Commendable though our recent progress in this arena has been, there remains much work to be done. In this paper, it will be shown that ergativity is a far more pervasive feature of natural language than even the most fanatic post-Europeanists have dared to dream; we will show that even phonological systems may be ergative. In fact, they cannot be otherwise.
Ergativity is defined as any morphosyntactic system which classifies transitive objects together with intransitive subjects, distinguishing both from transitive subjects. This phenomenon has been found to be quite common, and various mutations of it have been teased out of numerous languages; even the English construction "this bed sleeps six" has been (rightly) seen as exhibiting ergative characteristics.2
The core arguments of a clause may be categorized A, O, and S, where A represents the agent of a transitive clause, while O represents its patient; S represents the single argument of an intransitive clause. Morphosyntactic coding tends to group either A and S (nominativity) or S and O together, as symbolized in (1), where co-columnar items are treated identically:
(1)The presentation of nominal arguments in the roles of A, O, and S is governed by a complex set of discourse-related salience (focus) features such as given/new status, perceived importance, and topicality. Participants in A roles are nearly always given, and normally phonologically small; pronouns and zero anaphora are common. Typical O participants, on the other hand, are new, and often are presented with full noun phrases. They thus tend to be the most salient participants of the clauses in which they appear, that is, they attract the largest amount of listener attention. Of course, this is also true of S participants, since they are the only arguments of their predicates.
Nominativity Ergativity trans. A O A O intrans. S S
The phonetic analog of semantic salience is sonority, a measure which correlates closely with relative perceptual salience/prominence of phonetic segments. Vowels are much higher in sonority than are consonants, and thus, of greater perceptual salience.3
Just as nominal participants are presented as the arguments of predicates, so also are phonetic segments assigned to particular syllables. I will refer to "transitive" syllables when discussing those which take two phonetic arguments, while "intransitive" syllables are those which require only one.4 Having identified this correspondence, let us consider the following items from Hawaiian, an Austronesian language of Hawai'i:
These items unambiguously establish the syllable types V and CV for Hawaiian, and these two types account for all syllables which occur in the language.5
(2) i 'object marker' (3) ka 'definite article'
In (4), I demonstrate the parallelism between Hawaiian syllable structure and morphosyntactic ergativity:
(4)Hawaiian, therefore, is unambiguously ergative in its syllabic phonology: it utilizes identical V coding for both the single argument of a one-argument syllable and the more salient argument of a two-argument syllable.
Ergativity Hawaiian phonology trans. A O C V intrans. S V
Let us further consider a similar chart, (5):
(5)What (5) reveals is striking. While phonological ergativity was clearly exhibited by Hawaiian in (4), (5) makes equally clear that fact that phonological nominativity is impossible, since no language obligatorily codes intransitive syllables with only a C argument.6 In my 724 language sample, I found no such cases.
Nominativity Hypothetical phonology trans. A O C V intrans. S C
Thus, although morphosyntactic ergativity and nominativity have been found to occur with roughly equal frequency across the world's languages, phonological nominativity has been found not only non-occurring, but in fact impossible. Phonological ergativity, on the other hand, has been shown to exist in Hawaiian, and in fact to be the only possible system for any natural language.
2 Oh, by the way, I want to acknowledge my usual immense debt to Mr. Rob Norris for ideas contained in this paper. He and a certain tall, Indonesian-speaking UCSB linguist, who prefers to remain anonymous, should be held personally responsible for any mistakes which may be contained herein.
3 The sonority of vowels differs from that of consonants in varying degrees, depending on the specific consonants involved. This exactly parallels the varying salience of A participants, which may in a certain situation be nearly as topically salient as O's; but both cases are relatively rare exceptions.
4 Phonological ditransitives, which involve coda elements, are absent in the language to be treated here, and thus fall beyond the scope of the present work. The interested reader is referred to chapter 38 of my dissertation (Slater, to appear) for details.
5 Hawaiian actually allows tautosyllabic VV sequences, without restriction on which V's can co-occur; however, since this is true regardless of the presence or absence of an onset consonant, the analysis presented here is unaffected by what I will consider simply an elaboration of the V slot. In any case, Kalikimaka (1835) has shown that a VV sequence's length is precisely 1.61 times that of a single V, suggesting that the difference between the two types of nucleus is not phonological, but rather metric.
6 Czech liquids notwithstanding. The Czech pattern, as well as those of certain consonant-happy NW Native American languages, demonstrate various types of phonological split ergativity.
|The Boustrophedon-Plummerfeld Hypothesis--Jay Trones|
|Subliminal Linguistics--Trey Jones|
|SpecGram Vol CXLVII, No 2 Contents|