Introduction

Background

Flapping, or tapping, of /t/ involves the realisation of /t/ as a voiced alveolar flap. It is a well-known feature of American English that has been widely discussed in the linguistic literature. Flapping is a highly complex phenomenon, which is potentially influenced by a number of linguistic and extra-linguistic factors (stress, syllabification, phonetic surroundings, morphology, lexical frequency, as well as social and stylistic factors). There is no consensus on the exact conditions for flapping, and there are often inconsistencies in pronunciation dictionaries and other sources with regard to flapping in individual lexical items. One reason for this lack of agreement is the extensive variation in the use of flapping in certain contexts. This chapter explores aspects of this variation and attempts to bring some clarity to the situation by supplying quantitative data on actual usage. The main goal of the study is to estimate the relative frequency of the flap versus the plosive in three variable environments and identify the characteristics of any variation patterns that emerge. A second aim is to suggest possible explanations for the variation or lack of flapping in certain items (many with accentually parallel forms that regularly display flapping) that have not been satisfactorily accounted for in previous works.

American English /t/ flapping has been extensively described and discussed in numerous sources. Many previous accounts are purely theoretical and focus mainly on establishing rules describing the exact phonological conditions under which flapping applies, often by proposing various abstract mechanisms (for example Kahn 1976; Kiparsky 1979; Selkirk 1982; Nespor and Vogel 1986; Jensen 1993; Harris 1994; Davis 2005). Variation in the use of flapping has received less attention. The studies which have an empirical basis typically make use of experimental data, where subjects read wordlists or sentences (for example Zue and Laferriere 1979; Egido and Cooper 1980; Turk 1992; de Jong 1998; Steriade 2000; Riehl 2003; Fukaya and Byrd 2005; Kaplan 2008; Herd et al. 2010; Warner and Tucker 2011). A few studies have used corpora: Patterson and Connine (2001) use the SWITCHBOARD speech database, which is a collection of 2,400 telephone conversations, while Eddington (2006, 2007) and Byrd (1994) collect their data from the TIMIT corpus, where 630 speakers read the same ten sentences.