The basic types of lenition environments ('initial', 'intervocalic', 'final') need to be separately evaluated as they differ along parameters like word position (e.g., pre-consonantal vs. final codas) or stress relations. This paper argues that we need to recognise an additional such parameter: the length of the vowel preceding an intervocalic consonant. We show that a number of phenomena from varieties of English and German show lenition patterns which draw a distinction between reflexes found in post-short (vc) and post-long (vvc) environments. The theoretical consequence of our observations is that phonological theory needs to be able to account for the post-short vs. post-long distinction in the form of a parametrically-determined representational difference.

The paper stresses the need to distinguish between two subtypes of binary laryngeal systems, viz.[voice] languages versus [spread glottis] languages ("laryngeal realism"-Honeybone 2005). It criticizes the use of the primes H and L for this distinction in Government Phonology, and proposes an alternative representation, based on BackleyTakahashi (1998) and Nasukawa-Backley (2005). This feature geometric model assumes the same set of melodic components for obstruents and sonorants within a system but with a difference in the status of source elements across the language types. Therefore, accompanied by the mechanism of element activation, it is claimed to capture the cross-linguistic observations more adequately.

Two-way laryngeal systems are classified by Laryngeal Realism into [voice] languages (or “L-systems”, e.g., Slavic, Romance) and [spread glottis] ([sg]; or aspiration) languages (or “H-systems”, e.g., the typical Germanic pattern). More recently, Cyran (2014) has proposed Laryngeal Relativism (LR), claiming that phonetic interpretation is arbitrary, and as a result, two phonetically identical systems, even two dialects of a language, may turn out to diverge phonologically. His example is Polish: while Warsaw Polish represents the typical [voice]/L-system, he analyses phonetically identical Cracow Polish as an H-system (counter to Laryngeal Realism’s uniform classification of Slavic languages). However, in the “classical” version of [sg] languages (e.g., English), no laryngeal activity in the form of any kind of spreading is attested, which suggests the absence of any source element and, instead, a dominant role of obstruency (|h|). We, therefore, arrive at a three-way typology: h-systems, H-systems and L-systems. At the same time, arbitrary phonetic interpretation in LR predicts the existence of, e.g., h-systems with virtually no aspiration in the fortis series. We claim that this is indeed the characterisation of Italian. Using data from potential feature spreading situations, elicited in loanword and foreign accent settings, we show that Italian is an h-system, exhibiting no true laryngeal activity.

The variety of Basque characteristic of Getxo exhibits a form of coronal palatalisation that takes place intervocalically within and across words, triggered by a preceding [i] or [j]. This system in particular is interesting because it sets up a paradox as it both applies and does not apply at the word-level. The rule is sensitive to the leftward phonological context within the word-level and the rightward phonological context at the phrase level since in Getxo the trigger and target of palatalisation must come from the same word, yet the process only occurs if this sequence precedes a vowel-initial juncture: /in##V/, /il##V/. A previous solution involves stating palatalisation as a lexical rule and invoking a Duke-of-York Derivation to generate the masses of lexical exceptions attested largely in loanwords. This account misses a crucial generalisation, which is that, loanwords or not, there are no lexical exceptions across morphemes. We capture this generalisation and resolve the ordering paradox by relating palatalisation to the positional distribution of place features general in the language. This analysis involves positional underspecificaiton of nasals/laterals and a coronal default place of articulation. Underspecified nasals and laterals need place when they “become onsets” across word-boundaries, including through palatal spreading. In the reanalysis there are no “lexical exceptions” since these are underlyingly specified for place; neither is there need for word-level versus post-lexical phonology.

No abstract