If there is such a person as the average phonologist, he might have a conception of the relation between phonetics and phonology that comes close to the relation between social perceptions of crimes and a Code of Criminal Law. The Code's definition of various types of crimes and the penalty each type carries ultimately reflect, to put it crudely, the feelings of the people. Also, the Code's development will reflect social change. Criminal codes will typically incorporate the changing perceptions of the general public, and will now begin to include articles devoted to the use of the Internet, for instance. But at the end of the day, what counts in a law suit is what is in the Criminal Code, not the feelings of the people. So it is with phonology. It is easy to show that lexical forms are frequently related to functional (ergonomic) considerations, and that the way the grammar processes them into surface representations will amount to a reasonable articulatory task for the speaker, while equally the acoustic result will enable the listener to recognise these forms with reasonable ease. However, ultimately we say things the way we do because our lexical representations are the way they are, and our phonological grammar is the way it is.

It is often assumed that there is a sharp division between phonetic and phonological processes, but the two are often strikingly similar, as in the case of phonetic consonant–vowel coarticulation and phonological assimilation between consonants and vowels. Parallels of this kind are best accounted for if both types of phenomena are analysed within a unified framework, so similarities result from the fact that both phonetic and phonological processes are subject to the same constraints. A unified model of phonetics and phonology is developed and exemplified through the analysis of parallel phonetic and phonological assimilation processes. The model operates in terms of scalar phonetic representations to accommodate phonetic detail, but categorical phenomena can still be derived from the interaction of speech production constraints with constraints that motivate the formation of distinct categories of sounds for the purposes of linguistic contrast.

Through an investigation of glottalisation in several languages of North America, this article examines the role of perception in the grammatical distribution of segments. Specifically, we consider the hypothesis that perceptual factors on the timing of glottalisation in consonants determine the arrangement of such consonants phonologically. Our conclusion is in the negative. In the cases we study, the distribution of glottalisation appears to be governed by syllable structure, not by glottal timing. We interpret the patterns presented as evidence for a modular theory of phonetics and phonology rather than a theory where properties of articulator timing are directly referenced by phonological constraints. Patterns of glottalisation result from the interaction of phonological submodules that are largely independent of each other, though properties of particular modules may be influenced by phonetic factors.

This paper proposes a phonetically based account of phonological stop assibilation. We first review evidence for the phonological representation of sibilants proposed by Jakobson et al. (1951) and others. An examination of cross-linguistic data involving phonological assibilation reveals two asymmetries: (a) only high vocoids constitute assibilation triggers; (b) only following vocoids, not preceding vocoids, constitute assibilation triggers. It is then shown that high vocoid-conditioned phonological stop assibilation is motivated by the brief period of turbulence that sometimes occurs at the release of a plosive into a high vocoid. We argue that this turbulence is interpreted in phonology as the insertion of the feature [+strident] into the representation of the plosive, with the loss of the feature [−continuant] in the case of frication. Not only does this phonetic analysis explain the phonetic origin of assibilation, it also correctly reflects the cross-linguistic conditions on high vocoid-conditioned phonological assibilation.

This paper presents evidence for two claims: (a) that the underlying contrast between stops in Swiss German dialects is based purely on quantity and (b) that the duration of the stop closure is its sole reliable phonetic reflex, i.e. there is a geminate–singleton opposition acoustically manifested in long–short closure duration. Using production and perception data on initial, medial and final stops in Thurgovian, a dialect spoken in north-eastern Switzerland, we show that the pattern of phrase-medial contrast neutralisation supports both arguments: when the extra phonological length position of a geminate is not syllabifiable, the closure duration shortens and underlying geminates and singletons become indistinguishable. The perception data in particular make evident that closure duration is the crucial cue of the underlying contrast, because, in the absence of this phonetic correlate, listeners can no longer discriminate an underlying geminate from a singleton. The results bear not only on central issues concerning the representation of geminates but also on some intricacies of the phonology–phonetics interface.

In this paper, I propose an optimality-theoretic account of the generalisation that deletion processes that apply to intervocalic biconsonantal clusters canonically delete the first consonant (schematically, VC1C2V → VC2V). The approach to contextual neutralisation proposed here has two main components. First, I follow the licensing-by-cue framework (e.g. Steriade 1997) in identifying ‘weak’ elements as those without strong perceptual cues. Second, I argue that the constraints responsible for contextual neutralisation ‘target’ weak elements. This approach captures the deletion generalisation above, because the relevant targeted constraint prefers only the correct output VC2V (from which the weak consonant C1 has been removed), not the incorrect output VC1V. Intuitively, the representation containing a weak element (VC1C2V) is compelled to neutralise to a representation that is perceptually very similar (VC2V). The targeted-constraint approach is formalised by replacing the standard violation-based definition of OT optimisation with a new definition – which is equivalent except when ‘targeted’ constraints are involved – based on harmonic orderings. The approach is shown to extend to certain cases of (i) contextually determined feature neutralisation and (ii) phonological opacity.