The phonology of Welsh by S. J. Hannahs (review)

Hannahs’s book has solidly caught the attention of phonologists, as is evident in the number of published reviews it has engendered (Breit & Harris 2013, Mondon 2014, Czerniak 2015, Hammond 2015, Morris 2015). Each of these reviewers takes up the phonological details and analysis in the book, and on the whole they find much more to praise than to criticize. Whether appraising the opening sociohistorical discussion of Welsh (Ch. 1), the segment inventories with dialect variants (Ch. 2), a range of prosodic and segmental phenomena (Chs. 3–5), or even the closing invitation to scholars with regard to topics meriting further investigation (Ch. 7), the reviewers without exception declare H to have made a welcome and substantial contribution to the studies of Welsh and phonology with this text.

A portion of the book that reviewers appear to have found less comfortable, however, is to be found in Ch. 6, ‘Initial consonant mutation’.¹ The special status of this topic in a synchronic phonological description of Celtic languages is well known (reviewed by H on p. 120). H directly addresses the qualitative difference between those clearly phonological processes in Welsh, which may be accounted for in terms of optimality theory’s (OT) constraint types, maximizing faithfulness and minimizing markedness in preferred candidates (e.g. vowel mutation, vowel affection, and place and manner assimilation in consonants; see Ch. 4), and the phenomena referred to under the rubric of initial consonant mutations (ICMs), namely the grammatically and/or lexically conditioned alternation patterns that frequently reduce the phonological faithfulness of forms vis-à-vis their roots. They may involve the neutralization of distinctions (e.g. both /b/ and /m/ soft-mutate to /v/), or they may introduce marked segments with limited distributions (e.g. a series of voiceless nasal segments) in lieu of generally distributed phonemes (p. 123).

(1)

 

Clearly the ICMs had their origin in a phonetically motivated past state of affairs, but those environments as such no longer obtain and cannot be simulated theoretically without opting for a credulity-testing level of abstraction (Iosad 2010). H cites Green (2006) in this context, who concludes that a synchronic phonological analysis for the Celtic ICMs is not sustainable, arguing not merely that a phonology that could handle them would be excessively powerful, but furthermore that OT-phonologists themselves should not want these exercises in unfaithfulness and in rising markedness to turn out to be phonological in principle (H, pp. 133, 135). Assuring the reader that he ‘subscribes entirely’ (125) to the view that ‘ICM in Welsh … has become independent of the phonology’ (123), H nevertheless sets a three-part, phonologically oriented mission (125).

(i) To what extent is phonology involved in the mutations?

(ii) How are the alternations represented phonologically?

(iii) Are there aspects of ICM that can be related to phonological structure?

In light of the metaphor implied by the term mutation, the focus is understandably on the sounds that alternate within the ICM system, as follows (p. 126; orthographic representations in angled brackets).

(2)

 

From H’s perspective, only those initial consonants that have a distinct alternant in the table above are seen as participating in the ICM system at all (125), and those initials that have no correspondent in the nasal and/or aspirate rows appear in their radical (canonical) shape in the respective mutation contexts (see 3 and 4 below). In this way, the table serves to capture the range of observable alternations, and it systematizes them in a way that hints clearly at the phonological relations that hold between the majority of the sets. Appropriately, this table does not seek to offer insight as to the conditioning environments, since ‘[t]he initial mutations occur in various contexts; many are lexically determined, others morphologically or syntactically determined’ (127).

Whereas Green (2006) suggests lexical listing for all alternant forms, this step strikes H as injecting too much redundancy into the lexicon.² For H, the alternant patterns are to be extracted from the related full...