14 Different icons represent different language subgroups of the Austronesian language family.
15
All indigenous languages that fall on the island of Taiwan are dubbed Formosan languages, except Yami, which belongs in the Malayo-Polynesian family (Ross 2009).
The language under the present study, Kanakanavu, is located near the central part of Taiwan, with Saaroa and Tsou being in vicinity. Besides, It is noteworthy that in Taiwan alone, there are surprisingly more than 20 Austronesian languages (some are not shown in Map 2.2). This linguistic co-existence is believed to show an implication with regard to the position of Formosan languages, a topic to be addressed next.
2.2.2 Genetic relationship
A well-adopted hypothesis in recent years is that Taiwan is most likely the homeland of Austronesian languages (cf. Dyen 1965, Blust 1999, Li 2008, etc.). That is, the dispersal of Austronesian ancestors may have departed from Taiwan across the oceans, and these people eventually settled in different parts of the world (remember Map 2.1). This version of dispersal origin involves two assumptions: a) the more diverse the languages are in an area, the more likely this area is to be the homeland for the language family; b) The more proto characteristics reserved in the languages, the more likely the area is to be the homeland.
The subgrouping based on Ross (2009), as shown previously in Figure 1.3 in Chapter 1, added a whole-new layer for the Formosan languages—Nuclear
Austronesian, to which Kanakanavu belongs.
However, what concerns us in the present thesis is the subgrouping of Kanakanavu, which is grouped within the Tsouic family. The question to be asked is:
Is it substantially and adequately proven? As discussed in the literature, the answer is positive to some researchers while to others, it is not (cf. Chang 2006). We will briefly
16
address this issue in Chapter 5.
2.2.3 Typological characteristics
Formosan family exhibit several unique properties that are not shared in other languages. Two major characteristics are demonstrated as follows.
One peculiarity is the word order. Generally, the basic word order in Formosan languages is verb-initial, or predicate-initial (Li 2008:524). For instance, it is Verb-Theme-Agent in Tsou15, as shown in (1).
(1) Tsou (Li 2008:527)16, 17
pei’i ta ucey ’e ino.
AV.cook OBL taro NOM mother
‘The mother cooked taros.’ (glossing mine)
On the other hand, Kanakanavu, our target language in this thesis, is verb-initial as well, but the word order shows a Verb-Agent-Theme pattern, no matter the voice type.
See Data (2).
(2) Kanakanavu
tia miapacai Pi’i tutui iisua FUT AV.kill PN pig that
‘Pi’i is going to kill that pig.’
The other important feature in Formosan languages to be discussed here—voice
15 As Huang (2002:669) states, the predominant word order is AUX-V-O-A, or A-AUX-V-O, in AV clauses, but AUX-V-A-O in NAV clauses.
16 In the present thesis, we will adopt the following glossing abbreviations, in alignment with the corpus under study: AV=agent voice/actor voice, BV=benefactive voice, Ca=Ca-reduplication, COS=change of state, CV=circumstantial voice, IV=instrumental voice, LV=locative voice, NAV=non-actor voice. Other abbreviations will be mentioned in particular when needed.
17 Usually a Tsou clause will require the existence of an auxiliary preceding the verb.
17
system, has been a center of research in the literature since it displays highly complicated aspects in nature. The PAn voice system so far has not undergone dramatic development in terms of formal change18. That is, each voice type shares a similar form among the majority of Formosan languages. Below is a table showing the four morphemes in PAn.
Table 2.1 The voice system in PAn19 (Ross 2009)
Voice type
Actor voice Undergoer voice
actor patient location circumstance
Proto form
M-/-um- -en -an Sa-/Si-
What’s more interesting is the fact that the voice system may indicate certain syntactic, semantic, and even pragmatic traits, which happen to be our research goal in the present thesis.
Other characteristics, such as exclusive-inclusive 1st person plural pronouns, affixation of bound morphemes, and ergative alignment, are prominent in Formosan languages, though they do not concern us here.
With the general properties about Formosan family as the background knowledge, we may now turn in the next section to the previous Kanakanavu studies that we consider significant.
2.3 Linguistic studies on Kanakanavu
There are unfortunately very few studies on Kanakanavu language, partly
18 As a matter of fact, with that said, there are some complications concerning the development of PAn voice system. Some Formosan languages, such as Tsou, do not show similar morphological pattern with the one presented in Table 2.1. Is it that the languages (like Tsou) preserve the more ancient forms, or that the languages themselves have gone through certain innovative changes? Since it is not our focus here, suffice it to say for the moment that the generally reconstructed PAn voice morphology is basically retained in most Formosan languages.
19 These are in fact the forms in realis perfective.
18
because of the scarcity of native speakers and the fact that the tribe is located in a remote mountainous area, as have been mentioned in Chapter 1. Despite the adversity, we are still very lucky to have some a few pioneer researches, and based on these works, we may further examine and confirm what we know of so far, or adopt a new line of thinking which is more appropriate in pursuit of an integral description of the voice system.
Among all studies on Kanakanavu language, some are not published, while others are fortunately accessible to the public. These publications include Ogawa and Asai (1935:723-739), Tsuchida (1976:26-58), Mei (1982), Ho (1997), and miscellaneous works such as Szakos’s (2001) compilation on Kanakanavu stories and Kanakanavu folk songs in CDs (Li 2001), etc. We will now turn to each of the major works which pertain to the present study and some discussions are carried out below.
2.3.1 Ogawa and Asai
20(1935)
Ogawa and Asai’s (1935) The Myths and Traditions of the Formosan Native
Tribes (in Japanese) is perhaps one of the first pioneer studies in which a sketch of
Kanakanavu’s grammar is provided. As a matter of fact, the section on Kanakanavu in this compilation work is offered by Erin Asai (1894-1969). He described a great deal of the phonological system in this language, as well as a variety of morphological affixes and other parts of speech. Furthermore, a total of 7 texts are recorded, with rudimentary glossing and a general translation.Although Asai did not mention anything about the voice system, he did list several bound morphemes which happen to include what we call the voice markers.
See below.
20 Japanese scholars like Ogawa, Asai, and Tsuchida have made tremendous contributions to the studies and documentation of Formosan languages. See Li 2010 for a thorough discussion.
19
Ogawa and Asai (1935:724-727)
(a) [um-, -um-]: prefixes of a verb, e.g., um-aɭa ‘take’
(b) [-ʉnʉ]: denoting a passive meaning, e.g., tsʉʉɭa ‘see’>tsʉʉɭa-ʉnʉ ‘be seen’
(c) [-a, -ana]: denoting a place, e.g., tsau ‘person’>tsatsauwana ‘village’
(d) [si-]: denoting a tool, e.g., kəunʉ ‘food’>si-jakəʉnu ‘utensils’
The four affixes as analyzed above seem to match the core semantic instantiation of Austronesian voice system. However, types (c) and (d) were claimed to be more like a derivational affix that turns a nominal into another nominal on which the particular semantic meaning, place or tool, is imposed. That is to say, Asai did not deal with cases where a verb involves an additional core argument due to the affixation of -a/-ana or si-.
Despite the inadequacy of in-depth discussions of Kanakanavu’s voice system, Ogawa and Asai’s (1935) is still considered significant in that it contains a collection of texts from 12 Formosan languages with a brief description of grammar for each.
2.3.2 Tsuchida (1976, 2003)
Another Japanese linguist has been dedicating himself to Formosan studies for decades. Tsuchida Shigeru, born in 1934 in Tokyo, is well-known for his fruitful researches on the aboriginal languages spoken in Taiwan. His dissertation, entitled
Reconstruction of Proto-Tsouic Phonology and completed at Yale University in 1976,
undoubtedly sets a landmark in Formosan literature as it provides rich documentation of lexical items and phonological rules of the three ‘Tsouic’languages—Tsou, Saaroa, and, to our benefit, Kanakanavu. It is often cited by linguists working on these languages and/or the reconstruction of proto-forms.Tsuchida (1976) analyzes quite a few aspects of Kanakanavu grammar, ranging
20
from pronominal system, case markers,21 to perfective/imperfective/future markers.
What’s more important is that he tackles the issue of ‘focus system’ (in his terminology) and discovers four different types: agent focus, goal focus, location focus, and special focus.
Table 2.2 Kanakanavu’s focus system in Tsuchida (1976)
Focus type Form Role of the syntactic subject Agent focus
um-, -um-, etc.
agentGoal focus
-ini, -unu, -ənə
affected object Location focus-a, -an, -anə
locationSpecial focus
-ai, -i
goal/object of an actionThe above table shows an important piece of information about the voice system in Kanakanavu. Tsuchida discovers four different voice types, the first two of which, agent and patient (goal) voice, are generally uncontroversial. What’s actually more interesting is that LF, which corresponds to locative voice, is not as productive as AV and PV, since not every verb is capable of occurring with -a22, -an, or -anə. Another voice type, realized as -ai/-i, is proposed in his analysis. Although Tsuchida cannot but temporarily term this type ‘special focus’, he describes several properties that are not shared with PV. We will address this special voice type in Chapter 3.
In addition to his dissertation, Tsuchida also collects a total of 10 texts from four speakers, transcribing each story with glossing, and this compilation work
Kanakanavu Texts is published in 2003. He focuses more on the phonology of this
language in the beginning part, yet not much about the voice system is discussed.In fact, one more voice type, be it circumstantial, benefactive or instrumental, is
21 Tsuchida (1976:36) actually uses the term ‘relation markers’ for reference.
22 The suffix -a behaves more like relativizer. We will give some examples in Section 3.3.1.
21
expected to exist in the Philippine-type voice system, but it is not mentioned in Tsuchida (1976, 2003). We will bridge the gap of this missing type in Section 2.3.5 and in Chapter 3.
2.3.3 Mei (1982)
In his paper dealing with Kanakanavu’s pronouns and verbal inflection, Mei (1982) raises several in-depth issues concerning the voice system. The following table illustrates the different voice types.
Table 2.3 Kanakanavu’s focus system in Mei (1982)
Focus type Form
Actor focus
um-, -um-, mu-
Object focus 1-ini, -unu, -ʉnʉ
Object focus 2-ai
Time/Location focus -a, -an, -anʉ
As with Tsuchida (1976), Mei also distinguishes two types of patient voice, hence OF1 and OF2. He claims that OF2 is restricted to neutral aspectuality only.
Since verbs are often assumed to be neutral in subordinate clauses headed by if or
when, both OF1 and OF2 might be possible at a first glance. However, it is stated that
one major difference between the two types is that OF1 may only occur in if/when clauses, while OF2 is confined to main clauses. This statement may not seem to work with the data we have, but it is true that they both serve as a marker that selects a patient/theme to be the syntactic subject.Another crucial element discussed in this paper is the pragmatic factor in choosing AV and NAV (Mei 1982:221). Mei’s point of view opens a window that we
22
wish to see through in this paper. We will examine to what extent pragmatic factors like definiteness may play a role in the voice system in Chapter 4.
2.3.4 Ho (1997)
In a section of The Formosan Languages in Kaohsiung (in Chinese), Ho (1997) provides an organizational description of Kanakanavu’s basic structure. He also maintains that there are four voice types, similar to Tsuchida (1976) and Mei (1982).
Ho does mention a prefix, si-, which corresponds to the instrumental voice marker, but he does not include it in his discussion of the voice system nor does he give further examples that illustrate this specific construction in a clause (Ho 1997:241).
Nevertheless, three Kanakanavu texts, including one conversation, are transcribed in this book. Without assuming a professional knowledge background, this is certainly a good material from which one may start learning about this language.
2.3.5 Wu (2006)
Published in the journal University System of Taiwan Working Papers in
Linguistics, Wu (2006) marks a significant study in Kanakanavu literature as well. He
firstly proposes the existence of benefactive/instrumental voice in this language, in the form of se-, although the labelling benefactive is somewhat problematic. This problem will be addressed in Section 3.3.1.3, in Chapter 3.The issue Wu puts forward in his paper primarily deals with serial verb constructions, and the SVC constraint in Kanakanavu is claimed to differ from that in Tsou. In other words, Wu implies that the hypothesis that Kanakanavu and Tsou might have descended from the same ancestor language may not be as evident as they seem.
23
2.3.6 Other studies
Besides the above studies, other linguists have also attempted to figure out particular aspects in Kanakanavu. We will briefly discuss some of them as follows.
From a mophosyntactic point of view, Chang (2006) proposes six innovative characteristics, which seem to differentiate Tsou from the so-called Southern Tsouic languages23. It has been recognized as a crucial question as to whether the subgrouping is validated. We wish get back to this issue in Chapter 5.
Lan (2012) is the first M.A thesis working on Kanakanavu language. Her study probes the negative constructions and provides a large number of elicited sentences.
Another M.A thesis, Cheng (2013), investigates the modality in Marinax Atayal, with a typological comparison with three Formosan languages, including Kanakanavu.
Cheng and Sung (to appear) also discuss more extensively the expression of modality in Kanakanavu. Lan (2012), Cheng (2013) and Cheng and Sung (to appear) do not further elaborate on the voice system (since it is not their main topic), but they are without a doubt significant studies that help us understand the various structures and linguistic phenomena in Kanakanavu.
There are also efforts made to document the language in a descriptive fashion.
Szakos (2001) recorded and compiled dozens of Kanakanavu texts with several native speakers, though with a rather brief glossing. Last but not the least, CDs of Kanakanavu’s folk songs (Li 2001) are available for the public to appreciate their indigenous tribe.
Bearing these previous studies in mind, we will now review some relevant
23 Chang (2006) states that the innovative focus morphology, the loss of PAn perfective marker, the Focus Harmony Constraint on SVCs, the NAF-only causatives, the obligatory auxiliary constraint, and the emergence of 3rd singular nominative bound pronoun may constitute evidence against the
traditional Tsouic Subgrouping Hypothesis.
24
studies that investigate the relationship between Formosan voice system and discourse in the next section, since it constitutes our main topic in Chapter 4.
2.4 Formosan voice system and discourse
The Philippine-type voice system has long been studied in a discourse framework (Wouk 1999, Huang 2002, to name a few). The voice system in some languages is subject to factors like transitivity and topicality. Huang (2002), for example, explores Tsou and Seediq in terms of the use of voice both in narrative and conversational styles. It is found that AV clauses in Tsou exhibit a lower transitivity while NAV clauses suggest a higher one. In fact, Huang and Tanangkingsing (2011) analyze AV clauses with the (semantic) patient argument as Extended Intransitive Clauses (EICs), in which the patient argument receives an oblique case due to its inability to be tracked in the discourse. Therefore, by means of calculating the frequency rate of certain structures with the help of the corpus, it is made possible to understand how sensitive the voice system is in relation to pragmatics.
Other relations with grounding (Hopper 1982, 1986), individuation of patients (Hopper and Thompson 1980), and topicality (Givón1983, Cooreman et al. 1984) also indicate the nature of voice system in a language. In fact, we will see that the grammatical subject of a Kanakanavu clause indeed has a higher topicality, but only in certain aspects. Therefore, to examine any pragmatic implication in the use of the voice system, we will take advantage of the current theoretical discourse approach and present our results in Chapter 4.
2.5 Summary
Despite the fact that there are few studies on Kanakanavu, all the aforementioned
25
works will adequately serve a useful purpose in the analyses in Chapter 3, since we may further compare and examine the actual morphological realization of each voice, and then to look at the voice system from a multi-dimensional perspective. The next chapter will first give a morphosyntactic description of Kanakanavu’s voice system, with a small discussion of semantic roles. In Chapter 4, we will move further to the discourse aspects of the voice system.
26
Chapter 3 Voice Constructions in Kanakanavu
3.1 Introduction
The voice system in Kanakanavu generally exhibits a similar pattern as seen in many other Formosan languages. There have already been several discussions concerning this issue (cf. Tsuchida 1976, Mei 1982, Chang 2006, Wu 2006, etc.).
However, as mentioned in Chapter 2, none of these works has succeeded in presenting a clearer picture of the variant morphological forms and the syntactic structure of each voice construction. Therefore, one of the primary goals of the present chapter is to firstly lay out the morphosyntactic features of the voice system in this language, and then in Chapter 4, we will aim for the discourse implications of how people employ the voice.
A complication regarding this issue is perhaps the case of locative constructions, since the expected form, -an, turns out to be relatively unproductive comparing with the other voice types. Hence, we will look at this particular voice construction with special attention in this chapter, too.
We will first provide a sketch of Kanakanavu’s grammar in Section 3.2, and then we examine each voice construction in more details in Section 3.3. Section 3.4 offers an overview of the voice markers and those in non-indicative structures. The interaction of voice and tense, aspect, and modality markers will be presented in Section 3.5. Finally, an interim summary will be given in Section 3.6.
3.2 A sketch of Kanakanavu’s grammar
Before turning to Kanakanavu’s voice constructions, we will first provide a brief
27
introduction to the basic grammatical characteristics of Kanakanavu. Aspects such as phonology, word order, pronominal system, and discourse/case markers will be briefly discussed, most of which are based on previous literature.
3.2.1 Phonological inventory
There are at least eleven to twelve consonant phonemes in Kanakanavu. See Table 3.1. The consonant inventory generally reflects the typological tendency that a voiceless plosive/fricative phoneme is often predicted when the voiced counterpart exists, but not vice versa (Croft 2002:165). Note, however, that the phoneme /v/, for some reason, is an exception.24
Table 3.1 Consonants in Kanakanavu25 (cf. Tsuchida 2003)
manner position
Labial Dental/alveolar Velar Glottal
Plosive Voiceless p t k ʔ
According to Tsuchida (2003), /v/ was pronounced as a voiced bilabial fricative /β/ several decades ago. However, the voiced bilabial fricative is not retained anymore,
24 One possible explanation for the exception is that the phoneme /v/, which developed from /β/, was reconstructed as /b/ in PAn (Wolff 2010:149). Since the voiceless counterpart /p/ already exists in contemporary speech, the universal implication still holds.
25 The present paper will adopt the traditional transcription in Formosan literature, as shown below.
IPA Symbols adopted
28
since it is replaced with the labio-dental fricative /v/ in the present day, at least in the speech community of our informants26.
Affricate /ts/ and fricative /s/ are palatalized if they are followed by high front vowel /i/, as seen in examples given below.
cina ‘mother’ /tsina/ [tɕina]
sii ‘because’ [siː] [ɕiː]
We should also note that the /r-l/ distinction is documented in Tsuchida’s transcription, although they may not cause any distinctive meanings when one is substituted for another. Therefore, only the phoneme /r/ will be employed throughout the present study, unless a particular reference to the two phonemes is intended.
The vowels in Kanakanavu, on the other hand, are not as numerous as many other languages, since it has a basic vowel inventory of six.
Table 3.2 Vowels in Kanakanavu
frontness
height
Front Central Back
High i u u
Mid e o
Low a
Finally, Kanakanavu’s syllable structure, based on the Tsuchida (1976, 2003), is basically (C)V, but there are also instances where nasal sounds occur as a coda. Hence, the syllabic structure of Kanakanavu may be refined as follows:
Kanakanavu’s syllable structure: (C)V(Nasal)27
Kanakanavu’s syllable structure: (C)V(Nasal)27