The Tense, Aspect and Modality System in Murrinh-Patha

Abstract

Abstract This paper is an overview of the tense, aspect and modality (TAM) system in Murrinh-Patha, a polysynthetic language of the Daly River region of northern Australia. Our aim is to provide a detailed account of the range of TAM categories and their uses, highlighting along the way areas of interest for theoretical semantic analysis. We build here on earlier descriptions of tense, aspect and mood in Murrinh-Patha, most notably work by Street, and argue for reanalysis and refinement in various places, resulting in what we believe to be a more accurate and revealing analysis of the overall system. Keywords: TenseAspectMoodAustralian Aboriginal LanguagesSemanticsMurrinh-Patha Acknowledgements This work was conducted as part of the Marie-Curie project TAMEAL (Tense, Aspect, Modality and Evidentiality in Australian Aboriginal Languages), Grant Agreement PIRSES-GA-2008-230818-TAMEAL, and benefited from the financial support of the 7th Framework Programme of the European Commission. We also gratefully acknowledge financial support from the Arts Faculty at the University of Melbourne, and from the Australian Research Council for funding Rachel Nordlinger's fieldwork on Murrinh-Patha (DP0343354 Reciprocal across languages and DP0984419 Doing great things with small languages), as well as material support from the Faculty of Arts, Humanities and Social Sciences at the University of Western Australia. For comments and helpful feedback on earlier versions of this work, we are grateful to Joe Blythe, John Mansfield, Marie-Eve Ritz, two anonymous reviewers, other collaborators on the TAMEAL project and audiences of IPrA 2009, ALW 2010 and ALW 2011. We are particularly grateful to Joe Blythe for providing access to unpublished field data. Of course, all errors of fact and analysis are our responsibility. Notes 1Austin (Citation1998), although entitled Temporal Reference, Tense, and Aspect in Australian Aboriginal Languages, provides very little detailed discussion of the grammatical tense/aspect/modality systems, including only a brief overview of the categories in Jiwarli and Diyari. 2This introductory section draws in part on Nordlinger (Citation2011). 3Unlike many other northern Australian languages with similar verbal systems (see, for example McGregor Citation2002), in Murrinh-Patha and other languages of the Daly River region the two parts of the complex predicate form a single word from all perspectives: phonological, morphological, distributional, etc. 4The following abbreviations are used in the examples: ben 'benefactive/recipient marker', clf:anim 'noun classifier, animates', clf:human 'noun classifier, humans', clf:lang 'noun classifier, language', clf:thing 'noun classifier, neuter/residue class', clf:time 'noun classifier, times and places', clf:tree 'noun classifier, trees', clf:veg 'noun classifier, non-meat food', clf:water 'noun classifier, water', clf:weapon 'noun classifier, weapons', dat 'dative case', dm 'discourse marker', du 'dual', emph 'emphatic marker', excl 'exclusive', exist 'existential mood', f 'female', foc 'focus marker', fut 'future tense', FutIrr 'future irrealis', incl 'inclusive', loc 'locative', m 'male', neg 'negative', nfut 'non-future tense', O 'object', PImp 'past imperfective', pc 'paucal', pl 'plural', PstIrr 'past irrealis', rdp 'reduplicated', rr 'reflexive/reciprocal', S 'subject', sg 'singular'. Additional abbreviations are used in the description of the verbal template in Table 1 and are explained in the key provided there.The various classifier stem paradigms have traditionally been glossed with a number, in the absence of a complete analysis of their semantics. In this paper we provide an indicative gloss wherever possible (although note that this is mostly tentative at this stage) and include the number in brackets as well. Thus 'hands(8)', is the gloss for the paradigm denoting actions primarily done with the hands, traditionally referred to as paradigm 8. 5In these, and many other examples in this paper, morphophonemic processes have affected the surface form of the verbs (see Street Citation1987 for details). Thus, we include two lines of Murrinh-Patha text: the first line represents the surface form, and the second line represents the underlying morphological constituency. 8Note the presence of the future-tense marker on the adjectival nominal patha 'good'. It is not yet clear the extent to which TAM marking is possible on non-verbal elements in Murrinh-Patha, but it is found in a few examples in the corpus (another is given in example (104) below), and is known to exist cross-linguistically (e.g. Nordlinger & Sadler 2004). 6Many of the paradigms are similar in form, and some are very rare, making it hard to establish a definite number. Walsh (Citation1976) lists 33 paradigms while Street (Citation1987) lists 36 (although some of those listed in Walsh are not included in Street's list, and vice versa). Collaboration between Rachel Nordlinger, Joe Blythe and Nick Reid in 2005, in which we reconciled these two earlier sources and did further testing on the different paradigms, has led to the current number of 38—see Blythe et al. (Citation2007) for further details. 7These are sit(1), lie(2), stand(3), be(4), aloft(5), go(6), say/do(34), take(22) as well as the following three that have different semantics when used alone as a simple verb than they do when used in a complex predicate: feet(7) (which means 'action done with the feet' when used in a complex predicate, but something more like 'travel' when used as a simple verb); hands(8) (which means 'action done primarily with the hands' when used in a complex predicate, but 'say/talk (to someone)' when used as a simple verb), and the reflexive/reciprocal counterpart hands:rr(10), which means 'say/talk (to each other/oneself)' when used as a simple verb. 9This point will be important for the discussion of the non-future tense category in Section 4.1. 10The many cases of lexicalized, non-productive/non-transparent verbal meanings found in Murrinh-Patha could potentially be treated using the non-monotonic properties of Type Composition Logic (TCL (Asher Citation2011))—a formalism possessing: (i) a fully explicit semantic type system distinct from logical typing qua logical form (for instance, in TCL the set of unary predicates would comprise a wide range of semantic types (e.g. Colour, Shape, etc.), although they all possess type-logically similar logical forms); and (ii) an explicit, non-monotonic inference-based semantics to pragmatics interface, integrating e.g. world-knowledge and contextual information along with proper linguistic content in order to determine the interpretation of logical form (thereby offering principled interpretative mechanism for lexical as well as morphosyntactically complex expressions). 11For example, if α and β are the respective denotations of the classifier stem and lexical stem, then the denotation of the verb template would have the following form : [α] φ ∧[β] ψ (where [σ] τ indicates that σ has semantic type τ). 12In McGregor's (Citation2002: 29ff) analysis of such verb classifying systems, these meanings arise from the semantic feature of vectorial configuration; the other relevant features being Aktionsart and valency. 13This has important consequences for a formal semantic analysis of verb derivation in Murrinh-Patha for a theory of complex semantic types would then be required to handle the semantics of Murrinh-Patha verb derivation; the conjunctive types and dotted types implemented within Asher's (Citation2011 Type Composition Logic are formal possibilities we intend to investigate in future work. This would signal an important structural semantic difference from the account of Panyjima verb derivation offered in Caudal et al. (Citation2012), this volume), and which is based on semantic typing mechanisms within semantic dependencies (not mechanisms related to conjunctions of semantic types). 14Reduplication of lexical stems also encodes aspectual information marking repetition and iteration (see Street Citation1980), but we leave aside discussion of reduplication here as it does not appear to be part of the inflectional system, being only semi-productive and lexically idiosyncratic. 15Although large amounts of syncretism within paradigms means that not all paradigms distinguish all six categories formally. 16Subject agreement is complex in Murrinh-Patha, with the person/number classifier categories interacting with additional number markers to encode a four way number distinction (singular, dual, paucal (i.e. 3–10), plural) and a sibling vs. non-sibling distinction in the dual and paucal numbers. The basic patterns are represented here—for a more detailed discussion see Nordlinger (Citation2010a). The TAM category 'Exist(ential mood)' is discussed in Section 4.4 below. 17Some variations to these co-occurrences are possible in restricted contexts, and these are discussed in more detail below, with a fuller picture of the total system given in Table 5 in Section 6. 18It may be that this serialized construction is not an imperfective viewpoint marker expression per se, but rather creates a context in which an utterance can receive an imperfective viewpoint aspect interpretation, by ensuring that the clause at stake denotes an atelic event (i.e. contributes the proper event structure allowing an aspectually underspecified tense with a 'default' aspectual viewpoint to be interpreted as imperfective—we adopt here the terminology of Bohnemeyer and Swift (2004)). If this analysis is correct, sentences marked with the serialized construction and an underspecified tense such as the non-future could well be capable of inchoative, perfective viewpoint readings, given an appropriate change-of-state context. This is a common cross-linguistic property of the combination of atelic clauses with underspecified viewpoint tenses (cf. Bohnemeyer & Swift Citation2004). However, our current corpus provides no relevant examples, so this hypothesis will have to be checked by further fieldwork. 19Green (Citation1989) also notes that in Marrithiyel the event structure properties of the lexical verb stem can play a role in whether serialization occurs—some lexical verb stems have inherent atomic/punctual reference, and so can't appear with serialization, and others must. An examination of whether this analysis is true for Murrinh-Patha is yet to be undertaken. 20Taken from the book Mi Kilen produced by the OLSH School Wadeye. 21Taken from Kanhi-ka nanthi thay bere ngatha warda book produced by OLSH Wadeye school. 22Although he uses the term 'perfect', it is clear from Street's discussion that he intends this to mean 'past perfective'. 23Dyirbal is another Australian language with a non-future vs. future distinction (Dixon Citation1972: 55). The 'actual' categories in Kayardild (Evans Citation1995: 256–257) and Lardil (Hale Citation1997) also appear to share properties with the non-future tense described here, but further analysis is required to determine the interaction with Aktionsart for these categories. 24Another example, this time involving the classifier stem go(6) used in a clearly past perfective context, is given in example (71) below. 25As a reviewer correctly points out, the atelicity of the verbal predicate in these examples is shown explicitly by the reduplication of the lexical stem. 26In order to achieve such a backgrounded, imperfective reading of e 2, one would need to describe this event in the progressive: John was taking the 6 a.m. train. 27It is one of the great achievements of a theory like Segmented Discourse Representation Theory (Asher & Lascarides Citation2003) to have demonstrated how discourse structure and world-knowledge or other contextual inferences about the causo-temporal ordering of events in discourse could determine or even alter event structure information. For example, stative predicates involved in a causal sequence of events can thus be inchoatively viewed, cf. John laughed. And then Mary was gone. 28Malchukov (Citation2009) offers an explanation in terms of semantic coercion for this generalization that perfectively viewed events are apparently incompatible with a present temporal anchoring. While there is certainly something appealing about the idea that present and perfective are generally mismatched semantic categories, we believe this proposal to be in need of further refinements from a technical perspective (coercion is a notoriously problematic mechanism as it tends to overgenerate (cf. Bary 2009; Bonami Citation2002)), and an empirical point of view (as it is, Malchukov's analysis predicts that actual present narratives (i.e. non-historical presents) should not exist, but they do, in sports reports for example). 29We are here assuming a third category of aspectual viewpoints, namely resultative viewpoints, on top of Smith's (Citation1991) perfective and imperfective viewpoints (cf. Caudal & Roussarie 2006; Caudal 2012); it corresponds to the typical aspectual contribution of a perfect tense. Note that examples like (67) with a controlling agent show that these examples must be true perfect uses of the inflection, and not stative resultatives derived from the verbal combination with a present imperfective reading. Examples without a controlling agent, such as example (66), are ambiguous between the two possible readings. 30This pragmatic enrichment can be seen as a case of aspectual metonymy (or ellipsis): although the speaker only describes part of a given eventuality type by using an imperfective viewpoint (imperfective viewpoints being known to focus on an internal sub-event of a given event description: thus in John was washing the tray, the speaker only describes part of a 'John-wash-the-tray' event), contextual and world-knowledge based inferences make it clear that what she actually intends to describe is the whole of the associated event type, very much as if a perfective viewpoint tense had been used. Thus, this perfective-like description arises as the result of a pragmatic inference—in effect, some sort of implicature. See Caudal and Vetters (Citation2005) for a detailed analysis of such pragmatically enriched readings of the French imparfait. 31The use of the be(4) classifier may also contribute to the atelic nature of the predicate in this example. 32This is the same as in Ngan'gityemerri (Reid Citation1990). 33Note that in this table, the right two columns (headed nfut and PImp) represent the morphological categories, while the cell contents reflect the possible readings with different event types. 34We are grateful to Joe Blythe for pointing these examples out to us. 35A reviewer notes that it would be more systematic to label the distinction 'past irrealis' vs 'non-past irrealis' since this more accurately captures the range of functions and is also more consistent with the analysis of 'future' vs 'non-future' in the realis paradigm. While we take these points, we have stuck with the label of 'future irrealis' for this category since it reflects the formal association with the future realis (recall from above that these two categories are only ever distinguished in the third person, and in some paradigms not at all). 36We call 'foreclosed counterfactuals' those which convey possibilities that are no longer accessible at speech time (vs 'open counterfactuals'): if you had come/you should have come (foreclosed) vs if you come/you should come (open). 37Note that in example (104) the past imperfective inflection also appears on a pronoun (see also example (5) for an example of fut on an adjective). The extent of the marking of TAM on NPs in Murrinh-Patha is yet to be investigated, but note that it is not uncommon cross-linguistically for languages to encode TAM on nominals (see Nordlinger and Sadler (Citation2004) for a discussion). 38Related uses of the French imparfait are well-documented, as in (i), and are grammaticalized with idiomatic comparative constructions, which probably contribute a modal, counterfactual meaning (Berthonneau & Kleiber Citation2006).(i) Un peu plus et le train déraillait. A little more and the train derail-IMPF.3sg 'The train nearly derailed (but it didn't).'Russian past imperfective also has related uses—but it apparently does not even require an idiomatic construction to appear (see Grønn 2008).