Mál og saga stendur fyrir málþingi föstudaginn 23. október nk., daginn fyrir níunda Ólafsþing. Málþingið fer fram í stofu 285 í Sögu og hefst kl. 15:00. Þar flytja erindi þau Jayden Macklin-Cordes og Maria Copot frá Surrey Morphology Group við Háskólann í Surrey í Englandi, en Jayden Macklin-Cordes starfar einnig við ISLE Institut við Háskólann í Zürich í Sviss. Erindin verða flutt á ensku.
Heiti fyrirlestranna og lýsingar eru sem hér segir:
(1) Why do noun classes persist? Predictive processing and constituent order in a global sample
Jayden Macklin-Cordes (flytur erindið) & Maria Copot
The existence of grammatical gender is perplexing. Noun classes (i.e grammatical gender) are redundant, hard to learn, and readily lost but almost never innovated – yet around a third of the world’s languages have them (Allassonnière-Tang et al. 2021). Why do they persist?
Broadly, two explanations dominate. Noun classes may (i) help track and disambiguate referents, or (ii) support prediction, letting a listener narrow the search space of the mental lexicon before the noun arrives (Dye et al. 2018). Crucially, these make opposite typological predictions: prediction requires the class marker to precede the noun, while reference-tracking works best after it, linking back to a referent already introduced. The position of noun-class-marked constituents within the NP is therefore a testable signature of function.
We test this across 2,018 languages (WALS, Grambank, Autotyp), modelling whether four NP constituents – adjectives, numerals, determiners, and articles – carry noun class agreement as a function of their pre- or post-nominal position, using Bayesian phylogenetic regression to control for genealogy. Importantly, our model considers evidence from languages without noun class agreement, avoiding a survivorship bias (cases where noun classes are absent/lost are just as informative as cases where they are retained/gained). We then ask whether gain and loss of noun class marking is coupled to diachronic changes in constituent order along the branches of the phylogenetic tree.
Both analyses point to predictive processing as a functional explanation. Three of the four constituents favour pre-nominal marking – decisively so for articles and suggestively for numerals and determiners. The exception is informative: Adjectives are the most lexically contentful constituent, suggesting a division of labour: a reference-tracking role for the constituent that carries the most lexical content, alongside a predictive processing role for the more grammaticalised constituents. For all four constituent types, the rate at which noun class marking is gained and lost depends on constituent order rather than evolving independently of it. These results demonstrate how a minute, cognitive processing pressure can leave its imprint on global typology.
(2) Language ideology and the distribution of defectiveness and overabundance
Maria Copot (flytur erindið), Mari Aigro, Andrea Sims, Virve-Anneli Vihman
A canonical paradigm has one form per cell. Defectiveness and overabundance are two opposite deviations from the canon: a cell with no accepted form (a gap), and a cell with two or more (a doublet). A central question about both concerns their distribution in the lexicon: why are some lexemes defective or overabundant when others are not? Answers are generally sought in structure: it is assumed that some property in the organisation of the lexicon is licensing the gap or the doublet. This makes the research task about formulating a generalisation: the set of affected lexemes is treated as given, and the goal is to identify the structural properties they share. But properties that pick out exactly the affected lexemes, and no others, are notoriously difficult to find, which possibly points atthe importance of grammar-external factors. This talk asks whether social factors, and not structure alone, shape defectiveness and overabundance, examining how language ideologies, and individuals’ alignment with them, affect speakers’ acceptability judgments.
To that end, acceptability-judgment experiments are conducted on French verbal defectiveness and on Estonian overabundance, two systems with strongly codified but oppositely oriented standards: French ideology is markedly anti-variationist, Estonian comparatively variation-tolerant. The pairing poses a natural question: if an intolerance of variation can produce gaps, is a tolerance of it likely to license doublets? Judgments are examined for their dependence on the ideology encoded in the standard, on the individual speaker’s alignment with it, and on lexeme frequency. The results suggest that social factors are an integral part of how defectiveness and overabundance manifest, and potentially causative of difficulties in identifying conditioning structural factors, since structural factors are not the only thing at play. Additionally, we find that defectiveness and overabundance elicit strikingly similar acceptability profiles once speakers rather than dictionaries are consulted, a convergence that plausibly reflects the same cognitive mechanisms operating in both.
Um fyrirlesarana:
Dr Jayden Macklin-Cordes is a postdoctoral researcher at the ISLE Institute, University of Zürich. He investigates how languages and linguistic structures change over deep time, and what their evolution reveals about human cognition and culture. Using Bayesian phylogenetic methods, his work focuses on the Indigenous languages of Australia and New Guinea, and extends to large-scale comparison across the world’s languages.
Dr Maria Copot is a British Academy International Fellow at the Surrey Morphology Group (University of Surrey). Their research asks how knowledge of words is organised in speakers’ minds, and how that organisation interacts with memory, cognition and the social and communicative contexts in which language is used. The work approaches these questions empirically, combining behavioural experiments, computational modelling and large-scale statistical analysis to test claims from morphological theory, with a particular focus on paradigmatic organisation, defectiveness and overabundance.