The article looks at instances of specialisation for specific linguistic contexts in ‘command’ and ‘inference’ uses of will and must. It tests the feasibility of different motivations for this specialisation, such as statistical and construal pre-emption. It also proposes a new motivation for specialisation, polysemous pre-emption, i.e. whether a strongly entrenched polyseme of a given expression might pre-empt the use of an expression with a less strongly entrenched polyseme. The investigation uses corpus analysis and distinctive collexeme analysis to test the three motivations (statistical, construal, and polysemous pre-emption). The results show that all instances of specialisation with will and must could be explained through construal pre-emption and/or polysemous pre-emption, thus making recourse to statistical pre-emption unnecessary.

The English modals have been used as case studies in many domains of linguistic enquiry. Their diachronic development and patterns of synchronic variation in historical and contemporary corpora have been used to develop theories of linguistic representation, to further understanding of correlations between structure and use, and to investigate relationships between form and meaning. However, much of this research explores only the modals themselves: relatively little attention has been given to the study of modal collocations. In this article, we explore variation and change in collocational patterns of two modals (may and might) when they appear directly adjacent to the adverb well. Our analysis is corpus based, using quantitative data to explore macro-level trends in recent American English, and qualitative analysis to explore micro-level variation, particularly with regard to the development of concessive uses of may and might, and post-modal meanings more generally. We foreground the idea that modals show subtly different diachronic trends in specific collocations compared to perceived trends when looked at as an isolated class of auxiliary verbs.

While literature on English modality has usually focused on traditional modal and semi-modal verbs, to our knowledge, no attention has ever been given to the emerging be having to (BHT) construction. Through corpus analysis conducted on GloWbE, ICE, BNC and CLMET, this article investigates the semantic differences between have to and BHT that make them distinct in the English constructicon. We demonstrate that BHT conveys meanings of contingency, reluctance and inchoativity, and propose that its recent emergence may stem from a specific functional gap within the English modal system. While have to appears to be gradually grammaticalizing with future-oriented functions, BHT seems to be renewing the original (and less grammaticalized) dynamic functions of have to. Finally, we explore the productivity of the construction across different English varieties and the reasons for its lower frequency in postcolonial varieties. The hypothesis of negative retentionism proposes that a feature that was absent in the lexifier language at the time of contact may indeed be found to be less frequently used in the contact variety at a later stage due to colonial lag.

If supposedly homophonous words were acoustically distinct despite sharing phonemic form, theories of mental storage may have to account for the consistent differences with separate storage for each homophone. Previous studies of the homophonous functions or word classes of the English word like showed such subphonemic differences between functions, though some studies also found effects of utterance context alongside these. Schleef & Turton (2018) argued that all these function effects reduce to context effects, since function is not independent of context – for example, quotative like typically occurs before a pause and thus is typically subject to lengthening because of its position, not due to a lexicalised acoustic distinction between functions. Testing this argument with new data from a different regional variety to those used by Schleef & Turton, we only find differences that can be explained by context, in line with their argument. This casts prior findings of acoustic distinctions between like functions in new light, and introduces the need for further research (especially including the frequency of different functions).

This article offers a new reconstruction of the phonological history of pre-Old English, building on a potential parallelism between English, Frisian and North Germanic. Pivotal to the reconstruction is the development of PGmc *a, which is the target of eight different sound laws in the traditional theory. A combination of a conditional early fronting and rounding, followed by a gradual i-mutation impact, both with parallels in Frisian, and a relatively late seventh-century application of breaking before ‑rC can account for most of the attested spellings of instances with PGmc *a in the language of the early Épinal and Erfurt glossaries. This approach is much simpler than the traditional theory and allows parallelisms to be (re)established between the earliest stages of Old English, Old Frisian and Old Norse.

This article explores the extent to which listeners vary in their ability to notice, identify and discriminate variable linguistic features. With a view to improving speaker evaluation studies (SES), three types of experiments were conducted (noticing tasks, identification tasks and discrimination tasks) with regard to variable features using word- or sentence-based stimuli and focusing on three variables and their variants – (ING): [ɪŋ], [ɪn]; (T)-deletion: [t], deleted-[t]; (K)-lenition: [k], [x]. Our results suggest that the accurate noticing, identifying and discriminating of variants is somewhat higher in words than in sentences. Correctness rates differ drastically between variants of a variable. For (ING), the non-standard variant [ɪn] is more frequently identified and noticed correctly. Yet, for the variables (T)-deletion and (K)-lenition, the standard variants are identified and noticed more successfully. Results of the current study suggest that a more rigorous elicitation of identification and noticing abilities might be useful for a more complete understanding of the nature of social evaluation.

In the 1880s, Sievers proposed that in Old English words such as *feorhes, the loss of the post-consonantal *h caused compensatory lengthening of the vowel: fēores. Since there are no unambiguous traces of this sound change in later English, widespread analogical restitution of the short vowels was assumed (e.g. from feorh). The evidence for this lengthening is largely metrical. I argue that while Sievers is correct that words like <feores> often need to scan with a heavy initial syllable, this need not be explained by a general lengthening in the language at large. Indeed, the distribution of where heavy scansions are required in verse is typical for metrical archaisms: late prehistoric metrical values of words preserved for poetic convenience. Just as wundor ‘marvel’ can continue to be scanned as monosyllabic *wundr, or contracted hēan can scan as disyllabic *hēahan, so can light-syllabled feores continue to scan as heavy *feorhes. The same sets of poems that prefer non-epenthesized or non-contracted forms also prefer the heavy scansions of feores-type words. If heavy scansions of feores-words are seen as a matter of poetic convention, then the hypothesis of compensatory lengthening in the language generally is left without evidence and should be rejected.

Modal concord refers to the phenomenon where the co-occurrence of two modal elements with the same flavor and force (e.g. may possibly, must certainly) gives rise to the interpretation of single modality. Given their (arguably) equivalent semantics, constructions with modal concord and single modal (e.g. may, must) can function as alternative choices in different contexts of use – how do speakers choose between them, and how is the choice perceived? In this article, we take a ‘Register’ approach and report an experimental study of MC in US English, addressing their linguistic and social meanings with versus without situational context. The results show that (i) modal concord constructions differ from single modal ones in linguistic meanings, which casts doubt on the concord assumption, and (ii) modal concord has distinct social meanings from those of single modal constructions. Our findings suggest a correlation between the meaning strength of a linguistic expression and the social perception about the speaker. Context, manipulated via the single situational parameter of interlocutor relation (close vs. distant), did not interact with the linguistic or social meaning of modal concord constructions, the implications of which are discussed in relation to the multidimensional nature of conversational situations and the method applied.

The modal auxiliary form must plus perfect aspect (must have +V-en) has recently acquired the meaning of direct evidentiality in Multicultural London English, the new London dialect. Because the new meaning is a recent innovation we have a rare opportunity to witness its development at first hand, unlike earlier changes in the history of must. Our analysis supports the view that the classic definition of evidentiality in terms of information source is too narrow to explain the expression of evidentiality in spoken interaction, and that a broader definition in terms of epistemic authority is more appropriate. We argue that the direct evidential meaning is a coherent further step in the semantic changes undergone by must during its history. It represents a previously undocumented pathway in the grammaticalisation of evidentiality. It also supports the view that evidentiality is not a purely lexical phenomenon in English.

English employs a variety of comparative formation strategies. Theoretical and corpus-based research has established that their distribution depends on a variety of factors. In this article, we take an experimental approach to test analytic, synthetic and double comparative forms in relation to register in American and British English. We report on a rating study investigating the appropriateness and interpretation in terms of evaluativity of the three comparative forms. Our findings confirm the hypothesis that the comparative variants are not considered equally appropriate, but the effect is not as strong as would be expected under the hypothesis that frequency of occurrence is directly related to linguistic judgments. The analytic and double comparative alternatives exhibit lower appropriateness levels than the synthetic comparative. Analytic and double comparative forms are rated as less appropriate in formal than in informal contexts, which did not show an effect on the synthetic form. Furthermore, the analytic variant shows a different behavior in terms of the interpretation than the other forms in that a stronger effect of evaluativity is detected. Limitations and future directions are discussed. Our study is the first to provide experimental evidence for certain hypotheses emerging from corpus-based research.