The inclusory construction in Australian languages

HEADER FOR MPLAL EDITORS’ USE ONLY THE INCLUSORY CONSTRUCTION IN AUSTRALIAN LANGUAGES Ruth Singer Honours student in Linguistics, Melbourne University. In this article I make an initial attempt to describe the inclusory construction for some Australian languages as part of a continuing project. The inclusory construction enables the speaker to describe the structure of human groups in some detail when referring to them. While the construction has been described for a number of languages and mentioned by Blake (1987) in his cross-linguistic study of Australian Aboriginal grammars there has been no indepth investigation of the construction in Australian languages. Lichtenberk (2000) provides a good framework from which to start describing the construction he calls the inclusory construction. The construction is formalised in quite different ways in Australian languages to those previously discussed by Schwartz (1998a, 1988b) and Lichtenberk (2000). After describing the form of the construction for a number of Australian languages, I reassess Lichtenberk’s framework and its application to Australian inclusory constructions. Keywords: inclusory construction, nominal coordination, Australian languages. 1 INTRODUCTION Lichtenberk’s ‘inclusory construction’ is a common one among Australian languages (Lichtenberk 2000). The Inclusory Construction (IC) is best viewed as one with two main parts. One element refers to a group - the superset, while the other one or more elements refer to subset(s) of the superset. There are two basic forms of the IC. An example of the first, which I will call Type 1 is shown in (1)1. In Type 1 the superset is represented by an independent pronoun. The second form, which I will call Type 2, is shown in (2). In Type 2 the superset is referred to by a dependent pronominal such as a verbal affix or an enclitic. (1) Roper River Kriol, Ngukurr Dubala M. bin go. 3.dual <name> PST go M. and someone else went. (my data2) (2) Mara3 rna-na garriyi-mar the man(Nom) The man and I went. rniri-rlini we(ExDu)-went (Heath 1981:302) Schwartz (1988a, 1988b) presents a global overview of the IC but analyses the construction as derived from a coordination construction. I view ICs as an alternative to coordination constructions rather than as a type of coordination construction. By coordination constructions I mean groups of NPs in which elements of equal rank are conjoined together into a larger group which is prototypically a group of two. In contrast, the IC is an endocentric construction in which some elements of a larger group are referred to along with the larger group itself. Type 1 1 ICs are bold in the examples. This data was collected in Ngukurr, N.T. in 2001 on behalf of the Katherine Regional Language Centre. 3 Examples are presented in a standardised Australian orthography similar to that in Yallop (1982). ‘ny’ is used for a palatal nasal, ‘ty’ is used for palatal stops or similar sounds ‘rr’ is used for tapped or trilled rhotics and ‘r’ is used for approximant rhotics and ‘y’ is used for [j]. Where originals are in a similar practical orthography they are left unchanged. Unusual symbols are represented with the same symbol as in the source. The spelling of names of languages has not been changed. 2 1 HEADER FOR MPLAL EDITORS’ USE ONLY ICs have been described by Blake (1987) as the ‘inclusive construction’. He described Type 2 ICs separately although he does note the similarities between the two constructions. I will view both Type 1 and Type 2 as different forms of the same construction. A broad crosslinguistic survey of the Australian IC has never been attempted. I will start by describing the forms the IC takes in part 2. In part 3 I will look at its distribution within languages. In part 4 I will consider the status of the IC such as whether it is a constituent, whether it has a head, and similarities between the IC and other constructions. I will then consider in part 5 how Australian ICs fit into the typology for inclusory constructions proposed by Lichtenberk (2000). 2 THE FORMS OF THE INCLUSORY CONSTRUCTION IN AUSTRALIAN LANGUAGES Lichtenberk suggests two main parameters according to which Inclusory Constructions vary. His first parameter is phrasality - he distinguishes between ICs which are phrasal and those which are not. However, this parameter is not very useful for Australian languages. There is little evidence that the Australian type 1 IC is defined by its phrasality. In those Australian languages for which there are criteria to determine what is a noun phrase, it appears that the type 1 IC may consist of more than one NP in many cases4. In any case, for many Australian languages postulating the existence of noun phrases is problematic and thus whether the IC is phrasal or not is not a useful parameter. There are clearly two separate forms of the construction in Australian languages. The difference seems to lie in whether the pronominal used to refer to the subset is independent or dependent. In the Type 1 IC the superset and subset elements are typically adjacent and the superset form is a free pronoun. In Type 2 ICs the superset is represented by a dependent pronoun and thus there is no tendency for the superset and subset elements to be adjacent. However a free pronominal superset can sometimes occur which is coreferential with the dependent superset pronominal. This typically occurs adjacent to the subset term as in (3). (3) Nunggubuyu nurru=wa-ng, ma:gurn we[ExPl]killed.it [name] M. and us killed it (buffalo) nu-rru we[ExPl] (Heath 1984: 542) Lichtenberk’s second parameter in his typology of ICs applies well to the Australian IC. He distinguishes between implicit ICs such as those in examples (1) and (2) and explicit ICs which carry additional markers such as conjunctives or comitatives that indicate a connection between superset and subset terms. The types of extra marking that occur will be discussed in section 2.3. Any extra marking in the examples given in sections 2.1 and 2.2 will be ignored for the time being as will case marking of nominals. 2.1 The Type 1 IC For Type 1 ICs to occur, a language must have independent pronouns. In the Type 1 IC, the superset and subset terms are usually adjacent, forming a continuous string. However, it is possible for them to be separate. It is likely that discontinuous ICs only occur in languages for which discontinuous noun phrases are acceptable. However, the amount of data is too small at 4 This issue will be discussed in more detail later sections. 2 HEADER FOR MPLAL EDITORS’ USE ONLY this stage to state this with certainty. An example of a discontinuous IC from Diyari is shown in (4). From the two discontinuous ICs I have come across, they seem similar to discontinous NPs in that one element often occurs in topic position at the far left or right of the sentence. (4) Diyari ngali-tha yatha-rna warra-yi 1dlexclSA-01 say-PART AUX-PRES We said, my brother (and I) ngakarni nhiyi-yali 1sgDAT elder.brother-ERG (Austin 1981: 122) The only words which occur between the elements of Australian Type 1 ICs are nominal expressions such as modifying nominals, determiners, demonstratives and possessives. Number markers, conjunctives and comitatives may also occur between the words in the explicit form of the construction. The superset term occurs consistently either at the end or at the beginning of the construction depending on the language5. In example (5) the superset term occurs at the end of the IC. This example also shows that both elements of the construction can be pronouns. As in this case, an IC of two pronouns is usually used to expressive inclusive/exclusive distinctions lacking in the pronominal system of a language. (5) Yankunytjatjara nyunta ngali kuka-ku yanku-ku 2.sg(NOM) 1.du(NOM) meat-PURP go-FUT You and I might go for meat (Goddard 1985:100) The range of pronouns which may represent the superset is large. Dual pronouns are the most common but plural pronouns are not uncommon - such as in example (6). (6) Djapu raku-’rakuny-dji-rr-a nganapurr dead-REDUP-INCHO-UNM-IM 1plexcNOM Birandjitj-tja (name)+ABS-PRO We all, Frances included, were dying. (Morphy 1983:87) Most Australian languages have first, second and third person dual and plural pronouns. In some Australian languages special information is obligatorily encoded in non-singular pronouns such as generation harmony between members of the group and whether members are in the same moiety or not. For this reason the IC can carry much more additional information than a simple coordination construction which consists only of singular pronouns6. The subset term may be a possessive construction, coordination construction, kinship term, personal name, another pronoun or probably any term that can be used to refer to a human. It is not clear yet whether there is any limit to the number of subset terms. One is the most common number, but two may also occur. In example (7) there are two subset terms and they are place names which is unusual. The usual use of the IC with human referents may be a pragmatic tendency rather t h a n a s t r i c t rule. 5 Apart from Kriol and Kaytetye in which either order is acceptable. See Hale (1966) for more discussion of this. 6 3 HEADER FOR MPLAL EDITORS’ USE ONLY (7) Yankunytjatjara kutjupa tjut=a other many(ACC) maa-kati-ngi, away-take-PAST.IMPF Mimili-la Mimili-LOC-NAME Intalka-la tjana-la-kutu Indulkana-LOC-NAME 3.pl-LOC-ALL (They) took some others to Mimili, Indulkana and other places (around there) (Goddard 1985: 102) 2.2 Type 2 ICs There are a number of approaches we could take to the Type 2 IC. Similar constructions have been approached by Schwartz (1988a) and Bickel (2000) from the perspective of agreement. Bickel argues that there are a number of different types of agreement. The one commonly occurring in Indo-European languages he dubs ‘integrative agreement’ - agreement features of agreement markers and NPs match up exactly. He contrasts this with ‘associative agreement’ which occurs in other language families, in which “disagreement is systematically exploited as a constructional resource”. This includes the situation in the Type 2 IC in which the bound pronoun and free NP(s) are in an appositional or part-whole relationship with one another. Rather than treating Type 1 and 2 ICs as separate phenomena, I wish to approach Type 1 and 2 ICs together. In the Type 2 IC the superset element occurs as a dependent pronominal. Depending on the language such a pronominal occurs as an affix on a verb or else it may occur as an enclitic. Example (8) from Mara shows that the subset may be referred to by a number of separate elements, in this case all carrying dual prefixes that may serve to identify them as part of the one unit. (8) Mara ga-na-ngga there yarna-rniwi-tyintyi sleep+1.pl.excl.-did war-a du-the wuru-yimar du-that wur-munangga wurutya du-white two There we slept, (the two of us) and the two whites. (Heath 1981:376) In example (9) the superset is represented by an enclitic that attaches to the first element of the clause which in this particular example is the subset term. (9) Warlpiri Jungarrayi-jarra-npala ‘skin name’-dual-22 ‘You two Jungarrayi’ ‘You and a Jungarrayi’ (-22 means 2nd person dual subject pronominal enclitic) (Laughren 1977, from Nash 1986:212) This construction is ambiguous and has two possible interpretations as shown by the two different translations given. The Warlpiri type 2 IC will be discussed in more depth in section 2.3.3. 4 HEADER FOR MPLAL EDITORS’ USE ONLY 2.2.1 Type 2 ICs with reciprocals The inclusory construction can also occur with verbs bearing reciprocal affixes as shown in examples (10) and (11)7. The subset term Al-mardgu in example (10) - a matrimoiety name - is a good example of the way that any term that can be used to denote a human can be used as a subset term in the IC. (10) Kunwinjku Ani-ma-rr-en Al-mardgu 1.dual-marry-RR-NP (matrimoiety) I have to marry a woman of the Mardgu matrimoiety [lit.: We have to get married, me and an Al-mardgu woman] (Evans forth.:363) (11) Ngandi nyari-bu-ydhi-ni rni-yul-pula 1.du.excl.-hit-recip.-P.con m.sg.-Aboriginal-DUAL8 I and a [aboriginal] man were fighting (Heath 1978:291) 2.3 Case marking of Type 1 ICs Case can provide useful information about the structure of ICs. There are three main types of case marking that type 1 ICs can have. Firstly the subset and superset can each be marked for the same grammatical function - such as A (transitive subject), S (intransitive subject) or O (object). This may suggest they are both separate NPs in languages in which only one element of an NP usually carries the case marker for the whole NP. Secondly the elements of the IC may be marked for the same grammatical function but with a special pattern of case marking specific to the IC. Thirdly the subset may have a special type of marking that does not identify its grammatical function but links it to the superset. These three patterns have been found for a number of languages but only a few languages will be discussed here because of limited space. 2.4.1 Case marking pattern 1: Superset and subset marked for core grammatical function In Guugu Yimidhirr we find both superset and subset are marked for the same grammatical function, although the pronominals have a nominative/accusative case-marking paradigm and the nominals have an ergative/absolutive case marking paradigm. This pattern, shown in examples (12) and (13) is the same pattern of case marking as found in noun phrases, generic-specific constructions and part-whole constructions. It does not provide much information as to the structure of ICs. (12) Guugu-Yimidhirr Ngaliinh Dyaagi-ngun gambarr 1duexc+NOM Jack-ERG pitch+ABS Jack and I made the pitch balga-y make-PAST (Havidland 1979:105) (13) Guugu-Yimidhirr Bula ngadhu yumurr yuwal-inh dhada-y 3du+NOM 1sg+GEN+ABS child+ABS beach-ALL go-PAST Those two - my son included - went to the beach. (Haviland 1979: 105) 7 8 Heath (1984:544) discusses the occurrence of this in Nunggubuyu. Heath glosses -pula as ‘-AND ’ but I believe this is better understood as a collective plural suffix which is prototypically dual. See argument in section 2.3.2 5 HEADER FOR MPLAL EDITORS’ USE ONLY 2.4.2 Case marking pattern 2: superset and subset marked for core grammatical function but IC has special case marking pattern. In Kayardild the IC has a similar pattern of case marking to the Guugu Yimidhirr IC when it is in nominative case. This is illustrated in (14). (14) Kayardild nga-rr-a kajakaja warra-ja thaa-th 1-du-NOM daddyNOM go-ACT return-ACT Daddy and I will go (lit. ‘we two, including daddy, will go’). (Evans 1995:249) In non-nominative cases, free pronouns and possessive pronouns take exactly the same form. This leads to situations in which a pronoun plus a nominal could be interpreted as a possessive construction or an IC. In such a situation, the pronoun plus noun is always interpreted as a possessive construction as in (15). (15) Kayardild ngada kurri-ja bi-l-wan-ji ngamathu-y 1sgNOM see-ACT 3-pl-POSS-MLOC mother-MLOC I saw their mother (Evans 1995: 206) If an IC meaning is intended the associative case suffix is added to the subset nominal as shown in example (16) below. (16) Kayardild ngada kurri-ja bi-l-wan-ji ngamathu-nurru-y 1sgNOM see-ACT 3-pl-POSS-MLOC mother-ASSOC-MLOC I saw them, including mother (Evans 1995: 206) All elements of a Kayardild NP carry core case marking, as do elements of ICs. However, Evans (1995) does not consider the Kayardild IC to be a single noun phrase because it does not have a single semantic head. Instead he considers it to be a type of part-whole construction. He argues that like part-whole constructions the IC is double headed and thus a group of noun phrases. Although the IC is not very common in Kayardild it is interesting because it only has a special case marking pattern when it would otherwise be ambiguous with another construction. Goddard (1985) discusses ICs in Yankunytjatjara at some length. In Yankunytjatjara, Goddard distinguishes two main types of case: core case and local case. Core case identifies the core syntactic functions of nominals. Local case provides other types of information and includes allative, ablative and perlative case. Locative case is separate to these two types. In Yankunytjatjara NPs, only the last element carries the case marker(s) for the whole phrase as illustrated in example (17). (17) Yankunytjatjara papa tjapu tjurta-ngku mayi ngalku-rnu dog small many-ERG veg.food(ACC) eat-PAST The small dogs (puppies) ate the food. (Goddard 1985: 92) ICs carry core case markers on all elements which suggests that the elements are separate NPs. An example of an IC with core case marking only is shown in (18). (18) Yankunytjatjara9 9 The example is modified slightly because the original had two options for the subset term separated by a forward 6 HEADER FOR MPLAL EDITORS’ USE ONLY Tjirlpi-lu nyupali kati-ku-nti old.man-ERG.NAME 2du(ERG) take-FUT-MAYBE You and the Old Bloke might take (us) (Goddard 1985:101) However, when local case marker occur we find a different pattern of case marking. Although the locative case marker occurs on all elements10 the local case markers: allative, ablative and perlative only occur on the final element of the IC. In example (19) only the final pronominal superset carries the ablative marker. In example (20) only the final pronominal superset carries the allative case-marker. (19) Yankunytjatjara Yami-la tjana-la-ngurru Yami-LOC NAME 3pl-LOC-ABL from Yami and the others (“Yami mob”) (Goddard 1985:54) (20) Yankunytjatjara ....nyuntu-la ngali-la-kutu 2.sg-LOC 2.du-LOC-ALL ...towards you and I (Goddard 1985:54) Goddard argues that the IC makes two separate acts of reference and this is why superset and subset elements carry separate core case marking. He analyses the Yankunytjatjara IC as a complex NP - it contains more than one NP but behaves like a single NP. However, it is possible to analyse the Yankunytjatara IC as two NPs which are closely linked because they refer to the same argument but not fully unified - what Wilkins (1989) calls an ‘NP complex’ and contrasts with a ‘complex NP’. Rather than making a single complex act of reference it could be argued that the IC makes two separate, overlapping acts of reference in both Kayardild and Yankunytjatjara. This may provide a basis for arguing that the IC is always two or more NPs. 2.4.3 Subset term not marked for core grammatical function. In Diyari the ergative case marking suffix can mark a number of functions other than core grammatical functions. It can mark a nominal as being in instrumental case. It can also mark a nominal as being the subset of an IC as shown in (21). In example (21) the IC is in S function, not A, yet the subset term is marked with the ergative suffix. (21) Diyari ngali kanku-yali mawa-li11 ngana-yi 1DlexclSA boy-ERG hunger-ERG be-PRES We two were hungry, the boy and I. (Austin 1981: 239) Interestingly comitative expressions in Diyari referring to humans utilise the proprietive or locative case markers, not the ergative case marker. How the ergative came to be used for the IC is an interesting question. The distribution of case markers in Diyari comitative and inclusory slash. One option has been removed for clarity. According to Goddard the locative case marker may be acting as a stem-forming suffix rather than a case marker here but this is not important to the discussion - see Goddard (1985: 102). 11 Note that mawa-li (hunger-ERG) is not part of the IC. Austin describes this function of the ergative case marker: “abstract nouns functioning predicatively are inflected for ergative case and occur with the copula ngana- ‘to be’”(Austin 1981: 121). 10 7 HEADER FOR MPLAL EDITORS’ USE ONLY constructions may support the view that the IC is an independent construction rather than a derivative form of another construction. Discontinuous ICs are possible in Diyari as shown in (23) (repeated from (4)). Discontinuous NPs are also possible in Diyari. Even when separated from its superset as in this example, the subset term marked with the ergative suffix is not interpreted as an independent ergative NP but as a subset of the initial pronoun. (22) Diyari ngali-tha yatha-rna warra-yi 1dlexclSA-01 say-PART AUX-PRES We said, my brother (and I) ngakarni nhiyi-yali 1sgDAT elder.brother-ERG (Austin 1981: 122) Austin (1981) notes that the usual interpretation of a dual IC with a kin term subset is as a reciprocal kinship dyad as in (22) and (23). However, he notes that this is not an entailment and the IC may refer to the speaker’s kin and someone else. (23) Diyari pula kaku-yali wapa-yi 3slS elder.sister-ERG go-PRES The two sisters are going. (Austin 1981: 122) 2.4 Explicit inclusory constructions In many languages inclusory constructions lack any special marking. It is up to the hearer to detect that the inclusory construction is being used, based on a mismatch of agreement features, context and familiarity with the construction. Such constructions are dubbed ‘implicit’ ICs by Lichtenberk (2000). In some languages inclusory constructions may carry ‘extra’ marking indicating that a relationship exists between the superset and the subset terms. ICs with such extra marking are called ‘explicit’ ICs by Lichtenberk. Such marking may be obligatory or optional. The distinction between explicit and implicit ICs is Lichtenberk’s second parameter in his typology of inclusory constructions. This second distinction cuts across the distinction between Type 1 and Type 2 constructions - either can be explicit or implicit. This parameter would apply well to Australian ICs although it is difficult to decide how ICs with case marking should be treated. An IC with a special pattern of case marking points out the presence of an IC like the markers in explicit ICs. I have discussed the case marking of ICs in a separate section and will not place them in Lichtenberk’s typology as I have not decided how they fit in. Lichtenberk’s study focuses on a language, Toqabaqita, which does not have explicit ICs so he does not look at them in any depth. Schwartz (1988a) first discussed the existence of extra marking in some ICs calling the morphemes used to indicate a relationship between elements in an IC ‘connectors’. This is because although the extra morphemes used in explicit ICs are never specific to ICs they tend to operate quite differently within ICs than in their usual functional role. In Schwartz surveyed ICs in Indo-European, Afro-asiatic, Austronesian and other language groups. She found that all connectors used in ICs had a comitative function outside of the ICs. However, this is just one core function ‘connectors’ in Australian ICs may have, coordinating and number marking morphemes also occur with equal frequency. In addition, none of the languages I have looked at have ICs with a connector which is a free conjunction. For example, although these exist in Nunggubuyu, Arrente and Warlpiri they do not participate in ICs as connectors. I will use the term ‘marker’ instead of connector because in Australian ICs the extra term tends to occurs between superset and subset terms to link them like the languages Schwartz considered. Instead they are more likely to have a suffix on the subset term marking it 8 HEADER FOR MPLAL EDITORS’ USE ONLY as a part of a group, as going with something else. Inclusory constructions are contrasted with simple coordination of nominals. In English we associate a simple coordination with the use of the conjunction and between the two or more nominals to be coordinated. Although a similar construction is found in some Australian languages it is by no means the norm. In Australian languages coordination is often realised by simple juxtaposition of nominals. Alternatively coordination of nominals may involve affixation of a conjunctive affix to one or more of the conjoined nominals. Thus having a conjunctive element has no bearing on the distinction between an IC and a coordination construction. There are a range of markers which I have come across in Australian ICs. The main groups are nominal dual/plural suffixes and nominal conjunctive suffixes. Both types are suffixes which are suffixed to subset nominals in ICs. There is no clear boundary between these suffix types as they can all be used in the coordination of nominals. They are suffixed to one or both conjoined nominals and may be optional. I have divided the types of suffixes into three main groups based on how they are described in grammars. However, I actually wish to argue that these suffixes which appear in ICs all share a core function. As McGregor (1990) argues for -yoodoo in Gooniyandi, they all mark a nominal as belonging to a higher level of constituency. Thus in a sense they operate more like case markers than conjunctions or number suffixes - linking nominals upwards to higher levels of constituency rather than across to other equal constituents like the English and. As Dixon (1977) translates the suffix -ba in Yidiny they often mean ‘one of a group’ rather than ‘and’. This makes sense of why these suffixes often appear attached to a subset term of an IC, even when they are rare in nominal coordinations in a language. The single subset nominal appears to be alone but in fact is part of a larger group, referred to by the superset term. A suffix makes it clear that this nominal is not a separate participant in the clause but part of the same entity referred to by the superset term. I will first discuss ‘one of a group’ suffixes in Dyirbal and Yidiny to outline how these suffixes can operate. Then I will discuss Ngandi in which the suffix participating in ICs has been analysed as a conjunctive suffix. Many ICs have been described as having suppletive dual marking. I will suggest that these suffixes also share a common core of meaning with the others and fulfill a similar function in ICs. 2.3.1 ‘One of a group’ suffixes in Dyirbal and Yidiny. Dyirbal and Yidiny both have ‘one of a group’ suffixes and Dixon (1972, 1977) describes them as such. However, only in Yidiny does a ‘one of a group’ suffix participate in the IC. Type 1 ICs occur in Yidiny. Dixon (1977) writes “Coordination of nouns (with human reference) within an NP is achieved through the addition of -ba to each one” (p.145). Example (24) is an example of coordination in Yidiny and example (25) is an example of an IC in Yidiny. (24) Yidiny darnggidarnggi:ba yaburruba galing old woman-ba-ABS young girl-ba-ABS go-PRES An old woman (being one of a group of people) and a girl (being another member of a group) are going. (That is ‘a woman and a girl are going’, leaving open the question of whether they are going alone, or as part of a larger assemblage.) (Dixon 1977:177) 9 HEADER FOR MPLAL EDITORS’ USE ONLY (25) Yidiny nganytyi bunya:ba galing 1.non-sing woman-ba go-PRES (my gloss) A woman and I (and some others) are going. (Dixon 1977:178) Yidiny is an Australian language in which the form of the coordination construction and the IC are very similar. However the IC is still clearly an IC, and it seems that coordination of pronouns does not occur. The only way to link a pronoun and a noun apart from ICs is to use a comitative construction as shown in example (26). (26) Yidiny ngayu bunya:y galing 1.sg woman-COM go-PRES I am going with a woman (Dixon 1977:178) In Dyirbal there are two ‘one of a group’ suffixes: -gara meaning ‘one of a group of two’ and manggan meaning ‘one of a group of more than two’. An example of the use of -gara is shown in (27). (27) Dyirbal burbul-gara baninyu B.(name)-PAIR come-PRES Burbula and another person are coming (Dixon 1972: 230) Neither of the suffixes are attached to elements of the IC as shown in example (28). Thus the presence of ‘one of a group’ suffixes in a language does not predict their participation in ICs. (28) Dyirbal ngali bayi yarra baninyu 1.du NC12 man come-PRES (my gloss) man and I are coming (Dixon 1972:63) 2.3.2 ‘Conjunctive’ suffix in Ngandi In Ngandi a suffix -pula occurs which is usually suffixed to the second of two conjoined nominals. It is used in three of the four ICs in Heath’s (1978) Ngandi texts. Like -ba in Yidiny, the suffix -pula also functions as a dual number marker. One example was given in (11). Two more examples of Ngandi ICs are given in examples (29) and (30). (29) Ngandi rni-nyara-ng-bula MaSg-father-POSS-AND My father and I will go nyari-rrurdu-ng 1MaDuExcl-go-FUT (my gloss) (Heath 1978:128) (30) Ngandi rni-gorlokorndo-bula bari-ga-rrird-i, MaSg-(name)-AND 3MaDu-Sub-go-PPun Gorlokorndo and another man went (Heath 1978:259) In example (31) -pula does not occur: (31) Ngandi 12 NC means noun class marker. 10 HEADER FOR MPLAL EDITORS’ USE ONLY rnari-wo-tyaldya-rrurdu-ng rni-yul-mak-burkayi-yung, 2MaDu-both-together-go-FUT MaSg-person-good-really-ABS The two of you will go together, (you and the man, who is now) a well-behaved person. (Heath 1978:253) There are not a huge number of examples of the use of -pula in coordinations in Heath (1978). Simple juxtaposed lists of nominals are more common that coordinations using -pula which seems to mainly occur in coordinations of two nominals. The information in Heath (1978) suggests that -pula might be better understood as a suffix that indicates a nominal is ‘one of a group’ rather than a conjunctive. Another example of it in use is in example (32) in which it is underlined. Blake (2001) glosses -bula as -BOTH which would explain its use in dual number ICs, and is compatible with the meaning ‘one of a group of two’. (32) Ngandi ma-wurnrdan? VEG-black.plum balaka first nyarr-ma-ga-ma-ngi, we-VEG-DEFOCUS-get-PAST ma-berge?-bula. VEG-green.plum-BOTH We got black plums first, and then green plums as well. (Segmenting and gloss from Blake 2001: 423. Translation from original in Heath 1978:128.) 2.3.3 Number marking in ICs Two number marking suffixes are found in Gooniyandi: -yoodoo ~ yiddi (dual) and -yarndi (plural) (McGregor 1990). In general McGregor describes these suffixes as optional suffixes which mark a nominal as dual or plural. -yoodoo can also be added to non-singular pronouns to make them dual. Together with case suffixes they are analysed by McGregor as postpositions which “enter into constituency with noun phrases, forming postpositional phrases“. Usually only one per phrase occurs and they mark the number of the whole phrase not just the nominal to which they are attached. The Gooniyandi Type 1 IC in example (33) is analysed by McGregor as containing two NPs. The second one elaborates the first one. (33) Gooniyandi ngidiNP1 David-jooddoo NP2 We(Restricted) (name)-DU we two, including David (McGregor 1990:286) McGregor suggests that the first person pronoun has been omitted from NP2 but it retains the yoodoo suffix indicating that the phrase is a dual one. My analysis of the IC thus far has avoided explanations of the IC that suggest it is elliptical. Given its similarity to other Australian constructions such as the part-whole and generic-specific constructions I believe it is unnecessary and misleading to view the IC semantically as elliptical. It would be better to see the dual suffix as marking the nominal ‘David’ as being part of a dual group and thus being coreferential with the superset pronoun. Like the suffixes discussed in the previous section yoodoo can also appear to be acting as a conjunctive suffix as shown in example (34) below. 11 HEADER FOR MPLAL EDITORS’ USE ONLY (34) Gooniyandi wampi-yoodoo biddidhiya (name)-DU (name) Wampi and Amee (McGregor 1990:280) It is compatible with the descriptions given for the suffixes discussed in the previous sections that like -yoodoo they also be analysed like case marking suffixes that mark relationships between a nominal and a higher level of constituency. The Gooniyandi number suffix either marks a nominal as going together with other nominals as part of a constituent, or marks it as a noun phrase on its own and thus marks it as part of a non-singular constituent. This accounts for the fact that some of the suffixes discussed previously can also be number markers, making singular nominals into dual or plural nominals. It also accounts for the fact that most of these have a different interpretation when suffixed to a name then when suffixed to a common noun. As in example (27) from Dyirbal, repeated in (35) below. (35) Dyirbal burbul-gara baninyu B.(name)-PAIR come-PRES Burbula and another person are coming (Dixon 1972: 230) The single proper noun with a number marking suffix forms a noun phrase on its own of the number indicated. This analysis renders unimportant variation in whether number suffixes mark all or both nominals as these variations merely represent different patterns of marking analogous to different systems of case marking. The indeterminacy of the nominal dual and plural suffixes is illustrated by the Warlpiri type 2 IC. The IC is one of two interpretations of example (36). The pronominal enclitic attaches to the first word or phrase in Warlpiri so in these sentences it happens to attach to the subset term. The presence of the dual suffix -jarra creates the possibility of two possible interpretations as shown in the two translations given. (36) Warlpiri Jakamarra-patu=lpa=rnalu yanu J-SEVERAL=IMPF=1PL.S go-PST We including Jakamarra were going. We Jakamarras were going. (Laughren pers. comm.) Without the -jarra suffix on the ‘skin name’ Jakamarra, only the simple plural interpretation is possible. (37) Warlpiri Jakamarra=lpa=rnalu yanu J =IMPF=1PL.S went We Jakamarras were going. (Laughren pers. comm.) 3 THE DISTRIBUTION OF THE INCLUSORY CONSTRUCTION The use of the IC by speakers in each language depends on the use of the two main alternative constructions - coordinations and comitative constructions involving pronouns. In many languages, such as Yankunytjatjara (Goddard 1985) the coordination of nominals and pronominals together, in a similar way to the coordination of two nominals is not acceptable and 12 HEADER FOR MPLAL EDITORS’ USE ONLY instead the IC is used. In Wik Mungkan coordination of third person pronouns is an acceptable alternative to the IC but coordination involving first and second person pronouns is ungrammatical (Kilham et.al 1986). In other languages coordination of nominals and pronominals may be possible but uncommon. Languages in which there is a choice for speakers to use either a type 1 IC or coordination of a nominal and a pronominal include Djapu (Morphy 1983), Lardil (Hale 1966), Kriol (Sandefur 1979, Hudson 1983), and Torres Strait Creole (Shnukal 1988). Schwartz (1988a) describes ICs for non-Australian languages which are limited to particular grammatical functions but this does not seem to occur in Australian languages. In Toqabaqita the rules governing which type of IC can be used in each grammatical function are quite complex (Lichtenberk 2000). However, in Australian languages there do not seem to be many such constraints. Naturally, most examples found are of ICs in S or A function. However, example (20) above shows an IC in an oblique grammatical function in Yankunytjatjara. Koch (unpub.) has many examples from Kayteyte in which the ICs are in a range of core and oblique grammatical functions such as example (38). (38) Kaytetye aylekanthe-ketye mpwerne-ye-ketye weDU:EX:OM-AVERS wife's.brother-my-AVERS (frightened ) of me and my brother-in-law (Koch unpub.) The nature of the inclusory construction is such that the superset and subsets referred to are usually known to the hearer. Schwartz (1988b) wonders whether the construction is restricted to referring to humans, or to groups of close friends or kin. For Australian languages it is not clear whether there are strict restrictions to the type of referents that can occur in inclusory constructions. However, this may be because of the small number of examples of ICs documented for each language. An example was given of a Yankunytjatjara IC referring to place names (example (7)). However, that example is very unusual. In some languages the IC may only be acceptable with human referents, in other languages it may be more a result of pragmatic factors that the IC tends to be used for familiar groups of humans. The examples of the inclusory construction I have collected are from a wide range of Australian language groups. Both Pama-Nyungan and non-Pama-Nyungan languages may have the construction. Examples come from a wide area from Cooktown in northern Queensland across Northern Territory to Western Australia and southwards down to Central Australia. The inclusory construction may be particularly common in certain genetic or areal groupings of languages but it is difficult to say at this stage as it is likely many more languages have the construction than those mentioned here. 4 THE STATUS OF THE INCLUSORY CONSTRUCTION IN AUSTRALIAN LANGUAGES In many languages the structure of the inclusory construction is very similar to that of one or more other constructions in the language. Blake (1987), in his description of the construction, groups Type 1 ICs together with generic-specific and part-whole constructions. He describes them all as types of ‘superordinate constructions’ - one element being superordinate to the other. Although I have treated ICs as an independent type of construction. It is not clear whether ICs are really a separate type of syntactic construction in all languages. For example, Evans (1995) describes ICs as a type of part-whole construction in Kayardild. Dench (1995) does not describe the IC in Martuthunira as a unique construction either but groups examples of it together with a number of other structures, calling them all ‘adjoined noun phrase structures’ 13 HEADER FOR MPLAL EDITORS’ USE ONLY one noun phrase referring to a group together with other adjoined noun phrases which provide a fuller description of the group by specifying particular members of it. McGregor (1990) includes an example of an IC in his description of types of ‘juxtaposed phrases’. These three authors treat Type 1 ICs as no different to an elaboration of a non-singular nominal like that in example (39) below. In this example the superset is not a pronominal. Such constructions have not been considered here although they have similarities to ICs. (39) Martuthunira ngurnala-ngu-u ngarniyarrangu-u, pipi-thurti-i that.DEF-GEN-ACC family-ACC mother-CONJ-ACC pawu-thurti-i father-CONJ-ACC mimi-thurti-i uncle-CONJ-ACC ... his family; mother, father and uncle. (Dench 1995: 201) In some languages the IC is very common, but in others it is rarely used. For example, in Kayardild comitative constructions are commonly used in place of the IC (Evans pers. comm.). When the IC is rare in a language, the features that distinguish it as a unique construction may be difficult to detect. Alternatively, if rarely used, it may in fact not have any distinctive features but be a semantic variant of a more common construction. It is still possible to compare the IC in these languages with others in which it is clearly a unique type of construction because there are conceptual similarities between the constructions across languages. Where ICs are clearly a separate type of construction they may share features with a number of different constructions. In Yankunytjatjara we saw that the Type 1 IC stood out as a unique type of construction. Goddard (1985) groups inclusory constructions in Yankunytjatjara with a type of part-whole construction he calls the ‘personal’ construction. Both are described as complex NPs in which the constituents are not in a head-modifier relationship. Evans (forth) groups Type 2 ICs together with part-whole constructions in Bininj Gun-wok as two different ways that “nominal and verbal material enter into unificational constructions... giving two ‘takes’ on the domain of application of the verbal predicate... delineating two concentric levels (one narrower, one broader) at which the predication applies.” (Evans unpub.:361). It is likely that in many languages ICs are what Wilkins (1989) calls NP complexes - that is a group of NPs that tend to occur adjacently in texts due to discourse processes - rather than complex NPs which are act as single NPs. The question of whether the inclusory construction plays the role of a single NP (a complex NP) or is just a grouping of NPs (an NP complex) may vary from language to language. Earlier treatments of ICs worldwide - Schwartz (1988a, 1998b) - and in Australian languages - Hale (1966, 1973) - look towards finding a single syntactic structure underlying all inclusory constructions. However, despite being prevalent among Australian languages, it seems that the IC is not based on a single syntactic structure. It seems that cross-linguistically, in Australian languages similarities between inclusory constructions are semantic and conceptual rather than structural. There are also clear semantic similarities between part-whole constructions, inclusory constructions and generic-specific constructions. In inclusory constructions, the superset and its subsets do have a whole-part relationship but in part-whole constructions neither of the two entities is predictably more important than the other (Blake 1987). However, in the inclusory construction, the superset is clearly the central 14 HEADER FOR MPLAL EDITORS’ USE ONLY constituent. The superset term encodes the number of the whole construction. When it is a free pronoun the superset often has a fixed position at the beginning or end of the construction. Also, the subset terms can be removed with less effect on the meaning of the whole clause than the superset term. The fact that there is fixed ordering of elements most languages suggests that the IC is a type of constituent in many languages. Whether it is a constituent of equal rank to the NP or equivalent to a grouping of NPs seems to vary between languages. The superset term may be the syntactic head of this constituent in some languages. In other languages it may simply be the semantic head. 5 LICHTENBERK’S TYPOLOGY AND AUSTRALIAN INCLUSORY CONSTRUCTIONS Lichtenberk (2000) describes similar constructions to Australian ICs for Toqabaqita, a language of Malaita, Solomon Islands. In this language both Type 1 and Type 2 constructions occur. However, explicit ICs do not. Lichtenberk (2000) gives a brief typology of the construction he calls the Inclusory Construction based on two parameters. The first parameter distinguishes ‘phrasal’ from ‘split’ inclusory constructions. I have used a slightly different distinction because of the lack of Australian ICs which could be called a phrase. Although some ICs are clearly a type of constituent, it is possible that others are not really a constituent at all. In some languages the IC may be a group of NPs that have a tendency to cluster together in a clause. Where Lichtenberk uses the presence or absence of phrasality as his first parameter, I use the presence or absence of bound pronominals referring to the superset to make a parallel distinction for Australian languages. The only problem for my reinvention of Lichtenberk’s first parameter is the interesting and extremely common IC in the Oceanic language Yapese. The IC commonly takes the form shown in (40). (40) Yapese (Oceanic) Yow bea marweel Tamag 3.dual PRES.PROG work T.(man’s name) He and Tamag are working (Jensen 1977:195) Lichtenberk argues that this is a non-phrasal IC with an independent pronoun. There are two features of Yapese that need to be taken in to account in order to fit this example into my typology. Firstly Yapese has fixed VSO word order so Tamag is clearly identified as the subject. Also within the ‘verb complex’ the subject pronominal has a fixed position before the subject. More work is needed to decide whether this should be defined as a Type 1 or Type 2 IC. It is like a Type 2 IC because the subject marker is obligatory even when ICs do not occur it cross-references any subject nominal. It is clear which is the subject nominal because of the rigid word order in Yapese. This example suggests that more work is needed in defining what is meant by ‘dependent pronominal’ and in incorporating the role of fixed word order to make my typology apply to the range of ICs found outside of Australia. Lichtenberk’s typology needed some tinkering with in order to describe variation in the Australian IC adequately. However, it makes a great leap forward from previous typological studies of the IC by Schwartz (1988a, 1988b) who insisted on viewing the IC as a defective coordination of two nominals. Such a viewpoint has been proposed by Hale (1966, 1973) and Nash (1986) for Warlpiri ICs. However, it is not necessary to propose that ellipsis has occurred within Australian ICs. Similarities between the IC and other ‘superordinate constructions’ (Blake 1987) suggests it is more closely related to them than to coordination constructions 15 HEADER FOR MPLAL EDITORS’ USE ONLY although morphologically it may resemble coordination constructions. The form of the IC in Australian languages suggests that the construction is an alternative to NP coordination rather than a type of coordination. 16 HEADER FOR MPLAL EDITORS’ USE ONLY References Austin, Peter (1981) A grammar of Diyari, South Australia. Cambridge University Press: Cambridge. Bickel, Balthazar. (2000) On the syntax of agreement in Tibeto-Burman. Studies in language, Volume 24, pp583-609. Blake, Barry J. (2001) The noun phrase in Australian languages. In Jane Simpson, David Nash, Mary Laughren, Peter Austin and Barry Alper (eds) Forty years on: Ken Hale and Australian languages. Pacific Linguistics, Canberra. pp.415-426. Blake, Barry .J. (1987) Australian Aboriginal Grammar, Croon Helm, Beckenham. Dench, Alan C. (1995) Martuthunira, a language of the Pilbara region of Western Australia, Pacific Linguistics C-125, Canberra. Dixon, R.M.W. (1972) The Dyirbal language of north Queensland, Cambridge University Press, Cambridge. Dixon, R.M.W. (1977) A grammar of Yidiny, Cambridge University Press, Cambridge. Evans, Nicholas. (1995) A grammar of Kayardild, Mouton de Gruyter, Berlin. Evans, Nicholas (forthcoming) A grammar of Bininj Gun-wok. Goddard, Clifford. (1985) A grammar of Yankunytjatjara, Institute for Aboriginal Development, Alice Springs. Hale, Kenneth. (1966) Kinship reflections in syntax, Word, Volume 22, pp318-324. Hale, Kenneth. (1973) Person marking in Walbiri. In Anderson, S. and Kiparsky, P. (eds) A festschrift for Morris Halle, Holt, Rinehart and Winston, New York, pp308-344. Havidland, J. (1979) Guugu Yimidhirr. In Dixon, R.M.W. and Blake, B.J. (Eds) Handbook of Australian Languages Volume 1, Canberra: Australian National University Press. pp. 27-180. Heath, Jeffrey. (1978) Ngandi grammar, texts and dictionary, Australian Institute of Aboriginal Studies, Canberra. Heath, Jeffrey. (1980) Nunggubuyu myths and ethnographic texts, Australian Institute of Aboriginal Studies, Canberra. Heath, Jeffrey. (1981) Basic materials in Mara: grammar, texts and dictionary, Pacific Linguistics C-60, Canberra. Heath, Jeffrey. (1984) Functional grammar of Nunggubuy,. Australian Institute of Aboriginal Studies, Canberra. Hudson, Joyce. (1983) Grammatical and semantic aspects of Fitzroy Valley Kriol, Summer Institute of Linguistics, Work papers of SIL-AAB, Series A, Volume 8, Darwin. Jensen, John T. (1977) Yapese reference grammar. University Press of Hawaii, Honolulu. Kilham, Christine, Mabel Pamulkan, Jennifer Pootchemunka and Topsy Wolmby (1986) 17 HEADER FOR MPLAL EDITORS’ USE ONLY Dictionary and sourcebook of the Wik-Mungkan language, Summer Institute of Linguistics, SILAAB, Darwin. Koch, Harold (unpubished) Kaytetye grammar notes. Lichtenberk, Frantisek (2000) Inclusory Pronominals. Oceanic Linguistics, Volume 39, Number 1, pp1-32. McGregor, William (1990) A functional grammar of Gooniyandi, John Benjamins, Philadelpia. Morphy, Frances. (1983) Djapu, a Yolngu dialec, In R.M.W. Dixon and Barry J. Blake, (eds) Handbook of Australian Languages Volume 1, Australian National University Press, Canberra, pp1-304. Nash, David (1986) Topics in Warlpiri Grammar, PhD thesis, MIT. Sandefur, John R. (1979) An Australian creole in the Northern Territory: a description of Ngukurr-Bamyili dialects [Part 1]. Summer Institute of Linguistics, Work papers of SIL-AAB, Series B, Volume 3. Darwin. Schnukal, Anna (1988) Broken: an introduction to the creole language of the Torres Strait. Pacific Linguistics, Canberra. Schwartz, Linda (1988a) Conditions for verb-coded coordinations In Hammond, Michael, Edith A. Moravcsik and Jessica R. Wirth (eds) Studies in Syntactic Typology, Typological Studies in Language 17, John Benjamins, Amsterdam, pp53-73. Schwartz, Linda. (1988b) Asymmetric feature distribution in pronominal 'coordinations'. In M . Barlow and A. Ferguson (eds) Agreement in Natural Language: Approaches, Theories, Descriptions, Center for the Study of Language and Information, Stanford University, Stanford, pp237-249. Wilkins, David P. (1989) Mparntwe Arrente (Aranda): studies in the structure and semantics of grammar, PhD thesis, Australian National University, Canberra. 18