The lateral approximant in General British English (GB) is realised as light when occurring in the onset (leaf), and as dark in the rhyme (help, feel, google). Non-prevocalic positions are typical contexts for lenition, analysed in Element Theory as decomposition in weak positions. However, it is unclear how velarisation can be characterised as element loss if light [l] is represented as |A I|, while dark [ɫ] is represented as |A U|. Therefore, I propose that laterals in GB contain both the coronal |I| and the velar |U| element underlyingly (in addition to |A|), but because these cannot combine in a compound segment in English, they are both floating. Their association at the phrase level is determined by the apophonic chain (Guerssel & Lowenstamm 1996), mapped onto the structure of the syllable: |I| is attracted to the prevocalic position, |A| to the vocalic position and |U| to the postvocalic position. Darkening thus does not involve lenition of /l/, but partial interpretation in all positions. In contrast, I analyse vocalisation of dark [ɫ] as lenition, involving loss of |A| in weak positions. I integrate the lateral into the system of glides in English, and establish a typology of its behaviour across different accents.

Most speakers of Irish English use a dental stop for words containing <th>, a sound that is generally pronounced as [θ] and [ð], in other varieties of English (Wells 1982; Ó hÚrdail 1997). Alveolar stops [t,d] and dental stops [,] are articulatorily and acoustically similar, and thus it is unusual for a language to use them contrastively (e.g. Ladefoged 2001). Despite this, Irish English contrasts them and speakers of this dialect have no trouble distinguishing them. This raises the question as to whether speakers of a dialect which does not use this contrast can distinguish them. To investigate this, speakers of Irish English and American English participated in an identification task involving words produced by an Irish English speaker. American English speakers had a high accuracy but did significantly worse than Irish English speakers, and both groups did significantly worse when the contrast was in final position than when it was in initial position. A small-scale production experiment examined words with this contrast and the vowel /a/, with the finding that for speakers of both dialects, the vowel is longer in words ending in <th> than <t>. The findings are discussed in the context of linguistic experience, and the effect of surrounding consonants on vowel duration.

Research on complementizer selection has shown that the presence of a negative particle in a subordinate complement clause influences complement choice, leading to a relatively higher proportion of finite complementation patterns by increasing the complexity of the syntactic environment. Studies have also shown that different types of negation, namely not- and no-negation, increase the tendency towards more explicit complementation options (Rohdenburg 2015). The current study focuses on the effect of not- and no-negation on the complementation profile of the verb regret, which allows variation between finite that/zero-complement clauses and nonfinite (S) -ing clauses. The GloWbE corpus was used to create a data set of more than 4,000 examples from 16 varieties of English. The results of the analysis support previous findings that the presence of a negative marker in the complement clause increases the preference for finite patterns, especially in L2 varieties of English. However, contrary to the expectations of this study, no-negation was found to have a stronger effect on complement choice than not-negation.

The goal of this article is to uncover the system underlying three types of English relative clauses, and to characterise their distinctive uses in discourse: NP-integrated ones, namely restrictive and ‘a-restrictive’ relative clauses, and non-integrated ones, represented by non-restrictive relatives.

The area at issue is central, since understanding the functioning of these constructions requires reference to the fundamental interface between grammar (language system) and discourse (language use). The discourse functions of the three subtypes of relatives are claimed to be underlain by their intrinsic morphosyntactic and semantic properties. A major aim is to highlight the relative degree of ‘communicative dynamism’ of each subtype of relative clause, in terms of its respective contribution to the construction of discourse.

In doing so, the article focuses on the distinctive properties of presupposed as well as non-presupposed restrictive relatives, and of definite as well as indefinite NPs containing integrated relatives more generally. Along the way, it critically examines certain controversial conceptions of the structural and functional features of the constructions at issue, and, in particular, the claim that there is no essential distinction to be drawn between ‘integrated’ and phrase-external relative clause subtypes at all.

This article deals with the mechanisms that language users employ in historical periods to represent spoken language in writing. I focus on a set of features known as speech descriptors, which allow users to combine representation and evaluation of the speech (and, possibly, of the speaker), such as most disgusting in ‘he used most disgusting language’. The study shows that such speech descriptors in combination with the lemma language are used for a number of sociopragmatic purposes in the Old Bailey Corpus which includes legal materials, especially testimony, from 1720 to 1913. The descriptors can be used to avoid having to repeat a more specific, and possibly offensive and socially or morally inappropriate utterance; they can be used to put the spotlight on the evaluation and impact of the wording rather than the original speech itself; and they can help users mitigate, deflect, or disprove an accusation or guilt. Overall, the article demonstrates that studying the speech of the past must necessarily involve investigating the complex choices made by language users in terms of what to represent, what means to use to represent the speech, and how they evaluate the speech (and speaker).

Recent research into right-dislocated pronouns has provided details of the form and functions of lone pronoun tag (ProTag) constructions in Present-day British English. In this article, we present the first systematic investigation of ProTag constructions in an earlier variety, Early Modern English. Using as our corpus the dramatic works of Jonson, Marlowe and Shakespeare – writers already known to make use of tag questions in their works – we identified and analysed ProTag constructions. Our findings reveal that ProTag constructions in Early Modern English differ from their Present-day British English equivalents with respect to possible functions: in the earlier variety ProTag constructions could have a ‘Question’ function, the same as tag questions. We also found the relative frequency of demonstrative ProTags compared to personal ProTags to be significantly different: personal ProTags are far more frequently attested than demonstrative ProTags in our corpus of Early Modern English drama texts; this is the reverse of what has been found for Present-day British English. We propose that a key factor in the observed change is extension of the types of referents that demonstrative ProTags can have. This study offers a new perspective on ProTag constructions, their classification and development.