All of the contributions in this special issue respond to a somewhat paradoxical situation: lexis (or vocabulary) is probably the area of linguistics that is most accessible and most salient for a non-specialist audience, but at the same time it presents some uniquely difficult challenges for systematic scholarly linguistic analysis. This is especially the case for approaches that focus on statistics or quantification of data, such as are typical of modern work in sociolinguistics. For this reason, it often seems that lexis is the Cinderella that is excluded from the ball, a topic that modern sociolinguists tend to steer clear of because of the methodological difficulties. This special issue will try to investigate the background to this problem, and suggest some possible solutions.

The main papers in this issue all come from specialists in different aspects of broadly variationist research in sociolinguistics; this introduction will open the issue with the rather different perspective of a lexicographer and etymologist, in an attempt to place this specific problem in a slightly broader context, especially for people who (like the present author) are not specialists in sociolinguistics. What is it about words, and variation in the use of words, that is so salient for the general public? What are the levels of variation that dictionaries are attempting to record and reflect? What are the tensions and the limitations for dictionaries, and how do these compare with the challenges for sociolinguists? Ultimately, what is it about variation in word forms and word meanings that presents such challenges for empirical research, and what can be done to accommodate these difficulties?

The appeal of lexis for the general public is obvious. Paradoxically, while the ‘word’ is a unit for which theoretical linguists struggle to find an entirely satisfactory definition, it is the unit that has most salience for the general public and is the focus for most popular observations about language.Footnote ¹ In societies where the written word has high prestige, a folk-definition of a word as a unit written with spaces on either side is generally deeply ingrained (even if there is widespread puzzlement about how to regard forms written with an apostrophe, e.g. I'm, don't, and still more concerning compounds written variously as solid or with a hyphen or with a space between the elements, e.g. wingspan, wing-span, wing span). Since the concept ‘word’ is generally regarded (by non-specialists) as non-technical and unproblematic, the general public is generally very comfortable about discussing variation in the use of words, and when linguistic variation is discussed in general discourse the focus is typically on words. New words are frequently seen as an index of social and cultural change, not least the words of the year (or decade) that are frequently the focus of media attention, e.g. bromance, chillax, globalization, etc. Very often these words are identified by dictionary publishers, typically in response to eager requests from the media. Complex social changes (or perceived changes) are often summarized under a single word as label, e.g. chav, yuppie, Sloane Ranger, squeezed middle.

Meaning change is not always recognized as such by a non-specialist audience. Where change is dramatic, as in the case of gay in the meanings ‘happy’, ‘homosexual’, or (more recently) ‘foolish, stupid, socially inappropriate or disapproved of’, this is commonly apprehended as a change in meaning and discussed as such; compare Justyna Robinson's contribution in this volume for an interesting selection of comments on this topic from interview subjects, showing how attitudes to semantic change can be bound up with and influenced by attitudes to social change. In other cases, such as recent use of wicked or awesome in the sense ‘excellent’, non-specialist comment may identify that something is in the course of changing, but the change is as likely to be identified as ‘misuse’ or even ‘overuse’ than specifically meaning change.Footnote ² This may also apply to changes in function, for example changes in the use of like (in e.g. I was like ‘Oh my God!’, it was like all dark) are highly salient, but likely to be described with vague formulations such as ‘overuse/misuse/abuse of the word like’, ‘use of like not as a verb/not as an adjective’. Even observations on phonological change are typically identified with particular totemic words, which may be salient as being among the earliest in a lexical set to show a change, but may be prominent for other reasons, such as association with a particular figure in the media; sometimes a phonetic respelling is used as a shorthand for both a (supposedly) distinctive pronunciation and for semantic or syntactic variation (supposedly) associated with this pronunciation, as e.g. propa or bovvered.Footnote ³

There is a particularly strong association of lexis with regional identity. This has not gone unnoticed by commercial enterprises: mugs, tea-towels, coasters, t-shirts, and other items bearing words strongly associated with a particular locality are a mainstay of gift shops at tourist attractions and elsewhere. In Scotland there is currently an extremely popular range of such items printed with characteristic Scots words, in the form of a mock-up dictionary entry, with headword, pronunciation (rendered by phonetic respelling), definition, and an illustrative example. In other parts of Britain, as in many other English-speaking communities, it is clear that a similar market exists and that some entrepreneurs are beginning to spot this. Even more of a mainstay of the gift shop market at local interest tourist attractions are short wordlists of regional items with standard English glosses or commentary (in this article I will use ‘standard English’ to refer to the broadly standardized variety generally in use in broadcasting, print media, and in most formal contexts in daily life). In fact, in such wordlists many of the ‘words’ may in fact be respellings of standard English word forms, accompanied by their standard English spellings as glosses, or they may be items selected to illustrate what are really syntactic or morphological phenomena of a local variety, but the ‘dictionary’ format is seen as the most appealing and most understandable packaging for such data for a general public audience. See Jonnie Robinson's contribution in this volume for a more detailed investigation of some of the salient characteristics of books of this type, and for some examples.

On the other hand, lexis poses some significant challenges for linguists, especially in areas such as sociolinguistics where statistical or quantificational approaches often predominate. Exploring the reasons for this in detail could more than fill this journal issue, but several factors stand out.

Firstly, it is well known that the lexis of any language is almost infinitely extendible. New words can be and are introduced freely in everyday contexts; when they are formed by the productive word-forming processes of a language and are semantically transparent, the hearer/reader may well have no conception that a word is new, nor indeed may its user. Since it is impossible to have an exhaustive record of all language use, it is in fact impossible to have a certain observation of a first use of such a word, and research methodologies depend on identifying words not listed in dictionaries, not found in corpora, or not found in huge text collections such as Google Books.Footnote ⁴ Use of very large text databases suggests the likelihood in many cases of a related phenomenon: low-frequency words of transparent meaning and derivation may in fact show polygenesis, being coined independently on many different occasions, to meet the demands of particular communicational situations.Footnote ⁵ The other major source of new words, lexical borrowing from other languages (whether through loanwords, loanblends, or semantic loans), is also pervasive, especially in the contexts of bilingualism or multilingualism that are the norm for many English speakers; compare Jonnie Robinson in this volume for Punjabi words reported as either loanwords or code switches in fieldwork from contemporary Southall in west London.

To speak of the lexicon of a language is in fact a somewhat problematic abstraction. The common core of vocabulary that is shared by all speakers of a language may be quite small. The active and passive vocabularies of any individual speaker will reflect a wide variety of different factors, such as regional origin, education, profession, interests, and ultimately the sum total of that individual's life experiences.

The implications of these factors for linguistic research become clear when we consider the structure of the lexicon of a language. A useful starting-point for considering this is the network of meanings that forms the basis of a thesaurus. For any meaning in the thesaurus, there may be a number of different synonyms or near-synonyms attested in the ‘language’; some may be register-specific variants, while others may show less obvious constraints in their use. Few individual speakers will know all of these words, still less use them all. Additionally, many of these words will also, when used in different contexts, realize different meanings, located elsewhere in the thesaurus structure; that is to say, they are polysemous.

The Historical Thesaurus of the Oxford English Dictionary (HTOED) allows us to make semantic searches of the lexical resources of the full Oxford English Dictionary (OED). If we take the category ‘female sweetheart or girlfriend’ numerous synonyms are recorded as current in some variety of contemporary English, including (giving their OED labelling): fancy-girl, girlfriend, inamorata, lady-friend, lady-love, lass, woman, young lady, young woman, (slang) bride, jelly, (colloquial, now chiefly North American) girl, (British, colloquial) her indoors, (British, slang) bit of homework, (in Liverpool, slang) tart, (in Australia and New Zealand, colloquial) sheila, (slang, originally Australian) sort, (originally U.S.) baby, best girl, (slang, chiefly U.S.) (sweet) patootie, tootsy, (slang, originally U.S.) queen.

There thus appear to be numerous ways of saying more or less the same thing, although some of them are marked by the OED as being restricted wholly or in large part to particular localities or to particular registers (e.g. slang). Some of the questions that are harder for dictionary entries to answer are how many of these words would be used comfortably in front of the sweetheart, how many belong only to the in-group language of men, how many would be used in front of other members of the family, or among colleagues, etc. The illustrative quotations will give some further indications here, but crucially this is something that may be very variable, diachronically, diatopically (i.e. between different localities), and even between different individuals in a single time and place.

Here are some first clear openings for sociolinguistic research, if a suitable methodology can be identified. However, there is another crucial property of lexis that greatly complicates things. In addition to there being many different ways of expressing (with many different nuances) the concept ‘female sweetheart or girlfriend’, many of the words that realize this concept also sometimes realize other concepts. If we take one of the most basic of them, girl, it is clearly polysemous, showing the basic meaning ‘female child’ as well as ‘female sweetheart or girlfriend’, as well as numerous other meanings, including ‘young or relatively young woman’, ‘prostitute’, ‘female servant or domestic employee; maid; female employee’, ‘effeminate man’, as well as (still marginally current in Wexford in Ireland, but formerly more widespread) ‘child of either sex; a young person’. What I have placed between inverted commas here are separately defined and illustrated senses in the OED. Normally particular examples of use can be assigned confidently to one or other of these, although there may be difficult cases. However, the division into different numbered senses reflects an analysis of the evidence, and it is often the case that alternative analyses are available, or that what are presented as separate senses in the dictionary could alternatively be interpreted as showing different contextual nuances of an overarching more basic meaning. Dictionary senses are typically a pragmatic device intended to display the range of uses of a word; as linguistic units, they are problematic.Footnote ⁶

Additionally, there are often different nuances of meaning within a particular dictionary sense. For instance, what I have summarized above as ‘female servant or domestic employee; maid; female employee’ is in fact defined in a more detailed way by the OED as ‘A female servant or domestic employee; a maid (now chiefly hist.). Now more generally: a female employee. Freq. with connotations of social inferiority. Cf. boyn.¹ 1a.’ This definition stands at the head of a varied set of illustrative quotations from 1668 to the present day, and clearly summarizes a variety of different but related uses. Similarly, the sense that HTOED lists under the category ‘female sweetheart or girlfriend’ is defined by the OED: ‘colloq. A sweetheart, a girlfriend; a wife. Also (in later use) best girl: see bestadj. 2c. Freq. with possessive adjective. Now chiefly N. Amer.’ Distinguishing between these nuances of meaning requires close inspection of individual contexts of use.

The difficulties posed by lexis for any statistical or quantificational approaches to linguistics thus begin to become clear. The lexicon is not a fixed entity, but is almost infinitely extendible, and the lexicon of each individual is likely to be unique. Thus, in any situation where we want to measure variation, we cannot make any easy assumptions about what the available pool of variants will be. Even the very largest dictionary must be selective in what it includes, and, especially as regards non-standard lexis, this applies even to the OED. See Durkin (Reference Durkin, Markus, Upton and Heuberger2010a, Reference Durkin and Hickey2010b) for illustration of some of the practical difficulties in coverage of older dialect material, and see also Jonnie Robinson in this volume for practical illustration of the new lexical items that can be unearthed by a well-designed and extensive fieldwork project.

Additionally, lexis shows variation in various different dimensions. Semantic variation can be considered from two different perspectives, in technical terms (i) onomasiological, i.e. looking at what the word forms are that realize a particular meaning, and (ii) semasiological, i.e. looking at what the meanings are that are realized by a particular word form.Footnote ⁷ If we start from meanings, it is clear that there are different ways of saying the same thing. But if we start from word forms, it is also clear that the same word form can mean different things. Thus, beside Labov's classic definition of the sociolinguistic variable as showing two or more ways of saying one thing,Footnote ⁸ we also have to consider that words can be one way of saying two or more different things. (For a more detailed account of the history of this question in the context of sociolinguistic research see the introduction to Justyna Robinson's contribution in this volume.)

In practical terms, there are a number of potential pitfalls here. Firstly, a set of words that can sometimes realize the same basic concept are highly unlikely to be full synonyms, especially if we take into account their typical contexts of use, their connotative meanings, and the associative factors resulting from other meanings attaching to each word: compare again the set of words listed in HTOED under ‘female sweetheart or girlfriend’ given above. Furthermore, the polysemy shown by individual words in the group, e.g. girl, poses considerable problems for any approach that intends to make use of the massive resources of a corpus: we must find some way of disambiguating particular senses, which is likely to involve careful reading and coding of each individual example, thus greatly reducing the effectiveness of a corpus as a short cut to large volumes of data. If coding of each example is attempted, some of the senses of girl listed above are easy for any human reader to disambiguate, but others may pose greater difficulties to distinguish in context, particularly if one is working at the level of nuanced distinctions of meaning, and a clear and consistent methodology in coding examples is essential.

When a diachronic dimension is added, the problems multiply. Typical types of meaning change have been well described (broadening, narrowing, metaphor, metonymy, amelioration, pejoration, etc.), and considerable progress has been made in identifying typical mechanisms by which semantic change occurs.Footnote ⁹ However, the key problems remain that meaning change is largely unpredictable, it occurs singly rather than (like many phonological changes) affecting all of the items in a group at once, and it is highly susceptible to extra-linguistic social, cultural, and technological factors.Footnote ¹⁰

These then are the problems that the contributions in this special issue investigate, and to which they may offer some methodological solutions.

The contribution by Jonnie Robinson demonstrates the enduring potential of fieldwork based primarily on lexis. As noted above, words are by far the most salient linguistic units for the non-specialist general public, and this can be exploited in mass surveys of lexical usage, even when, as here, a good deal of phonological data is also collected. The scale of the survey described here, supported by the vast resources of the BBC, is enormous, but this scale also has implications for the nature of the survey, which had to be capable of being employed simply and effectively by a wide variety of different non-specialist interviewers. This piece introduces to this volume the concept of the ‘spidergram’, which also recurs (in an earlier and more elaborated form) in the contribution by Beal and Burbano-Elizondo. This relational diagram invites interviewees to think around a semantic field, and to supply a number of synonyms or near-synonyms if they are aware of them. In the BBC Voices survey, ‘elicited’ responses arising directly from the spidergram were then supplemented by any ‘spontaneous lexis’ encountered in the course of subsequent discussion around the topic.

The paper illustrates amply the extendibility and complexity of the lexicon. Even the single sample piece of fieldwork from Kirkby-in-Ashfield, Nottinghamshire given in the appendix has some interesting implications for the HTOED synonyms for ‘female sweetheart or girlfriend’ listed above. The category on the spidergram is wider, ‘female partner’, hence including ‘wife’ and ‘long-term (female) partner’, and possibly the responses given, missus, our lass, other half, better half, ‘by her name’, all refer to longer-term relationships, although it is interesting to note the presence of lass here, and especially the modification with our. Data of this kind also give the potential to help answer some of the questions raised above about the connotations and contexts of use of particular terms: here we have evidence that these particular words have been given as elicited responses by a group of ex-servicemen now working as fire-fighters in a particular location in a particular time (although we do not know, at least from this summary data, precisely which relationships they would consider each word appropriate to denote, nor in which contexts they would feel it appropriate to use each word).

Elsewhere in the article, as already noted, language contact is reflected by the presence of Punjabi loanwords or code switches among responses. Additionally there are lexical items that appear to be unrecorded in any dictionaries to date, e.g. jelly-legged, sealed, stooled (all with the meaning ‘drunk’). Just as interesting are pieces of evidence that supplement the existing picture from lexicographical sources in significant ways: for instance, the fact that tipteers ‘mummers’ is found here in contemporary use is an important postdating to the record of late-nineteenth-century currency (again from Sussex) given in the English Dialect Dictionary (and could be used eventually to help support inclusion in other dictionaries such as the OED, if other such examples could be found).Footnote ¹¹ The complicated notes on non-linguistic historical and cultural factors that need to be attached to items such as the cricketing term bodyline (alluding ultimately to the controversy surrounding the Ashes tour in 1932–3) show the susceptibility of lexis to extra-linguistic factors.

The paper by Beal and Burbano-Elizondo introduces an important diachronic perspective, looking at aspects of lexical change and semantic change. It takes as its starting point work by Glauser (1985) that drew on data from the Survey of English Dialects that was collected between 1950 and 1961. This was a traditionally conceived dialect survey, using a lexical questionnaire that took an onomasiological approach, i.e. collecting the words that speakers used to realize a particular meaning. The meanings concerned here are ‘boy’, ‘girl’, ‘son’, and ‘daughter’. From the (synchronic) data of the SED materials (to summarize drastically), the characteristic northern terms lad and lass appeared to be receding, especially in the meanings ‘son’ and ‘daughter’, and lass seemed to have receded further than lad and be likely to fall out of general use sooner. Beal and Burbano-Elizondo revisit the question using sets of data collected in Newcastle in 1994 and Sunderland in 2005, which are capable of yielding an apparent-time perspective, since they reflect speakers of different ages (and genders). They also exploit the potential of the Sunderland data, collected by Burbano-Elizondo using a Sense Relation Networks approach (from which the spidergram used in the BBC Voices project was derived), to yield information about the composition of semantic fields, and about fairly fine-grained semantic variation in the use of particular words. Thus they are able to adopt a semasiological perspective as well as an onomasiological one, looking at how the meaning of lass appears to be shifting towards a new central meaning ‘female sexual partner’. This gives a new perspective on Glauser's findings: lass may be declining in frequency in the core meanings ‘girl’ and ‘daughter’, but, rather than disappearing from use entirely, it may be showing semantic change in progress, shifting to a new (somewhat more specialized) meaning.

The final contribution in this volume, from Justyna Robinson, also looks at semantic change, from a semasiological viewpoint, i.e. looking at change in the meanings that are realized by particular words. The data looked at here are taken from a larger research project investigating the extent to which the methods and approaches of core sociolinguistic variationist research can be applied to the study of semantic change (see Robinson, Reference Robinson2010b).

In the example looked at in this volume, data on the use of gay were obtained from a carefully constructed survey, eliciting information on each speaker's range of active and passive senses for the word (or, more precisely, the senses in which they report that they use the word, as well as any other senses that they report being aware of in the usage of others), from a large group of interview subjects reflecting different ages, genders, and social classes drawn from the Sheffield area. At least three, and perhaps four, different senses of gay are in play here: ‘happy’, ‘homosexual’ (and perhaps ‘unmanly, effeminate’ as a distinct sense), and (as defined by the OED) ‘Foolish, stupid, socially inappropriate or disapproved of; “lame”’. The apparent time hypothesis, which has long been a mainstay of sociolinguistic work in areas other than lexis, suggests that it may be possible to observe change in progress in different age groups. This research suggests strongly that this is possible for lexis as well: the findings of the survey, subjected to rigorous statistical testing, show a pattern of change in progress. The sense ‘happy’ is most in use in the over 60 age group (although they are also frequent users of the ‘homosexual’ sense), and the result is especially striking for those who say that the word is in their active vocabulary. For those in the two middle groups, 19–30 and 30–60, the meaning ‘homosexual’ is clearly far and away the predominant one; the meaning ‘happy’ is one that people in the 19–30 group in particular may be aware of but do not themselves use. The 19–30 group may also show a distinct meaning ‘unmanly’, clearly linked to societal assumptions about homosexuality, but used without strong implications about sexuality. This may or may not be a transitional step to the meaning ‘lame’, which some people in the age group 19–30 report that they are aware of, and which is extremely common for the youngest group, aged up to 18, and for this group constitutes a real rival to ‘homosexual’ as core meaning of the word.

The value of this sort of work for observing semantic change in progress, and for learning more about the mechanisms of semantic change, is enormous. An important support for the hypothesis that these patterns of usage in different age groups do indeed show a historical change in progress is that they correlate well with real-time data, in this case the evidence of the OED and of scholars who have studied the recent semantic history of this word. If this were not the case, other hypotheses might have to be entertained at least as seriously as this one: for instance, age-grading could be in operation, i.e. younger speakers could be favouring the meaning ‘lame’ but then ceasing to use it so often as they become older. See Robinson (Reference Robinson2010b) for evidence that wicked may show just such a pattern in the same survey data, being used disproportionately frequently by teenagers in the sense ‘excellent’, although corpus data from the British National Corpus suggest that its currency in this meaning was higher 15 years ago than today. Thus in this instance age-grading may apply, and people may be decreasing their use of wicked in the meaning ‘excellent’ as they grow older.

The importance of taking account of significant extra-linguistic factors is shown by the example of skinny ‘low calorie’ (especially of coffee), examined in detail in Robinson (Reference Robinson, Allan and Robinson2011), drawing on data ultimately from the same fieldwork in the Sheffield analysis. In this instance, the apparent time hypothesis appears to be contradicted, because the recently innovated sense is not shown most frequently by younger speakers. Closer inspection of the data and consideration of extra-linguistic factors help explain the apparent anomaly: several branches of Starbucks serving (among other items) products designated as skinny had opened in the past several years; the most frequent users of this sense in the data were typically females from higher social strata working at the University of Sheffield, in close proximity to Starbucks branches.

In sum, the papers in this special issue all reflect and analyse the complex and multi-faceted nature of lexical and semantic change. They examine many of the key factors determining change, and, in particular, the methodological challenges involved in researching and charting change in progress. As a collection they do more than this, pointing the way towards new approaches to studying changes in word use and word meaning within the tradition of variationist work in sociolinguistics. Perhaps, Cinderella can and will go to the ball after all.