This paper provides new evidence and analysis of gradual and U-shaped phonological learning. Using a rich longitudinal corpus from Trevor (Compton & Streeter 1977, Pater 1997b) we demonstrate that some of Trevor's consonant harmony follows a statistically significant U-shaped trajectory, in contrast to the more typical S-shaped progression of his complex onsets. We then analyse these two developmental paths using an OT model of phonological acquisition (Hayes 2004, Prince & Tesar 2004, Tessier 2007, 2009), in which the learner's variation within and across stages is the effect of stored old productions rather than a variable grammar. The decrease in Trevor's faithfulness to place of articulation due to consonant harmony is attributed to the induction of a new constraint during the course of learning. Our analysis is paired with a computational implementation, showing how competition between old forms and the current grammar allows the model to derive both S-shaped and U-shaped patterns.

The term sonority projection refers to behavioural distinctions speakers make between unattested phonological sequences on the basis of sonority. For example, among onset clusters, the well-formedness relation [bn]>[lb] is observed in speech perception, speech production and non-word acceptability (Davidson 2006, 2007, Berent et al.2007, Albright, ms). We begin by replicating the sonority projection effects in a non-word acceptability study. Then we evaluate the extent to which sonority projection is predicted by existing computational models of phonotactics (Coleman & Pierrehumbert 1997, Hayes & Wilson 2008, inter alia). We show that a model based only on lexical statistics can explain sonority projection in English without a pre-existing sonority sequencing principle. To do this, a model must possess (i) a featural system supporting sonority-based generalisations, and (ii) a context representation including syllabification or equivalent information.

Sprung rhythm is a complex poetic metre invented and used by Gerard Manley Hopkins. We re-examine and amplify a seminal analysis of this metre by Kiparsky (1989). We coded the sprung rhythm corpus for stress, weight and phrasing, then used a computer program to locate every scansion compatible with Kiparsky's analysis. The analysis appears to be nearly exceptionless. However, it is incomplete in that it permits dozens or even hundreds of scansions for certain lines. We propose a Parsability Principle for metrics mandating that ambiguity of scansion be minimised, and suggest that under this proposal, the Kiparskyan system is not a possible metre. Our own revised analysis adds ten new constraints and is cast in the form of a stochastic maxent grammar. It produces an acceptably low level of ambiguity in metrical parsing, and is supported by a good match to the diacritics Hopkins employed to mark his intended scansion.