Introduction

Derived phonological contrasts (DPCs) occur when the set of distinctions in polymorphemic forms is greater than the set found in monomorphemic lexical items (Harris 1990). For instance, the past morpheme <ed> causes the vowel in the word ‘brewed’ to be appreciably longer than its counterpart in ‘brood’ in many speakers of Scottish English (Rathcke and Stuart-Smith 2016).

In structuralist phonological parlance, ‘brood’ and ‘brewed’ constitute a minimal pair; therefore, both vowels – were we to adopt this view – should qualify as fullyfledged phonemes. This option is probably not desirable because it misses at least two points. First, quite often, DPCs give rise to a very scarce functional load, which often correlates with unstable phonemicity. Second, many vowels involved in DPCs have a complementary distributional nature which would not be adequately captured if we allowed them to share the same status as vowel contrasts elicited with monomorphemic items. Now, turning to the generativist paradigm, it should be noted that, in the SPE framework, DPCs are analysed in exactly the same way as allophones: a rule including information on morpheme boundaries generates two different surface realisations from a single underlying entity. This approach is not entirely satisfactory either. As Harris (1990) explains, the historical behaviour of DPCs has been shown to closely resemble that of genuine phonemic contrasts. Once DPCs have emerged, an additional lexical category is potentially available for inclusion in underived morphemes, and the contrast can eventually penetrate the phonemic inventory. In this chapter, I adopt an intermediate view, according to which DPCs have a status of their own, sometimes called quasi-phonemic (Scobbie and Stuart-Smith 2008; Hall 2013).

My goal is to examine acoustic and perceptual data from various locations in the British Isles – Enniskillen (Ulster), Glasgow and Hull – in order to better understand the production and perception of certain DPCs found in English. This research falls within a broader framework that I call the Gradient Phonemicity Hypothesis (GPH), according to which a whole range of cognitive statuses is assumed to exist between allophony and phonemicity, and a difference between two sounds can be described by some measure of how typically allophonic or phonemic their relation is.