Conference Agenda

In this study, we investigate the complex interaction of phonological and analogical developments in the early English nominal inflectional patterns, using an agent-based model. The nouns affiliated originally with the Proto-Germanic ‘s’-stem declension (appearing as /r/ in modern languages, as in PDE child – children, cf. High German Kalb – Kälber ‘calf/calves’) show a three-way inflectional development, whereby the inherited allomorphy (the r-formative) is eliminated, retained or partly retained in their paradigms (Hogg & Fulk 2011: 58-61; Klein 2013; cf. Versloot 2014: 102-105 for similar patterns in Old Frisian). So far, no rationale has been mentioned that explains which lemma follows which specific trajectory.

The aim of this study is to provide an algorithmic interpretation of the developments in these paradigms, involving three interacting factors: apocope, analogy and relative frequency of paradigm cells. In order to do that, we developed an agent-based computational model of language change (Bloem, Versloot & Weerman, 2015) which correctly produces distributions of apocope and analogical paradigmatic levelling as attested in specific Old English lemmas. Frequencies from the Dictionary of Old English corpus (Healey et al., 2009) were used as model input, covering nearly all attested Old English text. The study demonstrates that the inflectional irregularities resulting from seemingly opposing impacts of phonological change and analogical levelling can be explained as a consequence of these factors’ interaction with lemmas’ individual case-number frequencies, resulting in varied frequency-dependent sub-patterns. These frequency variations relate to the meaning of the words and their use in the speech community, as in the case of hrīđer ‘cattle’ that is often used in the plural, or sige/sigor ‘victory’ that is much more common in the oblique cases in the singular. We also show that computational modelling and in particular exemplar-based modelling can be a productive avenue for analysing language change and thus have an added value for historical linguistic research.

References

Bloem, Jelke, Versloot, Arjen, and Weerman, Fred. 2015. An agent-based model of a historical word order change. Proceedings of the Sixth Workshop on Cognitive Aspects of Computational Language Learning: 22-27.

Healey, Antonette de Paolo, John Holland, David McDougall, Ian McDougall and Xin Xiang (eds.). 2009. Dictionary of Old English Electronic Corpus. http://www.doe.utoronto.ca/pages/pub/web-corpus.html.

Hogg, Richard M. and Robert D. Fulk. 2011. A Grammar of Old English. Vol. 2, Morphology. Chichester: Wiley-Blackwell.

Klein, Thomas. 2013. Zum r-plural im westgermanischen. NOWELE: North-Western European Language Evolution 66(2): 169–196.

Versloot, Arjen P. 2014. Die -ar-Plurale im Altwestfriesischen mit einem Exkurs über die sächlichen Plurale im Westfriesischen. Us Wurk 63: 93-114.

Several researchers (e.g. Gilquin & Gries 2009; Römer, Skalicky & Ellis 2020) have already emphasised the usefulness of combining corpus data with experimental data, as “different types of data present converging evidence to strengthen research hypotheses” (Römer, Skalicky & Ellis 2020: 304). This synthesis would benefit from a further comparison with agent-based modelling, which adds another way (or “domain”, as Livet, Phan and Sanders (2014) call it) to approach a linguistic phenomenon. In this study, the focus is on the role of analogy in the creation of innovations in World Englishes, more specifically Indian English and Hong Kong English. A corpus search in the International Corpus of English provided me with several innovations (e.g. discuss about X instead of discuss X) that have an existing analogous nominal and verbal form (e.g. a discussion about and talk about). I then created a forced-choice task where the participants were shown a sentence with an innovation and were asked to choose between the nominal and verbal analogy based on which form they found the most similar to the innovation. The agent-based model is a simulation of the interaction of the participants with these stimuli during the experiment and allows to compare the effect of different (sociolinguistic and language-internal) factors that might have influenced the participants’ choices, such as the variety, similarity in meaning, similarity in form, and frequency and priming effects.

References

Gilquin, Gaëtanelle & Stefan Th. Gries. 2009. Corpora and experimental methods: A state-of-the-art review. Corpus Linguistics and Linguistic Theory 5(1). 1–26. https://doi.org/10.1515/CLLT.2009.001.

Livet, Pierre, Denis Phan & Lena Sanders. 2014. Diversity and Complementarity of Agent-Based Models in the Social Sciences. (Trans.) Peter Hamilton. Revue française de sociologie. Paris: Presses de Sciences Po 55(4). 689–729.

Römer, Ute, Stephen C. Skalicky & Nick C. Ellis (2020). Verb-argument constructions in advanced L2 English learner production: Insights from corpora and verbal fluency tasks. Corpus Linguistics and Linguistic Theory 16(2). 303–331. https://doi.org/10.1515/cllt-2016-0055.

Discussion between presenters and audience about cases when to use (and when not to use) agent-based modelling in historical sociolinguistics, design choices and possible datasets.