in this section are one-place condition statives whose

single argument is a patient. There is a small class of stative verbs which have two semantic

arguments experiencer and theme) and one syntactic argument. They are referred to as

experience statives and are illustrated in (28).41

(28) Rimig-adn (ramig-an) ou na.
cold-EXP IsNOM PFT
'I am cold.'
A summary analysis of (28) is provided in (29). As seen in (29a), experience statives have

two semantic arguments. The first argument is an experience, the second argument (theme) is

incorporated into the predicate by the same rule which was described earlier in (6) and repeated

here as (30) (cf. Van Valin 1990:244). This rule incorporates the theme into the predicate in both

equational clauses (cf. 2.1.1) and experience stative clauses.

(29) a. General LS for experience statives: be' (x, [predicate']) x=experiencer,
predicate'-theme
b. Clause-specific LS for (28): be' (ou T, [ramig 'cold'])
c. Assign thematic relations: ou T *- experience
d. Create experience predicate by (30): be' + [ramig 'cold'] -- predicate
e. Assign semantic macroroles: undergoer +- experience 1 Macrorole
f. Assign syntactic functions: pivot <- undergoer-ou T
g. Assign case and prepositions: pivot-ou T'I' <- nominative case
h. Cross-reference verb: ramig 'cold' <- -an 'EXP'
(30) EQUATIONAL/ EXPERIENCE PREDICATE CREATION: (cf. Van Valin 1990:234)

be' + theme predicate


41Whereas patients are either animate or inanimate, an experience is defined as "An animate entity
whose registering nervous system is relevant to the predication" (Longacre 1976:27).









According to the default situation described in (10a. 1), experience statives should have two

macroroles since they have two arguments in their LS. However, the incorporation of one of the

two arguments into the predicate requires the default situation to be overridden and the verb to take

one less macrorole than is expected. In RRG, when the number of macroroles cannot be predicted

from the number of arguments in the LS, the number of macroroles is specified in the lexical entry

of the verb. This is formalized in terms of the feature [MR], with [+MR] meaning one macrorole

and [-MR] meaning no macroroles (cf. Van Valin 1990:227).42 Use of this feature would result in

the following partial lexical entry for the experience stative verb rimigadn 'cold': be' (x, [ramig

'cold']) [+MR]. One way to avoid specifying the number of macroroles in the LS of all experience

stative verbs is simply to say that the application of rule (30) results in the default situation in

(10a. 1) being overridden and only one macrorole. Note that although the default in (10 Oa. 1) is

overridden, the more general principle in (I a) is not violated since the number of macroroles is

less than the number of arguments in the LS.

Given that there is only one macrorole, according to (10b.2), it must be an undergoer. Thus,

like the condition statives described earlier in this section, experience statives have a single

macrorole which is an undergoer.

According to the Actor-Undergoer Hierarchy in (4), the theme argument should be the

undergoer. But the theme argument is incorporated into the predicate by rule (30) in (29d), so

there is no theme to function as undergoer; instead, the experience argument is linked to undergoer

(cf. (29e); Van Valin 1990:244). This is a marked linking between a thematic relation and a

macrorole in terms of the Actor-Undergoer Hierarchy in (4). Experience stative verbs are cross-

referenced by the suffix -an instead of the prefix m- (cf. (29h).43 Experiencers are not only




42Van Valin (1990:227) points out that it is never necessary to indicate that a verb takes two
macroroles.









semantically marked undergoers according to the Actor-Undergoer Hierarchy, but they also have a

corresponding markedness in terms of both frequency and morphology with -an being the marked

form.

Experience statives are similar to equational statives (cf. 2.1.1) in that both are two-place

statives and both involve incorporation of the theme (second argument) into the predicate, resulting

in a functional marked undergoer. However, experience statives differ from equational statives in

that the former are also formally marked in the verb morphology by -an, whereas the latter exhibit

no verbal cross-referencing since they are nonverbal.

As shown in Table 2.6, experience stative verbs are derived from a limited number of

adjective roots and noun roots. Table 2.6 also indicates that some adjectives can occur as either

condition stative verbs or experience stative verbs.44 For example, ramig 'cold' can occur as a

condition stative or an experience stative. The condition stative (meramig) focuses on the

stimulus and is used to describe something as being 'cold to touch', whereas the experience stative







43The morphophonemic processes involved here are quite complex (cf. Table 2.6 for other
examples). They include vowel harmony, nasal harmony, and nasal preplosion. Vowel harmony
operates in terms of the effects of root vowels on affixes. High vowels spread right-to-left as in
(28) where the second vowel i/ in the root /ramig/ replaces the first vowel /a/ when the suffix -an
occurs. See Kroeger (1992b) for further discussion of VH in Bonggi.

Nasal harmony operates from left-to-right. Vowels are nasalized following nasal stops. Nasality
spreads until it is blocked by a nonnasal consonant or the end of the word.

Word-final nasals are preploded following oral vowels. Final /m/ becomes [bm], final I/n/ becomes
[dn], and final /g/ becomes [kg]. For example, the final nasal of the underlying form I/m-korij/
'ST-dry' in (28) is preploded, resulting in the phonetic form [tj'korikg] because the preceding vowel
is oral. On the other hand, the final nasal of/mien/ ['nien] 'aunt' is not preploded since the
preceding vowel is nasal. See Boutin and Howery (1991) for further discussion of nasal preplosion
in Bonggi.

4The experience stative lupug-udn 'tired-EXP' is anomalous in two respects. Firstly, it is marked
by [-udn] instead of [-adn] like other experience statives. Secondly, lupug 'tired' is the only
adjective root that cannot occur as a condition stative; i.e. *mu-lupug 'ST-tired' is ungrammatical.
This root only takes animate pivots.









(rimigadn) focuses on the experience and is used to describe someone as 'feeling cold' (cf. Talmy

1985:99ff.).



Table 2.6: Experiencer stative verbs


ROOT CONDITION STATIVE EXPERIENCE STATIVE

Adjective ramig 'cold' me-ramig 'ST-cold' rimig-adn 'cold-EXP'

panas 'hot' m-panas 'ST-hot' penas-adn 'hot-EXP'

lupug 'tired' lupug-udn 'tired-EXP'

Noun umus 'sweat' umus-adn 'sweat-EXP'

puri' 'skin rash' puri-adn 'skin.rash-EXP'


To summarize what has been said regarding condition stative verbs: (i) their LS is predicate'

(x); (ii) the single argument 'x' is a patient; (iii) they have one macrorole which is an undergoer;

and (iv) these verbs are normally marked with m- which cross-references an undergoer pivot (e.g.

(21), (24); but cf. (25)).

To summarize what has been said regarding experience stative verbs: (i) their LS is be' (x,

[predicate']); (ii) the first argument is an experience and the second a theme which is incorporated

into the predicate; (iii) they have one macrorole which is an undergoer; and (iv) these verbs are

marked with -an which cross-references a marked undergoer pivot (e.g. (28); cf. also Table 2.6).

It was pointed out that condition stative clauses can have an optional adjunct which is an

antecedent cause effectorr) as seen in (26). These adjuncts are marked for oblique case with ga'

'from' (cf. 27f). Optional adjuncts can also occur in experience stative clauses as seen in (31). As

pointed out previously, the preposition ga' 'from' marks effectors which would be expected to occur

as actor in terms of the hierarchy in (4), but do not because there is no actor to which the

antecedent cause can be linked. Neither condition stative clauses nor experience stative clauses









have an actor, but for different reasons. In the case of condition statives (e.g. (26)), the absence of

an actor is due to the fact that the LS of these verbs has only one argument and no activity

predicate (cf. 27a). Thus, by (10), there must be a single macrorole which is an undergoer. On the

other hand, in the case of experience statives (e.g. (31)), the absence of an actor is due to the fact

that rule (30) applies which results in the default situation in (10a. 1) being overridden and only one

macrorole, an undergoer by (10b.2).

A summary analysis of(31) is provided in (32) (cf. (27) and (29)). Sentence (31) describes

an antecedent cause (ga' dolok 'from rain') and a resulting/consequent experience state (rimigadn

ou 'I am cold').

(31) Rimig-adn (ramig-an) ou ga' dolok.
cold-EXP IsNOM from rain
'I am cold because of the rain.'

(32) a. General LS for experience statives: be' (x, [predicate']) x=experiencer,
predicate'-theme
b. Clause-specific LS for (31): [from' (x, [be' (y, [predicate'])]
x=antecedent cause,
be' (y, [predicate'])=resulting state
y=experiencer, predicate'--theme
[ga' 'from' (dolok 'rain'), [be'
(ou T, [ramig 'cold'])]
c. Assign thematic relations: ou 'I' -- experience
dolok 'rain' <- effector (antecedent cause)
d. Create experience predicate by (30): be' + [ramig 'cold'] --) predicate
e. Assign semantic macroroles: undergoer <- experience 1 Macrorole
f. Assign syntactic functions: pivot <- undergoer-ou T
oblique adjunct <- effector (antecedent cause)-
dolok 'rain'
g. Assign case and prepositions: pivot-ou T'I' *- nominative case
effector (antecedent cause)-
dolok 'rain' <- oblique ga' 'from'
h. Cross-reference verb: ramig 'cold' <- -an 'EXP'
As mentioned above, experience statives should have two macroroles since they have two

arguments in their LS (cf. (10a. 1)). However, the incorporation of the theme argument into the

predicate by rule (30) in (32d) results in the default situation being overridden and the verb taking

only one macrorole, an undergoer by (10b.2). Because the theme argument is incorporated into the

predicate, it is not available to function as undergoer; instead, the experience argument is linked to

undergoer in (32e) (cf. (29e)). This marked linking between experience and undergoer results in









the experience stative verb in (31) being cross-referenced by the suffix -an (cf. (32h)). (32f)

assigns the undergoer (ou T) to the syntactic status of clause pivot according to the Pivot Choice

Hierarchy in (7).

Two types of stative verbs have been described thus far in this section: one-place condition

statives (e.g. (21) and (26)) and two-place experience statives (e.g. (28) and (31)).45 Some of the

one-place condition stative verbs in (24) can occur in clauses involving what has been called

possessor ascension in Relational Grammar. The claim within Relational Grammar is that the

possessor of a noun phrase (e.g. nya 'his' in (33)) can ascend to become a clause constituent (e.g.

sia 'he' in (34)). Sentence (33) illustrates a condition stative verb without possessor ascension,

whereas (34) illustrates the same condition stative verb in a clause with possessor ascension.

(33) N-doot (m-doot) guakng nya.
ST-bad spirit 3sGEN
'HE is unhappy.' (Lit. 'His spirit is bad.')

(34) Sia n-doot (m-doot) guakng.
3sNOM ST-bad spirit
'HE is unhappy.'
Summary analyses of (33) and (34) are provided in (35) and (36) respectively.46 In a RRG

analysis, (33) and (34) share the same general LS, predicate' (x), since they are both condition

statives (cf. (35a) and (36a)); however, their clause-specific LS is different (cf. (35b) and (36b)).

Because the macrorole assignment principles in (10) apply to verbs and not clauses, both (33) and

(34) contain a single macrorole which is undergoer. According to Van Valin (1990:251),

"[P]ossessor ascension is possible only if the possessed argument is an undergoer."










45Statives with an optional antecedent cause are included, e.g. (26) and (31).
46Cf. the analysis in (35) to that provided in (23).






62

(35) a. General LS for condition statives: predicate' (x) x=patient
b. Clause-specific LS for (33): doot 'bad' (guakng nya 'his spirit')
c. Assign thematic relations: guakng nya 'his spirit' <- patient
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-guakng nya 'his spirit'
f. Assign case and prepositions: pivot-guakng nya 'his spirit' <- nominative case (0)
g. Cross-reference verb: doot 'bad' <- m- 'ST'

(36) a. General LS for condition statives: predicate' (x) x=patient
b. Clause-specific LS for (34): be-at' (x, [predicate' (y)]) x=locative,
predicate' (y)=condition
state
y=patient
be-at' (guakng 'spirit', [doot 'bad' (sia 'he')])
c. Assign thematic relations: sia 'he' <- patient
guakng 'spirit' <- locative
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot *- undergoer-sia 'he'
direct core syntactic argument <- locative-guakng
'spirit'
f. Assign case and prepositions: pivot-sia hie' <- nominative case
guakng 'spirit' <- accusative case (0)
g. Cross-reference verb: doot 'bad' <- m- 'ST
Sentence (34) contains a locative adjunct, whereas (33) does not. (34) describes a state (sia

ndoot 'he is bad') and the location of that state (guakng 'spirit'). This is seen in (36b) where 'x'

refers to the location and predicate' (y) to a condition state. This provisional LS is constructed on

the analogy of be-at' (x, y), where x = the location and y = the located argument.47 As seen in

(36a), the general LS for one-place condition statives does not contain a location; however, (36b)

indicates that there is a location in the clause in question. Two clause-specific LSs are provided in

(36b). The latter is the same as the former with the exception that the Bonggi data are provided in

the latter. As always, the single argument of a condition stative verb is a patient which is linked to

undergoer in (36d).

The locative adjunct is assigned core syntactic status in (36e). The number of direct core

syntactic arguments need not be the same as the number of macroroles. Condition stative verbs

have a single macrorole. Therefore, they are intransitive in RRG terms, where transitivity is





47The LS in (36b) is provisional in that LSs are normally used to account for the relationship
between a predicate and its arguments, not adjuncts.









defined in terms of the number of macroroles a verb takes. Single-macrorole verbs are intransitive

and two-macrorole verbs are transitive (cf. Van Valin 1990:228). "The number of direct core

arguments a verb takes says less about its syntactic behavior than its macrorole number" (Van

Valin 1990:228).

Earlier in this section, a distinction was made between semantic arguments and syntactic

arguments. Recall that direct core arguments (core syntactic arguments) were defined in 1.3.1 as

arguments which are not preceded by a preposition (oblique case marker); for example, sia 'he' and

guakng 'spirit' in (34) are core syntactic arguments.48 On the other hand, odu 'sun' in (26) and

dolok 'rain' in (31) are syntactically oblique since they are preceded by the preposition ga' 'from'.

Other than the clause pivot, the following can occur as direct core syntactic arguments in Bonggi:

(i) semantic arguments with macrorole status (e.g. the undergoer diaadn 'me' in (19) of 1.3.1); (ii)

some semantic arguments without macrorole status (e.g. the locative argument bali nya 'his house'

in (15) (cf. (18f)); and (iii) some semantic adjuncts which do not belong to the LS of the verb (e.g.

guakng 'spirit' in (34)).

The distinction between semantic arguments and syntactic arguments can also be described in

terms of valency. Recall that experience statives have two semantic arguments experiencer and

theme) and one syntactic argument (the pivot). Thus, they have a semantic valency of two (e.g.

(29a)), but a syntactic valency of one (e.g. (28)). This difference in valency results from the

incorporation of the theme argument into the predicate by rule (30). On the other hand, condition

statives have a semantic valency of one (e.g. (35a)), but can have a syntactic valency of either one

(e.g. (33)) or two (e.g. (34)).49 Bonggi verb morphology is more concerned with semantic valency

than syntactic valency.




48As stated earlier, arguments in the LS of a verb provide a definition of core semantic arguments.

49Since syntactic arguments can be elided, both experience statives and condition statives can
have a syntactic valency of zero. The discussion here of syntactic valency ignores the possibility of
ellipsis.









The condition stative verbs which have been described thus far in this section (e.g. (21), (26),

(33), (34)) and those listed in (24) are derived from adjective roots. In fact, the majority of

condition stative verbs are derived from adjective roots. However, in a few cases, condition stative

verbs appear to be derived from noun and verb roots. One such case appears to be the condition

stative verb me-lou 'ST-embarrassed' in (37) which is derived from the noun root lou

'embarrassment'.

(37) Sia me-lou (m-lou).
3sNOM ST-embarrassed
'HE is embarrassed.'
A summary analysis of (37) is provided in (38) (cf. (23), (35)).

(38) a. General LS for condition statives: predicate' (x)
b. Clause-specific LS for (37): lou 'embarrassed' (sia 'he')
c. Assign thematic relations: sia 'he' +- patient
d. Assign semantic macroroles: undergoer +- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sia 'he'
f. Assign case and prepositions: pivot-ou T <- nominative case
g. Cross-reference verb: lou 'embarrassed' <- m- 'ST'
Like other condition stative verbs, me-lou 'ST-embarrassed' can occur with an optional

adjunct indicating antecedent cause as in (39) (cf. (26)). A summary analysis of (39) is provided

in (40) (cf. (27)).

(39) Sia me-lou (m-lou) ga' ku.
3sNOM ST-embarrassed from IsGEN
'HE is embarrassed because of me.'

(40) a. General LS for condition statives: predicate' (x) x=patient
b. Clause-specific LS for (39): from' (x, [predicate' (y)]) x=antecedent cause,
predicate' (y)=resulting
state,
y=patient
ga' 'from' (ku 'me', [lou 'embarrassed' (sia 'he')])
c. Assign thematic relations: sia 'he' <- patient
ku 'me' <- effector (antecedent cause)
d. Assign semantic macroroles: undergoer +- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sia 'he'
oblique adjunct <- effector (antecedent cause)-ku 'me'
f. Assign case and prepositions: pivot-sia 'he' *- nominative case
effector (antecedent cause)-ku 'me' <- genitive case
oblique ga' 'from'
g. Cross-reference verb: lou 'embarrassed' <- m- 'ST









Sentence (39) contains an antecedent cause (ga' ku 'because of me') and a resulting/consequent

state which consists of a condition stative verb (melou 'is embarrassed') and a patient (sia 'he').

Condition stative verbs do not have an antecedent cause for an argument (cf. (40a)). By (10a.2),

the verb has one macrorole since there is only one argument 'x' in the LS in (40a). Application of

(10b.2) to (40a) results in the single macrorole being an undergoer (cf. (40d)). The antecedent

cause cannot be linked to an actor because there is no actor. Thus, the antecedent cause, a subtype

of effector, is assigned oblique adjunct status (cf. (40d)), and marked by the preposition ga' 'from'

(cf. (40f)).

The condition stative verb me-lou 'ST-embarrassed' can also occur in clauses with two core

syntactic arguments (e.g. (41)). In (41), the nonpivot core argument, diaadn 'me' is semantically a

stimulus which is both similar to and different from the antecedent cause in (39). (39) and (41) are

similar in that both involve indirect causality. They are different in terms of the strength of the

causal relationship with antecedent cause being stronger than stimulus.50 Antecedent cause and

stimulus cannot cooccur in the same clause; that is, if a clause contains an optional adjunct

indicating antecedent cause (e.g. ga' ku 'because of me' in (39)), a nonpivot core syntactic

argument such as diaadn 'me' cannot also occur in the same clause.

(41) Sia me-lou (m-lou) diaadn.
3sNOM ST-embarrassed IsACC
'HE is embarrassed of me.'
Indirect causality (including antecedent cause and stimulus) is not associated with the LS of

verbs. It is not a semantic argument of a verb, although it can be incorporated into the LS of

clauses as in (27b), (32b), and (40b). Indirect causality occurs in syntax as either an optional









50Although the free translation for (39) and (41) attempts to capture the difference in the strength
of the causal relationship, Bonggi more clearly marks this difference than English.









adjunct (e.g. (26), (31), (39)) or a direct core syntactic argument (e.g. (41)). The difference in

syntactic form reflects a difference in semantic function.51

The condition stative verb me-lou 'ST-embarrassed' can occur in one other syntactic

environment; that is, sentences which contain a proposed phrase that is set off from the clause by a

pause (e.g. (42)).52

(42) Diaadn, nd-ou me-lou (m-lou).53
IsACC not-IsNOM ST-embarrassed
'As for me, I'm not embarrassed.'
The morphological behavior of the condition stative verb me-lou 'ST-embarrassed' in (37),

(39), (41), and (42) is the same as that of condition stative verbs derived from adjective roots (cf.

(24)). In each case, an undergoer-patient is cross-referenced by the verbal prefix m-. On the other

hand, their syntactic behavior is different. To begin with, me-lou 'ST-embarrassed' has a syntactic

valency of either one (e.g. (37), (39), (42)) or two (e.g. (41)), whereas most condition stative verbs

derived from adjective roots have a syntactic valency of one (e.g. (21), (26)). Except for a small

group of condition stative verbs which can occur in clauses involving possessor ascension (e.g.

(34)), the addition of another core syntactic argument in clauses with condition stative verbs

derived from adjective roots results in an ungrammatical sentence. Both melou 'embarrassed' and

condition statives derived from adjective roots can take activity verb complements (e.g. (43), (44)).

(43) Sia me-lou (m-lou) mpanu.
3sNOM ST-embarrassed walk
'HE is embarrassed to go.'



51Direct causality, like indirect causality, also has contrastive syntactic forms representing
different semantic functions (cf. 2.4).
52Note that (42) does not involve extraposition; diaadn 'me' is clause external.








(44) Sia mu-runggu' (m-runggu') mpanu.
3sNOM ST-sluggish walk
'HE is sluggish walking.'
In terms of derivational potential, me-lou 'ST-embarrassed' is distinguished from condition

statives derived from adjective roots. Activity verbs which are negated and in imperative mood can

be derived from the former (e.g. (45)), but usually not the latter.54 Condition stative verbs that are

derived from manner adjectives used to describe humans are an exception with respect to negative

imperatives. Thus, from (24), n-doot 'ST-bad', n-duruk 'ST-fast', me-lambat 'ST-slow', and mu-

runggu "ST-sluggish' can be negated (e.g. (46)).

(45) Dei ke-lou (k-lou)!
do.not embarrassed
'Don't be embarrassed!'

(46) Dei ku-runggu' (k-runggu') !
do.not sluggish
'Don't be sluggish!'
M-olok 'ST-scared' is another condition stative verb which is derived from a noun root (olok

'fear') and has the same morphosyntactic behavior as me-lou 'ST-embarrassed'. It can occur in

clauses with a syntactic valency of one (e.g. Sia molok. 'He is scared.'; cf. (37)). It can occur in

clauses with an optional adjunct indicating antecedent cause (e.g. Sia molok ga' ku. 'He is scared

because of me.'; cf. (39)). It can occur in clauses with a syntactic valency of two (e.g. Sia molok

diaadn. 'He is scared of me.'; cf. (41)). It can occur in clauses with a proposed phrase (e.g.








53Negated pronouns are restricted to first and second person nominative case. Phonologically,
negated pronouns are a single word. Syntactically, they function as the pivot and carry case.
Morphologically, they are odd in that pronouns are not normally inflected for negation; however,
nd- is an affix and not a simple clitic (3.2). The negation of pronouns is somewhat comparable to
the English contracted negative n't which Zwicky and Pullum (1983) show has all the signs of
being an inflectional affix and none of the properties of a clitic. Botolan Sambal, a language of the
northern Philippines (Luzon Island), also has a morphological negative which occurs with enclitic
pronouns (cf. Antworth 1979:50f.).

54Cf. 2.3 for a discussion of activity verbs.









Diaadn, ndou molok. 'As for me, I'm not scared.'; cf. (42)). Finally, it can occur in clauses with

activity verb complements (e.g. Sia molok mpanu. 'He is scared to go.'; cf. (43)).

I return to problems posed by stative verbs such as me-lou 'ST-embarrassed' and m-olok 'ST-

scared', after discussing possession and cognition statives in the next two sections.


2.1.4. Possession States

Sentence (47) illustrates an English possessive stative verb. The general logical structure for

possessive stative verbs is shown in (48a) and the clause-specific LS for (47) in (48b).

(47) He has a book.

(48) a. General LS for possession statives: have' (x,y) x=possessor, y=possessed
b. Clause-specific LS for (47): have' (he, book)
According to Foley and Van Valin (1984:48, 53) and Van Valin (1993:39), the first argument

'x' in the LS of a possession stative clause is a locative, while the second argument 'y' is a theme.

More recently, Van Valin (1994) uses the terms possessor and possessed to refer to these two

arguments with possessor = locative and possessed = theme. Whereas possessor and possessed are

semantically more transparent names, locative and theme are more commonly used as names for

thematic roles, especially if one is trying to keep the number of thematic roles to a minimum. The

number of thematic role names used to describe semantic relationships is not a problem according

to recent developments within RRG. These names are nothing more than mnemonics for argument

positions in LS (Van Valin, 1994). As pointed out in 2.1.1, the revised Actor-Undergoer

Hierarchy in (5) is based on the RRG position that thematic roles are not independently motivated.

As seen thus far in this chapter, thematic roles play no direct role in lexical representation. The

semantic properties of verbs are expressed by LS representations, not a list of thematic roles (cf.

Van Valin 1994). Throughout the remainder of this section, I use the transparent terms possessor

and possessed to refer to the two arguments in possessive states. At the end of 2.1.5 I return to

the relationship between these terms and argument positions in LSs.









The prefixes ki- and nd- mark verbs of possession. Id- marks the presence of possession, e.g.

(49), and nd- marks the absence of possession, e.g. (50).55 Both ki- and nd- occur with the

existential root ara, but only ki- occurs with possession stative verbs which are derived from noun

roots, e.g. (51).56

(49) Sia ki-ara oig.
3sNOM POS-exist boat
'HE has a boat.'

(50) Sia nd-ara pa saa.
3sNOM NEG-exist FPFT spouse
'HE is not married yet.'

(51) Sia ki-saa.
3sNOM POS-spouse
HE is married.'

A summary analysis of (49) is provided in (52).


(52) a. General LS for possession statives:
b. Clause-specific LS for (49):
c. Assign thematic relations:

d. Assign semantic macroroles:
e. Assign syntactic functions:


f. Assign case and prepositions:

g. Cross-reference verb:


have' (x,y) x=possessor, y=possessed
ara 'existential' (sia 'he', oig 'boat')
sia 'he' <- possessor
oig 'boat' +- possessed
undergoer -- possessor 1 Macrorole
pivot <- undergoer-sia 'he'
direct core syntactic argument <- possessed-oig
'boat'
pivot-sia 'he' <- nominative case
core argument-oig 'boat' +- accusative case (0)
ara 'existential' -- ki- 'POS'


The analysis presented in (52) is straightforward with the exception of (52d). By (10a. 1),

possessive statives should have two macroroles since they have two arguments in their LS (cf.

(52a)); however, according to (52d), Bonggi possessive statives have a single macrorole. A similar




55The prefix nd- comes from reduction of the negator nda' 'not'. Historically, nda' probably
developed into a critic nd=, which, in turn, was reanalyzed as a prefix. This is a case of
morphologization (cf. Anderson 1988b:352; Joseph and Janda 1988:195-96). The negator nda'
also reduces to nd- before first and second person nominative case pronouns, e.g. ou T, ndou 'not
I', aha 'you (sg)', and ndaha 'not you (sg)'.

56Possession stative verbs are derived from existential ara and noun roots. However, the use of ki-
to derive possession stative verbs from noun roots is not extremely productive. More frequently,
kiara occurs as in (49). The distinction between the use of ki- and kiara appears to correlate with
inalienable versus alienable possession.









situation was described for experience statives in 2.1.3, where according to the default situation

described in (10a. 1), experience statives should have two macroroles. Instead, I suggested that the

incorporation of one argument into the predicate requires the default situation to be overridden, and

results in experience statives having only one macrorole.57 There are two reasons why

incorporation was suggested as an explanation for experience verbs taking one macrorole instead

of the expected two by (10a. 1). First, incorporation provides a principled basis for overriding the

GENERAL MACROROLE ASSIGNMENT PRINCIPLES in (10). Second, incorporation

provides an explanation for a single macrorole without having to specify the feature [+MR] in the

lexical entry of the verb. Although incorporation might be a good explanation for experience

statives having only one macrorole, it cannot be used as an explanation for possessive statives

since they do not involve incorporation.

What sort of evidence is needed to show that a verb with two core syntactic arguments has

one macrorole instead of two? Van Valin (1990:240-48) uses case-marking as evidence in

Georgian. He provides the two universal case-marking principles in (53) (Van Valin 1990:241).

(53) a. If a clause contains a single macrorole argument, it is nominative.
b. The default case for direct core arguments which are not assigned macrorole status is
dative.
In Bonggi, case-marking does not provide sufficient evidence to distinguish verbs with two

core syntactic arguments and one macrorole from those with two core syntactic arguments and two

macroroles. This is due to the fact that distinctions between accusative and dative arguments are

neutralized in Bonggi, especially within core syntactic arguments (cf. Tables 1.2 and 1.3 in

1.3.1).58 Bonggi pronominal pivots occur in the nominative case, but there is no 'inversion' of

case like that described for Georgian (Van Valin 1990; 1993) and Latin (Michaelis 1993). On the





57Cf. the discussion following the analysis in (29), 2.1.3.
58This was stated slightly differently at the end of 1.3.1, where I pointed out that Bonggi does not
have unique dative case forms; instead, all nonpivot direct core arguments which are not actors
occur in accusative case.









other hand, as with inversion in Georgian, possessive statives in Bonggi are intransitive; i.e. they

have a single macrorole argument.59

Although the Bonggi case-marking system does not distinguish verbs with a single macrorole

from those with two macroroles, the syntax does. In terms of syntax, if an argument has macrorole

status, it must be able to function as clause pivot. Or stated another way, an argument cannot have

macrorole status if it cannot function as clause pivot. This is not meant to suggest that the clause

pivot must always be a macrorole; in fact, I show in 2.4 that this is not always the case. That is,

an argument can be the clause pivot without being linked to a macrorole, but an argument cannot

be linked to a macrorole unless it can also be the clause pivot. The macrorole assignment

principles in (10a) and (10b), repeated here as (54a) and (54b), are supplemented by the Bonggi-

specific principle in (54c).60

(54) GENERAL MACROROLE ASSIGNMENT PRINCIPLES:
a. Number: the number of macroroles a verb takes is less than or equal to the number of
arguments in its LS.
1. If a verb has two or more arguments in its LS, it will take two macroroles.
2. If a verb has one argument in its LS, it will take one macrorole.
b. Nature: for verbs which take one macrorole,
1. If the verb has an activity predicate in its LS, the macrorole is actor.
2. If the verb has no activity predicate in its LS, the macrorole is undergoer.
c. Pivot potential: if a verb has two macroroles, either of them can function as the clause
pivot.
By (54c), possessive statives only have one macrorole since only the possessor can be the

clause pivot, never the possessed item. By (54b.2), the single macrorole is an undergoer (cf.

(52d)). The pivot in (49), (50), and (51) is sia 'he', which is an animate possessor. The possessor

can also be inanimate, e.g. (55) and (56), where the possessor is oig na 'the boat'.






59Recall that in RRG, transitivity is defined in terms of the number of macroroles a verb takes.
Single-macrorole verbs are intransitive, two-macrorole verbs are transitive. The number of
macroroles a verb takes is more important than the number of direct core syntactic arguments it
takes (cf. Van Valin 1990:228, 1993:48).

60Van Valin (1993:48; 154, footnote 28) points out that the syntactic consequence of core
arguments in syntax not being assigned to a macrorole is that they cannot occur as pivot.








(55) Oig na ki-ara saat.
boat DEF POS-exist paint
THE BOAT has been painted.'

(56) Oig na nd-ara saat.
boat DEF NEG-have paint
THE BOAT has no paint.'
Since only the possessor can be the pivot in possession statives, and since pivots precede

nonpivots, the possessor must occur before the possessed entity. Sentence (57) is semantically ill-

formed if'paint' is interpreted as the pivot, since paint cannot possess a boat. However, (57) is

syntactically ill-formed if 'boat' is interpreted as the pivot, since the pivot must precede the

nonpivot.

(57) *Saat na nd-ara oig.
paint DEF NEG-have boat
*"THE PAINT has no boat.'


2.1.5. Cognition States

Cognition states are mental, internal, and nonvolitional, involving neither decision nor action

on the part of the undergoer. The construction has two arguments as illustrated by the English

cognition stative verb believe in (58). The LS for (58) is shown in (59). The first argument 'x' is

an experience, while the second argument 'y' is a theme (VV 1993:39).61

(58) He believes me.

(59) a. General LS for believe: believe' (x,y) x=experiencer, y=theme
b. Clause-specific LS for (58): believe' (he,me)
Bonggi has only two cognition stative verbs, pisiaa 'believe' and pandi 'know' which are

illustrated in (60) and (61). A summary analysis of (60) is provided in (62).

(60) Sia pisiaa diaadn.
3sNOM believe IsACC
'HE believes me.'


61Giv6n (1984:21) points out that thematic roles in possessive clauses have atypical values.








(61) Sia pandi uubm Sama.
3sNOM know language Bajau
I-HE knows the Bajau language.'

(62) a. General LS for believe: believe' (x,y) x=experiencer, y=theme
b. Clause-specific LS for (60): pisiaa 'believe' (sia 'he', diaadn 'me')
c. Assign thematic relations: sia 'he' +- experience
diaadn 'me' <- theme
d. Assign semantic macroroles: undergoer <- experience 1 Macrorole
e. Assign syntactic functions: pivot +- undergoer-sia 'he'
direct core syntactic argument +- theme-diaadn 'me'
f. Assign case and prepositions: pivot-sia 'he' <- nominative case
core argument-diaadn 'me' -- accusative case
g. Cross-reference verb: pisiaa 'believe' <- (0)

Cognition statives, like the possessive statives described in 2.1.4, have two core syntactic

arguments, but only one macrorole. Only the experience can be the pivot in cognition stative

clauses. Thus, by (54c) these statives have only one macrorole which, by (54b.2), must be an

undergoer (cf. (62d)). Cognition stative verbs are morphologically unmarked; that is, the root is

not cross-referenced (cf. (62g)).

Table 2.7 provides a summary of the types of states which have been described. They are:

equational (2.1.1), locative (2.1.2), condition (2.1.3), experience (2.1.3), possession (2.1.4),

and cognition (2.1.5). The second row contains the general LS for each subclass of state. The

third row lists the thematic relations (or mnemonic labels) which have been used to describe the

arguments in the LSs. The fourth row shows the number and nature of the macroroles occurring

with each subclass of state, and it summarizes the linking relationship between macroroles and the

thematic relations in row three. The fifth row lists the choices for clause pivot, the sixth lists the

syntactic category of the clause predicate, and the last row summarizes the verbal cross-

referencing.

Possession and cognition stative verbs are small closed classes, whereas condition stative

verbs form a larger, more open class. Experience and locative statives fall somewhere in between









in terms of class size. All stative verbs are uninflected for tense, and cannot occur in imperative

mood.62


Table 2.7: Summary of states


Eauational


Locative


Condition


p q


Exoeriencer


Possession


Cognition


LS be' (x,y) be-at' (x,y) predicate' be' (x, have' (x,y) believe' (x,y)
(x) [predicate'])

G-role x=locative x=locative x=patient x=experiencer x=locative x-experiencer
y=theme y=theme predicate'= y=theme y=theme
theme

macroroles undergoer undergoer undergoer undergoer= undergoer undergoer=
locativee -theme =patient experience locativee experience

pivot undergoer undergoer undergoer undergoer undergoer undergoer

syntactic nominal verb, verb verb verb verb
category deictic
of adverb,
predicate locative
preposition

verbal m- (for m- -an ki-, nd-
cross- verbal
referencing predicates
only)


The ability of different NPs to function as clause pivot is a well-known feature of Philippine-

type languages. Many introductions to languages like Tagalog suggest that any NP in a clause can

be the pivot.63 This is quite contrary to what is described here, where each stative subclass




62The cognition stative verb pisiaa can also be used as an imperative meaning 'believe!', in which
case it is an activity verb and not a state. In fact, there is a corresponding activity verb form -um-
+ pisiaa > mpisiaa (cf. 2.3). The root pandi 'know', however, cannot be used as an imperative;
thus, there is no corresponding activity form.

63Usually some term other than pivot is used; e.g. focused NP, subject, topic, or trigger.









strongly constrains the pivot possibilities. In fact, as seen in Table 2.7, each subclass of state has

only one pivot possibility.

I return now to the question of the relationship between 0-roles and LS argument positions

that was raised in 2.1.1 and 2.1.4. According to the revised Actor-Undergoer Hierarchy in (5),

0-roles are nothing more than mnemonics for argument positions. The relationship between 0-roles

and argument positions for the different subclasses of states is summarized in (63) (cf. (5); Van

Valin 1994).

(63) <
1 st arg of 2nd arg of Arg of state
predicate'(x,y) predicate'(x,y) predicate'(x)

Equational state: locative theme (cf. 2.1.1)
Locative state: locative theme (cf. 2.1.2)
Condition state: patient (cf. 2.1.3)
Experience state: experience theme (cf. 2.1.3)
Possession state: locative theme (cf. 2.1.4)
Cognition state: experience theme (cf. 2.1.5)
Existential state: locative theme (cf. below)
Emotional state experience theme (cf. below)
The LS for all states can be generalized as either predicate' (x) or predicate' (x,y) (VV

1993:36). Variations among stative subclasses result from variations within these two LSs. For

two-place statives, the first argument is either an experience or a locative, whereas the second

argument is a theme (cf. Van Valin 1990:226). Thus, thematic roles can be defined in terms of

argument positions. For example, patient is the argument in the LS configuration predicate' (x),

and experience is the first argument in either: (i) an experience state LS configuration be'

(x,[predicate']), (ii) a cognition state LS configuration believe' (x,y), or (iii) an emotional state LS

configuration love' (x,y).









If 0-roles are nothing more than mnemonics for argument positions, what is the function of

verbal cross-referencing in Bonggi and other Philippine-type languages?4 Recall that within

Philippine linguistics, the role of verbal cross-referencing is usually thought to encode the thematic

role of the pivot (cf. 1.3.1). If, as is usually claimed, verbal cross-referencing encodes 0-roles,

then 0-roles are certainly more than just mnemonics. To answer this question, we turn again to

Table 2.7. To begin with, equational states and nonverbal locative states are excluded because

they are nonverbal and therefore lack cross-referencing. Cognition states are also excluded

because they are not cross-referenced.

Possession stative verbs are marked by either ki- or nd- (cf. Table 2.7). These prefixes

indicate either the presence or absence of possession, not the thematic role of the pivot, which, in

possessive clauses, is always the locative/possessor argument. It was pointed out in 2.1.4 that

both ki- and nd- occur with the existential root ara. In fact, these prefixes can also indicate the

presence or absence of existence in existential clauses. Whereas only possessive clauses were

described in 2.1.4, (64) and (65) are examples of existential clauses (cf. (49) and (50)).

(64) Ki-ara lama dii bali nu.
POS-exist person at house 2sGEN
'There is SOMEBODY at your house.'

(65) Sia nd-ara kati'.
3sNOM NEG-exist here
'HE is not here.'
The LS of existential states is exist' (x,y), where x=locative/domain and y=theme/entity.

Existential states are closely related to locative states (cf. Van Valin 1994; Freeze 1992).

Although the verb morphology of existential and possessive statives is identical, there are

important differences between the two. Syntactically, the pivot always precedes the verb in




"The term "Philippine-type" language is used with both the linguistic and the political situation in
mind. Bonggi has features which are commonly associated with Philippine languages in the
linguistic literature, and thus, from the perspective of linguistics, could be called a Philippine
language. However, from the perspective of politics, reference to Bonggi as a Philippine language
is anathema to Malaysians. This is especially so since the Philippines refuses to relinquish its
claim to Sabah.









possessive clauses (e.g. (49), (50), (51), (55), (56)), whereas in existential clauses indefinite pivots

follow the verb (e.g. (64)) and definite pivots precede the verb (e.g. (65)). If we avoid transparent

labels such as possessor to describe arguments, and instead use more generic labels such as

locative, we find the circumstances given in (66).

(66) a. LS for possession states: have' (x,y) x=locative, y=theme pivot=undergoer-locative
b. LS for existential states: exist' (x,y) x=locative, y=theme pivot=undergoer-theme
Both possession states and existential states are two-place stative verbs with identical

thematic roles by (66). The difference is that existential states, like locative states, select the

second argument of predicate' (x,y) as the undergoer pivot, whereas all other stative verbs select

the first argument (cf. (63) and Table 2.7). If the function of verbal cross-referencing is to encode

the 0-role of the pivot, there is a contradiction with ki- cross-referencing the locative for possession

stative verbs (e.g. (49), (55)) and the theme for existential stative verbs (e.g. (64)).

Experience states are marked by -an which cross-references a marked undergoer pivot (e.g.

(28), (31)). This leaves condition states (e.g. (21), (24), (26), (33), (34), (37), (39), (41), (42),

(43), (44)) and verbal locative states (e.g. (15)) which are both marked by m-. For condition

states, the undergoer-pivot is a patient, whereas for locative states, the undergoer-pivot is a theme.

Furthermore, the class of emotional states mentioned in (63) has an undergoer-pivot which is an

experience.

Emotional states are a small class of two-place stative verbs listed in (67), and illustrated by

the verb musulaadn 'hate' in (68). A summary analysis of (68) is provided in (69).

(67) /m-/ /mV-/
m-ingisiadn ST-have.pity.on mi-siatadn ST-like
mu-sulaadn ST-hate
me-datadn ST-detest

(68) Sia mu-sulaadn (m-sulaan) diaadn.
3sNOM ST-hate IsACC
'HE hates me.'






78

(69) a. LS for emotional states: hate' (x,y) x=experiencer, y=theme
b. Clause-specific LS for (68): sulaadn 'hate' (sia 'he', diaadn 'me')
c. Assign thematic relations: sia 'he' +- experience
diaadn 'me' <- theme
d. Assign semantic macroroles: undergoer <- experience 1 Macrorole
e. Assign syntactic functions: pivot *- undergoer-sia 'he'
direct core syntactic argument -- theme-diaadn 'me'
f. Assign case: pivot-sia 'he' <- nominative case
diaadn 'me' <- accusative case
g. Cross-reference verb: sulaadn 'hate' m- 'ST
Emotional states are two-place statives with two core syntactic arguments, but only one

macrorole (cf. possessive states 2.1.4; cognition statives 2.1.5). The first argument in the LS is

an experience and the second a theme (cf. Van Valin 1990:244; 1993:79-80). Only the

experience can be the pivot, so according to (54c) these statives have only one macrorole which,

by (54b.2), must be an undergoer (cf. (69d)). Emotional states are cross-referenced by m- (cf.

(69g)). In fact, emotional states are interesting in that they appear to be doubly marked. Note that

all of the forms in (67) end in [adn], just like experience states (cf. Table 2.6). This is especially

interesting in that, by the Actor-Undergoer Hierarchy in (4), the theme argument is the unmarked

undergoer for both emotional and experience states. In the case of experience states, the theme

argument is incorporated into the predicate, and the experience argument is linked to undergoer

resulting in a marked linking between experience and undergoer in terms of (4) and a

corresponding markedness in morphology with -an. In the case of emotional states, the theme

argument is not incorporated into the predicate, and the experience argument is linked to

undergoer also resulting in a marked linking between experience and undergoer in terms of (4).









However, although a corresponding markedness in morphology is expected, instead, unmarked m-

occurs and all the verbs end in [adn], as if they were doubly marked with both m- and -an.65

Problems in analyzing two-place stative predicates were initially raised by the discussion of

melou 'embarrassed' and molok 'scared' at the end of 2.1.3. The analysis of possession states

(2.1.4), cognition states (2.1.5), existential states (2.1.5), and emotional states (2.1.5) has

shown that there is a unified analysis for stative verbs. Although some stative verbs can have two

core (nonoblique) syntactic arguments, they are all intransitive in RRG terms; i.e. they have only

one macrorole. Furthermore, there is only one possible syntactic pivot for each subclass of stative

verb, and the thematic role of the pivot is completely predictable.

I have also shown from the discussion of stative verb cross-referencing that the function of

cross-referencing in Bonggi cannot be to simply encode the e-role of the pivot. This still leaves

open the question of what function is fulfilled by cross-referencing. Two possibilities are suggested

by Table 2.7, either the highest-ranking macrorole is cross-referenced or the pivot is cross-

referenced (cf. Van Valin 1990:241).


2.2. Achievements

Achievements are situations which result from a nonagentive single change of state.

Achievements contain an underlying stative in their LS. The LS for achievements varies depending

upon the type of stative from which a particular achievement verb is derived. For example, (70)

shows the LS for achievement verbs which are derived from condition stative verbs (cf. Table 2.7).





65Perhaps the solution to the small leak in the data lies in emotional states being exceptions to
(54b.2). Only the experience can be the pivot; therefore, emotional states have only one
macrorole. The assumption here, based on (54b.2), is that the single macrorole must be an
undergoer. The exception would involve a LS with two arguments and no activity predicate, but
with a single macrorole which is an actor. Activity verbs can be derived from at least three of the
roots in (67); i.e. k-ingisiadn 'have pity on', ki-siatadn 'like', and ku-sulaadn 'hate' (cf. (45) and
(46)). The remaining word in (67) has a form ke-deta-an 'despised', suggesting the possibility of a
root other than datadn. Emotional states and experience states are similar in that only the
experience can be the pivot. The stimulus (theme) cannot be the pivot (cf. Talmy 1985:99ff.).






80

(70) LS for achievements with underlying condition stative verb: BECOME predicate' (x)
Achievements are derived from states by the addition of the logical operator BECOME which

indicates inchoative aspect (cf. Walton 1983:23; Van Valin 1990:223). Because achievements are

derived from states, states are considered basic. This section shows how the addition of the logical

operator BECOME to different subclasses of states described in 2.1 effects both their semantic and

morphosyntactic structure.

I begin by showing how achievement verbs can be derived from condition states (cf. 2.1.3).

For example, (71) illustrates an English stative clause and its LS, whereas (72) illustrates the

corresponding achievement clause and its LS.

(71) It is dry. dry' (it)

(72) It is getting dry. BECOME dry' (it)
The Bonggi clauses which correspond to (71) and (72) are (73) (cf. (21)) and (74). Whereas

the difference between states and achievements is indicated paraphrastically in the English

examples, the difference is indicated morphologically in Bonggi where m- occurs with condition

states and kin- (realized here as an infix /-mn-/) occurs with achievements.

(73) Sia ng-korikng (m-koring).
3sNOM ST-dry
'IT is dry.'

(74) Sia k-em-orikng fkm-koring).
3sNOM @-ACH-dryw
'IT is getting dry.'
Both the condition stative verb ng-korikng 'ST-dry' in (73) and the achievement verb

kemorikng 'ACH-dry' in (74) are derived from the adjective root koring 'dry'. A summary analysis

of (74) is provided in (75).








66The '@' is used to indicate the first consonant of a root or stem which is separated from the rest
of the root by an infix.








(75) a. General LS for achievements
with underlying condition state: BECOME predicate' (x) x=patient
b. Clause-specific LS for (74): BECOME korikng 'dry' (sia 'it')
c. Assign thematic relations: sia 'it' *- patient
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sia 'it'
f. Assign case and prepositions: pivot-sia 'it' <- nominative case
g. Cross-reference verb: korikng 'dry' -- km- 'ACH'
A comparison of (75) with (23) is instructive.67 The only difference is the addition of the

logical operator BECOME in the LS of the achievement verb and a concomitant change in verbal

cross-referencing. There is, however, no change in the assignment of thematic relations,

macroroles, syntactic function, or case. Thus, the single argument 'x', in both instances, is a

patient which is linked to undergoer.

Other achievement verbs derived from adjective roots which are used to form one-place

condition stative verbs are shown in (76). They, too, are cross-referenced by kon-, which is realized

in different ways, as seen by the arrangement of roots in (76).68







67(23) is a summary analysis of (21), which is repeated here as (73).

68km- is realized as /km-/ before vowel-initial roots and roots whose initial consonant is a bilabial.
Then an epenthetic vowel V/ is inserted to avoid impermissible syllable structures (cf. footnote
31). This vowel is subject to vowel harmony and vowel weakening. Bonggi is not unique in
treating bilabial consonants differently than other consonants in terms of affixation. For example,
see Kerr 1965:41, footnote 11; and Brewis (1992). If the initial consonant of the root is a bilabial
nasal, e.g. mula' 'old (things)' and mulak 'young', a rule of nasal deletion occurs deleting the nasal
portion of the affix. (Evidence for mula' 'old (things)' and mulak 'young' being nasal-initial roots is
found in causative constructions where the nasal occurs in the causative stems, e.g. pu-mula'
'make-old' and pu-mulak 'make-young'. Compare causative stems formed from vowel-initial roots,
e.g. p-ingad 'make-near', and p-odu "make-far'.)

In the elsewhere case, km- is realized as an infix, /-m-/, immediately following the root-initial
consonant. Then an epenthetic vowel /V/ is inserted to avoid impermissible syllable structures.
The infix /-m-/ cannot occur with vowel-initial roots and roots whose initial consonant is a bilabial,
so /k/ appears to occur in order to provide the right phonological environment for infixing a /-m-/.

In the past tense, the epenthetic vowel does not occur; instead, the vowel /i/ occurs in place of the
epenthetic vowel indicating past tense. Thus, the past tense form of kemorikng 'getting dry' (cf.
(74)) is kimorikng 'got dry'. I avoid past tense examples in this section in order to reduce the
morphological complexity in the examples.






82

(76) /km-/ /-m-I
kem-aal ACH-expensive d-em-alabm ACH-deep
kem-enta' ACH-unripe d-em-oot ACH-bad
kem-onsobm ACH-sour d-em-ama' ACH-dirty
kim-ingi ACH-crazy d-um-uruk ACH-fast
kim-iskidn ACH-poor g-em-angan ACH-tired
kem-odobm ACH-black g-im-ia ACH-big
kem-odu' ACH-far k-em-abul ACH-lazy
kem-omis ACH-sweet k-em-apal ACH-thick
kum-udap ACH-hungry k-em-aya ACH-rich
kum-utakng ACH-spoiled k-em-aya' ACH-lazy
kem-alus ACH-fine k-em-otul ACH-hard
kem-arabm ACH-forbidden l-em-ompukng ACH-fat
kem-panggar ACH-stiff l-em-anggu ACH-long
kem-pesa' ACH-broken 1-im-iug ACH-tall
kum-puti' ACH-white 1-um-uag ACH-baggy
kem-pagadn ACH-hard r-em-omuk ACH-rotten
kem-panas ACH-hot s-em-ega' ACH-red
kim-bisa ACH-strong t-em-ogi' ACH-pregnant
kum-buha' ACH-open t-em-ogobm ACH-diligent
kem-baru ACH-new t-em-ook ACH-ripe
kim-biru ACH-blue t-em-oyuk ACH-little
kem-basa' ACH-wet t-im-ihukng ACH-crooked
kim-bilug ACH-round t-um-ua' ACH-old (people)

/k-/
ku-mula' ACH-old (things)
ku-mulak ACH-young

Not all adjectives have the same status with respect to their potential for achievement verb

affixation. All of the adjective roots in (24) and (25) have corresponding derived achievement

verbs. That is to say, if an adjective root can occur as a condition stative predicate, it can also

occur as a derived achievement verb. However, if adjective roots cannot occur as condition stative

predicates, neither can they occur as derived achievement verbs.

2.1.3 pointed out that condition stative verbs such as ng-korikng 'ST-dry' in (73) (cf. (21))

have a single core syntactic argument which is semantically a patient, but such clauses can have an

optional adjunct which is an effector as was shown in (26) and described in (27). Similarly,

achievement verbs such as kemorikng 'ACH-dry' in (74) have a single core syntactic argument

which is semantically a patient (cf. (75a)), but such clauses can have an optional adjunct which is

an effector as shown in (77). The effector is marked with ga' '"from', indicating that it is

syntactically an oblique adjunct and semantically an effector.









2.1.3 also pointed out that ga"from' has a causative meaning, and only occurs with states

and achievements. Achievements, like states, express causality indirectly as reflected in the LS of

(78b) where one of the arguments of ga' 'from' is [BECOME predicate' (y)].

(77) Sia k-em-orikng (km-koring) ga' odu.
3sNOM @-ACH-dry from sun
'IT is becoming dry from the sun.'
A summary analysis of (77) is provided in (78) (cf. (27)).

(78) a. General LS for achievements
with underlying condition state: BECOME predicate' (x) x=patient
b. Clause-specific LS for (77): [from' (x, [BECOME predicate' (y)])]
x=antecedent cause,
predicate'(y)=resulting state,
y=patient
[ga' 'from' (odu 'sun', [BECOME korikng 'dry' (sia 'it')])]
c. Assign thematic relations: sia 'it' <- patient
odu 'sun' <- effector (antecedent cause)
d. Assign semantic macroroles: undergoer <- patient 1 Macrorole
e. Assign syntactic functions: pivot <- undergoer-sa 'it'
oblique adjunct *- effector (antecedent cause)-odu 'sun'
f. Assign case and prepositions: pivot-sia 'it' <- nominative case
effector-odu 'sun'+- oblique ga' 'from'
g. Cross-reference verb: korikng 'dry' -- km- 'ACH'
Achievement verbs can also be derived from the cognition stative roots described in 2.1.5.

For example, (79) illustrates the achievement verb kem-pandi 'ACH-know' (cf. (61)).69

(79) M-pagadn na kem-pandi (km-pandi) uubm Bonggi.
ST-difficult PFT ACH-know language Bonggi
'IT is difficult to learn Bonggi.'
A summary analysis of the downstairs (or complement) clause in (79) is provided in (80).

Like their stative verb counterparts, achievement verbs which are derived from cognition stative

roots have two arguments in their LS, both of which occur as direct core arguments in syntax, but

only the experience is linked to a semantic macrorole.






69The phonological processes resulting from the affixation of km- to the root pandi 'know' and
yielding the surface form kem-pandi 'ACH-know' are the same as those described in footnote 68
and exemplified in (76).






84

(80) a. General LS for know: know' (x,y) x=experiencer, y=theme
b. Clause-specific LS for (79): BECOME pandi 'know' (0, uubm Bonggi
'Bonggi language')
c. Assign thematic relations: 0 *- experience
uubm Bonggi 'Bonggi language' <- theme
d. Assign semantic macroroles: undergoer experience 1 Macrorole
e. Assign syntactic functions: pivot +- undergoer-0
direct core syntactic argument *- theme-uubm Bonggi
'Bonggi language'
f. Assign case and prepositions: core argument-uubm Bonggi <- accusative case (0)
g. Cross-reference verb: pandi 'know' <- km- 'ACH'
A third type of achievement verb has an underlying possession stative in its LS. The addition

of the logical operator BECOME indicates a change of possession (cf. Jolly 1993:277). As pointed

out in 2.1.4, the prefixes ki- and nd- mark verbs of possession. ki- marks the presence of

possession in both possession stative verbs (e.g. (49), (51), (55)) and achievement verbs with an

underlying possession stative, e.g. (81). nd- marks the absence of possession in both possession

stative verbs (e.g. (50), (56)) and achievement verbs with an underlying possession stative, e.g.

(82).70

(81) Sia ki-a-ardn (ki-ara-an) na saa.
3sNOM POS-exist-ACH PFT spouse
'HE just got married.'

(82) Sia nd-a-ardn (nd-ara-an) na saa.
3sNOM NEG-exist-ACH PFT spouse
HE is no longer married.'71
A summary analysis of (81) is provided in (83). A comparison of the analysis in (83) with the

analysis of a possession stative in (52) is instructive. The two analyses are the same with the

exception of the addition of the logical operator BECOME in the LS of the achievement verb in (83)




70Achievement verbs with an underlying possession stative are derived from the existential root
ara, never from noun roots (cf. 2.1.4).
71It is instructive to compare the meaning of the achievement clauses in (81) and (82) with the
stative clauses in (51) and (50). The stative clause in (50) is used when speaking about someone
who has never been married, while the achievement clause in (82) is used in speaking about
someone who has been married, but is either recently divorced or whose spouse is recently
deceased. Similarly, the stative clause in (51) is used when speaking about someone who has been
married for a while, whereas the achievement clause in (81) is used in speaking about someone
who has recently married. Thus, the achievement clauses are used to refer to a change in marital
status which is normally interpreted as being 'recent'.









and a concomitant change in verbal cross-referencing by the additional affix -an. There is no

change in the assignment of thematic relations, macroroles, syntactic function, or case. In both

instances, the possessor argument is the pivot.

(83) a. General LS for achievement with
underlying possession statives: BECOME have' (x,y) x=possessor, y=possessed
b. Clause-specific LS for (81): BECOME ara 'existential' (sia 'he', saa 'spouse')
c. Assign thematic relations: sia 'he' +- possessor
saa 'spouse' <- possessed
d. Assign semantic macroroles: undergoer *- possessor 1 Macrorole
e. Assign syntactic functions: pivot +- undergoer-sia 'he'
direct core syntactic argument <- possessed-saa
'spouse'
f. Assign case and prepositions: pivot-sia 'he' 4- nominative case
core argument-saa 'spouse' 4- accusative case (0)
g. Cross-reference verb: ara 'existential' -- ki- 'POS', -an 'ACH'72
Unlike other achievement verbs which are cross-referenced by a form of the infix -m-,

achievement verbs with an underlying possession stative are cross-referenced by the suffix -an.73

2.1.4 stated that possessors can be inanimate, e.g. (55) and (56). Sentence (84) illustrates an

inanimate possessor in an achievement possessive clause. The verb is marked with -an indicating

that this is an achievement. Example (84) is used if the boat is recently painted, while (55) is used

if a boat has an older paint job.

(84) Oig na ki-a-ardn (ki-ara-an) na saat.
boat DEF POS-exist-ACH PFT paint
THE BOAT has been painted.'

A fourth type of achievement has an underlying existential state in its LS. The addition of the

logical operator BECOME indicates a change in existential circumstance. The prefixes ki- and nd-

indicate the presence or absence of existence in existential clauses. Affixation in achievement





72ki- derives possessive states; -an derives achievement verbs from possessive states.

73The morphophonemic processes involved include nasal harmony, nasal preplosion, metathesis,
and vowel deletion. The vowel deletion rule which deletes a vowel before an identical long vowel is
required to derive [ki'aardn] 'have-GOAL'. The derivation is: underlying form #kiara + an#;
stress rule #kia'raan#; nasal preplosion #kia'raadn#; metathesis #kia'aardn#; vowel deletion -
[ki'aardn].









existential clauses is identical to that found in achievement possessive clauses. Sentences (85) and

(86) illustrate achievement existential clauses (cf. (64), (65)).74

(85) Ki-a-ardn (ki-ara-an) na lama dii bali nu.
POS-exist-ACH PFT person at house 2sGEN
There is SOMEBODY at your house.'

(86) Sia nd-a-ardn (ki-ara-an) na kati'.
3sNOM NEG-exist-ACH PFT here
IHE is not here.'
A fifth type of achievement has an underlying locative predicate in its LS. As pointed out in

2.1.2, locative predicates are either verbal or nonverbal, depending on what functions as the

predicate. In both instances, the LS contains the two-place abstract predicate, be-at' (x,y).75 The

addition of the logical operator BECOME to a locative predicate indicates a change in location. For

example, (87) and (88) illustrate two locative states, the former nonverbal and the latter verbal.

The corresponding achievement states are illustrated by (89) and (90) respectively. In (87) (cf.

(19)), the spatial deictic dii 'there' (cf. Table 2.3) functions as the predicate of the nonverbal state.

In (88) (cf. (15)), the predicate is derived from the locative adjective root ingad'near'.

(87) Sia dii.
3sNOM there
'SHE is there.'

(88) Sia m-ingad.
3sNOM ST-near
'SHE is near.'

(89) Sia kin-dii (km-dii).
3sNOM ACH-there
'SHE is going there.'

(90) Sia kim-ingad (km-ingad).
3sNOM ACH-near
'SHE is getting near.'




74The semantic contrast between the achievement existential clauses in (85) and (86), and the
stative existential clauses in (64) and (65) is interesting, and not reflected in the English free
translations. The achievement existential clauses are used to convey contra-expectation.

75Cf. Table 2.7; cf. also Jolly (1993:277).









The LSs and verbal cross-referencing for (87), (88), (89), and (90) are provided in (91a),

(91b), (91c), and (91d) respectively. In each case, the first argument (the locative) is unspecified.

The second argument (the theme) is both the undergoer and the syntactic pivot. There is no change

between locative states and achievements with underlying locative states in terms of the assignment

of thematic relations, macroroles, syntactic functions, or case (cf. (18)). The difference lies in the

addition of the operator BECOME to the LS of the achievement clauses and a concomitant change in

verbal cross-referencing.76

(91) a. LS for (87): dii 'there' (0, sia 'she') Cross-referencing:
b. LS for (88): ingad 'near' (0, sia 'she') Cross-referencing: m-
c. LS for (89): BECOME dii there' (0, sia 'she') Cross-referencing: km-
d. LS for (90): BECOME ingad 'near' (0, sia 'she') Cross-referencing: kin-
Like their locative stative counterparts, achievements with underlying locative predicates can

occur with a specified locative argument as in (92) (cf. (89)) and (93) (cf. (90), (15)). A summary

analysis of (93) is provided in (94) (cf. (18)).

(92) Sia kin-dii (km-dii) bali nya.
3sNOM ACH-at house 3sGEN
'HE is going to his house.'

(93) Sia kim-ingad (km-ingad) diaadn.
3sNOM ACH-near IsACC
'SHE is coming close to me.'







76The cross-referencing of achievements by km- is realized in different ways depending on both the
lexical category and the phonological shape of the root. kin- is realized the same for both
achievement verbs derived from locative adjectives (e.g. kim-ingad 'ACH-near' in (90)) and
achievement verbs derived from other adjectives (e.g. kem-aal 'ACH-expensive in (76)). The
phonological processes involved are described in footnote 68 and exemplified in (76). For many
adjective roots, km- is realized as an infix I-m-n-(e.g. d-em-alabm 'ACH-deep' in (76)) (cf. other
examples in (76)). However, km- is always realized as a prefix, never as an infix, for achievement
verbs derived from spatial deictics (e.g. kin-dii 'ACH-there' in (89)). In this case, an epenthetic
vowel NI /V/is inserted to avoid impermissible syllable structures (cf. footnote 68), and the nasal
portion of the prefix km- assimilates to the same point of articulation as the initial consonant of the
root (this follows the rule of nasal assimilation in footnote 31). Furthermore, /k/ is not simply
inserted just to provide the right phonological environment for infixation; it indicates direction
toward a location (goal) with these achievement verbs.






88

(94) a. General LS for locative state: be-at' (x,y) x=locative, y=theme
b. General LS for achievements
with underlying locative state: BECOME be-at' (x,y) x=locative, y=theme
c. Clause-specific LS for (93): BECOME ingad 'near' (diaadn 'me', sia 'she')
d. Assign thematic relations: diaadn 'me' <- locative
sia 'she' <- theme
e. Assign semantic macroroles: undergoer <- theme 1 Macrorole
f. Assign syntactic functions: pivot *- undergoer-sia 'she'
direct core syntactic argument -- locative-diaadn 'me'
g. Assign case and prepositions: pivot-sia 'she' *- nominative case
core argument-diaadn 'me' <- accusative case
h. Cross-reference verb: ingad 'near' <- kin- 'ACHI
It is instructive to compare the analysis of the achievement verb in (94) with the analysis of

the locative stative verb in (18). Both verbs have two arguments in their LS. In the stative

situation, (18a), both the theme (Figure) and the locative (Ground) are stationary, whereas in the

achievement situation, (94c), the theme (Figure) is moving with respect to the locative (Ground).

By (54c), locative statives and achievements with an underlying locative predicate only have one

macrorole, since only the theme can be the clause pivot, never the locative. By (54b.2), the single

macrorole is an undergoer (cf. (52d)). Thus, in both situations, following the Actor-Undergoer

Hierarchy in (4), the theme is linked to undergoer and the locative is not linked to a macrorole (cf.

(18d), (94e)).77 In both situations the pivot is an undergoer, and the locative is a direct core

syntactic argument (cf. (18e), (94f)).78 In the stative situation the verb is cross-referenced by m-

(cf. (18g)), whereas in the achievement situation it is cross-referenced by km-.

Bonggi achievement verbs with underlying locative predicates have a directional meaning.

For instance, the achievement verb in (92) indicates direction toward a location; that is, it is goal-

oriented. However, other achievement verbs indicate direction from a location; that is, they are

source-oriented. Two types of spatial deictics were described in 2.1.2: specific, illustrated by dii




77According to the Actor-Undergoer Hierarchy in (4), the theme is the unmarked choice for
undergoer since it is the rightmost argument. Furthermore, just because a verb has two arguments
in its LS, it is not required to have two macroroles (VV 1993:46-47).
78In (94g) accusative case is overtly marked since pronouns are distinguished according to case,
but in (18f) accusative case is not overtly marked since common nouns are not case marked (cf.
Table 1.3). Thus, case is lexically dependent here.









'there' in (87) (cf. Table 2.3), and nonspecific, illustrated by kuii "somewhere yonder' in (95) (cf.

Table 2.4). Given the nonverbal locative statives in (87) and (95), (96) and (97) are source-

oriented achievements derived these locative states. A summary analysis of (97) is provided in

(98).

(95) Sia kuii'.
3sNOM somewhere.yonder
'SHE is somewhere yonder.'

(96) Sia ti-dii (tm-dii).79
3sNOM FROM-there
'SHE is from there.'

(97) Sia tung-kuii' (tm-kuii').
3sNOM FROM.ACH-somewhere.yonder
'SHE is from somewhere yonder.'

(98) a. General LS for locative predicate: be-at' (x,y) x=locative, y=theme
b. General LS for source-oriented
achievements with underlying
locative state: BECOME NOT be-at' (x,y) x=locative, y=theme
c. Clause-specific LS for (97): BECOME NOT kuii' 'somewhere yonder' (0, sia 'she')
d. Assign thematic relations: sia 'she' <- theme
e. Assign semantic macroroles: undergoer -- theme 1 Macrorole
f. Assign syntactic functions: pivot +- undergoer-sia 'she'
g. Assign case and prepositions: pivot-sia 'she' <- nominative case
h. Cross-reference verb: kuii' 'somewhere yonder' <- tm- 'ACH'
The addition of the operator BECOME NOT to a locative predicate results in a source-oriented

achievement verb (cf. (98b)). Thus, by (98c), (97) is a source-oriented achievement verb. By

(54c), there is only one macrorole, since only the theme can be the clause pivot, never the locative.

By (54b.2), the single macrorole is an undergoer. By (4), the theme is linked to undergoer (cf.

(98e)). By (7), the undergoer is the pivot (cf. (98f)). The verb is cross-referenced by tm- (cf.

(98h)).

Differences between goal-oriented and source-oriented directional verbs include: (i) the

absence of NOT in the LS of goal-oriented directionals (e.g. (94b), (94c)), but the presence of NOT




79The vowel associated with the prefix tm- is epenthetic. It is inserted to avoid impermissible
syllable structures (cf. footnotes 31 and 68). The nasal portion /m/ does not occur with source-
oriented specific spatial deictics such as ti-dii 'FROM-there' in (96).









in the LS of source-oriented directionals (e.g. (98b) (98c)); and (ii) goal-oriented directional verbs

are cross-referenced by km- (e.g. (94h)), whereas source-oriented directional verbs are cross-

referenced by tm- (e.g. (98h)).

Perhaps, instead of these two prefixes, there are three: k-, t-, and m- with k- indicating goal-

orientation; t- indicating source-oriention; and -m- 'ACH' deriving a directional verb from a spatial

deictic. Note, however, that /m/ does not occur with source-oriented specific spatial deictics such

as ti-dii 'FROM-there' in (96). This is due to the fact that ti-dii 'FROM-there' and other source-

oriented specific spatial deictics can occur as locative adverbs as in (99) or locative prepositions as

in (100).

(99) Sia muhad ti-dii (tm-dii).
3sNOM leaves FROM-there
'SHE leaves from there.'

(100) Sia muhad ti-dii (tm-dii) bali nya.
3sNOM leave FROM-at house 3sGEN
'SHE leaves from her house.'
In both (99) and (100) the source interpretation associated with tm- is derivative from the

spatial/locative meaning of dii 'there' which is primary.80 Source-oriented spatial deictics which

are derived from the specific spatial deictics in Table 2.3 are shown in Table 2.8.81 The derived

spatial deictics are used to indicate a location away from the deictic center which is established by

the base form.











80Cf. Lichtenberk (1991:485) for a similar conclusion regarding To'aba'ita, an Austronesian
language in the Oceanic subgroup.

81Although source-oriented directional verbs (e.g. tung-kuii' 'FROM.ACH-somewhere yonder' in
(97)) can be derived from the nonspecific spatial deictics in Table 2.4, source-oriented spatial
deictics cannot be derived from the nonspecific spatial deictics in Table 2.4.