Zum Inhalt springenZur Suche springen

Form and meaning in English compounds: The role of prosody

Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 527200619 Funding period: October 2023-September 2026

Principal investigator

Prof. Dr. Ingo Plag

Postdoctoral researcher

Dr. Dominic Schmitz

Mercator fellow

Professor Melanie Bell (Anglia Ruskin University, Cambridge, UK)


An enduring issue in the study of compound words is the variable placement of stress in English noun-noun compounds: ‘left stress’, with main prominence on the first compound constituent (the modifier, N1), or ‘right stress’, with main prominence on the second constituent (the head, N2). For example, according to reference grammars of English, Oxford Street and opera singer have left stress, while Oxford Road and summer dress have right stress (e.g. Bauer et al., 2013, chapter 19). Various authors have attempted to formulate rules to account for the patterns described (e.g. Chomsky & Halle 1968, 17).

However, a large body of empirical work has more recently shown that compound stress assignment is not governed by deterministic rules but depends probabilistically on the distribution of properties across other compounds in the speaker’s mental lexicon, with factors such as informativeness, analogy, semantics, lexicalization and length playing a role (Arndt-Lappe, 2011; Bell, 2015a,b; Bell & Plag, 2012, 2013; Kunter, 2011; Kunter & Plag, 2007; Plag, 2006; Plag et al., 2007, 2008; Plag & Kunter, 2010; Plag, 2010). Yet despite this enhanced understanding, there is still no account that explains how these effects emerge from a speaker’s experience.

This project will address this gap by exploring to what extent the theory of discriminative learning can explain the patterns observed. We hypothesize that stress patterns emerge through an association between form and meaning in a process of discriminative learning. We will test this hypothesis using computational modeling of large amounts of speech data within the framework of the Discriminative Lexicon Model proposed by Baayen et al. (2019). By testing the hypothesis that such a general mechanism of association can account for apparently disparate effects, the project will make an important contribution to linguistic theory.