ABSTRACT. Intonation plays an important role in the interpretation of spoken discourse. Languages use pitch modulation, in combination with duration and intensity, to signal information structure and the intonational realisation of focus. However, the ways that languages, and language varieties, mark prosodic focus differs (Ladd, 2008).

Studies on English intonation have shown that, overall, listeners can accurately distinguish the location of focus, i.e. subject focus, on the basis of f0 (e.g., Rump & Collier, 1996). There is also evidence that listeners are less accurate in discriminating sentences with broad and object focus (Breen at al., 2010; Welby, 2003). Most of the perception research, however, has been conducted on established varieties of English (i.e., British English).

This study aims to fill a number of gaps in the literature by investigating how listeners of Indian English (IndE) and Australian English (AusE) perceive narrow vs. broad focus. This is important because we currently understand little about the ways prosodic marking influences semantic meaning in new and established Englishes, and how listeners cope with the intonational variability they encounter. IndE is of interest because it has one the highest populations of English speakers in the world (Crystal, 2004), and a unique influence of typologically distinct substrate languages on its intonational system. In Hindi and Tamil, for example, focus does not bring any change in accentuation (Patil et al., 2008; Keane, 2014), in contrast to post-focal deaccenting in AusE.

We report on a perception experiment conducted with 33 speakers of IndE (Hyderabad, India) and 21 speakers of AusE (Melbourne, Australia). Listeners completed a semantic task in a four-alternative forced choice paradigm to show how accurately they could distinguish focus type (narrow vs. broad) and location (subject, verb, object).

We fitted generalized linear mixed effects models using lme4 in R to test for significant effects. Both groups performed better when listening to narrow focus compared to broad focus (p < 0.05). In terms of overall categorisation of focus, AusE listeners did slightly better at the task than IndE (84% correct vs. 79%, p < 0.05). Patterns of response to different focus conditions varied according to listeners’ variety of English, in combination with the variety of English stimulus they responded to. AusE listeners performed substantially better (p <0.05) in the broad focus condition (71% correct) compared to IndE listeners (35.6% correct); both groups also performed better while listening to IndE stimuli. IndE listeners outperformed AusE listeners in object focus condition (90% correct vs. 75%, p < 0.05).

When taken as a whole, these results correlate with the way prosodic marking is used by speakers in production: AusE speakers rely on accent placement in combination with f0 cues, while IndE speakers rely on increase/decrease in f0, treating accent placement as optional. In line with more recent research on new Englishes (e.g Gussenhoven, 2014 - Nigerian English), the findings suggest that IndE is more challenging to fit into a single prosodic typology category, e.g. with head marking prominence like AusE, raising a broader theoretical question of typological distinction in English(es).

ABSTRACT. In many languages informational focus is marked intonationally in similar ways in declarative and interrogative utterances. In a study of Japanese and Korean (Ueyama and Jun 1998), utterance pairs like “I DON’T like horror movies” and “Do YOU like horror movies”? were compared and there was a strong trend in both constuctions for the word that carried informational focus to have a wider pitch span, and for following post-focal material to be incorporated in the same intonational constituent. The pitch span of post-focal material was also strongly compressed. Similar intonational patterns have been observed in Farsi WH-questions and declaratives, where the WH word or focused subject receives the major phrasal pitch prominence with subsequent material realised in a compressed pitch span (Scarborough 2007).

Previous studies of Mawng have shown that in a pragmatically-unmarked clause, an object NP generally follows the verb and the entire clause is often realised as a single major intonational phrase. However when the object or subject NP is in contrastive focus and precedes the verb, it is often realised as a separate intonational phrase, with a wider pitch span compared to following material in the same utterance . In sum, prosodic phrasing and phonetic pitch range manipulation tend to make the focal element “stand out” relative to the verb in a variety of OV, SOV, or SVO constructions. Broadly similar patterns are observed in studies of topic and focus in other Australian languages, including Jaminjung , and Djambarrpuyngu.

WH-question realisation in the Iwaidjan language Mawng was compared to focus-marking in the current study. Intonational phrasing patterns, accentuation patterns, and pitch span were investigated across three experiments. The first two manipulated object and subject focus and the third analysed WH-questions. Results suggest similar broad prosodic strategies:WH-words if present tend to be utterance initial and realised in the highest part of the speaker’s pitch range often with a strong rising prominence-lending pitch movement, similar to patterns observed when contrastive focus is realised on initial subjects or objects. However, material after the WH-word is usually incorporated into the same major intonational phrase. In the latter, there is more evidence of phrasing manipulation with many focused NPs realised as full intonational phrases with left dislocation.

In summary, the patterns of intonational realisation for WH-interrogatives and contrastive focus constructions are very similar in Mawng as in other languages, but with some minor differences. Initial position is always locus of pitch span widening and reset but in case of Interrogatives, speakers often use a wider pitch span on the WH- word with steeper local pitch downtrends within the same intonational constituent. We present these results in the light of recent developments in prosodic typology.

References

Ueyama, M. & Jun, S-A. 1998. Focus realization in Japanese English and Korean English Intonation. Journal of Korean Linguistics, 7, 629-45.

Scarborough, R. 2007. The intonation of focus in Farsi. UCLA Working Papers in Phonetics, 105, 19-34.

ABSTRACT. The study proposes a typology of the main semantic-pragmatiс functions of the new Russian quotative marker tipa based on the analysis of conversations in a student online forum (2013-2016). The classification builds on the parameters of (i) discursive embedding of the reported information, (ii) its source and (iii) reliability according to the actual speaker. The data shows uses which are currently specific to the younger generation, as well as those which are firmly entrenched in the language. Special attention is given to the fact that tipa represents two crosslinguistic tendencies of language change: the rise of new quotatives (Buchstaller 2014, Foolen 2008) and the grammaticalization of taxonomic nouns (e.g. sort of/kind of/type of constructions in English, French genre, espèce, Portuguese, Spanish, Italian tipo, Norwegian typ, sort, slag, etc; see, inter alia, Denison 2011, Rosenkvist & Skärlund 2013, Voghera 2013, Mihatsch 2007, Kolyaseva and Davidse forthc.). The analysis demonstrates the semantic-pragmatic shifts that quotative and related uses of tipa have resulted from, both typical for the grammaticalization path of taxonomic nouns documented crosslinguistically, and unattested so far in any other languages researched in this regard. The analysis also confirms that functioning of tipa squares well with the language-internal tendency revealed in studies on traditional quotative markers mol, deskat’ and de: “the decay of the system of markers of renarrative evidentiality in Russian language and the start of ‘regrammaticalization’ as modal markers of subjective quotation” (Plungian 2008).

References: Data: “Подслушано в МГУ” [Overheard at MSU], a student forum. URL: https://vk.com/overhear_msu.

Buchstaller, I. 2014. Quotatives: New Trends and Sociolinguistic Implications. Oxford: Wiley Blackwell. Denison, D. 2011. The construction of SKT. Plenary paper presented at Second Vigo-Newcastle-Santiago-Leuven International Workshop on the Structure of the Noun Phrase in English (NP2), Newcastle upon Tyne. Foolen, A. 2008. New quotative markers in spoken discourse. In: Bernt Ahrenholz, Ursula Bredel, Wolfgang Klein, Martina Rost-Roth, Romuald Skiba (eds.), Empirische Forschung und Theoriebildung. Frankfurt: Peter Lang. 117–128. Kolyaseva, A., Davidse, K. forth. A typology of lexical and grammaticalized uses of Russian tip. In: Russian Linguistics. Mihatsch, W. 2007. The construction of vagueness. “Sort of” expressions in Romance languages. In: Radden, G., Koepcke, K.-M., Berg, T., Siemund, P. (Eds.), Aspects of Meaning Constructing Meaning. From Concepts to Utterances. Amsterdam – Philadelphia: John Benjamins. 225–245. Plungian, V. A. 2008. O pokazatelyakh chuzhoy rechi i nedostovernosti v russkom yazyke: mol, yakoby i drugiye [On markers of other’s speech and unreliability in Russian: mol, yakoby, etc.]. In: B. Wiemer, V.A. Plungjan (Hrsg.), Lexikalische Evidenzialitäts-Marker in Slavischen Sprachen,München: Sagner, 285–311. Rosenkvist, H., Skärlund, S. 2013. Grammaticalization in the present: the changes of modern Swedish typ. In: Giacalone Ramat, Mauri, Molinelli (eds), Synchrony and Diachrony. A dynamic interface. Amsterdam: Benjamins. 313–338. Voghera, M. 2013. A case study on the relationship between grammatical change and synchronic variation: the emergence of tipo[−N] in Italian. In: Giacalone Ramat and Molinelli Mauri (eds.), Synchrony and Diachrony. A dynamic interface. Amsterdam: Benjamins, 283–312.

ABSTRACT. When we consider computer-based spelling standardisation, we usually think of it as a cleaning up process, but it can do more than that. Working with under-studied and under-resourced languages often means that there is no standardised orthography. The unavoidable spelling inconsistencies can be a nuisance, but they can also be a virtue, because they tell us something about how the native speaker transcribers perceive the sounds of their own language. Every so often we stumble over these insights by accident when we work through our data. Computer-based consistency checks allow us to look at these inconsistencies in the transcription in a much more organised manner. In this talk we discuss how computer-based consistency checks can be used to explore research questions in under-resourced languages.

The software company Appen has developed a range of scripts for spelling standardisation in their speech databases, which are large-scale collections of audio data with accompanying transcription and lexicons. Their Clients typically use these speech databases to train automatic speech recognition systems. The standardisation scripts identify spelling inconsistencies in the transcriptions, and the outputs of the scripts are used in various ways to standardise the spellings. Standardised spelling is crucial to the goals of automatic speech recognition.

In this presentation we demonstrate the use of these scripts using Yelmek, an endangered language spoken in the south of New Guinea, as a case study. There is no standardised spelling or writing tradition in this language, but there is a small digital corpus of recordings that has been collected during recent fieldwork. Most of these recordings have been transcribed by native speakers. Based on this data, we will show how a range of questions can be approached using Appen scripts, such as voicing contrasts in plosives, consonant deletion and /h/ insertion.

This shows how something tedious like spelling inconsistencies can be turned into something insightful for research on under-resourced languages, and will contribute to the stock of techniques that can be used when investigating other languages in similar situations.

ABSTRACT. Features of the sound change known as the California Vowel Shift (CVS; Eckert 2008) have been found throughout the American West (Fridland et al. 2016). One CVS feature, TRAP backing, is associated with California and Californian values both in the popular media (Pratt and D’Onofrio 2014) and in the ears of Californians themselves (Villarreal 2016). It remains to be seen whether TRAP backing indexes localness and/or local values, or remains indexed as Californian, in communities undergoing the CVS outside of California. The Midwestern state of Kansas provides an ideal test-ground for this question as TRAP backing is a recent change within Kansas (Kohn and Stithem 2015), and as perceptions of Kansas contrast starkly with the liberal, urbanized image of California. In light of these associations, this study investigates the local construction of meaning for a supra-local sound change by examining how listeners in Kansas perceive TRAP backing. Fifty-one university students participated in a perceptual task drawing from dialect recognition (Williams et al. 1999) and matched-guise (Campbell-Kibler 2007) methodologies, with stimuli read by young Kansan speakers. In each trial, listeners identified speakers’ regional origin and rated speakers on 14 affective Likert scales. Four critical stimuli belonged to one of two matched guises, which were acoustically manipulated (using Villarreal’s [2016] method) such that only TRAP F2 differed between matched guises. Conservative guises contained fronted (higher F2) TRAP and shifted guises contained backed (lower F2) TRAP. A Bayesian analysis of regional identification found that listeners identified the shifted guise as significantly more likely to be from California than the conservative guise (95% confidence interval for guise difference: [–0.167, –0.0112]), indicating some association among listeners in Kansas between TRAP backing and California; however, guise did not affect any other regional identification category. With respect to scales data, a principal components analysis revealed three principal components that together accounted for 61% of the variance in ratings. PC1, measuring “general prestige,” combined seven scales, including likeable, polite, and educated; PC2, measuring “Kansan-ness,” combined two scales: Kansan and small town; and PC3, measuring “innovativeness,” combined three scales: young, feminine, and fast. PC1 and PC3 significantly correlated with guise (p < .05), with shifted guises rated higher on “general prestige” and “innovativeness.” Conversely, PC2 significantly correlated with listeners’ regional identifications (p < .001), as stimuli rated high for “Kansan-ness” were most likely to be identified as being from Kansas and least likely to be identified as being from New York or California. These effects were apparently orthogonal; regional identification categories varied little on “general prestige” while guise categories varied little on “Kansan-ness.” These results suggest that in Kansas, TRAP backing is associated with California despite Kansan participation in the CVS. Instead of associating TRAP backing with local identity, as Californians do, this sound change appears to index prestige and youth in Kansas, perhaps motivating the spread of this sound change in the region. These results provide some clues to the rapid supra-local spread of the CVS, highlighting the local construction of meaning for a supra-local sound change.

ABSTRACT. Australian languages differ greatly in terms of their structure yet they are thought to share similar phonologies (Hamilton, 1996). There have been relatively few studies of the phonetic patterns found across Australian languages however (see Fletcher and Butcher, 2014 for an overview). As more linguistic data is collected there is a growing need for an Australian specific annotation method enabling time aligned labelling data to be synchronised to recordings. This paper reports a method for training an acoustic model to automatically segment Australian language speech recordings. The need for a specific model has emerged due to some typologically unusual phonetic features found across Australian languages. The current study uses the Munich AUtomatic Segmentation (MAUS) (Schiel et al., 2011) system to train an acoustic model using existing recordings from a modest selection of Australian languages. There has been previous success with segmenting non-Australian languages using a generic model derived from many sources (Strunk et al., 2014). This uses the SAMPA (Speech Assessment Methods Phonetic Alphabet) machine readable format. This does not work well for Australian languages however as there are some phonemes missing from this super-set, including the common retroflex series as well as lamino-dentals and pre-stopped nasals. These phonemes must use existing mapping for example apico-alveolar models for the dentals and plain nasals for the pre-stopped nasals. This is not always successful however as there are different acoustic transitions for these phonemes. This modified SAMPA model has been applied successfully to Australian language data although it required significant amounts of hand correction. To train a new Pan-Australian model significant levels of hand corrected labelled data were required. The data were drawn from a corpus of Australian languages gathered by Andy Butcher which was first automatically segmented and then hand corrected (See Table 1, see Figure 1). Additional recordings were added from data supplied by other researchers and hand-labelled at a phonetic level. The Butcher corpus is a phonetically balanced corpus of speech, in which all phonemes are recorded in word initial, word medial and word final position. This format is an ideal training set for a machine learning model, particularly when paired with short sentences and narrative.

The data were all standardised for labelling and then compiled into a database using the Emu-SDMS (Winkelmann et al., 2017) with tiers, Utterance, ORT (orthography), SAP (SAMPA) and PHO (Phonetic). Data used to test the model have not been included in the training set. Comparing this model to the SAMPA model, shows a significant improvement in the recognition of the segments. Early results also show that it is possible to train a Pan-Australian automatic segmentation model for MAUS that works for other Australian languages using a relatively small sample. The full workflow for this system is web accessible and can take orthographic input which when processed along with audio data returns a segmented and time-aligned annotation file (Kisler et al., 2017). This can then use further processing tools to generate morpheme, word and utterance tiers to make audio fully searchable. The aim is to make the model web accessible.

ABSTRACT. Digital technologies are being used by Indigenous language authorities for various purposes, including communication, documentation, preservation, revival, maintenance, teaching, and promotion. This project aims to investigate the ways in which the technologies support or inhibit these practices, as well as how they may in fact be configuring each other. In response to the limited opportunities to learn Indigenous languages at Australian universities, a template was developed for a Digital Shell, using WordPress and other open source tools. In 2016 a pilot program for teaching Bininj Kunwok (from West Arnhem Land) was developed and delivered in collaboration with the language owners (http://www.cdu.edu.au/centres/language-shell/). Since then, other language groups have expressed an interest in using the shell for sharing their own language and culture online. This presentation will report on how the Digital Shell is coming to life in languages in very different situations and for different audiences. The Ngukurr Language Centre is aiming to extend their face-to-face Kriol language awareness course in an online format. Kriol is a contact language widely spoken across the Katherine region of the Northern Territory, and many non-Indigenous people are keen to learn some language in order to facilitate interactions with Indigenous Kriol speakers. The Muurbaay Language Centre is working on using the Digital Shell to deliver Gumbaynggir language classes. As a language in revival in northern NSW, a number of Gumbaynggir people outside of the language area have requested opportunities to learn the language online. The collaborative nature of the Digital Shell project allows for exploration of the ways in which technology comes to life under the authority of the language owners. A comparison of these three different realisations of the Digital Shell provides a basis for research into how language authorities can use digital technologies for their own purposes in collaborative arrangements with technicians and language learners.

ABSTRACT. Increasingly, Aboriginal community members are seeking to both maintain the strength of the more widely spoken languages and revive those with few or no speakers (FATSILC, 2017). In this context, the role of documentation shifts from a fundamental focus on preservation of language and culture for posterity to an increased emphasis on the value of written, audio and audio documents as a resource for the present.

The purpose of this paper is to present outcomes from the first two phases of a project re-purposing older documentation to support the communicative teaching and learning of Mangarrayi at Jilkminggan, a remote Aboriginal community south-east of Katherine, NT. In this community only the most elderly individuals are fluent speakers of Mangarrayi.

In Phase 1, we asked younger adult community members what they would like to be able to say in Mangarrayi. The overall goal of this phase is to build a bank of words, word strings and language structures categorised thematically, with a focus on basic communication in day-to-day contexts. As heuristics we began with a framework developed for teaching European languages (Van Ek & Trim, 1998), informed by semantic domains specific to Aboriginal languages, and situations and contexts from Aboriginal language programs, for example Kaurna (Amery & Simpson, 2013).

Through workshops hosted at Jilkminggan community members worked together to identify contexts and themes in which to frame the teaching and learning of Mangarrayi for adults and children with varying background knowledge of the language. The process also involved the identification by community members of specific language relevant to selected themes. This enabled us to establish a list of core themes and topics from which we selected the four ranked as most important by a majority of participants: Family, Friends & Other People, Communication, Food & Drink, Daily Life, to begin eliciting specific expressions.

In Phase 2 we turned to the question of how older audio or video documentation might be practically repurposed for learning, as spoken exemplars of relevant words and expressions, by community members in a sustainable way. Through small group discussions and individual interviews, we asked community members how they might most easily reformat and derive useful audiovisual resources from older recordings. Specifically we involved middle aged adults, young adults and teenagers in a digital technology use survey and a conceptualisation task involving re-purposing of older resources (see Figure 1). The main findings from this phase were that, although internet and telephone access within the community are quite limited, many community members across a range of ages, considered technology to be “quite important” or “very important”. Whilst most community members had difficulty identifying technology or software that would allow them to extract the sound or images from a digitised video, many displayed more confidence in using programs such as Powerpoint to create resources using these. Overall results suggest that with specific training re-purposing of older digital resources could play a positive role in language revitalisation at Jilkminggan.

ABSTRACT. OLSH Thamarrurr Catholic College, a K-12 school located in Wadeye, NT, has run a bilingual enrichment program (May, 2008) with bi-literacy instruction (Murrinhpatha and English) since 1977 (Bunduck & Ward, 2017). Initial literacy instruction in the first three years of schooling is delivered through Murrinhpatha using a variety of resources developed by the Wadeye Literature Production Centre (LPC). This paper documents an ongoing project to develop a levelled reader series for early years literacy. We examine a number of linguistic and sociolinguistic challenges in developing this series.

The LPC currently has in excess of 300 texts for use in classrooms, however it is often acknowledged by both Indigenous and non-Indigenous teachers and assistant teachers that many texts are too difficult for students to read independently. Consequently, the LPC is developing a levelled reader series entitled Murrinh kardu mamay nukun. This series aims to match texts to students reading abilities and needs, inform targeted and individualised teaching/learning intentions and ultimately establish confidence in independent reading skills. Beginning readers benefit greatly from books they are able to read and that are appropriate for acquiring and monitoring reading strategies at the various levels of reading progression (e.g. Davidson, 2013). Along with quality instruction, quality targeted materials are fundamental for student learning.

The development of Murrinh kardu mamay nukun relies on the use of levelling criteria to determine the complexity of different texts. The levelling criteria used have been adapted from English-medium levelled text series (e.g. Fountas & Pinnell, 2006). The criteria relate to vocabulary, book format, text length, illustrations, content and concepts and text predictability (e.g. Rog & Burton, 2001). As a result of these criteria we identified appropriate frequently used ‘sight words’ (murrinh thamam thurran) for use in texts, identified pre-existing texts for use as levelled readers, levelled pre-existing texts and began production of new texts for the series.

There are a number of challenges in developing Murrinh kardu mamay nukun. These are mostly related to vocabulary and more specifically word length and lexical choice. It is necessary for authors to consider what types of vocabulary are decodable and accessible for students at various reading levels. The fact that Murrinhpatha is a polysynthetic language with long complex verb structures (Walsh, 2011) means authors must pay attention to factors such as subject choice which influence verb length and ultimately may make many verbs unsuitable for use in early levelled readers. We have also found that students' lexical choices do not always match vocabulary used in texts. This is due to authors’ use of murrinh yitthit (heavy Murrinhpatha) not matching students home language which may be considered murrinh parnturtparn (light Murrinhpatha) (Mansfield, 2014). This necessitates ongoing discussions around the style of Murrinhpatha to use in texts as well as consideration of how students interact with them.

The initial stages of development and implementation have led to increased opportunities for independent student reading, monitoring of student reading progression as well as contributing to the legitimacy of literacy instruction through an Australian language.

ABSTRACT. In this position paper I address some issues in the study of TAM in under-described languages, which are of particular relevance for Oceanic languages and the Pacific area in general. Firstly, I focus on a widespread problem of diagnosing different TAM categories and their interrelatedness, and secondly, I discuss some specific cases of subject marking as portmanteau TAM morphology.

Recently, many languages that have been previously characterised as having tense categories have been reanalysed as mood or aspect-prominent languages (Smith et al., 2007; Ameka & Dakubu, 2008). Bhat (1999) proposes a distinction between tense, aspect and mood-prominent languages to account for the possibility that different TAM categories can be the ‘basic’ one in a language. Oceanic languages are well-known for having the realis/irrealis distinction (Lichtenberk, 2016), whose cross-linguistic validity has often been questioned (e.g. de Haan, 2012; Cristofaro, 2012). I will contrast some analyses of binary temporal distinctions (e.g. future/non-future) with binary mood distinctions (realis/irrealis), and discuss the criteria commonly used for diagnosing the categories at issue.

Following the issues about realis/irrealis, I will present some challenges for semantic analyses of subject marking as portmanteau TAM paradigms. By focusing on a few Oceanic languages with realis/irrealis distinction in portmanteau subject markers (e.g. Nafsan, Wogeo), I will show that the realis paradigm is often morphologically less complex (see also Haspelmath, 2016). In addition to that, realis typically behaves as a ‘default’ marking of the verb, inconsistent with the definition of realis as referring to the actual world. I argue that what has often been labelled as realis is in fact subject marking alone unspecified for TAM, and that the typical realis interpretation of the ‘zero’-marked subject markers is pragmatically inferred through its paradigmatic contrast with irrealis (see Prince, 2016).

ABSTRACT. Complex TAM marking (i.e. the association of several exponents in the morphological marking of a given category, or any other conventionalised association between one or several morphological exponents, alongside some given syntactic element) is an extremely widespread phenomenon in non-Pama-Nyungan (nPN) languages. A number of broadly modal categories such as conditional counterfactuals, epistemics, deontics/hortatives/volitionals, predictives/futurates, but also so-called frustratives/avertives, and negative past event descriptions involve such complex markings.

This presentation aims to bring together a number of observations related to (i) the morpho-syntactic make-up of complex modal systems in a small sample of nPN languages and (ii) their semantic characteristics. Our sample comprises Iwaidjan (Iwaidja and Mawng), Guwinyguan (Bininj Gun-wok ((Evans 2003) and Anindilyakwa (Bednall & Caudal 2016)), Daly River (Murrinh-Patha (Nordlinger & Caudal 2012) and Ngan'gityemerri (Reid 2011)) and Jaminungan (Jaminjung, (Schultze-Berndt 2000)) languages. Notably, we suggest that contrary to expectations, many readings associated with such complex configurations are far from systematically bearing the hallmark of semantic composition, and that conventionalisation in the form/meaning relations plays a crucial role in organising the semantic profile of complex modal expressions in nPN languages.

            A significant number of nPN languages possess a number of distinct morphological modal paradigms, described using various terms in the literature (including e.g. ‘future’, ‘potential’, ‘optative’, ‘irrealis’). Such paradigms typically involve discontinuous morphs, with two exponents realised on two disjoint positions in the verb template (i.e. such morphs typically are (–EXP₁– ~ –EXP₂–). For example, in many Iwaidjan and Gunwinyguan languages, the leftmost of these two positions is a portmanteau exponent combining modal and pronominal information, whereas a suffixed exponent position encodes only TAM information. Significantly, many of these bi-exponent morphs are semantically polyfunctional, with the observed polysemy effects pointing to a semantic cluster of meanings – this is, for instance, often the case with so-called ‘futures’/ ‘irrealis’, which commonly have volitional, predictive/futurate, and interactional/deontic meanings (including with negation).  In Iwaidjan languages, prohibitives (Pym & Larrimore 1979:104) thus formerly overlap with negative predictives, jussives, hortatives; and positive uses of the same form cover prediction, volition, and mild requests.

1. Iwaidja (Iwaidjan):
   karlu                 ang-mana-man

neg                    1p-fut-take-Ø  

‘Don’t take it / you won’t take it/you shouldn’t take it’ (without NEG: ‘let’s take it!’)

Another common feature of this cluster is to incorporate a significant kind of interaction between so-called irrealis forms and negation. This correlates with e.g. the existence of dedicated negative irrealis forms, and even more frequently with the combination of (a) so-called frustrative interpretations of the irrealis with negatives, and/or (b) the use of negative irrealis forms to convey bona fide negative past events, cf.:

1. Anindilyakwa (Gunwinyguan):
   nara                  n-akəna      kenu-kwa-Ø                a-rmdak-akəna       angwarnda
   neg                    3m-that       irr.3m/2-give-pst      neut-many-that       neut.money
   He didn't give you all that money (JL, ANI_2016-07_06_002, 00:17:59-00:18:04).

Rather than rely on either a compositional semantic approach, or a Horn scale-based productive pragmatic account à la Van Linden & Verstraete (2008), we argue that such uses can be explained in terms of conventionalised invited inferences in the course of the development of some IRRealis categories, thus reflecting a ‘frozen’ stage in the evolution of the language.

We will demonstrate that all languages in our sample exemplify exactly the same pattern, and are (probably) amenable to the same analysis. More generally, we will provide a converging series of datapoints suggesting that conventionalised phenomena, the conventionalisation of implicatures/invited inferences and other conventionalised occurrences largely determine the form/meaning relation in the case of complex modal expressions.

ABSTRACT. In this talk we discuss the range of TAM categories and their meanings in Kunbarlang, a polysynthetic Gunwinyguan language of the northern Australia. We place Kunbarlang as a variant of restricted prefix-suffix systems, type 2 (R-IRR prefixes and different tense for IRR) in morphosyntactic typology of Australian composite mood marking and highlight issues of theoretical and typological interest.

The TAM values are constructed by combining the meaning of the subject pronominal prefix with the meaning of the verbal stem form. There are four series of subject prefixes related to TAM (they are portmanteaux of the TAM and the person and number of the subject). There are also four stem forms for each verb. One of the puzzles for an explanatory account of Kunbarlang TAM is that out of the 4×4=16 logically possible combinations only five or six are actually possible (depending on the speaker).

We begin by observing that there are two main classes of combinations: realis and irrealis. Two of the subject pronominal prefixes and two of the verb stem types are realis. The realis categories are the past, the present and the future in declarative clauses. The irrealis forms obligatory occur under negation, but also appear in other modal contexts, discussed below. Since the subject and the verb stem must agree in their realis value, this reduces the possible combinations by two (2×2×2=8). The values of the subject prefixes in the realis class are future and non-future; for the verb stems these are past and non-past. One of the four realis combinations, future+past, is ruled out because of the feature mismatch.

Turning now to the irrealis class, we find further restriction of the combinatorial possibilities. For some speakers, the subject and verb stem are in a one-to-one correspondence, i.e. for them only two combinations are possible. These are irr-past and irr-non-past, where the subject prefix and the stem bear the same value. Other (younger) speakers allow combining the non-past prefix with the past stem. The result is past tense reference, which leads us to conjecture that for them this prefix is only specified as irrealis, but not specified for tense. The three irrealis forms that are grammatical for these younger speakers are not isomorphic to the three realis forms, i.e. they are not negative polarity conditioned allomorphs of the realis forms.

When used under negation, the irrealis forms distinguish the past and non-past tense reference. Without the negative particles, however, these forms have modal meanings. The non-past have an epistemic possibility meaning and can occur in the antecedent of a conditional. More puzzling is the past class. On the one hand, those forms can have counterfactual meaning. On the other hand, they can refer to remote past habitual situations. We discuss this apparent semantic conflict of actualized (remote past habitual) vs. non-actualized (counterfactual) event feature, showing that typologically speaking, it is not particularly unusual.

ABSTRACT. Nominalisation suffixes in Wiru

Wiru, or Witu, is a language spoken in the Southern Highlands of Papua New Guinea and appears to be a member of the Trans-New-Guinea language family (Kerr 1967). In Wiru, there are a number of suffixes which create noun-like constructions which adopt standard nominal syntax and phonology. These suffixes commonly attach to transitive verbs and determiners, and, more rarely, to adpositions, and always appear word-finally; where two or more of these suffixes appear in a clause they each attach to a separate determiner. While brief references are made to these suffixes by Kerr in his Wiru-English dictionary (1972), they have never been otherwise examined. This paper describes these suffixes and examines their contexts of use.

The most common of these suffixes are –le, which refers to an action, and –mo, which refers to an utterance. Thus, iki ‘they say’ becomes iki-mo ‘the thing they say’ or iki-le ‘the thing they say to do.’ While the suffixes can appear in all moods, these two suffixes appear most frequently in reported imperative constructions.

There is also –pere, a ‘temporal’ nominaliser, and –ya, a more typical nominaliser meaning ‘thing’ or ‘object’. Although the suffixes each have a separate meaning, they do not appear to be derived from existing nouns, with the exception of –pere which also exists in a noun form meaning ‘time’, and they must attach to another word. The boundaries of this suffix group are somewhat fluid and it is possible that a number of other suffixes in Wiru could be categorised as similar nominalising suffixes. One such suffix, -ra, meaning ‘place’, could mitigate the notable absence of such a meaning in the existing Wiru suffixes (Foley 1986).

Nominalising suffixes not unique to Wiru, and two similar suffixes can be found in Kewapi, another Trans-New-Guinea language (Yarapea 2006). However, the two Kewapi nominalisers do not carry their own specific meanings. It is these meanings, as well as the ability to attach to verbs, determiners, and adpositions, which set the Wiru nominalising suffixes apart from many nominalising constructions.

References

Foley, W. (1986). The Papuan languages of New Guinea (Cambridge University Press, Cambridge)

Kerr, H. (1967). A preliminary statement of Witu Grammar (Masters Thesis, University of Hawaii)

Kerr, H. (1972). The Witu to English Dictionary (published online)

Yarapea, A. (2006). Morphosyntax of Kewapi (PhD Thesis, Australian National University)

Examples

(1) Ne eni-pere eni-le t-a mene wa a-ke-nea. 2SG that-TEMP that-ACT do-2SG.IMP must QUOT say-3PL.PST-NEUT ‘That was when they told me to do it.’

(2) T-a mene wa i-k-i-le odene t-a mene. do-2SG.IMP must QUOT say-PRS-2/3PL-ACT only do-2SG.IMP must ‘Only do the things they say to do.’ (3) Ani-ya mo-a mano. up-NOM get.2SG.IMP give.2SG.1.OBJ ‘Bring me the thing up there.’

(4) I -mo u –k -u This-SPE say-PRS-1SG I’m saying this [thing].

Abbreviations ACT Action nominaliser (-le) IMP Imperative NEUT Neuter NOM Nominaliser (-ya) OBJ Object PL Plural PRS Present PST Past QUOT Quotative SG Singular SPE Speech nominaliser (-mo) TEMP Temporal nominaliser (-pere)

ABSTRACT. First proposed by Matsumoto (1988) in an analysis of Japanese, General Noun-Modifying Clause Constructions (GNMCCs) are noun-modifying clause constructions (NMCCs) which have both relative-clause-like interpretations – in which the head noun functions as an argument or adjunct in the subordinate clause – and interpretations in which the head is neither an argument nor adjunct of the modifying clause. GNMCCs have been identified in Japanese (Matsumoto, 1988; 1997), Korean (Comrie, 1998; Matthews & Yip, 2017), and several other languages of Eurasia. This paper makes the case that Wiru, a Trans-New Guinea language spoken in the Southern Highlands Province of Papua New Guinea, has a GNMCC. Although the work of Matsumoto, Comrie and Sells (2017) has made much progress, our understanding of GNMCCs continues to be limited and there has not to date been an extensive analysis of a GNMCC in a Papuan language.

As the name suggests, GNMCCs are syntactic constructions in which a head noun is modified by a clause. NMCCs in Wiru precede the head noun and take the form of a well-formed sentence (compare (1a) with (1c)). To obtain the G for ‘general’, the same NMCC must be able to take not only arguments (1) and adjuncts (2) of the modifying clause as its head noun, but also nouns which have an ‘extended relation’ to the clause. For example, the NMCC in (3) expresses the content of the head noun – the noun itself is neither an argument nor adjunct of the clause. Comparing (1), (2), and (3) reveals that Wiru uses the same NMCC for argument, adjunct, and extended noun-clause relations, confirming the presence of a GNMCC in the language.

This paper focuses on two typologically interesting features of the Wiru GNMCC – duplication of the head noun, and matrix-verb agreement with the internal NP. Though in most cases the head noun is not explicitly stated within the GNMCC (1a), ‘doubling up’ is not only acceptable (1b), but in fact obligatory when the modifying clause contains a pronoun with the same referent as the head noun (4). In such cases, the head noun is semantically inert and is chosen from one of four possibilities – aroa ‘woman’, ago/ali ‘man’, or yarene ‘group’, depending on the sex and number of the referent.

When the pronoun in question is first-person, a mismatch arises with the third-person feature of the head noun. When a GNMCC noun phrase such as this is the subject of the matrix clause, the matrix verb must show agreement with the pronoun within the GNMCC (4a). Syntactic agreement with the head noun is unacceptable (4b). Thus, it is the internal and not the external NP which triggers agreement in such cases.

This paper makes a valuable contribution to our understanding of GNMCCs by making the first extensive analysis of a GNMCC in a Papuan language. Furthermore, it investigates features such as head noun recapitulation and internal NP agreement in Wiru that are of great typological interest.

ABSTRACT. The idea that switch reference(SR) should be regarded as a more pragmatically-oriented device has been gaining support in recent years. The two dominant theoretical approaches concerning this issue frame SR as a reference-tracking device (Haiman & Munro 1983, Foley & Van Valin 1984, Van Valin & LaPolla 1997) and as a system for signaling event-continuity (Stirling 1993). This paper, using the linguistic data from Wiru, a language spoken by approximately 15,000 people in Southern Highlands District of Papua New Guinea, further analyzes the functions of switch reference under these two approaches and argues SR should be considered as an agreement between eventualities with reference-tracking as a sub-function. In order to understand the functions of SR in Wiru, this paper begins with a comprehensive description of the verbal morphology of same-subject (SS) and different-subject (DS) marking. In canonical SR, a marker on the verb of one clause (also known as medial clause) is used to indicate whether its subject has the same or different reference from the subject of an adjacent, syntactically-related clause (also known as final clause). Morphologically, inside canonical SS systems, only person+number is marked after the verb stem in medial clauses and the tense of the final clauses is applied to all the related clauses. As for medial verbs in DS, both tense and person +number are marked. Compared with the three-way tense distinction of past, present and future in verbs of final clauses, medial verbs in DS only distinguish two (past and present) tenses. The functions of these canonical Wiru SR are tracking the references without any exception. However, Wiru also provides some interesting linguistic data where “unexpected” marking is displayed and SR is determined by changes in the situation other than a referential switch. Stirling (1993) has proposed a broader functionality of switch reference—event continuity, in which includes 6 pivots of SR. Among them, the referential (dis)continuity, agentivity value of important protagonist and location of the event are supported by Wiru data. Stirling states in agentivity that unexpected SS marking may occur when the syntactic subjects are not coreferent, but the subject of the final clause does not introduce a new agentive participant (e.g. in the case of impersonal constructions). If it does, DS marking will be adopted instead (examples are shown in the following paper uploads). In Stirling’s approach, SS is about agreement between aspects of eventualities, DS is about disagreeing in at least one of the eventuality parameters. This study presents a general switch-reference description of a very under-studied Papuan language, and further supports the two theoretical approaches where SR should be considered as a pragmatically-oriented device: SR as a reference-tracking device and SR as a system for signaling event-continuity. Although the first approach seems adequate to account for the canonical Wiru SR systems, many Wiru examples tend to suggest that SR is about congruence between eventualities, with reference tracking as a sub-function.

ABSTRACT. The present data come from fieldwork on Ende, a Pahoturi River language spoken in South Fly, Papua New Guinea. The phenomenon in question involves a nasal segment that shifts or disappears when the root occurs in an unaffixed context.

Zoll’s (1994) analysis for floating features and latent segments shows that language-specific rankings of general constraints account for and predict their divergent behavior. These features/segments share a common property of being restricted to a parseable environment. They differ in that some float in search of a better environment and others are fixed. They also differ in whether they are parsed as features or as segments. This predicts a typology of four types (Table 1). The languages Chaha and Yawelmani display three of the four types of latent phonology, but not floating segments. Zoll’s presentation implicitly suggests that everything that floats is a feature. Ende verbs exhibit nasal segments that float, filling this gap and completing the typology.

Some Ende verbs carry a floating nasal segment, N, that is left-aligned but floats rightward to land before an obstruent and agree with it in place, while maintaining syllabic sonority. N never appears word-initially, even on obstruent initial verbs (e.g., ‘to protect’) to avoid violating onset-sonority. In Table 2, N appears root-medially (a) and root-initially (b). If a verb only has a root-initial obstruent (c), an obstruent that follows a consonant (f), or no obstruent (g), N fails to appear, unless the verb has a suffix with an obstruent, in which case the obstruent is replaced (d, f, h).

The three families of Optimality-Theoretic (Prince & Smolensky, 1994/2008) constraints SEGMENT STRUCTURE, PARSE-FEATURE, and ALIGN account for the patterns in the Ende data, just as they did for Chaha and Yawelmani.

(2) a. ALIGN ([N], L, Stem, L): Align the left edge of N with the left edge of the stem. b. Rationale: N docks on the leftmost obstruent. (3) a. PARSE-N: N should be parsed. b. Ranking: PARSE-N » ALIGN-N c. Rationale: It is better to parse N further to the right than to align N with the left edge. (4) a. SEGMENT STRUCTURE: “a convenient cover term for the group of undominated constraints which render certain segment sequences impossible” (Zoll, 1994) b. Ranking: SEGMENT STRUCTURE » PARSE-N » ALIGN-N c. Rationale: N will fail to be parsed if it violates segment structure.

If N were a floating feature, it should appear word-initially, like other consonants in the language. However, this is banned. Further evidence for N’s status as a segment comes from Ende verbal reduplication. Verb roots appear with and without inflectional affixes (a-h). Table 3 shows that in non-affixed contexts, monomoraic roots reduplicate (i), while multimoraic roots do not (k). Monosyllabic roots with N do not reduplicate (m), suggesting the N contributes a mora.

The Ende data presented here fill a gap in the literature on floating and latent phonology. Floating segments exist and fit in Zoll’s analysis for floating and latent features/segments.

ABSTRACT. Free affix ordering, though it is typologically rare, has been documented in genetically and geographically dispersed languages including Chintang (Kiranti, Nepal), Murrinhpatha (non-Pama-Nyungan, Australia) and Tagalog (Austronesian, Philippines) (Bickel et al., 2007; Mansfield, 2015; Ryan, 2010). Free affix order may be cast as a ‘problem’ for morphological theory, since it undermines what had previously appeared to be a typological certainty: that affixation is rigidly sequenced, in contrast to the relative flexibility of ordering in words and clitics (Anderson, 1992, p. 261). But it can also be seen as a valuable opportunity for morphologists: when a grammar allows speakers to select between variant affix sequences, which forms do they use more frequently? In this study we use a substantial corpus of naturalistic Chintang speech data (Bickel, Stoll, et al., 2017) to investigate the quantitative distribution of variants.

Hypothetically, speakers of free affix order languages could select sequences at random (i.e. with uniform probability), or they could gravitate towards arbitrarily preferred sequences on a morph-by-morph or lexeme-by-lexeme basis. But what we find in Chintang is that preferred sequences exhibit patterns related to grammatical features, and independent of specific morphs or lexical stems. We call this CATEGORY CLUSTERING – where different exponents of the same grammatical feature (e.g., different subject agreement suffixes) share a common morphotactic position or ‘slot’. Category clustering has not been explicitly proposed as a morphological principle as far as we know, though it is an implicit norm in major theories such as A-morphous Morphology (Anderson, 1992), Distributed Morphology (Halle & Marantz, 1993), and Paradigm Function Morphology (Stump, 2001).

Data from the AUTOTYP typological database (Bickel, Nichols, et al., 2017), provides evidence that category clustering is strongly favoured worldwide in morphological systems that use fixed affix ordering. Figure 1 shows that for all agreement role features (subject, patient, transitive agent etc), there is a bias towards using the same morphological position (i.e. ‘n.slots=1’) for all exponents of the feature. This distribution would be highly unlikely without category clustering – i.e. in a null model where exponents of the same feature are equally likely to appear in any slot.

Chintang speakers exhibit category clustering in their probabilistic selection of variant wordforms. Verbs with multiple inflectional prefixes allow any ordering of prefixes. However speakers show a quantitative preference for SUBJ-NEG- over NEG-SUBJ-, and this preference is consistent, and maintains much the same distributional weighting, irrespective of the SUBJ agreement value selected (1, 2). This pattern is exhibited independently of lexical stems. Similarly OBJ-SUBJ- and OBJ-NEG- sequences are preferred over their inverses, and this preference is exhibited for all OBJ- and SUBJ- prefixes. Speakers thus demonstrate a probabilistic version of CATEGORY CLUSTERING.

In summary, we find evidence for category clustering as a typological bias in grammars with fixed affix ordering, and evidence for category clustering in Chintang, with free ordering. In Bresnan and colleagues’ formulation, “soft constraints mirror hard constraints” (Bresnan et al., 2001). We argue that this favours an interpretation of category clustering as a morphological processing bias.

ABSTRACT. See PDF

ABSTRACT. The languages of the Goldfields region of WA consists of a number of Wati and Ngadju family languages. Historical research undertaken by Wilf Douglas, Carl von Brandenstein, Geoffrey O'Grady, Daisy Bates, Norman Tindale and Doug Marmion was considered along with new research with current speakers of the Goldfields region in order to chart the characteristics of the languages and dialects of the region.

This paper presents the language family tree by defining the characteristics of each group of languages based on the findings of the last 6 years study in the region. The dialects of languages are identified along with creolised languages which evolved during mission settlements.

Undertaking the language studies in a highly contested native title environment impacts on the outcomes of the language studies as people identify by contemporary social groupings rather than linguistic perimeters. These factors are identified where they impact on the language family tree.

ABSTRACT. Ergative marking and function are generally adequately described in the grammars of the small minority of Pama-Nyungan languages written before 1930. Without the benefit of an inherited descriptive framework in which to place foreign ergative morphosyntax, missionary-grammarians engaged a variety of terminology and descriptive practices when explaining the newly encountered ergative structures.

The term ‘ergative’ to name the case marking the agent of a transitive verb, and the conception of an absolutive case, became established practices in the modern era of Australian grammatical description (Silverstein 1976; Dixon 1972). This occurred without recognition that the same terminology and conception of syntactic case had previously been employed in the pre-academic era of Australian linguistic description. This paper presents the earliest usages of these terms and concepts.

Wilhelm Planert’s grammar of Diyari (1908) and of Arrernte (1907), both published in German, uses the terms ‘ergativ’ and ‘absolutiv’ with current reference to describe these syntactic cases. Both works are significant within global history of ergativity in giving the earliest paired oppositional use of the terms ‘ergative’ and ‘absolutive’. Planert’s use term ‘ergative’ with modern reference, which was followed by Carl Strehlow (1908), had first been used by W.Schmidt (1902), when describing the Papuan languages Kai and Miriam (Meryam Mir), and the Pama Nyungan language Saibai (Western Torres Strait). The term was, however, first employed by G.Taplin (1872) in a grammar of Ngarrindjeri, but with reference to peripheral case function.

Dixon, R. M. W. 1972. The Dyirbal language of north Queensland. Cambridge: Cambridge University Press. Planert, W. 1907. Australische Forschungen I. Aranda-Grammatik. Zeitschrift für Ethnologie 39. 561-566. Planert, W. 1908. Australische Forschungen II. Dieri-Grammatik. Zeitschrift für Ethnologie 40. 686-697. Schmidt, W. 1902. Die Sprachlichen Verhältnisse von Deutsch-Neuguinea [Part 2]. Zeitschrift für afrikanische, ozeanische und ostasiatische Sprachen 6. 1-99. Silverstein, M. 1976. Hierarchy of Features and Ergativity. In Dixon, ed. 1976, 112-171. Strehlow, C. F. T. 1908. Einige Bemerkungen über die von Dr. Planert auf Grund der Forschungen des Missionars Wettengel veröffentlichte Aranda-Grammatik. Zeitschrift für Ethnologie 40. 698-703. Taplin, G. 1872. Notes on a comparative table of Australian languages. The Journal of the Anthropological Institute of Great Britain and Ireland 1. 84-88.

ABSTRACT. The semantics of one tense-aspect-mood-polarity (TAMP) category in the Oceanic language Vera’a (ISO 639-3: vra), spoken on Vanua Lava in North Vanuatu, is investigated. Cognate TAMP forms in the closely related languages Mwotlap (François 2003:166) and Vurës (Malau 2016:494 – 505) have been analysed as essentially “[…] underspecified for tense, aspect and mood […]” (François 2010:505) and presenting a state-of-affairs (SOA) as ‘detached’ from reference time, a view extended to respective Vera’a forms by François (2009). Its underspecification is meant to account for the wide variety of uses of the TAMP marker in question, ranging from generic or future SOAs to chains of SOAs situated in the past. The respective TAMP marker in Vera’a, glossed TAM2 here, covers roughly the same range of uses as its Mwotlap or Vurës cognates, yet we propose here to analyse it as bearing specific aspectual meaning, namely presenting a SOA with regards to its left boundary in the sense of Sasse’s (1991) model; different interpretations are pragmatically inferable from this basic meaning. Firstly, outside of any discourse context, the use of TAM2 always situates a SOA as simultaneous with or posterior, but never anterior, to coding time (CT) (Fillmore 1997) and expresses its inception or imminence (cf (1)). The same is true for its use in narrative discourse with regards to a reference time (RT) other than CT that is established by the preceding discourse (cf (2) & (3)). The fact that it often occurs in chains of SOAS is explained here by means of the implicature that a previous SOA ends before the inception of a new one (cf (2) & (3)); the implicature is defeasible (cf first two clauses in (3)). Some irrealis uses of TAM2 are explainable as extensions of its inceptive semantics, so that an imminent not yet existing SOA is interpreted as desired (cf (4)) or in fact remains unrealised (cf (5)). As for generic SOAs, these are generally expressed by a different (stative) TAMP marker in Vera’a. Where TAM2 is used, this implies a hypothetical frame, so that its use can be paraphrased as ‘whenever SOAX, then it follows that SOATAM2’. Although some details remain to be worked out, our analysis provides a straightforward and explicit (and hence testable) account of a wide, yet clearly restricted, range of related uses of the TAMP marker at hand. We argue that both François’ (2003, 2009) and Malau’s (2016) accounts in terms of underspecification of TAMP values are designed primarily to account for its functional breadth, yet fail to capture existing restrictions on its use and interpretation, and describable interrelations between its senses.

ABSTRACT. This paper is part of larger project seeking to contribute to theoretical discussion of aspectual particles (also called ‘phasal quantifiers/adverbials’, ‘scalar focus particles’, among other terms) such as ‘already’, ‘still’ and their negative counterparts (Loebner, 1993, Mittwoch, 1993, Krifka, 2000, Moosegard Hansen, 2008). Such discussion has primarily focused on European languages (with some exceptions, see eg. van Baar, 1997). Here, we propose to analyse one tense-aspect marker in Jaminjung among a wider set, the clitic =guji. This clitic has non-temporal uses marking a referent as first in a series of referents, as exemplified in (1):

(1)        pigipigi=yirram=biya       walyang=guji     wirib     gamurr   buliki    birang

pig two=SEQ                 in.front=FIRST dog       middle   cow       behind

“..the two pigs in front, the dog in the middle, the cow behind.’’

The clitic –guji can also mark a first event in a temporal series and is often followed by a clause containing another clitic, =biyang, expressing sequentiality (‘and now/and then’, see Ritz and Schultze Berndt, 2015):

(2)        narrng=guji        gani-buwa=nggu,                         jarrg=biyang ba-mili

stuck=FIRST     3SG>3SG-POT-IV.bite=2SG.OBL             pull=SEQ IMP-IV.get/handle

“Let it bite properly first, then pull quickly.”           

Another frequent use is translatable by English ‘already’:

(3)        aa, nganjuga=guji

ah! 2SG>3SG-IV.take.PST=FIRST

“Ah, you took it already.”

Some clauses include both =biyang and =guji, where =biyang re-sets the time under discussion and its combination with =guji brings about a retrospective meaning:

(4)        jalig=marlang=biyang     gurlbany-gi=guji             ga-rdba-ny

child=GIVEN=SEQ        ground-LOC=FIRST       3SG-IV.fall-PST

“The child has already fallen on the ground .”

(5)        ngiya=guji=biyang         barr                   gan-ardgiya-ny

PROX=FIRST=SEQ      hit.against          3SG>3SG-IV.throw-PST

“..here [in this picture] he has already smashed it (by dropping it).”

Finally, =guji can be used to mean ‘for the first time’, a meaning also corresponding to German erst (‘first/first time’):

(6)        jungulug=guji    marndaj digirrij-mindij=binji        gurru-wu-rum yawal

one=FIRST       all.right die-TIME=ONLY            2PL-POT-IV.come run.many

“One time for the first time when I'm dead all right, you'll come running!”

Interestingly, German erst is also used as the dual of ‘already’ (see below) meaning ‘only’, ‘not until’. Jaminjung uses another clitic =gung either alone or in combination with girrang ‘yet’ to express the meaning of ‘still’ (and with negation to convey the meaning ‘not yet’).

The English aspectual particle already and its German equivalent schon have been semantically analysed as asserting that the proposition expressed by the clause holds at a time t and as presupposing that ~p was true before t (Loebner, 1993). Krifka (2000) developed this analysis treating ‘already’ as focus sensitive (the focus can be the whole sentence or a subpart of it), expressing the presupposition of a restriction of a set of valid alternatives. The inference is that p occurred earlier than expected. On the other hand, erst implies that p occurred later than expected. As discussed in Moosegard Hansen (2008), it is the aspectual meanings of such markers that has attracted most interest in scholarly works, despite the fact that they typically have a wide range of meanings. As a result, Moosegard Hansen (2008:5) calls for more in-depth analysis of the relations between their aspectual and non-aspectual meanings. The present paper attempts to contribute to this research area by examining the contexts of use of the clitic =guji, taking account of a range of factors that influence its interpretation such as combination with other clitics/particles, rhetorical function of the host clause, tense, aspect (see discussion of schon ‘already’ in relation to perfective and imperfective aspects in Löbner, 1993:196) and general contextual information. The paper aims to uncover any relations between the different uses of =guji and to relate these to the representations of aspectual markers found in the literature.

ABSTRACT. The Australian Aboriginal language Ungarinyin (Worrorran) has six conventional strategies for expressing properties about the internal temporal organisation of an event, as illustrated in (2).

The verbal suffix -yirri/-erri in (2a) indicates continuative aspect and the suffix -wa/-ba in (2b) marks iterative aspect (Rumsey, 1982; Author, 2015). Most Ungarinyin verb phrase constructions consist of a combination of a coverb and an inflecting verb as schematically represented in (1). The coverb consists of an open class of minimally inflecting verbs, while the inflecting verbs consist of a closed class of about 15 roots (also cf. McGregor, 2002). Within this verb phrase structure, iterative aspect is marked on the coverb, while the continuative suffix appears on the inflecting verb.

1. [[coverb verb] [inflecting verb]]

A third type of strategy Ungarinyin uses for signalling properties about event structure is shown in (2c), where the choice of the inflecting verb (-yi-/-e- ‘be’) itself gives rise to a progressive aspectual meaning.

In addition to these morphological strategies, Ungarinyin has three less clearly ‘grammatical’ ways of signalling aspect-like meanings, such as the lexical strategy in (2d), in which the adverb bija ‘already’ signals the completion of an action, and the reduplication strategy in (2e), in which the coverb, but never the inflecting verb, is repeated to give rise to a progressive meaning. Finally, the intonation contours under (2f) show that the final intonation unit in this example receives a high maintained pitch, which equally signals an ongoing event.

These highly diverse strategies raise questions about the extent to which the six ways of signalling properties about temporal event structure constitute a grammatical system. While especially the latter three strategies are not restricted to Ungarinyin, in this paper I demon­strate that the interactions between the types and loci of marking in the language suggest a high degree of systematicity. This observation has more general typological implications for the analysis of grammatical vs. extra-grammatical expression in verbal categories.

ABSTRACT. Nafsan (South Efate) is an Oceanic language of Vanuatu, spoken by an estimated 6,000 people. Previous research includes an outline of the Nafsan phonological system (Thieberger, 2006), and discussions of the relationship between the sound systems of Nafsan and neighbouring language varieties (Clark, 1985; Lynch, 2000). However, there are several remaining challenges in understanding the segmental and prosodic patterns of the language, particularly relating to vowel contrasts and their implementation in different contexts. Furthermore, as for most languages of Vanuatu, the phonology of Nafsan has not yet been the subject of phonetic research. This talk will present findings from a new phonetically-based study of Nafsan phonology, with a focus on the acoustic and durational correlates of vowel distinctions in the language. Compared to languages spoken further to the north in Vanuatu, and the general preference for CV syllable structures among Oceanic languages (Lynch et al., 2002), Nafsan exhibits complex phonotactic structures, with syllable onsets including a range of heterorganic consonant clusters. These appear to have arisen from a historical and potentially ongoing process of medial vowel deletion, and there may be some variation in current production patterns (e.g. Thieberger, 2006). There is some evidence for both long and short vowels, though the status of vowel length is not yet clear. Given that there may be complex relationships between vowel length, syllable structure, vowel deletion, and other prosodic phenomena such as stress, a more comprehensive understanding of Nafsan vowel distinctions is necessary for research on other aspects of Nafsan phonology to proceed. Data were collected with Nafsan speakers in Erakor on the island of Efate in Vanuatu. Five speakers participated in the vowel study which is the focus of this talk. A set of monosyllabic words containing /i, e, a, o, u/ in CV, CCV, CVC, CCVC and VC syllables was prepared, and following spoken prompts, participants produced each word three times in isolation and three times in a frame. Audio recordings (96kHz, 24-bit) were made using a Zoom H6 recorder and Countryman headset microphone, and automatic segmentation of the speech signal was performed using the Munich Automatic Segmentation System (Strunk et al., 2014; Kisler et al., 2016). A hierarchical database was constructed using the EMU Speech Database Management System (Winkelmann et al., in press), and the acoustic and durational characteristics of vowel tokens produced in the frame were queried and analysed using the emuR package in R (Winkelmann et al., 2016; R Core Team, 2016). Duration measures provide compelling evidence for a length contrast for all vowels (e.g. Fig.1), but also show that vowel duration is influenced by syllable type. Results for first and second formant frequencies at vowel midpoints provide supporting evidence for five distinct vowel qualities (e.g. Fig.2), and also show that long vowels tend to be more peripheral in the vowel space. In addition to offering a detailed description of the Nafsan vowel system, these findings provide a better understanding of some of the factors influencing vowel realisation which need to be considered in ongoing prosodic research.

ABSTRACT. This talk discusses the issue of tonal domain specification in the Himalayas, particularly within Tamangic languages, a sub-branch of the Tibeto-Burman family. Using data from Ḍanḍagaon, a Western Tamang village along the historic Trisuli River Valley trade route in Central Nepal, I demonstrate that this variety of Tamang diverges from the typology of Tamangic phonological word-tone, and that the two of the four tones underspecified for pitch and two of the five grammatical suffixes underspecified for tonal domain are evidence of multiple levels of underspecification working in concert in the tonal system of Ḍanḍagaon Tamang (DT).

Following Yip (2002), a tonal domain is a language-specific space that is defined by syntactic structures that limits tone assignment processes to occurring within this domain, and could be the phonological word, the syllable, or the morpheme. The Tamangic language family tonal domain has been described as the (phonological) word (e.g., Hildebrandt, 2007), having been described for Gurung, (e.g. Mazaudon, 1978; Sprigg, 1997) as well as Manange (e.g. Hildebrandt, 2005). Regarding Risiangku Tamang, Mazaudon reports that the tone of the phonological word is ‘entirely conditioned by the tone of the root’, (2005: 83-84) and ‘[t]he pitch characteristics of the whole word are determined by the lexical morpheme it contains’ (2012: 146).  Similarly, in Ḍanḍagaon, the four contrasts that are found on monosyllables are also present on disyllables that are (synchronically) monomorphemic, evidence of a morphemic tonal domain, shown in (1). As such, we would expect that the two morphemes comprising a bimorphemic disyllable would each maintain their respective lexically specified tones, represented in (2).

Through acoustic and perceptual examination and comparison of monomorphemic disyllables and monomorphemic monosyllables inflected with five grammatical suffix morphemes to create bimorphemic disyllables, however, we see that some grammatical suffix morphemes are underspecified for tonal domain, exhibiting the same tonal realizations observed on monomorphemic disyllables represented in (3). This tonal domain underspecification is not unlike what Mazaudon (2005) reports for Risiangku Tamang and other Tamangic languages, specifically that (all) grammatical affixes are ‘devoid of distinctive tones’ and that ‘the melody simply extends, or rather deploys itself over the number of syllables available’ (83). 

Further examination of DT bimorphemic trisyllables (i.e. disyllables inflected with grammatical suffix morphemes) with comparison to monomorphemic tri- and tetrasyllables sheds light on a preference to prosodic foot structure in polysyllabic words, shown in (4). This final piece of evidence motivates the definition of the DT tonal domain to be defined as the phonological word, with consideration to the prosodic foot.

The combination of the tonal domain underspecification and the metrical behavior of tone observed in polysyllables described in this talk is evidence that the representation of tonal systems is multipart and not always straightforward. This study contributes to the exceptionally under-described Himalayan phonology, and further identifies the question of what level of importance the prosodic foot has in the tonal realization across the Tamangic languages.

ABSTRACT. Prosody is one of the few areas given little attention in Jeffrey Heath’s (1984) otherwise extremely detailed descriptive grammar of Wubuy. Hore (1981) presents a more extended analysis, based on his own fieldwork. Heath (1984:32) claims that stress accent is realized as a high pitch contour, while Hore (1981:11) claims it is the combined effect of increased intensity, pitch and duration, but that raised pitch is the most significant of these correlates. The study presented in this poster represents the first attempt to subject these claims to acoustic phonetic investigation, in the hope that it will lead to a better evaluation of those authors’ analyses of the metrical system of Wubuy. 50 words of between three and seven syllables were recorded from three fluent speakers, with 5 tokens of each word in initial position and 5 tokens of each word in final position within a sentence frame. While a full statistical analysis is still underway, early results suggest that intensity (RMS) may in fact be a better indicator of word-level prosodic prominence in Wubuy than fundamental frequency, despite the fact that fundamental frequency has received more attention from phonologists interested in the phonetic correlates of prosody (e.g. Ladd 1996). In the example of murrarbu ‘crab’ shown below, the fundamental frequency remains fairly level, while the penultimate syllable has a much stronger intensity peak than adjacent syllables. In dhimburru ‘northeast wind’, shown below, the most prominent syllable appears to be the initial. While there is a slight rise in the fundamental frequency throughout the word, the intensity is highest in the initial syllable. However, the interactions of the acoustic correlates of stress with phonemic vowel length at the word level, and with superordinate prosodic units at the intonation level, will require further investigation.

Heath, J. (1984). Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Hore, M. (1981). Syllable Length and Stress in Nunggubuyu. In Work Papers of SIL-AAB (Vol. 5, pp. 1–62). Darwin: Summer Institute of Linguistics. Ladd, D. R. (1996). Intonational Phonology. Cambridge: Cambridge University Press.

ABSTRACT. According to Talmy’s (1991, 2000) typology of motion events, satellite-framed languages (e.g., Germanic, Slavic) map the path onto a satellite (a prefix, a particle) and the manner onto the verb, while verb-framed languages (e.g., Romance) express the path in the verb and the manner, in a secondary element, e.g., a gerund (cf. (1) and (2)). Talmy’s typology became the focus of scholarly work within cognitive typology (Ibaretxe 2004), psycholinguistics (Ellis 2013; Slobin 1991) and, to a lesser extent, generative linguistics (Acedo-Matellán 2016), among other fields. However, scant attention has been paid to the advantages of a constructional approach to Talmy’s findings, one notable exception being Narasimhan (2003). In this presentation, I report on the theoretical findings from an elicited narratives task based on a 5-min. extract from Chaplin’s City Lights in which the expression of motion in three languages was dealt with, namely German, Polish (both satellite-framed) and Spanish (verb-framed). My results provide evidence that a constructional approach to motion events is more accurate than the classical lexicalization patterns approach for several reasons. It has been previously observed that a phrasal level of representation is necessary to deal with the compatibility of non-motion verbs with path satellites, since these are not specified in the verb’s semantics (Goldberg 1996, Narasmihan 2003, see ex. (3)). Here I take this line of reasoning one-step further and I show that a constructional approach is relevant not only from the perspective of argument structure. Following Croft (2007), I prove that using fine-grained constructional frames as a tertium comparationis is more appropriate when dealing with cross-linguistic generalizations. Specifically, it allows for elucidating both general trends that go beyond Talmy’s binary typology, as well as contrasts within a typological group (intra-typological variation). On the one hand, I show that the choice of lexicalization patterns depends on the conceptual properties of the constructional frame (Givón 1980). Speakers of all three languages showed greater preference for satellite-framed than verb-framed patterns in their speech about caused-motion events—a pattern that was reversed for events rendered from a self-motion perspective. Caused-motion constructions are conceptually more complex and they usually involve information about how a force is applied to a figure to initiate the movement (Talmy 1988). In other words, this kind of events are typically associated with an additional manner component (force-dynamics), which, in turn, increases the occurrence of manner/path conflated descriptions in each language (see Table 1). On the other hand, Talmy’s typology should include more construction types than previously discussed. For example, it is important to distinguish between prefixed constructions, as in Polish, and particle-constructions, as in German. Among other things, prefixes are more lexicalized with the verb than particles and, consequently, (i) the accumulation of more than one bounded path in a single clause is highly constrained in Polish but not in German; (ii) particles show a greater compatibility with high-manner verbs. In short, a constructional approach allows for dealing with generalizations which would remain invisible under a two-way classification of languages as a whole.

ABSTRACT. In canonical inflection (Corbett 2009), each cell of a paradigm is occupied by a unique form. More specifically, lexical material remains constant while the inflectional material of each cell is unique. Deviations from this canonical ideal – either where lexical material does not remain constant (e.g. suppletion) or where inflectional material is not unique (e.g. syncretism) – have been referred to as ‘lexical splits’ (Corbett 2015): that is, a paradigm can be partitioned (or ‘split’) between those cells which adhere to the canon and those that deviate from it.

At their simplest, lexical splits result in a paradigm being cleaved in two (e.g. English go vs went), but often things are much more complex. For instance, in the case of verbal inflection in Skolt Saami (Finno-Ugric, Finland) the convergence of multiple morphophonological rules leads to as many as seven unique inflectional stems (Feist 2015) (Table 1).

One approach to dealing with this degree of complexity is to look at the individual factors which lead to the stem alternations. In Skolt Saami, the contributing factors are consonant gradation, vowel height alternations and palatalisation. Taking this perspective, we see that the six-way split in forms is only possible through the combination of a three-way split in consonant grade, a two-way split in vowel quality and an alternation between plain and palatalised stems (Table 2).

After presenting the inflectional system of Skolt Saami in this way, I demonstrate how the same approach can be applied to even more complex inflectional systems where seemingly impenetrable complexity can be more readily described by breaking things down into component splits. This approach can be applied equally to splits in lexemes (i.e. internal to the paradigm), as seen in Skolt Saami, and splits across lexemes (i.e. inflectional classes), which I illustrate with data from the notoriously complex inflectional class systems of Oto-Manguean (Feist & Palancar 2015).

I argue that morphological complexity can be more easily understood if we adopt analyses which peer below the surface in this way. Seeing the component parts of a complex split as individual entities can, for instance, license straightforward explanations for their existence, explanations which might be totally obscured if we were to look only at the combined effect of the splits. For example, in Skolt Saami, consonant gradation was once a phonological process determined by the shape of suffixed morphemes, while palatalisation was associated with a high second-syllable vowel. The loss of word-final consonants and the contraction of unstressed vowels have removed the conditioning environments of these processes rendering them purely morphological, yet these diachronic changes provide a logical explanation for the apparent random distribution of forms across the paradigm. Viewing the processes independently of each other thus helps us to disentangle the complexity of the synchronic system.

ABSTRACT. The form-frequency correspondence hypothesis says that when two grammatical patterns that differ minimally in meaning (i.e. that form a semantic opposition) occur with significantly different frequencies, the less frequent pattern tends to be overtly coded (or coded with more coding material), while the more frequent pattern tends to be zero-coded (or coded with less coding material). Salient examples are the pairs of categories in (1), where the first member is more frequent and coded with shorter material (or with zero) in all languages.

(1) singular plural (book – book-s) present future (go – will go) 3rd person 2nd person (Spanish canta – canta-s) nominative accusative (Hungarian ember – ember-t) affirmative negative (go – don’t go) allative ablative (to – from) positive comparative (small – small-er) predicative verb nominalized verb (go – go-ing)

The idea that cross-linguistically systematic coding asymmetries like those in (1) are related to frequency of use is found in Greenberg (1963) and Croft (2003: Chapter 4), but the precise causal mechanism has not been widely discussed, and alternative views are still common (e.g. Haiman 2008 on iconic motivation, or the generative literature that still often appeals to abstract markedness). In this presentation, I would like to present results from a larger research project that studies form-frequency correspondences in grammar and I argue that the scope of the observation is much broader than has generally been recognized. In particular, it comprises the pairs of forms given in (2).

(2) noncausal causal (Swahili ganda – gand-isha ‘freeze’) causal noncausal (Maltese fetaħ – n-fetaħ ‘open’) inalienable alienable (Nyulnyul nga-lirr – ja-n yil ‘my mouth/dog’) alienable inalienable (Paamese ani – a-vat ‘coconut/head (no possr.)’ singular plural (English carrot – carrot-s) plural singulative (Welsh moron – moron-en) P-argument A-argument (Dyirbal yarra – yarra-ŋgu ‘man’ A-argument P-argument (Dyirbal ŋadya – ŋayguna ‘I’)

These pairs seem to contradict the idea of a general form-frequency correspondence, but when looking more closely at the types of elements found in the first and the second line, one sees that the form-frequency hypothesis is strikingly confirmed: While for some kinds of meanings (such as ‘freeze’), the causal verb is clearly less frequent, the situation is the opposite for other kinds of verbs (such as ‘open’) – and analogously for the other cases. Form-frequency correspondence also explains many other cross-linguistically systematic patterns that are normally seen as unrelated because grammatical marking and lexical marking are rarely seen together, e.g. the fact that some relations tend to be expressed by cases (e.g. ‘at’, ‘in’), and others by adpositions (e.g. ‘under’, ‘behind’, cf. Zwarts 2010), or that perfects tend to be expressed periphrastically (Dahl 1985). I also briefly address the issue of a causal mechanism, arguing that the causal chain is “frequency –> predictability –> shortness of coding (or zero coding)”, and that an explanation of the ultimate sources of the frequency asymmetries is not necessary. This explanation is shown to be superior to alternative explanations in terms of semantic complexity that are still widespread.

ABSTRACT. Early differences in children’s vocabulary related to socioeconomic status (SES) have been attributed, in part, to quantitative and qualitative differences in linguistic experience (e.g. Fernald et al., 2013; Hirsh-Pasek et al., 2015; Hoff, 2003; Huttenlocher et al., 2010). Such differences have recently been traced back to 6 months, as observed during short recording sessions at home (e.g. Vanormelingen & Gillis, 2016). In this longitudinal study, we examined quantitative aspects of infants’ early linguistic experience at home, to assess SES differences and their potential association with later spoken vocabulary skills. Based on previous research, we predicted that higher SES, measured as maternal education, would be associated with 1) greater quantity of infants’ linguistic experience and, later 2) infants’ larger spoken vocabularies.

Fifty families participated, from working- and middle-class Australian English-speaking backgrounds, representing two levels of maternal education (higher ≥ bachelor degree, lower ≤ bachelor degree). At child age 6 to 9 months and 12 to15 months, we obtained day-long digital audio recordings in families’ homes on two days. Using the Language Environment Analysis system (LENATM; LENA Foundation, 2009), we extracted 12-hour estimates of three aspects of linguistic experience (adult words, conversational turns and child vocalisations). At child age 12 to 15 months and 19 months, parents reported on children’s expressive vocabulary using the Australian English Communicative Development Inventory (OZI; Kalashnikova et al., 2016).

As predicted, despite considerable individual variation, higher maternal education was associated with greater quantities of linguistic input at 6 to 9 and 12 to15 months. On average, compared to the lower maternal education group, the higher maternal education group spoke 4375 more words per day to infants at 6 to 9 months (p = .014), and 3080 more per day at 12 to 15 months (p = .022). Importantly, they also had 110 more back-and-forth conversations per day with their infants at 6 to 9 months (p = .005), and 126 more per day at 12 to 15 months (p = .013). Group child vocalisation differences were not statistically significant.

At 12 to 15 months and 19 months, we found no significant 2-group differences in OZI scores (though maternal education differences between narrower educational strata were evident at 12 to 15 months). Overall, conversational turns best predicted later vocabulary, accounting for 11-27% of variance (p < .05).

Our findings show that SES differences in linguistic experience are evident in the first year of life, earlier than previously shown. Verbal interactions appear to be key for infant vocabulary development, with potential applications in home language enrichment programs.

ABSTRACT. Traditional dialectology data collection used to rely on non-mobile older rural men as the basis of ‘authentic’ dialect studies (Chambers and Trudgill, 1998). In today’s highly mobile, young and urban Australia, this model no longer applies. By introducing sociolinguistic methods of data collection and analysis (performing chi-squared tests on native speakers from different age groups, states, gender, high school setting, and urban/rural environments), linguists can gain a richer understanding of dialect use (see Labov, 1972). Australian English is a variety of English that has been regarded as having limited forms of regional variation (Moore, 2009). Despite this, regional lexical variation was first explored through a traditional dialectology approach in the 1980s, and overwhelming evidence of lexical regional variation across Australia was uncovered (Bryant, 1992). This paper uses this early work as a starting point to examine language variation and change over time in Australian English by revisiting, comparing and expanding dialectology methods. The efficacy of newer data collection techniques and implementation (Vaux, 2003) are compared with older, more traditional techniques (Bryant, 1985). Findings of this survey include highly statistically significant differences (p<0.01) in reported production of lexical items between age groups and across states. This work is part of a larger thesis which examines native Australian speakers’ folk linguistic ideologies and compares them to reported usage of regional variation to better understand how Australians perceive and report variation across Australia. 

ABSTRACT. The present study is situated within the developing field of corpus-based short-term diachronic research in English. The focus is on grammatical change, and follows in the footsteps of the growing body of studies that use quantitative methods to identify and explain diachronic variation in the frequencies of grammatical categories across varieties of English (e.g. Leech et al. 2009; Collins 2015). The study adopts a multivariate design and is based on frequency findings for 68 linguistic features known to be sensitive to diachronic/stylistic variation. A minimum frequency of one token per 1000 words was set as a threshold for a feature to be included.

Using parallel Australian, British and American corpora sampled in 1931, 1961, 1991 and 2006, we seek to shed light on questions relevant to the endonormativisation of Australian English. These include: Is there a reflection in the 1931-1961 corpus-based grammatical frequencies of the British-oriented outlook that dominated Australian culture and politics in the first half of the 20th Century? Is there a shift in the 1961-1991 Australian English frequencies away from British English that is suggestive of an Australian move towards linguistic independence? Do the 1991-2006 confirm or disconfirm the presence of American-derived ‘Global English’ influence?

The eight million-word corpora representing British and American English are all members of the extended ‘Brown-family’. For Australian English we used the threefour recently-compiled, temporally-parallel, ‘AusBrown’ corpora (AusBrown30s, AusBrown AusBrown60s, AusBrown90s, AusBrown06) each comprising around 240,000 words (divided evenly among each of thecontaining three genres: fiction, learned and press reportage material). In order to achieve a complete matching of corpora, 60,000-word Texts selections were made were selected from each of the three genre categories across the ninesixteen corpora in order to achieve a generic balance..

The frequencies for the 68 grammatical features in the database were normalised and log-transformed (to compensate for large frequency differentials), and average frequencies calculated for the three3 varieties (and the three3 genres) at each time period. Multidimensional scaling was then used to plot similarities and dissimilarities across time periods/varieties, in each genre.

The frequency findings provide support for Collins and Yao’s (fc) claim that the (grammatical) endonormativisation of Australian English begins to gather momentum in the 1961-1991 period – coinciding, arguably, with a rise in nationalistic feeling in this country – with divergences from both of the two ‘supervarieties’ becoming more marked by 2006., particularly in the press.

References Collins, Peter. 2015. Grammatical Change in English World-Wide. Amsterdam: John Benjamins. Collins, Peter and Xinyue Yao. Forthcoming. Colloquialisation in Australian English: a cross-varietal and cross-generic study of Australian, British and American English from 1931 to 2006. English World-Wide. Leech Geoffrey Leech, Marianne Hundt, Christian Mair, and Nicholas Smith. 2009. Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge University Press.

ABSTRACT. Ngarinyin (WORRORRAN), a Non-Pama-Nyungan language from Kimberley region of Western Australia, Australia has three tenses (present, past, future), four moods (indicative, imperative, optative, irrealis) and three aspects (continuative, iterative, punctual) (Rumsey 1982, Rumsey 2001). Ngarinyin aspect marking is split syntactically and morphologically. Although all aspect markers are suffixes, continuous is indicated on inflecting verbs (syntactically free); whereas iterative and punctual are indicated on coverbs (syntactically constrained) (Rumsey 1982).

I begin with an overview of aspect in Ngarinyin. I ask the question why iterative and punctual aspects are the only morphological marking allowed on Ngarinyin coverbs (which are otherwise do not inflect)? Why are iterative and punctual aspect not marked on inflecting verbs like the continuative aspect?

Although two out of the three Ngarinyin aspects are coverb suffixes, the continuative aspect (-yirri~ -nyirri) which is marked on inflected verbs is the most common. Iterative (-wa~-ba) is reasonably common, but has in many instances become lexicalised. Punctual (-wini~-bini) is very rare. I discuss the disparity of Ngarinyin split aspectual marking contexts in terms of:

a) genetic (comparison with related language Worrorra (Clendon 2000)), b) areal (comparing with similar disparities in unrelated Northern Territory Non-Pama-Nyungan languages e.g. Jaminjung (Schultze-Berndt 2012)) c) diachronic (How did the categories develop?)

I also mention the broader context as it relates to coverbs and inflecting verbs and propose two different paths of development of these word classes in Ngarinyin.

ABSTRACT. TAM values are typically expressed in Ngumpin-Yapa languages (see McConvell & Laughren 2004) by verbal inflectional suffixes or by the interplay between verb forms and auxiliary morphemes. When present, auxiliaries host enclitic pronouns, as illustrated by Walmajarri (Ngumpin) examples from Hudson (1978) in (1). Of the two Yapa languages, Warlpiri and Warlmanpa, Warlpiri shows the greater divergence from Ngumpin languages in both its verbal inflections and its auxiliary morphemes. By comparison with both Warlmanpa and Ngumpin languages, the Warlpiri verb inflectional paradigm of verbal inflections lacks a number of distinctions including the punctual/iterative contrast (except on one dependent verb). Past and non-past tense forms are mostly derived from earlier participial forms, while potential and – with the exception of the northwest dialect – future modal forms are either absent or have been reanalysed, as illustrated by comparison with Walmajarri in (1) to (3) and with Warlmanpa in (4) and (5).

In Warlpiri these semantic distinctions are expressed by a richer – and different – inventory of auxiliary forms than in the other languages. A comparison between Walmajarri and Warlpiri irrealis clauses illustrates this point. In (2), Walmajarri expresses the past versus present modalised irrealis by contrasting verbal suffixes, while in (3) Warlpiri expresses this semantic contrast by means of the presence or absence of an imperfective auxiliary morpheme lpa (discussed by Legate (2003)) in combination with a unique irrealis verbal inflection cognate with the Walmajarri past irrealis form. (The enclitic lpa also marks imperfective with the past tense verb form cognate in all but one conjugation with a participial form). In eastern Warlpiri the present (1c') versus future (1d') contrast is marked exclusively by auxiliary morphemes. These combine with the non-past verb form, also derived from an ancestral participial form, cognate with the simple past form of some Ngumpin conjugations. The irrealis verb form in the Walmajarri sentence (2b) which contrasts with the past irrealis in (2a) functions as the imperative form in all NGY languages. In Warlpiri, it is exclusively dedicated to the imperative function.

In addition to the absence of some NGY verbal inflections, Warlpiri lacks auxiliary morphemes such as Walmajarri pa and nga in (1) and (2) which occur in some of the other languages: pa in Mudburra, nga in Warlmanpa and Djaru. An auxiliary morpheme nga, often glossed 'dubitive' (dub), which typically encliticizes to the auxiliary complex including the pronominal enclitics, is present in Warlmanpa and most Ngumpin languages, but is absent from Warlpiri as shown by the comparison between the Warlmanpa sentences in (4) and their Warlpiri translations in (5). (The Walmajarri counterpart of dubitative =nga is =rta.) 

I will discuss how Warlpiri disambiguates distinctions marked by verbal inflections in Warlmanpa and Ngumpin languages by its richer inventory of complementizers which combine with the auxiliary aspectual morpheme lpa and the tense/mood ka (cf. (1c') with (5a&b)). I will focus particularly on how the interplay between the scope of modal operators, aspectual values and speaker perspective is expressed.

ABSTRACT. In a worldwide cross-linguistic study based on a sample of 868 languages, Gil (to appear) examines whether Tense-Aspect-Mood (TAM) marking is optional or obligatory (see map below). Languages with optional TAM marking are those in which there are some basic declarative affirmative main clauses with no grammatical expression of any TAM categories, while languages with obligatory TAM marking languages are those in which all basic declarative affirmative main causes contain a grammatical expression of at least one of the three TAM categories. For the purposes of the study, the grammatical marking of TAM may be either bound or free; moreover, it may be either dedicated, expressing nothing but TAM concepts, or portmanteau, combining the expression of TAM with that of other concepts, such as voice, or agreement features such as person, number and gender. Worldwide, both language types are common: while optional TAM marking occurs in 377 languages in the sample, constituting 43%, obligatory TAM marking occurs in 491 languages, or 57%.

Inspection of the Gil (to appear) study reveals a striking distributional pattern: in general, large language families that have spread over an extensive geographical region, are associated with obligatory TAM marking. For example, the rate of occurrence of obligatory TAM marking is 90% in Bantoid, 100% in Berber and Semitic, 83% in Indo-European, 100% in Uralic, 94% in Altaic, 95% in Trans-New-Guinea (excluding Timor-Alor-Pantar), and 92% in Pama-Nyungan. The correlation between obligatory TAM marking and large families spread over extensive geographical regions represents a particular case of a more general tendency for grammatical complexity to correlate with socio-political complexity; some other manifestations of this tendency are briefly surveyed.

However, one large language family bucks the general trend: in Austronesian, the rate of occurrence of obligatory TAM marking is just 27%. Why should this be the case? This paper argues that the surprisingly low rate of obligatory TAM marking is due to a qualitatively different mode of expansion of Austronesian languages, in comparison to the other large language families. The crucial difference is associated with the spread of Austronesian into the Indonesian archipelago that took place some 3500 years ago. In Gil (2015) it is argued that the languages of Mainland Southeast Asia, the Indonesian archipelago and western New Guinea constitute a linguistic area that was already in place before the intrusion of Austronesian languages, and that one characteristic of this linguistic area was optional TAM marking. Whereas the Austronesian languages of Taiwan and the Philippines are mostly obligatory TAM marking, those of the Indonesian archipelago are almost exclusively optional TAM marking. The loss of obligatory TAM marking by the incoming Austronesian languages is thus accounted for in terms of language contact, and assimilation to the pre-existing typological profile of the languages of the Indonesian archipelago. Compared to most other language dispersals, the spread of Austronesian into the Indonesian archipelago is thus characterized by a greater degree of language contact, in which particular linguistic features, such as, in this case, TAM marking, undergo horizontal diffusion, independent of genes and cultural packages. In conclusion, this paper considers two other cases of language spread involving the replacement of obligatory with optional TAM marking: the intrusion of Trans-New-Guinea into the Timor-Alor-Pantar region, and the emergence of Indo-European lexifier creoles and creole-like varieties such as Singlish and Melanesian Pidgin.