Mismatches in weight criteria across weight-sensitive processes within individual languages present difficulties for theories of moraic structure, particularly regarding coda weight. Previous accounts, which stipulate that codas are variably moraic to account for the typological variation in the weight status of CVC for primary stress, make incorrect predictions for the status of CVC in other weight-sensitive phenomena, including tone, word minimality and secondary stress, among others. This article proposes a theory of Uniform Moraic Quantity coupled with a new syllable weight metric as a solution, which captures CVC’s flexible weight status while maintaining the cross-linguistic moraicity of codas and avoiding the incorrect predictions that frustrate the standard variable-weight approach.

It is well-known that English variable word-final coronal stop deletion (CSD) is less likely to occur when the final coronal stop instantiates the inflectional suffix -ed. It is sometimes hypothesised that the reason for this effect is to avoid the homophony between past and present tenses that would result from the suffix -ed being deleted. This reasoning suggests another hypothesis: that CSD should also be disfavoured when it would create homophony between two distinct lexical items, such as bald and ball. In this squib, we test that hypothesis on data from a corpus of Philadelphia English. We find no evidence that probability of CSD is affected by homophony avoidance between lexical items. This weakens the case that homophony avoidance is at play in disfavouring CSD in the -ed case, and may have implications for the theory of homophony avoidance in phonology in general.

This study tests the hypothesis that alternation patterns with strong lexical support are more robust than those with no, or weak, lexical support. Focusing on three alternation patterns in Korean with varying productivity and generality, we measured lexical support in two ways. First, we conducted an acceptability-rating experiment investigating Korean speakers’ judgements on non-words with and without violations of the phonotactic constraints motivating the alternations. In addition, we performed a simulation of learning a maximum entropy (MaxEnt) Harmonic Grammar from a dictionary corpus. The results of the experiment and computational modelling confirmed the hypothesis by showing that if an alternation is robust, its associated phonotactic constraint is learned with a high weight from the MaxEnt simulation, and it affects the participants’ well-formedness ratings for non-words. Consequently, the results of this research support the claim of a tight link between alternations and phonotactics.

It is well-attested that floating tones can associate across a word boundary, but it is typologically unusual for floating weight units to do so. The Nuer language presents a floating suprasegmental component (FSC), which is part of lexical morphemes, and includes a unit of quantity and a High tone. This component is located at the left edge of nouns and is realised primarily across a word boundary on a preceding vowel. This article examines the FSC through a phonological analysis and a production study with eight speakers. These investigations reveal how the FSC interacts with the specifications for vowel length and tone of the adjacent context. Specifically, the weight unit of the FSC lengthens a preceding word-final short vowel, and its High tone combines in a compositional manner with tone of this preceding context. Comparisons with related languages suggest that the FSC developed out of a word-initial vowel /a/.

Phonotactic learning has been a fertile ground for research in the field of phonology. However, the challenge of lexical exceptions in phonotactic learning remains largely unexplored. Traditional learning models, which typically assume all observed input data to be grammatical, often blur the distinction between lexical exceptions and grammatical words, consequently skewing the learning results. To address this issue, this article innovates a categorical-grammar-plus-exception-filtering approach that harnesses the discrete nature of categorical grammars to filter out lexical exceptions using statistical criteria adapted from probabilistic models. Applied to naturalistic corpora from English, Polish and Turkish, the learnt grammars showed a high correlation with the acceptability judgements in behavioural experiments. Compared to benchmark models, the model performs increasingly better with data that contain a higher proportion of lexical exceptions, reaching its peak in learning Turkish non-local vowel phonotactics, highlighting its ability to handle lexical exceptions.