麻豆传媒高清

Forrest Panther, Linguistics Department

He aha ng膩 r艒p奴 kupu o te reo M膩ori? I te nuinga o ng膩 reo, e m膩rama ana ng膩 r艒p奴 "kupuingoa" me "kupumahi". Engari, i te reo M膩ori, ka taea te kupu kotahi hei kupuingoa, hei kupumahi hoki; i te rerenga: "e kai ana ia i te kai", ko 'kai' he kupingoa, he kupumahi hoki. Kotahi, e rua r膩nei ng膩 kupu 鈥榢ai鈥� i ng膩 hinengaro o ng膩 kaik艒rero M膩ori? E ai ki 膿tahi kairangahau, kotahi te kupu, engari, ki 膿tahi atu, e rua. Ka rapu m膩tou i t膿nei p膩tai, ki te whakam膩tau me te t膩tari whakahua.聽

What are the parts of speech in te reo M膩ori? In most languages, the categories of "noun" and "verb" are clear. However, in M膩ori, one word can be both a noun and a verb; in the sentence "e kai ana ia i te kai", "kai" is both a noun and a verb. Are there then one or two words "kai" in the minds of M膩ori speakers? According to some researchers, it is one, but according to others, it is two. We will investigate this question with experimentation and phonetic analysis.

Project Team

Forrest Panther 鈥� PI
Jen Hay 鈥� AI
Heeju Hwang 鈥� AI

Royal Society of NZ Marsden Fund: 2025 鈥� 2028 $852,000

Perceptual experts like forensic scientists train for years to excel in their domain of expertise. Yet some people perform at expert level in tasks like fingerprint-matching without any relevant training or experience. These 鈥榮uper-matchers鈥� are untrained perceptual experts and seem to have a unique perceptual intelligence that enables specialised processing of sensory information that they harness to excel. Trained perceptual experts like forensic scientists also process sensory information in a unique way to novices, but it is unclear whether these abilities arise after years of training or experience, or whether they may also be the result of a pre-existing perceptual aptitude. This project will investigate the cognitive mechanisms that define perceptual experts. We will investigate whether untrained experts possess a unique perceptual intelligence in how they process and sample visual information, comparing their abilities with trained forensic science experts. Using behavioural and eye-tracking experiments, we aim to understand the information processing strategies of both groups during visual comparison and determine whether super-matchers' and forensic scientists' abilities transfer to other domains. The research promises to deepen the understanding of human perception and cognition, and could influence training programmes in forensic science and other domains, like medicine or border security.

Project Team

Bethany Growns 鈥� PI
Caleb Barr 鈥� Research Assistant

Royal Society of NZ Marsden Fund, Fast Start: 2025 鈥� 2028 $360,000

In te reo M膩ori (M膩ori), some syllables are pronounced with more emphasis than others 鈥� they are stressed. Currently, we know very little about stress in M膩ori. We also cannot be sure which syllable is stressed for all words in contemporary M膩ori, due to various sound changes which have occurred because of contact with English. The proposed project addresses this. It will examine the distribution and acoustic properties of stress, how it has changed over time, and how it is perceived by speakers. This project will further the documentation of M膩ori and will help address prevailing Eurocentrism in stress research.

Project Team

Kirsten Culhane 鈥� PI
Forrest Panther - AI

Royal Society of NZ Marsden Fund, Fast Start: 2025 鈥� 2028 $360,000

Changes to speech and language with aging have focused on physiology, speech intelligibility, changes in cognitive processing, and declines in hearing. However, sociolinguistics has shown that we often can signal our place in the social world through choices we make during speech production. We will examine how speech changes in advanced age, and the contributions of speech physiology and sociolinguistic behaviours to these processes. This will advance our understanding of how speakers construct their linguistic persona in old age, and how combined physiological and sociolinguistic factors constrain speech into older age.

Project Team

Lynn Clark 鈥� PI
Megan McAuliffe 鈥� AI
Elena Sheard 鈥� Post Doctoral Fellow
Elsa Wright 鈥� Research Assistant
Eliza McIntosh 鈥� Research Assistant

Royal Society of NZ Marsden Fund: 2024 鈥� 2027 $660,000

This project measures the degree to which machine-assisted writing changes the grammatical production of individual writers. Our hypothesis is that greater intervention (e.g., grammar suggestions as opposed to grammar checking) leads to a greater reduction in individual and group-level variation. Further, we hypothesize that writers in central locations like the US are less influenced than writers in more peripheral locations like New 麻豆传媒高清. This is because NZ English is a less-common dialect, under-represented in digital settings. For example, the models used to assist the writing process have likely never encountered code-switching between English and te reo M膩ori. The essential question is whether the use of generative artificial intelligence will lead to the extinction of less-common dialects like New 麻豆传媒高清 English. This project thus systematically measures the amount of change which machine-assistance creates in the production of individual writers. To accomplish this, we collect samples from participants both with and without machine-assistance and use established methods from computational linguistics and forensic linguistics to compare these samples. The larger issue at stake here is human creativity in the use of language. As human writers, we can navigate an almost infinite number of syntactic pathways through a sentence. The reason that there is robust grammatical variation across dialects, registers, and individuals is precisely because of individual and group-level creativity in the use of grammar. What happens to this creativity when writers use generative AI? We simply do not know.

Project Team

Ben Adams 鈥� PI
Jonathan Dunn 鈥� International collaborator
Andrea Nini 鈥� International collaborator
Gan Qiao 鈥� Post Doctoral Fellow

Royal Society of NZ Marsden Fund: 2024 -2027 $660,000

New 麻豆传媒高清ers who don鈥檛 speak te reo M膩ori are nonetheless regularly exposed to it. It is heard in songs, in the media, and at the opening of formal events. Recent work shows that this regular exposure leads them to build up some detailed implicit knowledge of the language, including knowledge of the language鈥檚 words and sound structures. But does this implicit knowledge also extend to knowledge of the grammar? We will use this population to understand the detail and extent of what can be implicitly learned about grammar from the ambient environment. Past research on the earliest stages of grammar learning has focussed on infants, or on very artificial learning tasks. Both of these research areas face considerable limitations. Working with infants presents obvious methodological challenges. And in artificial grammar learning experiments, there are natural limits on the amount of exposure that participants can realistically be trained on. The unique linguistic landscape of New 麻豆传媒高清 provides an excellent laboratory within which to study implicit language learning in a way that surpasses these restrictions. We use this valuable real-world context to fully explore the limits of implicit grammar learning with much more ecological validity than has been previously possible.

Project Team

Jeanette King 鈥� PI
Forrest Panther 鈥� AI
Jen Hay 鈥� AI
Wakayo Mattingley 鈥� Post Doctoral Fellow

Royal Society of NZ Marsden Fund: 2024 鈥� 2027 $870,000

This project explores the personal writings of Scottish immigrants settling in New 麻豆传媒高清 during the late nineteenth century, examining the role of language in reflecting and constructing their ethnic and national identities. It will focus in particular on their use of Scots spellings and uniquely Scottish words, as well as their adoption of New 麻豆传媒高清 colloquialisms and te reo M膩ori. Scottish immigrant letters will be quantitatively analysed to identify the frequency of these features and which factors influenced their occurrence. Combining these results with sociocultural insights derived from historical scholarship, this research will consider how this reflects the competing roles of immigrants鈥� ethnic loyalties and their emerging, post-colonial identities, and how that influenced processes of assimilation and adaptation that characterise early dialect formation. Firstly, letters produced by Scottish settlers will be digitised and semi-automatically transcribed using specialised software. The resulting text files will then be compiled into a purpose-built corpus, through which instances of Scots or New 麻豆传媒高清 English and M膩ori features will be extracted, along with information on the migrants鈥� background (such as gender, profession and location of settlement). This will be statistically analysed to explore writers鈥� use of each type of feature overall, in relation to one another, and in relation to their background attributes. This will uncover firstly the extent to which Scottish immigrants retain Scots features and which factors made them more likely to, suggesting the strength of the 鈥榟omeland鈥� on their behaviour. Secondly, the frequency of new features and which factors encouraged their adoption will be explored, with higher rates suggesting a desire to integrate into the local community. Finally, comparing behaviour across both sets of features and across different writers can highlight similarities and differences within this group, indicating the role of their multinational identities in shaping settler language use in nineteenth-century New 麻豆传媒高清.

Project Team

Sarah van Eyndhoven

Royal Society of NZ Rutherford: 2024 -2026 $400,000

Many languages contain complex words, which are composed of multiple meaningful parts (as in, e.g. 'catlike'). But are all complex words equally easy to dissect into their parts, or do speakers think of some complex words as 鈥榤ore complex鈥� than others? And can the ease with which a word can be dissected affect how it is pronounced? In M膩ori, for example, the words 鈥榳hakaiti鈥� and 鈥榳hakairo鈥� are made up of two parts, both with 鈥榓i鈥� spanning the boundary between the parts. The 鈥榓i鈥� is often pronounced differently in these words, probably because speakers think of the former word as more clearly made up of two separate parts than the latter. The pronunciation of both words also varies across speakers. This project looks at how variation in this kind of perceived dissectability can affect the pronunciation of M膩ori words.

We collect measures on how dissectable M膩ori words are and investigate how degree of dissectability affects aspects of the pronunciation of vowel sequences, including aspects such as tongue position, pitch, amplitude and timing. This furthers the documentation of te reo M膩ori and feeds directly into a current debate in linguistic theory.

Project Team

Jen Hay 鈥� PI
Jeanette King 鈥� AI
Forrest Panther 鈥� AI
Kirsten Culhane 鈥� Post Doctoral Felllow

Royal Society of NZ Marsden Fund: 2023 鈥� 2026 $647,000

How does the brain produce words? It is increasingly understood that variation in language is key to answering this fundamental question. In M膩ori, there is a word ending that attaches to verbs when they are used in the passive voice. There are 17 forms of this ending, and there is much cross-verb and cross-speaker variation. For example, the verb 'tapa' "to name" receives the ending form '-ia': 'tapaia' "is named". However, another possible form is '-ina': 'tapaina' "is named". The patterns of these alternating forms are very poorly understood, and there is almost no documentation of the distribution of these form variations. I will undertake text-based and experimental analysis to document and model this and other types of variation to determine the patterns involved. Factors that I will analyse include individual, and dialect variation, as well as changes in the use of these forms over time.

Project Team

Forrest Panther 鈥� PI
Penny Harris 鈥� PhD scholarship student

Royal Society of NZ Marsden Fund Fast Start: 2022 鈥� 2027 $360,000

Sociolinguists have long studied relationships between individual sounds and social meaning. The way the speaker pronounces the vowel in 鈥榗at鈥�, for example, conveys social information about that speaker. However recent work shows that sounds don鈥檛 work in isolation of each other, but rather, as coherent subsystems. For example 鈥� if a speaker pronounces /a/ in an extreme way, we can make predictions about their likely pronunciation of a range of other vowels. It is likely that these patterns of co-variation may reflect clusters of speakers with shared social characteristics, and subsystems of sounds that carry shared social meaning. However, this can only be established through close perceptual work. With large-scale experiments using recordings of naturally-occurring speech, we investigate whether patterns of co-varying sounds can carry social meaning, and whether this social meaning is accessible to the listener. Our approach is completely bottom up - allowing both the co-varying variables considered and the potential social meanings to arise directly from the data. The question of whether social meaning can be carried by patterns of co-variation in speech is fundamental to our understanding of how social meaning accrues to linguistic forms, and thus - ultimately 鈥� essential to understanding how language works.

Project Team

Kevin Watson 鈥� PI
Jen Hay 鈥� AI
Lynn Clark 鈥� AI
Elena Sheard 鈥� previous Post Doctoral Fellow
Ruey Fern Tan 鈥� Research Assistant
Marlee Rosoff 鈥� MA scholarship student

Royal Society of NZ Marsden Fund: 2022 鈥� 2026 $660,000