.. Nash, David. 1982. Warlpiri preverbs and verb roots, pp.165-216 in _Work Papers of SIL-AAB_, Series A Volume 6 in Memory of Lothar Jagst, ed. by Stephen M. Swartz. Berrimah, N.T.: SIL-AAB. .m1 7 .m4 3 .he 1 /David Nash/WARLPIRI VERB ROOTS AND PREVERBS .he 2 //Revised July 1981 .fo //% .hyn .fr f .ce CONTENTS .sp .in 10 .nf 0. Introduction 1. Some issues relevant to preverbs 1.1 Typological parallels 1.2 Morphological structure of suffixed compounds 2. Preverbs in Warlpiri 2.1 Semi-productive preverbs 2.2 Lexical preverbs 2.3 Semantic compositionality 2.4 Productive preverb-verb combinations 2.4.1 Adverbial preverbs 2.4.2 Dative-adjunct preverbs 2.4.3 Derived preverbs 2.5 Etymology of preverbs 2.6 Examples: verbs in /yi-nyi/ 'to give' 2.7 Summary of preverb properties 3. Classification of Warlpiri roots 3.1 Roots with many preverb-derived stems 3.2 Roots with a few preverb-derived stems 3.3 Roots with no derived stems 3.4 Etymologically complex roots 4. Comparison of Warlpiri inventory with others 5. Conclusion References Appendix .in -10 .fi .sp .hyn This paper considers the range of verbal words in the Warlpiri language of the west-central Northern Territory, and compares the inventory of verb roots in Warlpiri with that of some closely related languages. This necessarily involves the related category of "preverb", discussed in detail for Warlpiri in section 2. .ft I first visited Lajamanu (Hooker Creek) only in November 1977, soon after Lothar Jagst's untimely death. Though I had never met him, I was made welcome by his widow, and later became aware of the high esteem in which he is held in that community. It is thus a honour to be associated with his memory in this volume. .br Funding for this research has been provided by National Science Foundation grant number BNS-7913950 (Kenneth Hale, Principal Investigator). The part of this paper focussing on preverbs in Warlpiri is an extension of section 2.6 of Nash (1980:42-55), and was presented to meetings of the Central Australian Linguistic Circle (at Alice Springs, 23 August 1980) and the Australian Linguistic Society (at Monash University, Melbourne, 26 August 1980). A revised version was presented at the Seminar on Australian Aboriginal Languages at M.I.T., 9 January 1981. I am grateful to participants in these meetings for their help, and particularly to Ken Hale and Mary Laughren. .br Warlpiri data is almost entirely from Ken Hale's field notes and the files of the Warlpiri dictionary project. I am indebted to the many workers who have contributed to this project, (in particular the supervisors Pam Harris, then Mary Laughren), for on their shoulders the classification presented here is built. .br Transcription is in the practical orthography, except that hyphens are used more flexibly. I also use on occasion a "2" to indicate that the preceding morpheme is reduplicated. Glosses for pronominal clitics used a numerical code to indicate person and number, following Hale (1973a), and an added s 'subject', o 'object'. The abbreviations ERG (Ergative), ABS (Absolutive), DAT (Dative) are used for the syntactic cases of Warlpiri and the other languages under consideration: more precisely, for the abstract cases linked to argument positions of lexical items by the particular language's Linking Rules (see Nash 1980, Chapter 6 and the references there). The notation [E] indicates that the form or usage is confined to the Willowra (Lander River) and Hanson River dialects, i.e. the east of the area where Warlpiri is spoken. .ft .sp The paper shows how preverbs bring out a covert distinction between two sorts of Warlpiri verb root -- (i) those which occur in a number of lexical stems (i.e. in combination with a number of "lexical preverbs"), and (ii) those which do not. It is shown that roots of type (i) are in general to be regarded as more ancient than roots of type (ii), as shown for instance by the much higher correspondences between roots of type (i) and verb roots in the languages to the north and west of Warlpiri. Another aspect of the preverb-verb combination is that it has given rise to new verb roots, synchronically simple as the result of reanalysis of a semantic-morphological disparity. .sp .ne 4 0. INTRODUCTION The primary elements of certain complex verbs may be referred to as "preverb" (following Hale 1973b:453-55 for Warlpiri) and "verb stem", so that the complex verb of the relevant type has the following structure: .sp .ne 4 .li 3 [ preverb [root-inflexion] ] V V -----stem----- .sp In Warlpiri, such a structure is justified phonologically and morphologically, though the units are not always semantic units. As will be seen below, particularly in section 2.3, the lexical entry for a complex verb may assign a meaning to the "theme" consisting of preverb and root, which combines semantically with the tense/aspect inflexional suffix. .sp Another definition of "preverb" relies on the Auxiliary category (Capell's "catalytic auxiliary") found in Warlpiri and its neighbours. The Auxiliary constituent occurs only initially or in second position in the sentence, and so it is possible in these languages to propose the following: .sp .in 10 .ll 60 A preverb is a morpheme which combines with a verb to form a unit which may occur at the beginning of a sentence, immediately preceding the Auxiliary (and may also occur in other contexts as well). .in -10 .ll 65 .sp Such a definition is applicable in Warlmanpa, Walmatjari, and probably also in Jaru, Gurindji and Warumungu. In discussing languages further afield, the term "preverb" may be used more generally, as will be seen below. .sp The topic "simple and compound verbs: conjugation by auxiliaries in Australian verbal systems" was one of the foci of the A.I.A.S. 1974 Biennial Conference (see Dixon, ed. 1976:613-768). Perhaps because of a confusion concerning the terms "auxiliary" vs. "catalyst" (mentioned in the introduction, Dixon, ed. 1976:14), the Warlpiri language was classified by the rapporteur in terms of its "catalytic auxiliary" (Capell 1976:617-18). .ft Applied to Warlpiri, "auxiliary" refers to the complex containing a tense/aspect base (possibly zero) and pronominal clitics cross-referencing subject and object, as described in detail in Hale (1973a). .ft However, Warlpiri also has complex verbs parallel to those of Walmatjari, described by Joyce Hudson in the same session, and these in turn are similar to the complex verbs in a number of languages of northern Australia: see papers by Heath, Kofod and Sharpe in Dixon, ed. 1976. Thus, the earlier notion that .sp .ll 60 .in 10 the structuring of a verb from an invariant verb root and an auxiliary verb is a phenomenon not shared by a very large number of Australian languages (Sharpe 1972:vii) .sp .ll 65 .in -10 depended on a restricted definition of "auxiliary". With the greatly increased amount of grammatical information that has become available over the last decade, it is possible to perceive an areal continuum in the structure of the verbal word, as summarised by Dixon (1980:280-81, section 9.4.3 Number of verbs): a number of Australian languages, particularly in the east of the continent, have hundreds of mono-morphemic verb roots, whereas the same range of verbal ideas is expressed mostly by complex verbs in a number of other Australian languages, particularly in the west of the continent. .sp Heath (l976:735) distinguishes between two complex verb structures in different north-east Arnhem Land languages, both of which might be described as "an uninflected main verb bound to a following inflected auxiliary" (Heath 1978:88 on Ngandi and "most languages in the area"): .sp (i) [root - thematising suffix - suffixes] .br (ii) [verb] [pronominals-auxiliary-suffixes] . .sp Heath finds (i) in Mara, Alawa, Warndarang and Yugul, and (ii) in Nunggubuyu, Ngandi, and the Yuulngu languages (such as Ritharngu). Merlan (to appear:8-9) distinguishes two types of complex verb within Mangarrayi. These Arnhem Land languages fall into the "complex verb" area of the Australian continent. Dixon exemplifies the distribution with Dyirbal (hundreds of mono-morphemic roots) from the east on the one hand, .ft The Yirr'-Yorront language of south-west Cape York Peninsula, i.e. not far from Dyirbal in continental terms, has numerous compound verbs, and these show structural properties quite similar to those of Warlpiri preverb-verb combinations -- see Alpher (1973:391-401). .br Similarly Gumbaynggir, Gabi, and the West Torres Strait language, which have "a set of productive prefixes, yielding a considerable number of compound verbs". (Dixon 1972:17). .ft and mentions on the other Warlpiri, Walmatjari, Gurindji (Kuurrinyji) and languages of "Kimberleys and Daly River". He notes that Diyari (north-east South Australia) has over 200 verb roots: Austin (1981a:71) records 248 verb roots in that language. It can be added that Yidiny (293 roots, according to Dixon 1977:207), Ngiyambaa (Donaldson 1980), Lardil (Hale 1981), and the Arandic languages of central Australia (K. Hale, p.c.), also exemplify the "eastern" property of having hundreds of verb roots. On the other hand, Warlpiri's eastern neighbour Warumungu has around 50 verb roots (Simpson 1980); Jaru, with "only forty odd verbs" (Tsunoda 1978:79) follows the pattern of its neighbours Warlpiri and Walmatjari. In his account of preverbs in Jaru, Tsunoda (1978:199) mentions parallels in the Pama-Nyungan languages Malngin, Wandjira and Walmatjari, and the non-Pama-Nyungan Gidja, Miriwung, Wunambal and Ungarinjin. .sp The qualitative statistical situation in Nyangumarda is typical: .ll 60 .in 10 by dictionary count, simple, i.e. monomorphemic, verbs number barely one hundred, while complex verbs are many times more numerous; by text count, the scales are tipped the other way: simple verbs far outnumber complex. (O'Grady 1970:49) .ll 65 .in -10 .sp In this respect, Warlpiri is numerically intermediate between the languages to its north and north-west (with no more than 45 or so verb roots) and the languages to its south and east (with many hundreds of verb roots), for the current dictionary shows around 115 roots. .ft Languages to the west of Warlpiri include a number of Western Australian languages with hundreds of verb roots. Jingili, a neighbour-but-one to the north-east of Warlpiri, also has hundreds of verb roots. Pintupi (a south-western neighbour of Warlpiri), has well over a hundred roots, and inspection of the Pintupi Dictionary shows many complex verbs similar to Warlpiri. .ft It is striking that even the language most closely related to Warlpiri -- Warlmanpa -- has just over 40 verb roots. I propose that the discrepancy is in part attributable to Warlpiri's recent acquisition of a number of additional verb roots. .sp Near the other numerical extreme is the Daly River language Malak-malak, which exhibits "conjugation by 6 auxiliaries". There are a couple of complications to a straightforward comparison between the "auxiliaries" of Malak-malak and the verb roots of languages to its south, since each auxiliary has a number of stems that supplete with the tense/aspect categories, and it is unclear how the semantic categories governing auxiliary choice are related to the putative etymologies of the auxiliary roots. However, it is interesting to note at this stage the semantic range covered by each auxiliary, particularly as indicated by its meaning when it occurs as an independent verb: .ft Summarised from the detailed discussion in Birk, 1976:47-74. Note too Birk's observation that "the semantic function of the auxiliary is relatively slight in relation to its other functions." (p. 74) .ft .sp .nf .ne 7 conjugation meaning 1 (none identifiable; dependent only) 2 movement, distal, copula 3 movement, iterative 4 sitting, proximate 5 lying, distal 6 standing, proximate .sp .fi Another morphophonemically complex situation occurs in Ungarinjin in the Kimberleys, but the verbal inventory shows much more similarity with those of Pama-Nyungan languages to the south: .sp .ll 60 .in 10 The set of roots on which all compound verbs are based (i.e. those whose inflected forms can serve as auxiliary verbs) is closed and quite small, numbering fourteen. Of these fourteen, all but two, /-wu/ and /-yinde/ ... sometimes also occur in isolation as "simple verbs". (Rumsey 1978:108) .sp .ll 65 .in -10 The roots used exclusively as simple verbs number "several thousand" and fall into 7 conjugations (though the "vast majority" of non-compounding roots fall into the conjugation containing 8 of the 14 auxiliaries) -- Rumsey (1978:151). .sp The wider Australian correspondences proposed thus comprise the following terms used by other workers: .sp .nf .ne 7 LANGUAGE Warlpiri, Jaru root preverb Walmatjari Root 2 Root 1 Ungarinjin auxiliary simple root Malak-malak auxiliary verb Alawa auxiliary particle Yirr'-Yorront verb root prefix .fi .sp Of course, the properties of the categories vary from language to language -- in the foregoing discussion my aim has been to draw out the similarities. .sp .ne 4 1. SOME ISSUES RELEVANT TO PREVERBS .sp 1.1 TYPOLOGICAL PARALLELS .sp .ne 3 Indo-European: .sp The term "preverb" has long been used by Indo-Europeanists, dating at least from Varo (praeverbum). The term refers to particles which appear following a noun (like a postposition) and simultaneously before the verb, and which, in early Indo-European, were typically spatial adverbials (such as "up, down, away"). These forms are the antecedents of German (separable and inseparable) verbal prefixes. .ft I am grateful to J. Schindler for this information. .ft Tsunoda (1978:199) makes a comparison between preverbs in Jaru (and neighbouring languages) and the German "Vorsilben", i.e. separable (verbal) prefixes. Russian verbal prefixes are also similar -- Pesetsky (1979) argues that in morphological structure they are the sister of the inflected verb root (rather than being a sister of the bare root). Williams (1981:270) draws attention to compound Latin verbs like "ludificio" 'to make game, to play'. .sp English: .sp English phrasal verbs (verb-particle constructions) are similar to preverb-verb constructions in two ways (hence Alpher's (1973:403-6) and Capell's (1976:624) use of the term "phrasal verbs" for certain complex verbs in Cape York languages). A phrasal verb has a meaning that is not simply a composition of the verb and the preposition (particle); and the verb and particle may form what has been called a "discontinuous constituent" in that other words, such as pronouns, can intervene between the root and the particle. .sp .ne 3 Hindi: .sp Hindi forms compound verbs consisting of an uninflected verb root (called the "main verb") followed by an inflected verb root (called by some the "vector verb" or "auxiliary"). It is interesting to compare the semantics of this basic set of Hindi verbs with that of the Australian verb roots classified in section 3, and so I list them here. .sp Only a fraction of Hindi verb roots may be the second (vector) verb in a compound verb, and it is not completely clear which verbs are to be classified as possible vector verbs. All 23 authors listed by Hook (1974:19-20) in his comprehensive study of Hindi compound verbs agree that the following 6 roots are vector verbs (using Hook's orthography): .sp .in 10 .nf baiTh 'sit down' de 'give' jaa 'go' le take' paR 'fall' uTh 'get up, rise' .fi .sp .un and Hook (1974:119-20) would add the following 18: .br .nf aa 'come' cal 'walk, go' choR (de) 'leave behind' cuk 'run out, be used up; be paid off' Daal 'throw' dhar 'put; throw; keep' khar@ ho 'stand up' le jaa 'take away' maar 'strike; kill' mar 'die' nikaal 'take out' nikal 'come out; go out; turn out' paa 'find, get' (anti-vector) rah 'stay, remain' rakh (de) 'put down' sak 'can, be able' (anti-vector) .fi .in -10 .sp .ne 3 Georgian: .sp Merlan (1980a:10) points out a similarity between German verbal prefixes and the "preverbs" of Georgian, particularly "in earlier stages of the language". Georgian preverbs typically mark aspect; certain Warlpiri preverbs, such as the "quantifier" preverbs (section 2.4.1), have an aspect-marking function. .sp .ne 3 Algonquian: .sp Bloomfield (1940:103), no doubt taking the term from Indo-European studies, dubbed as "preverbs" the pre-verbal particles found in Algonquian languages of North America. .sp .ll 60 .in 10 Certain particles, _p_r_e_v_e_r_b_s, freely precede verb stems... /kees/ "completion," /pes/ (used only as a preverb) "hither"... Particles and even longer words are often _i_n_c_l_u_d_e_d between the members of such compounds: M[enomini] nekees-pes- [teh-wenah]-neewaaw "but I did see him on the way," with /teh wenah/ "however" included. .sp .ll 65 .in -10 Goddard (1979:60) assigns a structure to the complex verb, and he apparently faced analytical problems similar to those posed by Warlpiri preverb-verb combinations: .sp .in 10 .ll 60 The basic structure of a [Delaware] verb form may accordingly be represented by the following diagram: .sp .ne 4 .nf |prefix| | or |-(preverb(s) )root(+medial)+final-endings |change| ---basic stem------ -------compound stem------------ .fi (the plusses are stem-internal; the dashes separate the stem from the affixes; the space is a word boundary, though indicated by a hyphen in the transcription used [by Goddard].) .ll 65 .in -10 .sp In other words, Goddard places a stem-internal word boundary between the preverb and the following verb stem. He does this because "independent words may occur between a preverb and the rest of a compound stem" and "preverbs and basic stems--with a prefix or change, if any--are each treated phonologically as an independent word" (1979:59), yet the compound stem is the unit to which a prefix or change is applied. This analytical paradox is addressed in the next section. .sp 1.2 MORPHOLOGICAL STRUCTURE OF SUFFIXED COMPOUNDS .sp I propose to analyse Warlpiri compound verbs as consisting of a preverb compounded with an inflected verb stem, even if the preverb-stem combination merits its own lexical entry. Thus the part of a compound verb preceding its inflexional suffix, a unit I am calling a "theme", is not a morphological constituent of the complex word of which it is a part. .sp The situation is in these respects parallel not only to the Algonquian complex verbs mentioned at the end of 1.1, but also to certain English derived nominals as in the following example: (Cf. Levi 1978:66-74) .sp .ne 2 .nf [transformational grammar]-ian -----compound stem------- N .fi .sp The paradoxical nature of such a structure is that two words appear to make up one stem (which thus takes stem affixes). There is, however, another way of analysing these structures. In this view, the morphological structure is simpler, such as in: .sp .ne 2 .nf [transformational] [[grammar]ian] A N N ----related lexical entry----- .fi (where A=Adjective, N=Noun) and the meaning of the complex form, if it is to be related compositionally to the meaning of its parts, employs units which do not always correspond to morphological structure. Thus, in the above example, "transformational grammar" is not a morphological constitutent in "transformational grammarian", but it is involved (in some unspecified manner, relying on the fact that there is another lexical entry "transformational grammar") in deducing the meaning of "transformational grammarian". Which structure is justified for so-called adjective-noun compounds in English has been the subject of recent research in morphology (e.g. Levi 1978, Allen 1978), which is relevant here because the Warlpiri preverb-verb combination has a similar structure. Of course, "-ian" is a derivational suffix, but this does not affect the parallel with inflexional suffixes. English examples such as "white-washed" raise the same problem (Williams 1981). .sp It is difficult to see any alternative to having as the head of a lexical entry an entity, the "theme", which is not a morphological constituent of a word in which it occurs. The "theme" is readily identifiable within Williams (1981) theory of lexical relatedness: it is that part of a word remaining when the inflexion (the "head" of the word, in Williams' theory) is deleted. Williams shows how this resolves "# inside +" paradoxes in English and Latin, and also mentions (1981:279) the application to Russian and Warlpiri. In addition, Warlpiri phonology motivates the right-branching compound verb structure. Processes of vowel assimilation, and stress assignment, operate generally in domains beginning at each morphological "[" and extending rightward to the next "[" (passing over any number of "]") -- see Nash (1980) for exemplification, for instance in reduplications and nominal compounds. Given that an inflected compound verb has structure: .sp [preverb [root-inflexion]] .sp the phonological processes of vowel assimilation and stress apply without any additional restrictions on them. This would not be so if compound verbs had the structure: .sp [[preverb-root] inflexion] .sp See also section 2.3. .sp .sp .ne 3 2. PREVERBS IN WARLPIRI This section presents details of the preverb category in Warlpiri. An understanding of the ways preverbs combine with verbs is required to understand the following section which distinguishes Warlpiri verb roots on the basis of their interaction with preverbs. For further details, particularly of phonology and morphology, see Nash (1980) and the references there. .sp .ne 3 2.1 SEMI-PRODUCTIVE PREVERBS A typical Warlpiri preverb is /pirri/ 'scatter', occurring in such verbs as the following (with the meaning of the root given in square brackets after the meaning of the complex verb): .ft A further verb with preverb /pirri/ is /pirri-ma-ni/ 'ABS alight, set down'. If this involved the sense of /pirri/ being discussed, the root would need also to have the case-frame ABS, thereby ruling out /ma-ni/ 'ERG get ABS' and 'ERG cause ABS...', and leaving only /-ma-ni/ 'ABS make noise'. However, it is more likely a lexicalised combination: the Willowra dialect synonym is /tarda-ya-ni/, involving /ya-ni/ 'ABS move, go', and the /pirri/ in this particular stem has to be seen as merely homophonous with the distinct /pirri/ referred to in the text. Thus this verb occurs in combination with an adverbial preverb (see 2.4.1) in the textual example: .sp .nf .ne 2 Kala-lu yaarl-yaarl-pirri-ma-nu yalumpu-ju. Aux-they on top- 2 -alight-Past that-Part 'They set down over the top of those (men)' .fi .sp (Big Willy Japanangka, text, page 7; transcribed by Mary Laughren.) .ft .sp .nf pirri-kiji-rni 'ERG scatter ABS (by throwing ABS), ERG broadcast ABS' [throw] pirri-mati-rni 'ABS (eyes) bulge out' [go in line] pirri-nguna-mi 'ABS lie down individually, scattered' [lie down] pirri-panti-rni 'ERG pierce, squash ABS and send flying' [spear] pirri-parnka-mi 'ABS run away and scatter' [run] pirri-ya-ni 'ABS disperse' [go] pirri-yi-nyi 'ERG distribute ABS (to DAT)'[give] .fi .sp Another is /rdilyki/ (eastern alternant /tilykirr/) 'breaking', as in: .sp .nf .in 10 rdilyki-pi-nyi )'ERG break ABS' [act on] tilykirr-pi-nyi [E]) rdilyki-ya-ni 'ABS break' [go] rdilyki-paka-rni ) 'ERG smash ABS to pieces' [hit] tilykirr- [E] ) rdilyki-kati-rni 'ERG step on ABS breaking it' [step on] rdilyki-luwa-rni 'ERG hit ABS, breaking it' [hit] rdilyki-paji-rni 'ERG break ABS' [cut] rdilyki-panti-rni 'ERG break off (spear) in ABS' [pierce] .in -10 .sp and two further compounds which are apparently specialised: .sp rdilyki-kiji-rni 'ERG (horse) throws ABS' [throw] rdilyki-wangka-mi 'ABS (cooked meat) break into pieces' [speak] .fi .sp A third is /milki/ 'demonstrating, so as to give an example', as in: .sp .nf .ne 3 milki-purla-mi 'ABS demonstrate a call; ABS shout to make presence known' [shout] milki-purra-mi ) 'ERG light warning fire (ABS)' milki-yingki-rni) [burn; make fire] milki-wangka-mi 'ABS demonstrate speech' [speak] milki-yirra-rni 'ERG show ABS to DAT(?)' [put] .sp .fi As may seen in some of these examples, the preverb-verb combination sometimes has a more specialised meaning (optionally) than a simple composition of the meaning of the preverb and of the verb root. However, these "semi-productive" preverbs apparently occur with any verb root with which it would make sense to combine them, even if the root occurs in no lexical preverb combinations (see section 2.2), and for this reason I call them "semi-productive". Each has its own lexical entry, and is associated with a rule specifying how to combine its meaning (definition) with the meaning of a verb with which it is combined. .sp These preverbs may only occur with a verb stem which is a lexical entry: a verb root, or a lexical preverb-verb combination, but not a stem which is itself a (non-lexical) preverb-verb combination (cf. productive preverbs, 2.4); .ft The distinction drawn here may need modification. That is, there could be certain preverbs which can occur with one or two preverb-verb combinations, but, for no apparent semantic reason, fail to combine with other verbs. In other words, it may be possible to have a lexicalised preverb-preverb-verb combination, although the few candidates that have come to my attention have proved analysable in other ways. .ft Furthermore, like all non-derived preverbs, they do not take any nominal or verbal inflexion, and do not occur uncombined with a verb root. Examples include: .ft Note the /-pa/ augment (for a small number of morphemes, the augment is instead /-ku/), which must be present when the morpheme is an independent word (including when it hosts enclitics). .ft .nf .in 10 .sp jaa open (of entrance) jaalypa whispering jamparl chomping jurnpu pile up kanginy(pa) misperceiving karlirr(-karlirr) swerving larra splitting manyu in play milki demonstrating paarr(pa) into flight parri mashing pirri distributing, scattering raa clearing rdilyki ) breaking tilykirr [E]) rdungkurr dislodging, removing tiirl, tirl splitting, cracking turluny bent double walyi spilling wanyany adhering warurrka noisily wirrpiyi gathering (people) warlka feigning .sp .fi .in -10 An example of a semi-productive preverb in combination with a stem which is a lexical preverb-verb combination is: .sp .nf /kanginy-purda-nya-nyi/ 'ERG fail to hear ABS properly' (cf. /purda-nya-nyi/ 'ERG hear ABS') .fi .sp .ne 3 2.2 LEXICAL PREVERBS Some preverbs have an identifiable meaning, but occur with only a couple of verbs. An example is perhaps /ngirrily/ 'teasingly' as recorded in: .sp .nf .ne 2 ngirrily-nyina-mi 'ABS flit around (people (DAT))' [sit, be] ngirrily-ngirrily-wangka-mi 'ABS bait, talk sarcastically' [speak] .sp or /ngardaly/ 'turn over', recorded in: .sp ngardaly-kiji-rni 'ERG turn ABS over' [throw] ngardaly-wanti-mi 'ABS turn over, roll over' [fall] .fi .sp Others may occur with only one or two verb roots, and affect the meaning of the root to such an extent that it is difficult to assign an isolable meaning to the preverb. An example is /wurru/, which combines only with two roots: .sp .ne 2 .nf wurru-ka-nyi 'ABS stalk DAT, ERG stalk DAT' [transport, go with] wurru-marda-rni 'ERG keep ABS out of sight, ERG hide ABS' [have] .sp .fi Apparently /wurru/ has something to do with seclusion (cf. the adverbial preverb /wuruly(pa)/ 'into seclusion'). However, with our current knowledge of Warlpiri, there is no basis on which to predict that the notion of "cause another to accompany" will be played down or omitted from the contribution of /ka-nyi/ to /wurru-ka-nyi/, leaving its contribution as just that of a general verb of motion (though perhaps it is relevant to the sense that the movement of the stalker accompanies, albeit at a distance, the movement of the quarry). Hence /wurru-ka-nyi/ requires its own lexical entry: the combination /wurra-ka-nyi/ is the head of a lexical entry distinct from the lexical entry headed by /ka-nyi/, and there is probably no lexical entry headed simply /wurra/. .ft Nash (1980:50) incorrectly stated that the synonymous /wurru-ka-nyi/ and /yura-ka-nyi/ were the only occurrences of /wurru/ and /yura/. A second verb involving /wurru/ is given above; and /yura/ is also found in /yura-yura-ya-ni/ 'ABS go along out of sight'. .ft .sp Another example is /yirri/, found in: .nf yirri-ka-nyi 'ERG make ABS itch' [transport] yirri-kiji-rni 'ERG startle ABS; ERG set ABS onto DAT (e.g. to sick dogs onto) [throw] .fi .sp A lexical preverb is the only type of preverb which occurs as part of a lexical entry with a verb root -- all other preverbs are in some way "productive", and have their own lexical entries rather than occuring with particular roots in lexical entries. A number of the properties of lexical preverbs follow from this, e.g. (i) the fact that an Auxiliary may not occur between a lexical preverb and a root; (ii) the fact that a lexical preverb cannot occur following its associated root (or removed from it); (iii) the fact that the only preverb-verb combinations which occur with "semi-productive" preverbs are combinations involving lexical preverbs (section 2.1), which otherwise only occur with verb roots in their scope. .sp .ne 3 2.3 SEMANTIC COMPOSITIONALITY A preverb usually goes to form a compound verb which is a hyponym of the root verb -- the preverbs adds a component of meaning to the meaning of the root. Sometimes, as noted above, the combination may have a specialised meaning which is even more particular than the "sum" of the meanings of the preverb and verb root, but of course this does not detract from the combination from being a hyponym of the root. .sp But hyponymy does not always hold for lexicalised combinations: it can fail to hold in a variety of ways. If a root has multiple senses, the more general senses tend to show up in verbs compounded from that root. Often it appears that one component of the meaning of the root is "washed out", or "diluted" .ft I owe this term to Merlan (to appear). .ft leaving other aspects of the root's meaning intact. I illustrate this in 2.6 with a presentation of all complex verbs involving the root /yi-nyi/ 'to give'. Further examples include the one involving /ka-nyi/ mentioned in 2.2 above, and a number of verbs built on /nga-rni/ 'ERG ingest ABS' which "wash out" the component 'swallow' from its meaning, but retain 'act on by mouth'. .ft I am indebted to Mary Laughren for this observation, and the detailed semantic analysis of /nga-rni/ verbs. .ft For example, /jamparl-nga-rni/ 'to chew', or 'to crunch with mouth' are actions which do not necessarily involve swallowing. Similarly, /purda-nya-nyi/ 'ERG hear ABS' retains the notion of perception in the root /nya-nyi/ 'ERG see ABS'. .sp There are indications that failure of compositionality is not limited entirely to the lexical preverbs. For instance, a derived or productive preverb might enter into a lexicalised combination which is no longer a hyponym of the root verb (perhaps because the meaning of the combination has changed over time, or because the meaning of the root has changed). A case in point may be the root /karri-mi/ 'ABS stand' which enters into some combinations to which its contribution appears to be no more than "to be", as in: .sp .nf .ne 2 jaamalamala-karri-mi 'ABS yawn' rdilypirr-karri-mi 'ABS be bitten, wounded' (cf. rdilypirrpa 'hole, wound') rdumurdumu-karri-mi '1. ABS (heart) beat; 2. ABS worry about DAT' .sp manyu-karri-mi 'ABS play' .sp .fi It is interesting to note that Warlpiri Sign Language verbs apparently may involve the same sort of dilution of root meaning. Kendon (1980:107) records that: .sp .in 10 .ll 60 To sign "play" one signs a compound, the second sign of the compound being the sign for /karri-mi/. Here, thus, the root structure of a complex spoken expression is translated into sign simply by compounding the sign versions of the roots of which the compound is constructed. .in -10 .ll 65 .sp Most of the roots listed in section 3.1, but no others, are susceptible to this dilution and generalisation when they combine with preverbs -- I indicate in the list in section 3.1 which roots particularly evidence dilution. It is rare that the change in meaning is so great that the case frame of the preverb-verb combination does not contain that of the root verb. Leaving aside the superficial instances analysed as homophonous roots (/-nga-rni/ 'ABS move', and /-ma-ni/ 'ABS make noise'), the only examples are /wurru-ka-nyi/ 'ABS stalk DAT; ERG stalk DAT' (mentioned in section 2.2), and the handful of etymologically "incorporated object" verbs of bodily function. .sp In fact, certain "linking irregularities" (see Nash 1980:Chap.6) may be attributed to the persistence of the case-frame of the root verb, as is seen in the pair: .sp /kanginypa-nya-nyi/ 'ERG fail to see ABS' .br /kanginypa-karri-mi/ 'ABS fail to perceive DAT'. .sp From the meaning of these two verbs, and the semantic roles of the arguments, one would expect that these two verbs would have the same case frame -- probably 'ERG--ABS'. Yet, it seems, since /karri-mi/ has its subject argument linked to the 'ABS' case, the compound /kanginypa-karri-mi/ does also. .sp Note that there are few known instances of a difference in the linking of the subject argument between a verb root and a preverb combination involving that root. The few examples involve a root with ERG-linked subject, which occurs with a "lexical" preverb to form a compound verb with ABS-linked subject. .ft The other instances of differences between case-frames of root and compound result from the general addition of an argument to a case-frame (as with dative-adjunct preverbs), or merging of semantic representations (as with the infinitive preverbs discussed below). .ft The instances are as follows, with the frame of the root given in square brackets: .sp .nf .in 5 /kuumarri-yi-nyi/ 'ABS moan, groan' [ERG give ABS to DAT] /wurru-ka-nyi/) 'ABS stalk DAT' [ERG transport ABS] /yura-ka-nyi/ ) .in -5 .fi .sp The generalisation is also illustrated (negatively) by the pair of verbs just quoted, since both verbs involve a "semi-productive" preverb, not a "lexical" one, and thus the case-frame of the root persists in the compound. This situation should not be surprising: unless a preverb-verb combination merits its own lexical entry, its case frame (and meaning) is known only through general rules (albeit "lexical rules"). .sp There is another way to observe semantic composition of certain preverb-verb combinations: when the meanings (and case-frames) of two verbs fuse to form one. This happens in two ways in Warlpiri: .br (i) The Infinitive of any verb V may combine in preverb-like fashion with the root /ya-ni/ 'ABS go' (and certain other motion verbs, such as /parnka-mi/ 'ABS run') to form the compound verb /V-INF-ya-ni/ 'go along V-ing'. .br (ii) The Infinitive of one of the four verbs of stance may combine in a preverb-like fashion with the root /yirra-rni/ 'ERG place ABS (at ALL)' to form the compound verb /V-INF-yirra-rni/ 'ERG put ABS in V-ing position'. .sp Notice that in (i), the derived verb has the case-frame of the modifier, the verb V acting as preverb -- thus /paka-rninja-ya-ni/ 'ERG go along striking ABS' -- whereas in (ii) the verb has the case-frame of the head verb. In terms of case-frames, the combinations are as follows: .sp .nf (i) (ERG-)ABS(-DAT) + ABS --> (ERG-)ABS(-DAT) ERG-DAT + ABS --> ERG-DAT (ii) ABS + ERG-ABS --> ERG-ABS .fi .sp It is possible to make the generalisation that the argument of the "head" (finite) verb which bears the role of Theme (usually linked, as here, to the ABS case) corresponds to the subject argument of the "modifier" verb (the Infinitive-preverb) and takes the case of the subject of the "modifier", while the remaining arguments are not affected. The notion "subject" employed is the usual one in Warlpiri (that relevant to person-number marking, and infinitival control). .sp The generalisation extends, with respect to the "going" construction (i), to the Warlpiri verbs with case-frame ABS-ABS ("cognate object" verbs, see Nash 1980:196-97) or ERG ("incorporated object" verbs, see Nash 1980:201), and to the following lexicalised combinations: .sp [parntarri-nja]-wanti-mi 'ABS stoop' .br (related to /parntarri-mi/ 'ABS crouch', /wanti-mi/ 'ABS fall') .sp [yula-nja-ku]-ma-ni 'ERG make ABS cry' .br (related to /yula-mi/ 'ABS cry', Purposive /-ku/ and the Causative /-ma-ni/; pointed out to me by Ken Hale) .sp .ne 5 2.4 PRODUCTIVE PREVERB-VERB COMBINATIONS 2.4.1 ADVERBIAL PREVERBS The adverbial and dative-adjunct preverbs are distinguished by their ability to combine with a verb which is itself a non-lexical preverb-verb combination. But first note that they can, like semi-productive preverbs, combine with a lexical preverb-verb combination, as /jaala/ and /warru/ do in: .sp .nf jaala juurl-juurl-pi-nyi 'ABS jump back and forth' warru kupaly-(w)anti-mi 'ABS swarm, flock around' .sp .fi The greater combinatory power of adverbial preverbs is exemplified in the following sentence: .sp .nf Ngapa wiri-ngki-nganpa pina lani-ma-ni wirlinyi-kijaku. rain big -Erg -111o back afraid-Caus-Npast hunting-NegPurp 'The big rain will deter us from hunting' .fi .sp Here, the productively formed /lani-ma-ni/ (cf. /lani/ 'fear') is in combination with the adverbial preverb /pina/ 'back, in return'. Another example, this time with the adverbial preverb sentence-finally and separated from the verb, is: .sp Jinjin-(y)u-ngu-lpa-npa-ju ngaju-ku warrarda. -give-Past-Imp-2s-1o I-Dat always 'You told me [to get stuff for you] all the time.' Similarly, /jurnta/ (treated in 2.4.2) and /yarda/ in: .ne 3 .sp Pirlangkiti-ji jurnta-kuju-rnu yarda. blanket-me(Dat) -throw-PAst again 'He threw off (from me) my blanket again' .fi .sp The ability to combine with a complex verb apparently goes hand in hand with another property showing that the "adverbial" and "dative-adjunct" preverb is rather loosely connected with the verb stem with which it combines, viz. the ability of "adverbial" and "dative-adjunct" preverbs (but not the "semi-productive" or "lexical" preverbs) to follow rather than precede the accompanying verb stem, thus: .sp .li 2 [(preverb)[root-inflexion]] [preverb] V .sp In fact, these preverbs sometimes occur even further separated from the main verb, under conditions which I do not understand. As the example of /warrarda/ above shows, such separated preverbs, to the extent that they occur, often occur sentence-finally. .sp Other adverbial preverbs include: .ft Nash (1980:47-48) includes "semi-productive" and "adverbial" preverbs under the general label of "adverbial" preverbs. .ft .sp .nf .in 10 jaala back and forth jaantaku protruding jaarl blocking jangkardu in opposition to DAT japara eating while going along jarrwara wrong direction, astray jawirri V and leave behind jayirrpa snatching jiwirlki-2 with appendage bouncing juka jutting upwards julyurl(pa) in(to) water, fire juru-juru rolling, sliding juul(pa) up to a point, halfway laja carrying nganjini on arrival pina back, in return pirriki in pity pulpurru in opposition to DAT [Lander] rarra dragging rdalji motionless wajili running wapal(pa) V in search of wapirdi on arrival, approach of warlka feigning warru(kirdikirdi) around, in a circle wuruly(pa) in seclusion yaarl(pa) on, over top of DAT yungka blithely .sp .un and the quantifier preverbs: .sp jarnku, palju [Lander] each kutu anything, anywhere, anyhow puta some, partly muku all, completely warrarda always, consistently yarda more, again .sp .in -10 .fi The degree of productivity separating "semi-productive" from "adverbial" preverbs may not be anything more than a difference of degree, and may well be an artefact of incomplete data. At this stage we cannot do much better than observe that "adverbial" and "dative-adjunct" preverbs have been recorded in combination with verbs that are preverb-verb combinations, and that "semi-productive" preverbs have not; and that it is generally "semi-productive" (and "lexical") preverbs that enter into combinations whose semantics is not entirely compositional. .sp Adverbial preverbs, and preverbs in general, are distinguished from adverbs in two respects: .sp (a) the inability of preverbs to take any nominal inflexions -- an adverb, such as /yaruju/ 'quickly', is like a Nominal, in that it may take an Ergative suffix, as in: .sp .ne 2 Yaruju-rlu-ju paka-rnu. 'He hit me quickly.' quickly-Erg-me hit-Past .sp Note, however, that it may be that some adverbial preverbs listed above are also to be classified as adverbs, in which case the next property, (b), is the sole deciding one. .sp (b) the ability of preverbs to form a unit with the following verb root, which may be a pre-Auxiliary unit in the sentence, vs. the inability of a "true adverb" or any Nominal to do so (unless it acts as a derived preverb). .sp .ne 3 2.4.2 DATIVE-ADJUNCT PREVERBS This sub-category of preverbs is called "dative-adjunct" since its members introduce an extra argument, always linked to DAT case, in the functional structure of the verb with which they combine -- see Nash (1980:48-49,200-01,222) for details. A further distinguishing property is that most (the exception is /jurnta/) dative-adjunct preverbs cannot host a directional enclitic (whereas all other preverbs, except perhaps for some lexical preverbs, may host a directional enclitic). .sp One dative-adjunct preverb /jurnta/ 'away from DAT, removal from (with adverse effect on) DAT' is exemplified in a sentence quoted in 2.4.1. Another is exemplified in the following sentence, which also shows that a dative argument may be added to a clause by one of these preverbs even if the clause already contains a dative selected by the verb root: .sp .nf Ngarrka-ngku-rla-jinta karli karnta-ku man -Erg -Dat-Dat boomerang woman-Dat .sp kurdu-ku marlaja-punta-rnu child-Dat Causative-deprive-Past .sp 1.'The man took the boomerang away from the child, because of the woman'; 2. 'The man took the boomerang away from the woman, because of the child'. (The sentence has at least these two readings, according as to whether the dative-marked nouns are construed with the oblique argument of /punta-rni/ 'ERG take ABS away from DAT', or with the dative argument added by the preverb /marlaja/.) .sp The dative-adjunct preverbs are: .sp .in 10 .ne 2 .nf jirrnganja) Comitative, with (dependent) DAT yirrkirnpa) jurnta Adversative, away from DAT .ne 2 kaji [Yuendumu]) Benefactive, for DAT ngayi [E] ) .ne 2 marlaja Causative, made possible by DAT (marlangka -- restricted version) piki(piki) in jeopardy of DAT .fi .in -10 .sp In the languages to the north-west of Warlpiri with "particle-auxiliary" constructions, there is no equivalent of the dative-adjunct preverb, i.e. there are no particles which introduce an extra argument into the clause (F. Merlan, p.c.). .sp There are two adverbial preverbs mentioned above in section 2.4.1 which also introduce a dative argument, but which are not to be classified as true dative-adjunct preverbs. Not only do they host directional enclitics, but also the dative argument introduced by both these adverbial preverbs controls an Objective Complement (in /-kurra/) and not an Obviative Complement (in /-rlarni/) (a further distinction overlooked by Nash (1980:49n7)). .sp .ne 3 2.4.3 DERIVED PREVERBS Preverbs may be derived productively from Infinitives and Nominals in combination with a couple of verb roots: see Nash (1980:42-45) for an account. The Inchoative root /-jarri-mi/ and Causative root/-ma-ni/ combine productively with Nominals, and form a complex verb with a structure quite similar to that of a preverb-verb combination, e.g.: .sp .nf jukurr-ma-ni 'ERG dream about ABS(?)' (cf. jukurrpa 'dream') wikiny(pa)-ma-ni 'ERG taper ABS' (cf. wikinypa 'ledge, etc.') pinpin-ma-ni 'ERG flatten, smoothe ABS' .fi (cf. pinpinpa 'flat, smoothe') .ft It should be noted that those Nominals which require the augment /-pa/ also require it when combined with /-jarri-mi/: it can only be dropped in combination with /-ma-ni/. Thus /*pinpin-jarri-mi/ is apparently ill-formed, and only /pinpinpa-jarri-mi/ is well-formed. The explanation for this presumably has to do with their basically nominal status, as non-derived preverbs drop /-pa/ freely. .ft .sp The Inchoative and Causative may be separated by a directional enclitic from the nominal with which they combine. Normally a nominal may not host a directional enclitic, and the fact that it may just in this context supports the analysis of this construction as involving a denominal preverb. An example is the directional clitic /rra/ in: .sp .nf wiri-rra-jarri-nja-ya-ni big-forth-Inch-Inf-go-NPast 'it goes growing forth' .sp .fi These Nominal-Verb constructions are comparable to "noun incorporation", as found to a limited extent for instance in Yidin (Dixon 1977:465-73). Warlpiri also has a few instances of preverbs derived from nouns, other than the productive Inchoative and Causative combinations, and these fit the pattern of "object incorporation". Examples are all verbs of bodily function, such as: .sp .nf /kuntul-pi-nyi/ 'ERG cough' (cf. /kuntulpa/ 'cough' (N) /ngungkurru-pangi-rni/ 'ERG snore' (cf. /pangi-rni/ 'ERG scratch,dig ABS') .fi .sp Note the persistence of the ERG-linking of the subject of these two one-argument verbs -- compare the discussion in 2.3. .sp Infinitives as preverbs are discussed in section 2.3. The productive power of Infinitives and Nominals as derived preverbs is best seen as a property of those particular verb roots which enter such combinations -- hence they differ in the type of productivity from the combinations of 2.4.1 and 2.4.2. .sp .ne 3 2.5 ETYMOLOGY OF PREVERBS Preverbs are evidently a long-established category in Warlpiri, well integrated into the structure of the language. .sp The morpheme structure of preverbs conforms fairly closely to that of nouns (and words) in Warlpiri (see Nash 1980:67-81). There are two respects in which a preverb's form is less restricted than that of a nominal root in modern Warlpiri: (i) a preverb, unlike any other Warlpiri morpheme, may end in a consonant -- any coronal sonorant non-glide consonant; (ii) a few preverbs consist of a single syllable, either an open syllable with a long vowel, or a closed syllable with a short vowel -- both types not possible for nominal roots, but which were probably possible nominal shapes at an earlier historical stage. In general, the morpheme structure of a preverb conforms to the template: .sp .nf n CV(V)(C) (CV(C)) , n a non-negative integer .sp .fi A few preverbs may be derived from nominals which no longer exist independently, through a freezing of a derivational process mentioned in section 2.4.3. Some others are cognate with present-day verb roots, possibly from a common nominal source, as in: .sp .nf parnti-mi 'ABS smell, give off odour' parnti-nya-nyi 'ERG smell ABS, perceive by smell' (cf. the Nominal /parntirr-parntirrpa/ 'smell of corpse') .sp .fi However, it is also possible that preverb /parnti/ derives from a verb root. Compare verb derivation by stem-compounding, reported for Diyari's eastern neighbour Yandruwandha by Austin (p.c.), and attested by one compound Diyari verb (Austin 1981a:69). .sp A few preverbs may even be derived from something like stylised onomatopoeia. For example, /la(a)rr/ 'humming, whining', the only non-productive preverb recorded with the root /purla-mi/ 'ABS shout to DAT', giving /la(a)rr-purla-mi/ 'ABS hum, whine'; and /yapajaa/, the cry sometimes made by people in pain, is used in the complex verb /yapaja(yi)-ma-ni/ 'ABS groan in pain'. .ft Heath (1976:736-7) observes for Ritharngu that some "roots" (sc. preverbs) may be used as a contraction for the full verb. "The root form is usually pronounced emphatically, like an interjection, and is definitely expressive". Alpher (1969) had noted similar particles in Yirr'-Yorront, calling them "ideophones", and later "interjections" (Alpher 1973:58). .ft .sp .ne 13 2.6 EXAMPLES: VERBS IN /YI-NYI/ 'to give' .sp The root /yi-nyi/ 'to give' enters into numerous preverb combinations of a variety of types. Here I classify all recorded combinations according to the type of preverb, and demonstrate the varying degree of semantic compositionality (discussed in 2.3). This presupposes an analysis of the meaning of /yi-nyi/, which has not yet been achieved. I ask the reader's indulgence on this point, and adopt as a working definition the suggestion (due to Mary Laughren) 'X causes control over Y to pass to Z'. .sp .ne 3 2.6.1 PRODUCTIVE PREVERBS .sp The following are a sample of the combinations possible of /yi-nyi/ and the productive preverbs. Note that each combination is a hyponym of /yi-nyi/: .sp kutu-yi-nyi ERG give DAT anything at all nganjini-yi-nyi ERG give ABS to DAT on arrival yarda-yi-nyi ERG give DAT some more ABS pina-yi-nyi ERG give ABS back to DAT .sp .ne 3 2.6.2 SEMI-PRODUCTIVE PREVERBS .sp The first example of a semi-productive preverb in combination with /yi-nyi/ is probably a hyponym, and the second is not, the full sense of /yi-nyi/ is not present even though there is a controlled transfer involved. .sp pirri-yi-nyi ERG distribute ABS to DAT (/pirri/ generally means 'distributing, scattering') .sp rduul-yi-nyi ERG light ABS (fire), ERG fan ABS (fire) (/rduul/ generally means 'rapidly expanding' or the like, as in /rduul-pardi-mi/ 'ABS explode', /rduul/nga-rni/ 'ABS flare up') .sp .ne 3 2.6.3 LEXICAL PREVERBS .sp There are several sorts of lexical preverb found in combination with /yi-nyi/. .sp (a) First, those which form a hyponym of /yi-nyi/, and, it appears, involve a preverb derived, perhaps only in the historical sense, from a nominal: .sp .nf ngunyungunyu-yi-nyi ERG give DAT what one has so far (and go for more) (ngunyungunyu 'one load', ngunyungunyu-ngarrirni 'ERG advise, propose, suggest') .sp yajarri-yi-nyi ERG exchange ABS freely (with DAT?) (yajarri 'gift', yajarri-mi (with recip.) 'ABS exchange gifts') .sp yirdiyi-yi-nyi ERG make room for DAT (yirdiyi 'road, route, way, space') .sp jantajanta-yi-nyi ERG pass ABS around to DAT, ERG lend ABS, ERG share ABS (jantawarra 'distributed, for everyone', jantawarra-ma-ni 'ERG share ABS', jantalypa 'quartered kangaroo') .sp .fi (b) Second, those formed with a preverb which occurs only with one other root, namely /pi-nyi/ 'ERG act on ABS'. The combination with /pi-nyi/ apparently has the same meaning as the combination with /yi-nyi/. The meaning of the compound is not a hyponym of /yi-nyi/, unless a quite abstract transfer of control from one entity to another is conceived as being involved in the meaning of the combinations. .sp .nf jakuru-yi-nyi ERG announce one's own departure (to DAT?), ERG take leave (of DAT?) jinjin-(y)i-nyi ERG request favour of (ABS/DAT?) .sp kintil-(y)i-nyi ERG close incision in ABS with stick for purpose (/kintilpa/) .fi .sp There are two other preverbs found only with /yi-nyi/ and /pi-nyi/, but for these there is no connexion of meaning between the the combinations with the two different roots: .sp .nf marri-yi-nyi ERG brandish ABS, ERG feint with ABS (marri-pi-nyi '1. ERG covet ABS, 2. ERG bury ABS (deceased relative)' ) .sp yaru-yi-nyi (with recip.) ERG go in a big group (yaru-pi-nyi 'ERG wipe ABS, ERG spray ABS', yaru '1. covered with water, 2. light in weight' ) .sp .fi (c) Similar to (b), preverbs which are generally found only with /yi-nyi/ and /pi-nyi/, or only with /yi-nyi/. These all have the phonological shape: .sp .nf CV(C)CVrr-(y)i-nyi .sp .fi where the CV(C)CV element is often relatable etymologically to a nominal. The compound verb generally has two arguments (not three). The first argument, linked to ERG case, is typically a human acting in control on another entity represented by the second argument, linked to ABS case. The action generally involves an instrument (optionally expressible in the ERG/INSTR case), the type of which is inherent in the verb's meaning, and which is in some cases etymologically perceptible in the CV(C)CV element. .sp .nf palyarr-(y)i-nyi ERG anoint ABS [E] (palya 'wax, sticky substance'; palyarr-wanti-mi 'ABS fall and shatter') .sp jurdurr-(y)i-nyi ERG cover ABS (jurdu 'dust, dust cloud', jakarr-pi-nyi 'ERG cover ABS') rdilypirr(y)i-nyi) ERG make hole in ABS, [E] tilypirr(y)i-nyi) ERG pierce ABS .sp pirlirr-(y)i-nyi) ERG beat in 1 1/2 time pirlirr-pi-nyi ) (Warlmanpa: pirlipi 'clap sticks') .sp purlturr(y)i-nyi) ERG stick spear part way in ABS purlturr-pi-nyi ) .sp yimirr-(y)i-nyi) ERG frighten ABS with empty yimarr-pi-nyi ) talk or noise (yimi 'speech, message') .sp .fi Two further preverbs which fit with the ones in this subsection, but which occur with a larger number of other verb roots are: .sp .nf rdungkurr-(y)i-nyi ERG dislodge ABS rdanjarr-(y)i-nyi ERG provision ABS .fi .sp Both have a discernible meaning: /rdungkurr/ 'dislodging' (found with /kiji-rni/ 'throw', /yirra-rni/ 'put', etc.), and /rdanjarr/ 'a lot of' (found with /ka-nyi/ 'carry', etc.). .sp One line of speculation about the origin of these /rr/-final preverbs is to relate them to the 1st conjugation roots in /rri-mi/ listed in 3.3(a), and discussed in 3.4, at least in proposing that there was once a denominal verbalising suffix */-rri/ which formed verbs of the "zero" conjugation (the fore-runner of the present-day 1st and 3rd conjugations -- see Hale n.d.). In combination with certain nominals of form CV(C)CV, at least, the */-rri/ suffix form a transitive "applicative" verb, with the meaning that the instrument represented by the given nominal be used characteristically. Thus, at an earlier stage, a verb root such as: .sp *palya-rri 'to use /palya/' .sp existed. These underwent reanalysis, given the phonological, morphological and semantic possibility of reinterpreting the form as being derived from /yi-nyi/, and thus created a preverb morpheme such as /palyarr/ which subsequently became available for combination with certain other verb roots. Note that vowel assimilation does not penetrate the preverb in modern Warlpiri -- so that /rdilypirr-(y)i-nyi/ has Past tense /rdilypirr-(y)ungu/ not */rdulypurr-(y)ungu/. .sp Naturally, this proposal calls for further supporting evidence, and an account of the relation, which appears problematic, of the postulated */-rri/ to other suffixes reconstructed as */-rri/. It is offered at this stage as a suggestion to explain the otherwise remarkable set of preverbs just listed. .sp (d) Finally, preverbs which occur only with /yi-nyi/, and which do not form a hyponym of /yi-nyi/, but nevertheless generally form a transitive verb with ERG-linked subject which is typically a human in control of an activity. .sp juta-yi-nyi ERG hit ground to determine presence of yams rdingan-yi-nyi ERG curse ABS by boning rdalpu-yi-nyi ERG carry ABS under arm, as coolamon julyamarda-yi-nyi ERG throw ABS (dirt) up (at DAT?) wakawaka-yi-nyi ERG use arms and kill ABS kuru-yi-nyi ERG warm ABS over fire kardi-yi-nyi ERG heat ABS (also kardu-yi-nyi, cf. kardu-ma-ni 'ERG manufacture ABS') .sp rdirri-yi-nyi ERG start ABS, ERG start to V-INF-Purposive (also found in /rdirri-yarnka-mi/ 'start on journey') .sp A further, probably related, set of examples is: .sp .ne 4 mirrmirrparlu-yi-nyi) ERG make ABS shine mirrmirrmarli-yi-nyi) mirrmirrmarli-yi-nyi ERG (cold weather) cause ABS to shiver .fi .sp .ne 3 2.7 SUMMARY OF PREVERB PROPERTIES .sp The types of Preverbs and their various properties discussed in section 2 are as follows: .sp .nf I II III IV PREVERB TYPE directional? Aux? V-PVB order? V/PVB separate? INF-ya-ni yes no no no N-jarri-mi) yes mostly no no N-ma-ni ) not N-V yes yes yes (yes,irrelevant) semi-productive yes yes no no lexical no? no no no adverbial yes yes yes sometimes dative-adjunct no yes yes sometimes .sp .fi .ne 3 Abbreviations of properties: .sp -directional?: Can the preverb host a directional enclitic? .br -Aux?: Is the order PVB-AUX-V possible? (i.e., can an Auxiliary occur between preverb and verb). .ft The Auxiliaries observed to occur in the intervening position all have zero "complementiser" and monosyllabic or zero "base" -- i.e. they begin with /ka-/ 'Present', /-lpa-/ 'Imperfect', or consist only of person-number clitics. See Nash (1980:chapp.2,7). .ft .br -V-PVB order?: Can the preverb occur immediately following (as well as preceding) the verb it is combined with? .br -V/PVB separate?: Can the preverb be separated from the verb (by words of the sentence, other than by directionals and the Auxiliary)? .sp .ne 3 3. CLASSIFICATION OF WARLPIRI ROOTS All the known Warlpiri verb roots are listed in this section. They are arranged according to a gross property of combination with preverbs, viz. the degree to which they occur in lexical preverb combinations (as described in the previous section). Beyond this there are, of course, many finer classifications not considered here, at least within groups 3.1 and 3.2, according to which roots combine with a particular semi-productive preverb -- for instance, /pirri/ 'scatter' selects verbs with a certain motional component in their meaning. .sp The citation form of the root is the non-past inflected form, which suffixes to indicate conjugation membership (except that /nga-rni/ is the sole member of the 4th conjugation and not a member of the 2nd conjugation). The gloss assigned is extremely skimpy, and is designed to give a rough idea of the basic meaning of the root, and to present the case-frame(s) it selects. .nf .in 10 .sp .ne 4 .un 10 3.1 ROOTS WITH MANY PREVERB-DERIVED STEMS >janka-mi ABS burn ERG (fire) burn ABS jarri-mi ABS become ... >karri-mi ABS stand nguna-mi ABS recline >nyina-mi ABS sit >pardi-mi ABS arise parnka-mi ABS run purra-mi ERG (person) burn ABS wangka-mi ABS speak, ABS make characteristic noise ABS speak to DAT wanti-mi ABS drop wapa-mi ABS move about -jirri-rni ERG act forcefully on ABS kiji-rni ERG throw ABS ERG put decoration ABS on DAT luwa-rni ERG hit ABS with missile marda-rni ERG hold ABS >ngarri-rni ERG scold ABS ERG tell ABS to DAT ERG signify ABS paji-rni ERG cut ABS paka-rni ERG strike ABS panti-rni ERG pierce ABS -parri-rni ERG act on ABS yirra-rni ERG place ABS at LOC,ALL ERG create design ABS on DAT >ka-nyi ERG transport ABS >nya-nyi ERG perceive ABS ERG look about for DAT >pi-nyi ERG impact on ABS >yi-nyi ERG give ABS to DAT >nga-rni ERG ingest ABS ji-ni ERG scold ABS ma-ni ERG get ABS -ma-ni ERG cause ABS become... >ya-ni ABS go .sp .in -10 > -- these roots are often diluted in meaning in combination with a preverb (see section 2.3) .sp .ne 4 3.2. ROOTS WITH A FEW PREVERB-DERIVED STEMS .in 10 jiti-mi ABS descend kampa-mi ABS burn ERG (fire) burn ABS karrka-mi ABS proceed kulpa-mi ABS return home -para-mi ERG follow ABS pura-mi ERG follow ABS purla-mi ABS shout to DAT -yalpi-mi return (from hunting) and... yarnka-mi ABS start on journey ABS grab for DAT yula-mi ABS cry kati-rni ERG weigh down on ABS pangi-rni ERG scratch ABS parda-rni ABS wait for DAT -pirri-rni ERG act on ABS... punta-rni ERG take ABS away from DAT yingki-rni ERG set fire to ABS yurrpa-rni ERG grind ABS -nga-rni ABS move -ma-ni ABS make noise .sp .ne 6 .in -10 3.3. ROOTS WITH NO DERIVED STEMS (COMBINE ONLY WITH PRODUCTIVE PREVERBS) .fi Some roots combine with one preverb which may not have "productive" status: such preverbs are given in square brackets after the root concerned. .sp .ne 6 (a) FROZEN DERIVED FORMS (see 2.6.3(c), 3.4) .sp POLYSYLLABLE+rri-mi .sp .nf .in 10 kawarirri-mi ABS wander ngarlarri-mi ABS laugh ngurntirri-mi ABS scold DAT parntarri-mi ABS crouch [maarr- 'crouch flat'] wapirri-mi ABS conceal DAT yajarri-mi (+recip) ABS exchange gifts yurirri-mi ABS stir .sp .un 10 .ne 5 (b) FROZEN PREVERB-VERB COMBINATIONS (see 3.4) CVC+pi- (cf. root /pi-nyi/, in 3.1 above) karlpi-mi ERG (temperature) cause suffer ABS tirlpi-mi ERG flake ABS jampi-rni ERG lick ABS marnpi-rni ERG touch ABS parnpi-rni ERG touch ABS wirrpi-rni ERG eat ABS [avoidance] [yilyi- 'slurp'] yarrpi-rni ERG build fire ABS yirrpi-rni ERG insert ABS .sp 2 .ne 3 CVC+ji- (cf. root /-ji-ni/, 3.1 above) larrji-rni ERG scratch ABS nyunji-rni ERG kiss ABS parlji-rni ERG wash ABS rdirrji-rni ABS start a fight walji-rni ERG pluck ABS winji-rni ERG pour ABS .sp 2 .ne 3 OTHERS waraparnpi-mi ABS sing out ABS announce DAT wuurlparra-rni ERG dehair ABS tirlarri-rni ERG split ABS .sp .in -10 .ne 3 (c) DERIVED ROOTS .sp The following two roots also occur as preverbs, and have etymologically related nominals: .sp .in 10 pali-mi ABS die [jipirr-jipirr- 'flicker'] parnti-mi ABS smell .in -10 .sp The following two roots are the only ones in the third conjugation which consist of more than one syllable. The element */ra-/ indicates an additional ancient root of that conjugation (Hale n.d.). The eastern Warlpiri /mapara-nyi/ corresponds in meaning to /mapa-rni/ in other dialects, and is probably related to nominal stems of the form /maparnpa/ 'healing powers', etc. found in various Australian languages. .sp .ne 2 .in 10 mapara-nyi ERG anoint ABS [E] yurlpara-nyi ERG send ABS [E] .sp .ne 3 .un 10 (d) POSSIBLE RECENT LOANS FROM ARANDIC -yalpi-mi (see section 3.2) yampi-mi ERG leave alone ABS ERG reject ABS yarlki-rni ERG bite ABS yilya-mi ERG send ABS to DAT .sp .ne 4 .un 10 (e) OTHER ROOTS WITH NO LEXICAL PREVERB COMBINATIONS .sp japirdi-mi ABS threaten DAT jarnti-mi ABS limp jija-mi ABS succumb to DAT jirrti-mi ABS hang out kapati-mi ABS be uneasy in DAT karli-mi ABS flow [jirrmily(pa) 'tear'] lirri-mi ABS swell nyurla-mi ERG knead ABS papi-mi ABS ignite rdipi-mi ABS gather rdipi-mi ABS encounter DAT turlka-mi ERG pinch ABS wipi-mi ABS radiate wirli-mi ERG poke ABS wirnpirli-mi ABS whistle wirnti-mi ABS (women) dance yirdi-mi ABS be frightened yuka-mi ABS enter yulka-mi ABS cherish DAT .sp jaja-rni ERG eat off ABS japi-rni ERG ask ABS about DAT jarnti-rni ERG scrape ABS jiti-rni ERG tease ABS kardi-rni ERG fetch, get ABS (water) kipi-rni ERG winnow ABS kulpa-rni use maja-rni ERG straighten ABS mapa-rni ERG anoint ABS mati-rni ABS go in procession mila-rni ERG select (best one of) ABS mirri-rni ERG erase ABS ngaja-rni ERG void ABS nganti-rni ERG erect ABS ngarlki-rni ERG block ABS ngarrmi-rni ERG ritually increase ABS paja-rni ERG taste ABS parnta-rni ERG withdraw-from-fire ABS payi-rni ERG ask ABS about DAT pirrki-rni ERG trim ABS rdanpa-rni ABS accompany DAT wanja-rni ERG take ABS (sip) off DAT wardi-rni ERG straighten ABS wari-rni ERG tie ABS warri-rni ERG seek DAT warrka-rni ABS climb up, ERG ride ABS yaja-rni ERG enlist ABS yarli-rni ERG soak ABS yipi-rni ERG pick out pimple ABS yingki-rni ERG set fire to ABS yirnti-rni ERG capsize ABS yunpa-rni ERG sing song ABS .in -10 .fi There are a number of verb stems in Warlpiri which are synchronically analysable as preverb-verb combinations, but which show some phonological amalgamation, typically cluster simplification at the preverb-root boundary (see Nash 1980:79), and possibily extra stress reduction on the root. Examples include a subclass of verbs with root /yi-nyi/, namely those with /rr/-final lexical preverbs, listed in 2.6.3(c), such as /rdilypirr-(y)i-nyi/ 'ERG wound ABS'. Note, for instance, that the Past inflexion of /rdilypirri-nyi/ is /rdilypirrungu/, not */rdulpurrungu/ -- the combination has not undergone the reanalysis discussed in 3.4. Other similar combinations are those with glide-initial roots /wapa-mi/, /ya-ni/, /wanti-mi/. .ft A more solidified, reanalysed combination is that of an Infinitive with what was the verb root /*yan-/ 'to go', which is undoubtedly the historical source of the modern Inceptive and Progressive derivational affixes -- see Nash (1980:41). This is different from the combination of /ya-ni/ with an Infinitive preverb discussed in 3.4. .ft .sp .ne 3 3.4 ETYMOLOGICALLY COMPLEX ROOTS Within group 3.3 there are verb roots which are apparently etymologically complex, but in modern Warlpiri are unanalysable roots. (Compare the analysable roots mentioned at the end of section 3.3.) .sp In particular, in 14 of these roots (but in none of those of groups 3.1 and 3.2) it is possible to identify two monosyllabic root-final formatives (*-pi and *-ji) which are probably related to the modern transitive roots /pi-nyi/ and /ji-ni/. .ft As far as I know, Ken Hale was the first to make this suggestion; he drew my attention to it with the example /tirlpi-mi/. .ft Supporting evidence includes: .sp (i) the case-frames of the relevant modern roots are ERG-ABS, even the two which are in the predominantly intransitive conjugation (viz. /karlpi-mi/ and /tirlpi-mi/), with 1 exception out of 14 (/rdirrji-rni/); .br (ii) the non-final part of the root has phonological shape CVC, where the second C is [+son,+cor, +cons] (except in /jampi-rni/ 'ERG lick ABS' -- possibly involving epenthesis or assimilation, and note common Australian /jawa/ 'mouth'), i.e. a possible word-final consonant at an earlier stage of the language (before /-pa/ augmentation -- see Hale 1973b); and the only /rrj/ and /rlp/ clusters in Warlpiri verb roots are confined to these 14 roots. .sp Notice that the roots just discussed are the historical product of reanalysis. The reanalysis brings about a congruity between morphological and semantic structure: the erstwhile "theme" (a semantic unit which corresponded to no morphological constituent) is reanalysed as a simple root (which is both a meaning unit and a morphological unit). .sp .ne 3 .nf *[tirl] [pi-TENSE] [tirlpi-TENSE] --stem-- ==> ---stem----- ---theme-- .fi .sp A third formative evident in group 3.3(a) in */-rri/, which is probably related to the derivational affix of the same form in the southern Pilbara languages, Western Australia, (where it has the function of a Passive marker -- Austin 1981b), and the wider range of possible cognates gathered by Sutton (1976:303-4) including forms in /-rri-/ which mark Inchoative and Middle (Refl/Recip). Evidence includes: .br (i) the 7 roots are all intransitive (i.e. have a subject in the Absolutive (unmarked) case); .br (ii) the non-final part of the root has phonological shape CV(C)CV, the most common word shape; .br (iii) the only textually attested verbal reduplications (Nash 1980:136-47) which copy part (not all) of a verb root are those which copy the first two syllables of /ngarlarri-mi/, /parntarri-mi/, giving /ngarla-ngarlarri-mi/, /parnta-parntarri- mi/ respectively (though /wirnpi-wirnpirli-mi/ may also be well-formed); .br (iv) the only verb roots which contain two different high vowels (i.e. both /u/ and /i/) are /ngurntirri-mi/, /yurirri-mi/ (~/yururri-mi/) and /nyunji-rni/), all three of which are proposed to be etymologically complex. (The root /winji-rni/ has been recorded in Past inflexion as /winjurnu/ as well as the usual /wunjurnu/. However, /winjurnu/ violates the Warlpiri stem prohibition "*/iCu/ unless C is /p,w/", whereas /nyunjirni/, etc. run counter to a generalisation which holds only verb roots.) .sp At what may well be a greater etymological depth, it is possible to perceive one more verbal formative in Warlpiri -- the /*na-/ of stance verbs isolated by Merlan (to appear). There are only two verb roots in Warlpiri that have this final syllable: /nyina-mi/ 'ABS sit, be' and /nguna-mi/ 'ABS recline'. Further, these two roots are the only roots of the /-mi/ conjugation ("1st", in Warlpiri literature, properly the "zero"-conjugation) which take an Imperative and Irrealis other than in /-ya/ -- these two roots prefer the Imperative in /-ka/ (and Irrealis /-karla/), the ending of the "2nd" conjugation, and of inceptives formed from 3rd conjugation roots (Nash 1980:41). .sp .ne 4 4. COMPARISON OF VERB-ROOT INVENTORIES The Appendix contains Tables 1-4 presenting all the known verb roots of four Pama-Nyungan languages of the so-called Western Desert type in the west-central Northern Territory and adjacent Western Australia: another member of the Ngarrka subgroup Warlmanpa, and three languages of the Ngumbin subgroup Jaru, Gurindji, and Walmatjari. .sp The roots in the Tables in the Appendix should be compared with each other, and with the list of Warlpiri roots in 3.1. The reader will see that the majority of the roots have cognates, and the majority of those that do not nevertheless have a corresponding root with fairly similar meaning. .ft The covert classification of Warlpiri verb roots as presented in section 3 strictly should also be applied to the languages with which I am comparing Warlpiri. To do so requires a considerable wordlist of preverb-combinations of various types, and this is not available to me for the neighbouring languages. It is possible to report at this stage that half the verb roots of Warlmanpa have not been recorded in non-productive preverb-verb combinations, including those which for other reasons one might suspect are a comparatively recent development in the language; and there is a similar covert classification of the Walmatjari roots (for instance, /yinpa-/ and /marra-/ do not occur in preverb-verb compounds -- Hudson 1980:2). Thus the cline of verbal inventory size is repeated, though less steeply, for the covert "core root" inventory. .ft .sp There are 13 verb roots found in all five languages, and there a further 5 roots found in four of the five languages, shown in Table 5. .sp .ne 6 .ce TABLE 5 -- COMMON VERB ROOTS .nf gloss Walmatjari Jaru Gurindji Warlmanpa Warlpiri burn kampa- kampa-rn kampa kupa-ka kampa-mi (hit luwa- luwa-rn luwa- la-ka luwa-rni (spear la-nta la-n bite paja-rra paya-n paya- piya-ka paji-rni sit,stand karri- karri-rn karr-y ka-rra karri-mi fall wanti- wanti-ny wan-y wa-nta wanti-mi say ma-nyja ma-ny ma-ny -ma-nta -ma-nta eat nga-nyja nga-ny nga-ny nga-nyja nga-nja give yu-ngka yu-ng jayi-ng yu-ngka yi-nyi carry ka-ngka ka-ng ka-ng ka-ngka ka-ngka see nya-ngka nya-ng nya-ng nya-ngka nya-ngka hit pu-ngka pu-ng pu-ng pu-ngka pi-nyi get -ma-nta ma-n ma-n ma-nta ma-ni go ya-nta ya-nta ya-n -ya-nta ya-ni .sp scold ---- ju-n ju-n jin-ka ji-ni put ---- yaa-n yuwa-rr ya-ka yirra-rni void ngaja-rra ---- ngaya-rr ngaya-ka ngaja-rni sing yinpa- yunpa-rn yinpa- ---- yunpa-rni leave wanyja-rra warnta-n wanyja-rr winyja-ka ---- .fi .sp There are also 4 roots found in 3 of the 5 languages: .sp .nf .ne 4 gloss Walmatjari Jaru Gurindji Warlmanpa Warlpiri cut ---- kuma-rn kuma- kuma-ka ---- cry lu-ngka lu-ng ---- lu-ngka ---- scratch ---- pangi-n ---- pangi-ka pangi-rni wet ---- kurnta-n kunyja- kinyja-ka ---- .fi .sp The roots in Table 5 account for the following fraction of all verb roots in each language: .sp .ne 3 .nf Walmatjari Jaru Gurindji Warlmanpa Warlpiri 16 17 18 17 17 42% 44% 60% 43% 16% (55% of 3.1) .fi .sp Notes: .br 1. There is sometimes a variation in meaning from language to language, as is evident in the slightly fuller glosses given in the Appendix. .br 2. Warlpiri has /paja-rni/ 'taste' and /paji-rni/ 'cut, bite', and it is probably the former which is cognate with the root 'bite' in the other languages. .br 3. Warlmanpa /jin-ka/ means 'burn', not 'scold', but I have listed it with roots of similar form above. .br 4. Certain differences between the languages are the result of variations in citation form: for instance /pu-ngka/ is the Imperative of Warlpiri /pi-nyi/ (the NonPast citation form); /ji-ni/ has Past /ju-nu/, etc. .br 5. A further member of the Ngumbin subgroup, Mudbura, also appears to have verb roots corresponding to those in Table 5. Since the sources of information on this language are much less extensive, I have not included it in the table. .sp If the exercise is repeated with the the Warumungu roots given by Simpson (1980), or the Pintupi/Luritja roots given by Hansen & Hansen (1977), the number of apparent cognate roots is smaller (even though these two languages are also of Western-Desert type, and have fifty or more verb roots). Specifically, of the 18 most common roots listed above, 7 are found in Pintupi (kampa-, ma- 'get', nya-, patja-, pu-, ya-, yu-) and only about 4 in Warumungu (nya- 'see', pu- 'bite; dance', -ma- '(unproductive trans. suffix)', yungku- 'burn (intr.)'). .ft This observation may hold implications for genetic subgrouping: the current classification has three co-ordinate Western Desert Type subgroups for these languages: Wati (Pintupi, etc.), Ngarrka (Warlpiri, Warlmanpa), and Ngumpin. The evidence of these verb roots links Ngarrka more closely with Ngumpin than with Wati. .ft However, the meanings of the common roots, if not their form, usually correspond to roots (rather than compounds) in Warumungu. .sp It is evident that Warlpiri's verbal inventory differs much less from that of the related Western-Desert type languages than appears from a simple count, at least with respect to those of its own subgroup (Ngarrka), and the one to the north and north-west (Ngumpin). When one makes the comparison instead with the covert class of Warlpiri verb roots which are presumably of more ancient vintage in the language, viz. the "core roots" of 3.1 above, which have entered into a number of lexical preverb-verb combinations, then the similarity between Warlpiri and its neighbours increases. Conversely, we see that there is an historical explanation for the diversity of combinatory frequencies of different Warlpiri verb roots. .sp It is also interesting to compare the meanings of the recurring roots with those of the 24 compounding Hindi roots, listed at the end of section 1: about half correspond. .sp .ne 3 5. CONCLUSION .sp My proposal for history of certain frozen Warlpiri combinations in sections 3.3(a) and especially 3.3(b) is the Warlpiri parallel to Merlan's (1980b, to appear) proposal for certain Pama-Nyungan languages, particularly Yidin and Dyirbal. She has shown that a number of disyllabic verb roots have second syllable putatively cognate with an ancient monosyllabic verb root. Donaldson (1976:767) had suggested a parallel history for certain verbalisers in Ngiyambaa, and Dixon (1980:418-9, but see 256) suggested for Pama-Nyungan verbs that "perhaps the majority of modern polysyllables" are fused compound structures. Thus the Warlpiri situation bears out Merlan's observation that an archaic complex verb structure comparable to those productive today in many non-Pama-Nyungan languages is discernible in the history of many Pama-Nyungan languages. .sp The reanalysis evident in the history of these etymologically complex verbs is of the type that minimises disparity between morphological structure and semantic units. The change is in the direction of restoring compositional semantics of complex words. It appears that such a disparity, as in the structure: .sp .nf [preverb [root-inflexion] ] ----theme---- .fi .sp is not uncommon (Williams 1981), but that in instances where the "theme" requires its own lexical entry, with properties separate from those of the root and preverb, it is not uncommon for reanalysis to give rise to the structure: .sp .nf [[preverb-root] inflexion] .fi .sp and for there to thus arise a new root. .bp .ce APPENDIX: VERB ROOTS IN NEIGHBOURING LANGUAGES .sp The verb roots of four Pamam-Nyungan languages to the north and west of Warlpiri are given in this appendix. .sp As for the Warlpiri roots in section 3, the glosses given to the verb roots in these languages are extremely skeletal, and serve no more than as a hint at the meanings of the root. .sp Where known, a case-frame is provided, using the abbreviations: .nf .sp ERG - Ergative ABS - Absolutive DAT - Dative LOC - Locative ("Accessory" in Walmatjari) .fi .sp Some roots (including all those which have case-frames both with and without an ERG) are given more than one gloss, and thus the root appears more than once (once for each gloss. .sp The conjugation membership of each root is indicated in the given suffix. .sp Table 1 gives the roots of Gurindji (Kuurrinyji) with an English gloss on the same line. For Gurindji, the suffix is a particular consonant (or zero) which is an abstract conjugation marker. The roots are basically as provided by Patrick McConvell (p.c., and to appear) and Norman McNair (p.c.), and I have added case-frame information and slightly amended some glosses in the light of data from Hale's Gurindji Notes (1959). .ft Norman McNair kindly provided me with a list of the 26-7 (eastern) Gurindji verb roots that he has recorded at Wave Hill (Kalkaringi and Dagaragu), and all but one occur in the eastern Gurindji list provided by McConvell. McNair identifies some of the roots provided by McConvell as being from the westernmost Gurindji dialect (Malngin). I have indicated dialectal information in square brackets. .ft .sp Note the generalised use of /luwa-/ in the "avoidance language" of Gurindji (McConvell to appear:19). Used in the avoidance vocabulary, /luwa-/ thus has any case-frame, and, as Tsunoda (1978:219) observes for Jaru, where /luwarn-/ is also the general avoidance verb, "this verb can be combined with _a_n_y preverb." .sp Table 2 gives 36 of the 39 Walmatjari verb roots recorded by Joyce Hudson. It is based on her list (1978:43) and I have added the case-frame markers to the glosses from information in her discussion of transitivity. The Imperative (Hudson's Irrealis) form of the verb is given to show conjugation membership. .sp The Jaru roots listed by Tsunoda (1978:85) are in listed Table 3. I have recast the forms into the Warlpiri orthography. Tsunoda (1978:3) refers to two dialects: "N" Nyininy (to the east, neighbouring Malngin) and "W" Wawarl (to the west, neighbouring Walmatjari). .sp .nf .ce TABLE 1 -- GURINDJI VERB ROOTS .in 10 ja-n ERG copulate with ABS [W?] jama-rr ERG grind ABS (seeds) with stone jayi-ng ERG give ABS to DAT jiya- ABS burn jiya- ERG deprive of ABS ju-n ERG scold ABS ka-ng ERG take ABS kampa- ERG cook ABS karr-y ABS sit, be karrwa- ERG have, keep ABS kaya-rr ERG kick ABS kuma- ERG cut ABS [Malngin] kunyja- ERG spray ABS luwa- ERG hit ABS with missile [Malngin] general avoidance verb ma-n ERG get ABS ma-ny ABS say minyja- ERG wet ABS nga-ny ERG eat ABS ngaya-rr ERG void ABS nya-ng ERG see ABS pa-rr ERG hit ABS papa-rr ERG weigh down ABS paya- ERG bite ABS pu-ng ERG act on ABS [wu-ng W?] tuma- ABS grow wan-y ABS fall wanyja-rr ERG leave ABS wuya- ERG throw ABS [?W only] ya-n ABS go yunpa- ERG sing ABS yuwa-rr ERG put ABS .bp .in -10 .ce TABLE 2 -- WALMATJARI VERB ROOTS -jarri- ABS become ... -ka-rra ERG [place] ABS -kuji- ERG cause ABS to ... -ma-nta ERG [..do..] ABS jula- ERG tell, call ABS kampa- ERG cook ABS ka-ngka ERG carry ABS kanyji- ERG stamp, tread on, dance ABS karla- ERG dig, scrape ABS karri- ABS stand karrpi- ERG tie, bind ABS kirra-nyja ABS sit la-nta ERG pierce, spear ABS lu-ngka ABS cry, ERG cry for DAT luwa- ERG hit ABS with missile ma-nyja ABS speak, say; ERG speak to DAT manyji- ABS burn; ERG burn ABS mapa- ERG spread on, rub ABS marra- ERG scold ABS, talk angrily, argue nga-nyja ERG eat ABS ngaja-rra ERG give birth to ABS nguna- ABS exist, be nya-(ng)ka ERG see ABS paja-rra ERG chop, bite ABS papaja-rra ABS call out pu-ngka ERG hit ABS with implement, hand ruka-nyja ERG forget LOC tarra-ngka ERG throw ABS turta-ngka ABS get up (also turta-pu-ngka) waa-ngka ERG follow ABS wanti- ABS fall wanyja-rra ERG leave behind ABS warnta- ERG get ABS ya-nta ABS go yinpa- ERG sing, curse ABS yu-ngka ERG give ABS to ABS yuka-nyja ABS lie down [and two others in the "zero" class] .bp .ce TABLE 3 -- JARU .in 10 jampun ERG kiss ABS jan ERG copulate with ABS jun ERG scold ABS kamparn ERG burn ABS kang ERG carry ABS karrirn ABS sit (N only) karrun ERG hold ABS kayan,kayin(N) ERG kick ABS kumarn ERG cut ABS kurntan ERG wet ABS lan 37562 ERG spear ABS lung ABS cry luwarn ERG shoot ABS; general avoidance man ERG get ABS many ABS do, talk marran ERG tell ABS ngany ERG eat ABS nyanan ERG miss ABS (target) nyang ERG see ABS nyin, nyinang ABS sit pangin,panan(W) ERG scratch ABS parntiny ABS smell (N only) payan,payin(N) ERG bite ABS pulang ABS call out (N only) pung ABS burn pung ERG hit ABS wajparn ERG throw ABS wantiny ABS fall warntan ERG leave ABS (N only) wawang ERG wipe ABS wawarn ERG search for ABS (N only) wayan, wayin (N) ABS become wayirn ERG tie up ABS wurarn ERG stop ABS (a fight) yaan,yayin(N) ABS be yaan,yayin(N) ERG put ABS yan ABS go,come yingkarn ERG smooth ABS yung ERG give ABS yunparn ERG sing ABS yuwarn ERG send ABS .in -10 .bp .ce TABLE 4 -- WARLMANPA .sp These are taken from the list of 44 Warlmanpa verb roots in the appendix to Nash (1979). .sp .in 10 -ja-rra ABS become -jiya-ka ERG burn ABS -ma-nta ABS make noise -ya-nta ABS go jama-ka ERG grind ABS jin-ka ABS burn, ERG burn ABS jutpu-ngka ABS run ka-ngka ERG carry ABS ka-rra ABS sit, be kapu-ngka ERG chase, follow ABS karla-ka ERG spear, poke ABS kinyja-ka ERG wet ABS kipa-ka ERG twist together ABS kiya-ka ERG throw ABS kuma-ka ERG cut ABS kupa-ka ERG (fire) burn ABS la-ka ERG hit ABS with missile lamarta-ka ERG hold ABS lu-ngka ABS cry ma-nta ERG get ABS marta-ka ERG have, hold ABS murla-ka ERG copulate with ABS nama-ka ERG crush, weigh on ABS nga-ka ERG tell, scold ABS nga-nyja ERG ingest ABS ngarta-ka ERG trim ABS ngaya-ka ERG void ABS nguka-ka ERG drink ABS nya-ngka ERG see ABS paka-ka ERG strike ABS palu-ngka ABS die pan-ka ABS go pangi-ka ERG scratch, dig ABS partu-ngka ABS rise piya-ka ERG cut, bite ABS pu-ngka ERG hit, act on ABS wa-nta ABS fall waka-ka ABS climb wang-ka ABS speak wayi-ka ERG seek DAT winyja-ka ERG leave behind ABS ya-ka ERG put ABS at ALL yila-ka ABS drip, leak; ERG melt ABS yu-ngka ERG give ABS to DAT .bp REFERENCES Abbreviation: .sp .nf AIAS Canberra: Australian Institute of Aboriginal Studies, and, for publications since mid-1976, Atlantic Highlands, N.J.: Humanities Press .sp .na .fi ALLEN, Margaret. 1978. 'Morphological investigations.' University of Connecticut (Storrs) dissertation .sp ALPHER, Barry J. 1969. 'Yir-Yoront ideophones'. Paper presented to the Arizona Academy of Sciences, Spring 1969. 3pp. mimeo., Tempe, Arizona: Anthropology Department, Arizona State University .sp ----------- 1973. 'Son of Ergative'. [A grammar of the Yir-Yoront language]. Ithaca, N.Y.: Cornell University dissertation .sp AUSTIN, Peter. 1981a. A grammar of Diyari, South Australia. Cambridge Studies in Linguistics, 32. Cambridge: Cambridge University Press .sp ---------- 1981b. 'Case-marking in Southern Pilbara languages'. 32pp. ms. Presented at seminar on Australian Aboriginal Languages, M.I.T., 9th January .sp BIRK, D.B.W. 1976. 'The Malakmalak Language, Daly River (Western Arnhem Land)'. Pacific Linguistics series B, no. 45. Canberra: Department of Linguistics, Research School of Pacific Studies, Australian National University .sp BLOOMFIELD, Leonard. 1940. 'Algonquian.' In Cornelius Osgood (ed.), Linguistics Structures of Native America, pp. 85-129. Vking Fund Publications in Anthropology. Number Six. Reprinted 1963, Johnson Reprint Corporation .sp CAPELL, A. 1976. 'Rapporteur's introduction and summary', in Dixon ed. 1976:614-25 .sp DIXON, R.M.W. 1972. The Dyirbal Language of North Queensland. Cambridge Studies in Linguistics, 9. Cambridge: Cambridge University Press .sp ---------- (ed.) 1976. Grammatical Categories in Australian Languages. Linguistic Series No. 22. AIAS .sp ---------- 1977. A Grammar of Yidin. Cambridge Studies in Linguistics, 19. Cambridge: Cambridge University Press .sp ---------- 1980. Languages of Australia. (Cambridge language surveys). Cambridge, London, etc.: Cambridge University Press .sp DONALDSON, Tamsin. 1976. Wangaybuwan, in Dixon ed. 1976:763-8 .sp GODDARD, Ives. 1979. Delaware Verbal Morphology: A Descriptive and Comparative Study. (Outstanding dissertations in linguistics series reprint of April, 1969 Harvard University dissertation.) New York & London: Garland Publishing, Inc. .sp HALE, Kenneth L. n.d. 'Walbiri conjugations'. Mimeo. M.I.T. .sp ---------- 1973a. 'Person marking in Walbiri'. In Stephen Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, pp.308-344. New York: Holt, Rinehart & Winston, Inc. .sp ---------- 1973b. Deep-surface canonical disparities in relation to analaysis and change: an Australian example, pp. 401-58 in Current Trends in Linguistics, Volume 11, ed. by Thomas Sebeok. The Hague: Mouton .sp ---------- 1981. Preliminary vocabulary of Lardil. Typescript, M.I.T. .sp HANSEN, K.C. & L.E. 1977. Pintupi/Luritja Dictionary. Summer Institute of Linguistics, Australian Aborigines Branch, Berrimah. 2nd edition, Alice Springs: Institute for Aboriginal Development .sp HEATH, Jeffrey. 1976. 'North-east Arnhem Land'. In Dixon, ed. 1976:735-740 .sp ----------- 1978. Ngandi grammar, texts, and dictionary. AIAS .sp HOOK, Peter Edwin. 1974. The Compound Verb in Hindi. The Michigan Series in South and Southeast Asian Languages and Linguistics No. 1. Ann Arbor: Center for South and Southeast Asian Studies, The University of Michigan .sp HUDSON, Joyce. 1978. The core of Walmatjari grammar. AIAS .sp ---------- 1980. 'A tentative analysis of illocutionary verbs in Walmatjari'. Linguistics C2, A.N.U., June. 30pp. ms. .sp KENDON, Adam. 1980. 'The sign language of the women of Yuendumu: a preliminary report on the structure of Warlpiri Sign Language.' Sign Language Studies 27:101-112 .sp LEVI, Judith N. 1978. The Syntax and Semantics of Complex Nominals. N.Y., etc.: Academic Press .sp McCONVELL, Patrick. (to appear). 'Neutralisation and degrees of respect in Gurindji'. In Jeffrey Heath (ed.), Language and Kinship In Australia. Oceania Linguistic Monographs .sp MERLAN, Francesca. 1980a. 'Tense, aspect, and the Georgian perfect'. 24+pp. ts. .sp ---------- 1980b. 'Review of R.M.W. Dixon, A Grammar of Yidin.' Mankind 12.3(June),270-1 .sp ----------- (to appear). On the prehistory of some Australian verbs. Oceanic Linguistics 18.1. 38pp. ms. .sp NASH, David. 1979. 'Preliminary vocabulary of the Warlmanpa language, with grammatical preface'. Mimeo., M.I.T. .sp ----------- 1980. Topics in Warlpiri Grammar. Cambridge, Mass.: M.I.T. dissertation .sp O'GRADY, Geoffrey N. 1970. 'Nyangumarda conjugations'. In S.A. Wurm and D.C. Laycock (eds.), Pacific Linguistics Studies in Honour of Arthur Capell, pp.845-64. Canberra: Pacific Linguistics .sp PESETSKY, David. 1979. 'Russian morphology and lexical theory'. General examination paper, March 1979. 92pp. typescript. M.I.T. .sp RUMSEY, Alan Lee. 1978. 'A Grammar of Ungarinjin with special reference to the structure of discourse'. University of Chicago dissertation .sp SHARPE, Margaret C. 1972. Alawa Phonology and Grammar. Australian Aboriginal Studies No. 37, Linguistic Series No. 15. AIAS .sp SIMPSON, Jane. 1980. 'Preliminary vocabulary of Warumungu'. Mimeo., M.I.T. .sp SUTTON, Peter. 1976. 'The 'having' affix and other morphemes in fifty Australian languages.' In Dixon, ed. 1976:297-305 .sp TSUNODA, Tasaku. 1978. 'The Djaru language of Kimberley, Western Australia'. Melbourne: Monash University dissertation .sp WILLIAMS, Edwin. 1981. 'On the notions "lexically related" and "head of a word"'. Linguistic Inquiry 12.2(Spring):245-274