Dance Compounds in OED3 in the Light of Diachronic Big Data Analysis

Abstract

This paper reports on a research project that seeks to explore, firstly, how dance compounds have been described in the Oxford English dictionary and, secondly, to what extent their treatment is likely to be enhanced with textual big data. It begins by presenting a historical sketch of the dictionary, now consisting partly of the second edition and partly of the revised third edition, followed by an outline of the Google Books archive. The research methodology is then explained and the results are discussed and illustrated with examples. Since frequency has become a key parameter in historical lexicography, the compounds with their frequencies are listed in an appendix with the proviso that the figures may be unreliable. Despite the problematic nature of the quantitative data, the qualitative findings indicate potential improvements not only for the entries awaiting revision, but also for those which have already been updated. Once revised, OED3 will no doubt surpass its predecessor in accuracy and richness of detail, but, in the meantime, it should be examined for consistency between traditional principles and modern working practices.

1. Introduction

This paper reports on lexicographical research with a twofold aim. Firstly, it sets out to examine the way dance compounds have been described in the third edition of the Oxford English Dictionary (henceforth, OED3), the largest historical dictionary of English. Secondly, it seeks to investigate to what extent the lexicographical treatment may be enhanced with big data, Google Books being the main diachronic resource applied for this purpose. The concept of dance was selected due to its universality: compounds recorded in the dictionary denote forms of dancing typical of the anglophone world, as well as those practiced in a variety of cultures, including Congo dance ‘any of various dances… influenced by Kongo dance traditions’, lion dance ‘a traditional Chinese dance in which the dancers are masked and costumed to resemble lions’, shawl dance ‘a dance originating in the East, in which a shawl or scarf is waved’,¹ and Spanish dance ‘the traditional dance form of Spain, of gypsy origin… frequently involving the use of castanets’ (cf. Podhajecka, 2010).

Compounds are combinations of two or more words, one of which functions as the head and the other(s) as the modifier. Straddling the borderline between syntax and morphology (Bauer, 1998, pp. 135−136), they fall into two major groups: endocentric compounds, whereby the head is inside the compound (e.g., country dance is a type of dance), and exocentric compounds, whereby the head is outside the compound (e.g., redhead ‘a red-haired person’).² This notwithstanding, classifying compounds consistently is fraught with difficulty (cf. Bauer, 1983; Fabb, 1998; Haspelmath, 2002; Lieber, 1992; Marchand, 1969; Spencer, 1991; Štekauer et al., 2012), which refers, for instance, to differentiating between a compound (e.g., a blackbird) and a phrase (e.g., a black bird). Lieber (2005, pp. 376−377) remarks that stress, spelling, lexicalised meaning, the unavailability of the first element to syntactic processes, and the inseparability of the compound are taken into account, but, as she admits, none has been considered fully satisfactory. Given such labile criteria, distinguishing compounds from phrases is often impossible (Bauer, 2019, p. 45).

Another problem concerns semantic transparency. Even though each compound is compositional in meaning, the relation between its constituents is not always overtly expressed (ten Hacken 2016, p. 1). The data assembled by Pham & Baayen (as cited in Schäfer, 2017, p. 149) demonstrate that while some semantic types are transparent (e.g., firebomb), others are partially opaque (e.g., ragtime), and still others are fully opaque (e.g., jackass). According to Stockwell & Minkova (2001, pp. 13−14), opaqueness stems from phonetic change on the one hand (e.g., Lord as a combination of loaf ‘bread’ and OE weard ‘guard’) and generalisations or extensions of meaning on the other (e.g., bull’s eye ‘the centre of a target’ originally denoted ‘the central protuberance formed in making a sheet of blown glass’). Every now and then, semantic-etymological conundrums come into play; in contrast to blueberry, strawberry is opaque, because its meaning is partially independent of the meaning of its elements (Davis-Secord, 2016, p. 74; cf. Jarema, 2006, pp. 48−49; Liberman, 2005, p. 81). Interpretation and predictability of meaning is a complex process, hence, its becoming subject to a broad range of studies (see, e.g., Körtvélyessy et al., 2022; Schäfer, 2017; Štekauer, 2005).

The extent to which languages use compounding differs, but Germanic languages have been prolific in the formation of compounds (Bybee, 2015, p. 189). They are found in Old English, some based on Latin (e.g., handbōc ‘handbook’) and others on Scandinavian models (e.g., bātsweġen ‘boatman’), and Middle English with the pervasive impact of words of Anglo-Norman and Latin origin (e.g., gentil man ‘gentleman’), as has been attested by Durkin (2014, pp. 101, 196, 288).³ Their productivity gained momentum from Early Modern English onwards and nominal compounds, in particular, became an extremely productive type throughout the entire Modern English period (Nevalainen, 1999, pp. 352−354). The significance of compounding as a source of new lexical resources is reflected by the facts and figures derived from OED3: in November 2023, the dictionary recorded 215,042 compounds, by far the most frequent word-formation type, followed by derivatives (155,679), borrowings (88,341), and words produced by conversion (14,292).⁴

2. Outline of the research project

2.1. OED3

The OED has a long history whose beginnings may be traced back to two papers read by Richard Chenevix Trench, the Dean of Westminster, before the Philological Society (Trench, 1857). They expressed the need for a fully exhaustive dictionary of English compiled on historical principles. From the end of the eighteenth century, as Considine (2016, p. 170) notes, scholars believed that ‘individual words could… be seen as having histories of development, from birth to death’, so what the dictionary was expected to establish was ‘the biographies of each word in the language’ (Simpson, 2014, p. 17). Herbert Coleridge was appointed the first editor and, after his death in 1861, Frederick Furnivall stepped in (see, e.g., Winchester, 2003). However, it was James Murray, a Scottish schoolmaster with his ‘patient and scholarly temperament’ (Landau, 2001, p. 80), who became the dominant figure behind the project.⁵ Appointed an editor in 1878 and a full-time editor in 1885, Murray collaborated with three subeditors: Henry Bradley, William Alexander Craigie, and Charles Talbut Onions, all of whom exerted an influence on the shape of the dictionary (see, e.g., Brewer, 2007; Gilliver, 2016; Mugglestone, 2005; Murray, 1995).

The contract signed between the Philological Society and Oxford University Press stipulated that it be a four-volume dictionary compiled over a period of ten years. The first edition titled А new English dictionary on historical principles (Murray et al., 1884−1928) was published in instalments. Not only was it much greater than had originally been envisaged, but it also took much longer to produce. It sought to show the development of English vocabulary from Anglo-Saxon to contemporary times for British English and other world Englishes. The dictionary included close to two million quotations documenting usage, many of which were sent to Oxford, in response to Murray’s appeals to the public, by thousands of voluntary readers around the (anglophone) world. In this way, it became one of the first crowdsourced lexicographical projects.⁶ The editors, however, had long been aware that the endeavour would be ‘ended, but not complete’ (Mugglestone 2005, p. 198). In 1933, it was republished as the Oxford English dictionary (henceforth, OED1) together with a one-volume Supplement. In 1957, Robert Burchfield was appointed editor responsible for compiling a second Supplement, which was issued in four volumes between 1972 and 1986 (see, e.g., Brewer, 2009a; Gilliver, 2016; Ogilvie, 2013).⁷

In 1984, a new project was founded seeking to bring the dictionary into the modern age. This was a pragmatic move; Marghanita Laski, a prolific contributor to Burchfield’s Supplement, warned as early as 1972 that the dictionary was in danger of becoming ‘a magnificent fossil’ (Schäfer, 2003, p. 253). The undertaking was supervised by Edmund Weiner and John Simpson who consolidated the first edition, both of the Supplements, around 5,000 entries from the New English Words Series (Weiner, 2009, p. 401), and three volumes of Oxford English dictionary additions series (Simpson, 1993; Proffitt & Simpson, 1997) into one dictionary. Published in 1989 (Simpson & Weiner, 1989) in twenty volumes, it came to be known as the second edition (henceforth, OED2). The long-awaited electronic version was launched in 1994.

Soon thereafter, plans for a new edition were mooted. Since 2000, OED2 has, therefore, been under a top-to-bottom revision aimed at ‘pulling the still largely Victorian and Edwardian second edition into the twenty-first century’ (Brewer, 2022, p. 379) and leading to the third edition (henceforth, OED3). The revision was initiated and implemented by John Simpson and, since 2013, continued by Michael Proffitt as editor-in-chief (Simpson & Proffitt, 2000−). The agenda is voluminous: ‘Definitions and semantic analyses of words are being re-written, entries re-dated, etymologies newly analysed in the light of the last hundred years and more of scholarship, and hundreds of thousands of quotations have been added, along with numerous other changes’ (cf. Weiner, 2009, p. 400).⁸ ‘Lexicographical thrills and spills’ experienced in wrestling ‘the venerable OED out of its Victorian straitjacket and into the new, elegant garb it displays today on the Internet’ have been engagingly recounted by Simpson (2016, p. 343).

Given the technological development over the last few decades, OED3 has benefitted both from the power of the computer and ‘the almost limitless supply of evidence available’ (Weiner, 2015, p. 235). Apart from in-house reading programmes, the editors have at their disposal the Making of America collection of nineteenth-century material, Chadwyck-Healey’s Literature Online, and the academic journals archive JSTOR (Gilliver, 2016, p. 566). Other major databases include Early English Books Online and Eighteenth-Century Collections Online (Considine, 2016, p. 174), abbreviated to EEBO and ECCO respectively. The World Wide Web and communication channels, such as Twitter (Gilliver, 2016, p. 573); the Sketch Engine, a corpus query system with text analysis tools (Connon Martin, 2019); as well as online corpora, such as the two-billion-word Global web-based English corpus (Salazar, 2023, p. 29), are also handy sources of data. Advanced computational initiatives, moreover, have been pursued to facilitate the lexicographers’ work.⁹

The availability of fully-searchable databases and corpora is all the more significant since frequency has become a key parameter in historical lexicography. The evidence of word usage, as Simpson (2008, pp. 120−121) indicates, ‘allows the editors to select words and senses of words for inclusion according to their frequency of usage’, which means that well documented words are admitted and those for which the evidence is scant are omitted. Allan (2010, p. 170), too, affirms that ‘The OED’s policy on the inclusion of lexical items and senses of lexical items is based on consistent criteria related to frequency’. Each headword in OED3 is now provided with two kinds of frequency, historical frequency and modern frequency,¹⁰ albeit the latter would better suit a learner’s rather than a historical dictionary (cf. Jackson, 2002, p. 140).

Thanks to the wealth of lexical resources, the expertise of OED3 staff, and the help of a multitude of ‘collaborative contributors’ (Simpson, 2004, p. 194), the dictionary has grown considerably. OED1 recorded 252,200 entries, 414,800 defined word-forms, and 1,861,200 citations. Nearly a hundred years later, OED3 has doubled in size: in November 2023, it included 518,827 entries, 878,485 meanings, and 3,840,299 quotations.¹¹ According to Williams’ estimates, the revision of OED2 is half-way through, with 50% of the entries having either been revised (43%) or newly added (7%),¹² although the ‘patchwork of entries from different periods and by different editorial staff’ (Allan, 2012, p. 20) has resulted in synchronisation problems.

2.2. Google Books

Google Books is a diachronic archive encompassing a multilingual collection of texts from forty university libraries around the world (Michel et al., 2011, p. 176). Having been scanned and processed with optical character recognition (OCR), they were made available in a searchable format in 2004. A large proportion of the texts are now in the public domain, but, by dint of the settlement agreement with authors and publishers reached in 2008, Google Books also includes in-copyright books. Consequently, it offers access to materials that no single institution could physically possess, realising the promise of the modern ‘Library of Alexandria’ (Roberts, 2017). The archive is otherwise called a database (Miller, 2008, pp. 487−488), a library (Pechenick et al., 2015), or a repository (Thelwall, 2022, p. 45); the terminology is largely a matter of convention.

In order to avoid confusion, it needs to be highlighted that Google Books is used here with reference to the overall digitisation project rather than the Google Books corpora compiled by Mark Davies, totalling thirty-four million words for British English and 155 million words for American English,¹³ or the 2019 database of Google Books Ngram Viewer calculated at over a trillion words for English alone (Tyrkkö & Mäkinen, 2022, p. 16).

With forty million scanned books, comprising a number of words which would be difficult to approximate even roughly, Google Books, a type of big data (Aiden & Michel, 2013, p. 18), dwarfs any other resource available for historical scrutiny. In fact, most existing historical corpora are too small for large-scale investigation, so ‘they can only scratch the surface’ (Svensén, 2009, as cited in Brinton, 2016, p. 206; cf. Nevalainen, 2013, p. 38).¹⁴ Despite the gargantuan size, the composition of Google Books is messy (cf. Davies & Chapman, 2016, p. 132). Among other things, the quality of OCR is often poor; the metadata are notoriously unreliable (e.g., Ryan & Weiss, 2012); many search results are displayed in the form of text snippets; and hands or fingers occasionally mar the content of the page (Brown, 2017, p. 67). In view of these weaknesses, Google Books has not been recognised as ‘the greatest humanistic project of our time’ (Somer, as cited in Howard, 2017), but it remains ‘a valuable heuristic tool’ (Steggle, 2015, pp. 24−25).

Apart from Google Books, I used HathiTrust and Internet Archive, two digital libraries (see, e.g., Fishman, 2023, pp. 75−76), mainly to authenticate the sources. Both provided me with a number of full-text versions that exhibited the compounds in a wider context and, equally importantly, helped to correct the metadata. In complicated cases, I also availed myself of WorldCat, the world’s largest library catalogue.

2.3. Research methodology

The study included two basic steps. Firstly, I identified the compounds with the dictionary’s advanced search option, dance having been selected as ‘headword’ and ‘compound’ as the preferred word-formation type. The procedure returned 176 results, of which forty (e.g. dance hall) had to be ignored and 136 turned out to be valid noun-noun (N+N) or adjective-noun (A+N) compounds. Each entry was then examined in terms of its internal features.

Secondly, I conducted extensive searches for the compounds in Google Books. To this aim, I used the Google search engine with its spectrum of search facilities, of which selecting specific time spans (e.g. 1700−1848 for plantation dance first recorded in 1848) was of primary importance. Both the singular and plural forms, and, where applicable, different spelling variants of the modifiers (e.g. Congo or Kongo for Congo dance) and the heads (dance or danse for fairy dance) were considered. A range of Boolean expressions (e.g. “plantation dance” AND “slaves”) was employed in order to refine the searches. This step also included checking the frequency of each pattern generated by Google’s algorithms. A careful analysis of the corpus of citations collected for each compound ensued; it was time consuming inasmuch as each occurrence had to be scrutinised for the right form and meaning.¹⁵ The results were then compared with OED3 entries to ascertain any discrepancies between the two.

In what follows, the results of my research are discussed and illustrated with examples. The revised compounds are marked with the letter R in superscript, whilst the unrevised ones are marked with the letter U. The right-pointing arrow ► indicates recommended changes.

3. The discussion

OED1 treatment of word-formation patterns, including compounds, has been diligently studied by Kastovsky (2000). This paper omits his discussion in order to focus on OED3, which is not only an online version, but has recently migrated to a new platform.¹⁶ The slightly abridged and reshaped information from the OED website pertains to the display of compounds and other derived words:¹⁷

On the former OED website, compounds were sometimes treated as main entries and sometimes as subentries within the entry for one of the parent words. For example, kitchen garden n. was treated as a main entry, kitchen towel n. as a subentry within kitchen n.1, and trench kitchen n. as a subentry within trench n. … This created an inconsistent experience for users and limited the ways that compounds and other derived words could be viewed, searched, and grouped. To overcome this, we’ve made some changes to our new website: Compounds and derivatives that were formerly subentries are all treated as main entries. For example, kitchen towel n. is now a main entry with a meaning… and is therefore comparable in display and functionality to kitchen garden n. All the compounds and other words derived from the entry’s headword are listed in the compounds and derived words section… For each compound or derivative that has been upgraded to a main entry, the publication history is explained in the publication details, e.g., at kitchen towel n.: ‘Originally published as part of the entry for kitchen n.1.

As may be inferred from the above, compounds are described in two major ways: as main headwords and elements of the parent headword’s compounds section; one may puzzle over the fact that main headwords are also displayed in compounds and derived words sections. In my research sample, main entry status has been allocated to 108 dance compounds (e.g., Ghost Dance), whereas twenty-eight are listed in the compounds sections (e.g., ballet dance in the entry for ballet). With just a couple of exceptions (dew dance and torch dance), they are followed by dated citations. Taking into account the dictionary’s publication history, the compounds include three categories: fully updated entries (e.g., belly dance), newly published entries (e.g., lap dance), and entries awaiting a fully-fledged revision (e.g., apache dance); the status of the first two is comparable. In November 2023, as many as eighty-four compounds in my sample were either fully revised or newly added, and fifty-two remained to be updated.

Contrary to what is published on the website, thirty-one dance compounds (e.g. cakewalk dance, community dance, homecoming dance, matachin dance, subscription dance, and wonder dance) are treated neither as main headwords nor elements of the compounds sections. Identified with the ‘search in definition’ option, such latent compounds are not hyperlinked, which limits their usability,¹⁸ but are also supported with citations. It is unclear whether the first signifies the earliest recorded occurrence, a principle of utmost importance in historical lexicography.¹⁹

Before moving on to further analysis, an important disclaimer needs to be made: frequency should be approached with caution, because Google Books is not a corpus and this has implications for the quantitative data. First and foremost, the search results are verifiable manually only for a proportion of the queries (a maximum of 300 sources), which means that the frequency cannot be corroborated in any way. By the same token, the syntactic and semantic complexities are impossible to be sifted through, so the number of relevant hits remains unknown. Since searches for the plural forms (e.g., snake dances) often return inaccurate results by providing patterns consisting of a noun and a verb (N+V), frequency has been limited here to the number of hits for the dictionary’s canonical form (i.e., the singular form), which stands in contrast to the method of calculating frequency in OED3.²⁰

The distribution of the compounds across Google Books is shown in Appendix  1. The first sixteen rank positions are taken by compounds whose frequency exceeds 100,000; for eight, it is even counted in hundreds of thousands. They cover two types: those whose heads are modified by broad qualitative adjectives (e.g., folk dance, modern dance, and round dance) and those suggesting a specific use, feature, origin, or location (e.g., dinner dance, tap dance, and school dance). Apart from school dance, listed in the compounds section of school, all the high-frequency items function as main headwords. Two of them, folk dance, illustrated with barely two quotations, and tap dance, covering noun and verb forms in a single entry, deserve special attention during the revision process.

Compounds of the first type constitute a challenge due to their widespread use. Evidently, not every occurrence will reflect contemporary dance^R ‘a style of expressive dance which developed in the mid 20th cent., originally combining elements of both classical ballet and modern dance, now encompassing a wide variety of dance styles’ or modern dance^R ‘a free expressive style of dance developing in the early 20th cent. as a counter to classical ballet’, if only because the present soon becomes the past. A history of dancing from the earliest ages to our own times (Vuillier, 1898, p. 289), for instance, treated the waltz, the galop, and the polka as the then contemporary dances. No wonder the first attestation for modern dance is dated to as early as the eighteenth century (Anonymous, 1763, p. 7) and that for contemporary dance to the end of the nineteenth century (Grove, 1895, p. 21), pertaining to dance forms unrelated to the compounds. It is worthy of attention that the two terms are considered synonymous (e.g., Adair & Burt, 2017, p. 179; Craine & Mackrell, 2010, p. 307), so they should be cross-referenced to each other in OED3.

As for the second type, Spanish dance^U is perhaps the most difficult compound to deal with. OED3 defines it as ‘the traditional dance form of Spain, of gypsy origin and characterised by elaborate heel-work and frequently involving the use of castanets’, but the definition is not entirely satisfactory. Dancing styles in Spain, a country inhabited by a variety of peoples throughout its history, were influenced mainly by the Byzantine, Moorish, Gypsy, and Jewish cultures (Llano Kuehl-White, 2012, p. 3). Four genres of dances have been distinguished: folk dances performed in colourful regional costumes; eighteenth-century classical dances characterised by elegance and balletic form (e.g., bolero); Andalusian dances performed to the sound of castanets (e.g., malagueña, seguidilla, fandango, and farruca); and flamenco, a dynamic solo dance (Vittucci & Goya, 2003, pp. xv−xvi). Rhythmic footwork with the heels, called zapateado in Spanish (cf. zapateado in OED3), is peculiar to Andalusian and flamenco dances, so only these genres are legitimate.²¹

Apart from the most frequent compounds, ninety-eight fall into the frequency range 100,000−1,001, accounting for 72% of the sample. I would tentatively suggest that jazz dance, victory dance, scalp-dance, and buffalo dance, among others, be upgraded to main headword status; torch dance illustrated with citations; clog dance followed by more than one citation; and slam dance separated into a nominal and verbal entry respectively.²² On the other end of the frequency spectrum (200−0), we find compounds whose usage has been significantly restricted; the frequency of rapier sword dance and cottager’s dance falls below 10.²³

As regards the revision process, Williams (2021) tells us that

entries are prioritized for revision or addition based on a number of factors, including evidence of significant semantic development, token frequency in linguistic corpora, and frequency of online lookups, and supplemented by ad hoc prioritization, topical coverage, and special initiatives, including those aiming to expand the coverage of particular varieties of World English. (p. 61)

One might wonder whether my findings are indicative of these priorities.

Let us first consider potential obstacles, including the use of punctuation marks, homonymic forms, and syntactic patterns resulting in semantic ambiguity.

Before entering the house, Adonio stopped at some neighbouring cottages, where playing on a pipe, he began to make his dog dance. (Anonymous, 1770, p. 8)

I have been sorry to observe some of your amusements, such as your bellie dance. (Stuart, 1839, p. 1)

Sir! Do you presume to insult me by saying that I didn’t see the table dance, and by doubting my word when I tell you nobody moved it or could move it? (Anonymous 1860, p. 30)

Old Dobson is a slow coach; and as to the fellows, they are a bad set: that beast Dance leads them all. How do you like Dance? (Monroe, 1854, p. 18)

They were as stupid and indifferent to his warnings and oncoming judgments and imminent peril, as drunken men are stupid, or crazed by wine, dance, and laugh before the horror-stricken beholders, who see the lion about to leap upon them… they stand fascinated upon its track. (Anonymous, 1898, p. 100)

Each of the above passages includes the correct form, but not the correct meaning. More exactly, in two cases, dance turns out to be a verb (cf. to see sb do sth and to make sb do sth); in two others, we are dealing with an alleged spelling variant (bellie)²⁴ and a homonym (Dance) respectively; and, in one, a comma between wine and dance means that there is no compound here.

In the sections below, I focus on the treatment of the dance compounds in OED3. Attention has been paid mainly to those features of the entry that are likely to be improved.

3.1. Etymology

The etymological references in OED3 include condensed information on the compounding process, which, for round dance, goes as follows:

round dance^Rn.

Formed within English, by compounding.

< roundadj. + dancen.

A few headwords exhibit a little more idiosyncratic treatment. Slam dance is labelled as a noun and a verb, but the etymology describes only the nominal compound (cf. skirt dance), whilst touch in touch dance is either a noun or a verb. Only the first constituent had been given in gumboot dance, Irish dance, rapier sword dance, Scottish dance, and wag-tail dance; most of these etymologies were first corrected and then unexpectedly deleted, perhaps for the purposes of the revision. The reason the entry for Scottish dance disappeared from OED3 is unclear.

In the case of twenty-two headwords, the etymology includes both the standard information and a comment or a note. Their quality varies from a comparison with a single corresponding form (e.g., country dance) to a handful of cognates in related languages (e.g., Spanish dance), to high-quality informative notes (e.g., bear dance). For as many as thirty compounds, however, no etymology has been attached, of which nineteen represent headwords awaiting revision (e.g., buffalo dance, cushion dance,²⁵ and tap-dance) and eleven are revised, but undefined (e.g., ballet dance, Maypole dance, and touchdown dance).²⁶ This is doubly disappointing: facing undefined headwords with no etymology, users are left to their own devices. Since a neat and tidy treatment of headwords has been one of the objectives of the revision, it is to be hoped that all compounds in OED3 ultimately become main (i.e., defined and etymologised) headwords.

Occasionally, etymologies of unrevised headwords will have to be corrected. In OED3, shake dance (1968), for instance, is regarded as a back formation from shake-dancer (1956). My research shows that the earliest attestation of shake dance is dated to 1908 (Reagan, 1908, p. 71), whilst shake dancer has been traced to 1936 (Honigberg, 1936, p. 38).

3.2. Lexical variants and the cross-reference structure

OED3 covers a range of variant spellings attested throughout the history of English. This does not, as a rule, apply to compounds, the forms of whose modifiers are given under the parent entry. In the case of Gypsy dance, it would be expedient to check the spellings in the entry for Gypsy. Google Books provides evidence for five patterns: Gypsy dance, Gypsey dance, Gipsy dance, Gipsey dance, and Jipsy dance, of which the first is the most widespread (15,400) and the last the least frequent (2).

Lexical variation, viewed as an umbrella term including synonymy and near-synonymy, is represented abundantly in OED3, or so it seems at first sight. Lexical variants include forms used in different varieties of English, in different registers, and in different time periods, many of which ‘differ in their denotations, connotations, or both’ (Nevalainen & Tieken-Boon van Ostade, 2006, p. 300). The identification of variant forms of dance might look like a task for dance specialists, but Google Books can be quite helpful. If we search for the Boolean expression “[dance compound]” AND “also called”, for instance, we will come across various suggestions that may be subject to further analysis:

The Sun Dance was also called the Medicine Lodge in English and sometimes the Willow Dance because of the use of willow wreaths in the ceremony (as well as to confuse whites who were anxious to suppress it). (Liberty, 2007, p. 76)

Morris dance, also called the moresque because the blackened faces of the dancers resembled the Moors, is a survival of early weapon dances, which were not war dances but an ancient form of religious worship. (Hamilton, 2015, p. 149)

Bear Dance… which was also called the ‘Woman Step Dance,’ was an annual courtship-type ceremony in which unmarried women were allowed to take the initiative in selecting unmarried men as dance partners. (Hittman, 2013, p. 69)

Women’s fancy dance is also called the shawl dance or butterfly dance. (Prichard, 2023, p. 10)

Another major opposition that helped define Laban’s choreography was that between Ballet and Modern Dance, or what he also called Free Dance [Der Freie Tanz]. (McCaw & Lehner, 2023, p. 43)

The main dilemma is specifying the area of usage of the terms or, in the case of cross-cultural contexts, overlapping areas of usage. Another is the expertise of the authors, which may differ markedly. Nonetheless, even this minuscule sample appears to be a useful source of information: morris dance is not cross-referenced to moresque in OED3; the definition for the second sense of bear dance fails to make it clear that the dance was ‘ladies’ choice’ (Romero & Gray, 2008, p. 330); and modern dance was also termed free dance, especially in Central Europe (cf. Lehikoinen, 2006, p. 31; Love, 1997, p. 37).

OED3 records various synonyms of the dance compounds or their senses, but only for main headwords. They include morphologically related items (e.g., ice dance = ice dancing); full forms (e.g., sun dance = sun dance ceremony) and abridged forms (e.g., rag dance = rag); national variants (e.g., kissing dance = cushion dance) and regional variants (e.g., tea dance = thé dansant); synchronic forms (e.g., wag-tail dance = waggle dance²⁷) and diachronic forms (e.g., frog dance = † frogs’ hornpipe). Some compounds are also cross-referenced to analogues, that is, lexical items preceded by ‘cf.’ (e.g., table dance) and ‘see’ (e.g., chicken dance).

The definitions occasionally introduce hyperonyms of the dance terms (e.g., Grandpapa dance^R ‘a kind of country dance’), hyponyms (e.g., set dance^U ‘a quadrille, country-dance, or the like’), or both (e.g., Scottish country dance^R ‘any of various traditional Scottish dances, as a jig, reel, or strathspey… this type of dancing as a genre’).²⁸ The dictionary’s advanced search facility allows for a systematic grouping of compounds under a selected hyperonym, although the results tend to be somewhat inconsistent. For example, national dance is used solely in the definition of csardas; street dance appears twice (b-boying and body-popping); and erotic dance returns nine hits (e.g. hootchy-kootchy and ballum rancum). Intriguingly, lambada is defined as erotic dance, but the second sense of freak dance (‘A type of dance in which partners are in close proximity and move in a sexually provocative way, typically with a female dancer gyrating her buttocks against the crotch of a male dancer’) is not. Erotic dance, despite its relatively high frequency in Google Books (9,810), has not been accorded headword status in OED3, just like national dance and street dance.

An advanced search for dance as a verb in the category ‘headword’ displays twenty verbs converted from dance compounds, of which thirteen are revised and seven are awaiting revision.²⁹ It may be instructive to trace the way these forms have been linked.³⁰ Firstly, there are no cross-references between unrevised nouns and verbs, of which the latter are undefined (e.g., folk dance v.). Secondly, no link has likewise been offered in the case of some updated headwords (e.g., belly dance, disco dance, and swing dance). Thirdly, six entries demonstrate a different treatment: freak dance n. 2 refers the user to freak dancing n.; freak dance v. to freak dance n. 2; morris dance v. to morris v. 1; slow dance n. to slowie n. 2 and vice versa; and swing dance n. to swing n. 2. Taxi dance n. and taxi dance v. are both cross-referenced to taxi dancer n., whilst touch down n. is the only compound referring to the corresponding verb and vice versa. Of the three types of treatment, the first and second are unhelpful, as users looking up the compounds may be unaware of the existence of homographic forms. The third is an attempt to cover a wider semantic network (e.g., freak dancing, slowie, taxi dancer), but it also lacks regularity.

Upon careful analysis, the area of incongruity turns out to be even broader. Firstly, in a proportion of cases, the links are unidirectional rather than bidirectional. The first sense of fairy dance^R, for example, is cross-referenced to fairy ring^R,³¹ but not the other way round, which is also exemplified by the pairs given below:

Irish dance^Rn. = Irish step dance^Rn.

kissing dance^Un. = cushion-dance^Un.

pillow dance^Rn. =cushion-dance^Un.

Romaic dance^Rn. = Romaika^R n. 2

Secondly, there are a handful of compounds whose modifier developed several senses, one of which refers, explicitly or implicitly, to a dance. OED3 uses two strategies in this respect, depending on the status of the compound in the dictionary macrostructure. Main headwords are treated as synonyms of the parent headwords or their senses:

Congo dance^Rn. =Congoadj. 1. Any of various dances… influenced by Kongo dance traditions.

a. death dancen. = Dance of Death at dancen.6.c. An allegorical representation of Death leading men of all ranks and conditions in the dance to the grave… Also called Dance of Macabre, French danse macabre: see Littré.

jazz dance^Rn. =jazzn. 3.c. Any of various styles of dance performed to or associated with jazz music.

rag dance^Rn. =rag,n. 2.b. A dance performed to ragtime music; a dancing party at which ragtime music is played. Now chiefly historical.

Other compounds are cross-referenced to the compounds and derived words sections instead. This treatment is problematic, as the definition of the parent headword would shed more light on the meaning of an undefined compound than other compounds. For shimmy dance, for instance, these would be shimmy shiver, shimmy shake, shimmy fox, and shimmy damper, three of which are undefined and, hence, entirely uninformative.

minuet dance^Rn. See minuetn. Compounds. (‘a stately dance for two in triple time’)

rumba dance^Rn. See rumban. Compounds. (‘an Afro-Cuban dance; a ballroom dance imitative of this, danced on the spot with a pronounced movement of the hips. Also: dancing in this style’)

rock’n’roll dance^Rn. See rock’n’rolln. & adj. Compounds. (‘a style of dance performed to this music’)

shimmy dance^Un. See shimmyn. 2 Compounds. (‘a lively modern dance resembling a foxtrot accompanied by simulated quivering or shaking of the body… Originally U.S.’)

Three entries conform to neither of the above patterns: morris dance^R has no references to morris^R, even though the latter’s first sense is synonymous with the former; reel^R is actually part of the definition of reel dance^R (‘a lively dance, a reel; (also) a piece of music to which such a dance may be performed; also figurative’), and swing dance^R is cross-referenced to the entry for swing^U, but there is no dance sense in it. The last is clearly an omission to be corrected in the course of the revision; as has been argued by Pittman et al. (2015, p. 421), ‘Swing is an umbrella term for a variety of dances, such as West Coast Swing, East Coast Swing, Jive, Jitterbug, Shag, and Lindy Hop’.

Thirdly, some potentially corresponding forms are not in OED3 at all. I identified four distinct cases: ball dance (1706) appeared 150 years before ballroom dance^R (1850) came into vogue; calumet dance (1717) was an earlier lexical variant of the first sense of pipe dance^R (1778); demon dance (1818) could be considered a variant of devil dance^R (1849); and discotheque dance (1964) was attested simultaneously with disco dance^R (1964). The major difference concerns the frequencies of the unrecorded compounds, that is, ball dance has 12,000 hits in Google Books, calumet dance comes second with 6,500 hits, demon dance is third with 4,540 hits; and discotheque dance is the lowest in the ranking with 1,300 hits.

Cross-reference structure is a powerful device ensuring immediate access to different components in the microstructure, macrostructure, and megastructure of a dictionary, including semantically related lexical items (cf. Hartmann & James, 2001, p. 32). This study suggests that the editors should make the effort to guarantee that dictionary users be satisfied with the results of information retrieval.

3.3. Definitions and sense distinctions

Despite definitions being a core element of a reference work, relatively little is known about the principles of defining in historical lexicography. Defining is challenging, since it is ‘a multi-dimensional discipline, and no defining manual is able to codify all possible features without losing the spontaneity of lexicographical creation’ (Silva, 2000, p. 83). OED1 compilers succeeded in working out some patterns that ‘established a rhythm (by repeated formulas and a consistent, predictable definition shape for each word-category)’ (Silva, 2000, p. 83), many of which have been employed to date. Simpson (2014, p. 34) confirms that, whilst the shape of the dictionary has been modified, ‘the editorial philosophy remains constant. The structure to which new information is added today is in essence the structure designed by James Murray and his colleagues over one hundred years ago’.

It is worthy of mention that semantic development from the perspective of a diachronic resource is far from a clear-cut concept. As Durkin (2016a) remarks,

most statements about the development of words over time are abstractions from a complex set of data, open to challenge on numerous points of detail when investigated closely; the historical lexicographer has constantly to chart a difficult course between cataloguing a confusing welter of fine details and multiple possibilities in which even the most dedicated reader may easily become lost. (p. 242)

The welter of details and multiple possibilities is indeed a serious quandary which calls for cautious and systematic evaluation of findings, to the best of one’s ability, in the course of research. Simpson (2008, p. 121) reveals that the editors ‘are instructed to select from the wealth of illustrative documentation’, focusing on the variety of genres in which a term has been used, the major spelling variants, the geographical and chronological spread, and citations providing information not included in the definition.

Let us examine the defining practices in OED3 (and sometimes the lack thereof) through the prism of my sample. It includes thirty-four undefined compounds, of which approximately 30% come from updated entries, that is, ballet dance, love dance, Maypole dance, medicine dance, minuet dance, nautch dance, nosegay dance, rock’n’roll dance, rumba dance, and touchdown dance. The lack of a definition is a serious drawback in a historical dictionary, forcing the users to establish the meaning on their own. This is not always an easy task. As Fabb (1998, p. 66) points out, ‘popcorn is a kind of corn which pops; once you know the meaning, it is possible to see how the parts contribute to the whole – but if you do not know the meaning of the whole, you are not certain to guess it by looking at the meaning of the parts’. If we hypothesise that only a proportion of the compounds are self-explanatory, which is clearly the case, leaving opaque headwords undefined must be seen as a user-unfriendly technique.

Which of the compounds, then, are semantically opaque? We could probably eliminate those whose modifier developed a dance sense, that is, ballet dance, minuet dance, nautch dance, and rumba dance. Most of the remaining items seem to be partially transparent; one would deduce that love dance expresses affection for the partner; Maypole dance is performed around the Maypole; rock’n’roll dance is danced to rock’n’roll music; nosegay dance is performed by girls with bunches of flowers; and touchdown dance denotes a dance by a player who scored in a game of rugby or football. Google Books evidence shows that these interpretations are incorrect in two cases: love dance imitates a form of courtship, also in the world of animals, but there is considerable cultural variation in the form of the dance; and the now obsolete nosegay dance, a translation from French, was defined as ‘a kissing dance’ (Cotgrave, 1611).³² Apart from these two compounds, one headword that would definitely require a definition is medicine dance. In OED3, it is cross-referenced to the entry for medicine, which comprises seven senses and a number of subsenses, but the users are ill-equipped with knowledge of what they are looking for. Closer analysis brings us to sense 5.a (‘among some Indigenous peoples of North America: magical power, esp. for healing or protection… something similar among other peoples’), which appears the most relevant, so the compound may be defined as follows:

medicine dance, n. In some Indigenous cultures, particularly among Native American peoples, a ceremonial or spiritual ritual believed to have healing or protecting powers, led by a medicine man or woman who often goes into trance to facilitate physical or spiritual healing.

Of 102 defined compounds, sixty-nine explain meaning by means of a defining formula (e.g., taxi dance); eighteen include at least two different senses or subsenses (e.g., partridge dance); nine are defined by synonyms (e.g., pillow dance); five by a definition and a synonym (e.g., Irish dance); and one is cross-referenced to the quotations section (bull dance), an outdated method.

The entry for trance dance has two senses, of which the first subsense is given below:

trance dance^Rn. 1.a. Originally: any of various ritual dances in which participants enter a state of trance; esp. one of the Balinese sang hyang religious dances, during which gods are thought to occupy the bodies of the dancers (usually young girls). Also: a secular or theatrical performance of such a ritual dance.

As might be expected, Google Books sources include details absent from the definition. To provide a specific example, in some cultures trance dance is believed to have healing powers (e.g., by removing an evil spirit from a sick person) and it often takes the form of a group dance. Trance rituals performed in Bali, to which the above subsense refers, are well documented in English (e.g., Reilly, 1998, pp. 231−233), but they were preceded by those of dancing dervishes, as is shown by an antedating citation.³³ After all, the earliest occurrence of dancing dervish in OED3 is dated to 1687.

Nearly 70% of the compounds are defined concisely, following more or less the same pattern, but the informative value of the definitions varies. Let us compare examples of dances performed by native Americans or, in line with OED3 modern terminological policy, Indigenous peoples of North America. Two entries include a regional label (North American) which seems redundant in the context of the defining strategy employed.

bear dance^Rn. 2. Any of various ritual or ceremonial dances practised by Indigenous peoples in North America and Northern Asia. Also (more generally): the ceremonial event at which this dance is performed.

dog dance^Rn. A ceremonial dance performed by some Indigenous peoples of North America.

grass dance^Rn. A ceremonial dance of some Indigenous peoples of the Great Plains, typically performed in clothing featuring long, brightly coloured fringes that sway with the dancers’ movements.

partridge dance^Rn. b. North American a ceremonial dance inspired by the partridge (partridge n. I.1.b), performed by some Indigenous peoples of North America.

pipe dance^Rn. North American. Among some Indigenous peoples of North America: a ceremonial dance involving or relating to a peace pipe or calumet.

prairie chicken dance^Rn. A ceremonial dance of some Indigenous peoples of North America, inspired by the courtship displays of the prairie chicken.

Five definitions use some as a hedge, which might be responsible for a lack of precision. In his analysis, Kamiński (2019, p. 70) remarks, however, that hedges are ‘useful in large-scale dictionaries compiled on historical principles (such as the OED), as the full account of the history of word meanings would be impossible to achieve without them’. This is particularly pertinent in the description of dances performed by various tribes of Indigenous peoples, which is the reason my own definition of medicine dance also resorts to a hedge. None of the definitions alludes to the fact that, for tribal communities, ‘dance has been considered a transcending experience, a way of bonding with nature and gods’ (Mini, 2015, p. 62).

Of the six definitions given above, pipe dance, prairie chicken dance, and partridge dance are defined more or less accurately by referring to pipe, prairie chicken, and partridge, even though not every user will be able to conjure up the behaviour of partridges and prairie chickens. The second sense of bear dance is vague, but the vagueness is compensated for by the sociohistorical comment added to the entry; the definition of grass dance concentrates on the performers’ fringes but fails to explain what they stand for; and that for dog dance is the most ambiguous. This apparently stems from the fact that dog dance was performed in diverse ways: dancers made use of a dog skull, a dog liver suspended from a pole, a feast of dog flesh, and a dog dressed with ribbons hung by the neck to a gallows. It would be worth adding that, to the Hidatsa tribe, a North American people of the Plains, dog dance was a sacred dance performed to thank the spirits for the warriors’ strength (Stotter, 2009, p. 56; cf. Fuhrer, 2014, p. 15).

Speaking of sense distinctions, dictionary editors are well known for their inclination to be either lumpers or splitters, the former ‘packaging senses up together’ and the latter ‘unpicking them in fine – perhaps over-fine – detail’ (Mugglestone, 2011, pp. 44−45). The dance compounds offer limited material, but, every now and then, Google Books data seem sufficiently strong to suggest a semantic distinction. Let us consider the examples below:

Apache dance^Un. A violent dance for two people in which the partners are dressed as apaches; cf. Apache n. 2.

➣ Apache dance, n. 1. A ritual dance performed by the Apache tribe inhabiting New Mexico; cf. Apache n. 1; 2. (also with lower-case initials) A violent dance in which the male dancer kicks and tosses the female dancer; cf. Apache n. 2.

Note: Apache dance n. 2 resembled a pantomime, in which the apache man (pimp) simulated brutally dragging the female performer (prostitute) by the hair, kicking her, and throwing her across the dancefloor. Imported from Paris into American cabarets and dance halls in the early 1900s, it received its name from primitive connotations with male Native Americans’ ritualised sexual domination over submissive females.

fan dance^U, n. A solo dance in which the performer uses a fan or fans, esp. to conceal her nudity.

➣ fan dance, n. 1. A ceremonial dance in the Far East, esp. Korea, China, and Japan, in which women skillfully wave one or more fans; 2. A solo dance in which the performer uses a fan or fans, esp. to conceal her nudity.

gypsy dance^R, n. a. A style of traditional Spanish Romani music, typically played on the guitar and accompanied by singing, dancing, and often rhythmic backing, such as clapping, castanets, etc.; cf. flamenco n.; b. a style of Hungarian Romani folk music, played predominantly on string instruments, typically in a polyphonic style; (also) a piece of classical music based on or influenced by this style.

➣ gypsy dance,n. A lively dance performed by Gypsy women or in a Gypsy-like fashion, characterised by graceful movements of the arms and hands, and swishing flashy skirts, often with the use of tambourines; b. a style of traditional Spanish Romani music, typically played on the guitar and accompanied by singing, dancing, and often rhythmic backing, such as clapping, castanets, etc.; cf. flamenco n.; c. a style of Hungarian Romani folk music, played predominantly on string instruments, typically in a polyphonic style; d. a piece of classical music based on or influenced by this style.

slow dance^R, n. A dance in which a couple move in coordination with each other to music of a slow tempo, esp. while holding each other closely in a romantic or intimate manner. Also: a song or piece of music (esp. one regarded as romantic or sentimental), with a slow tempo, to which such a dance may be performed.

➣ slow dance, n. A dance in which a couple move in coordination with each other to music of a slow tempo, esp. while holding each other closely in a romantic or intimate manner; b. a song or piece of music (esp. one regarded as romantic or sentimental), with a slow tempo, to which such a dance may be performed.

In the first case, three changes are suggested: adding an earlier sense referring to the Apache tribe, correcting the existing definition, and providing a sociohistorical note elucidating the form and origin of the dance.³⁴ The line of argumentation has been borrowed from Shope (2016, p. 61), who quotes an apache dancer in New York: ‘Indians, and in fact, every savage tribe… dance wildly, madly, before they go to war’. Such stereotypical images of Native Americans are also found elsewhere, because ‘the evidence we have of them comes mostly from Euro-American sources with Western biases’ (Fuhrer, 2014, p. 13).³⁵ That the colonised were depicted almost exclusively by the colonisers poses a range of new obstacles for the revision process, particularly in the domain of world Englishes. Among other things, not only were the Indigenous peoples deemed inferior, but the indigenised varieties of English were approached with hostility and contempt ‘jarring to our twenty-first century sensibilities’ (Dolezal, 2006, p. 701).

In the second, there is every indication that fan dance appeared first, chronologically speaking, in the culture of the Far East and in a more decent form than its Western counterpart. As Campbell (1877, p. 560) observed, ‘The beauty hides her smiles behind her fan, and to see a troupe of Japanese girls dance the fan-dance is something worth remembering’.

In the third, introducing a subsense for the dancing performances seems a necessity. Wedeck (1973, p. 316) maintains that a Gypsy dance ‘means a dance as performed in a typically and distinctively Gypsy manner’ (cf. Charnon-Deutsch, 2004; Piotrowska, 2013). The only problem to be resolved was whether the dancing subsense should come first or second, but the antedatings helped me to establish the order of subsenses. For the time being, the earliest source referring to the dance is an eighteenth-century travel account (Townson, 1797, p. 367),³⁶ whilst that referring to music is a composition titled ‘Gypsey dance’ or, in another copy, ‘Gypsy dance’ (Anonymous, 1825, p. 74). Of course, both Roma dances and music had been described in English much earlier, though without using the compound (e.g., [Croxall], 1721, p. 5).³⁷

In the last, differentiating between the two subsenses of slow dance, for dance and music respectively, appears to be in line with OED3 editorial policy (see, e.g., supper dance, reel dance, trance dance; cf. chicken dance).

Whenever the defining formula permits no explanation of the headword’s meaning and the quotations provide little assistance, a sociohistorical note clarifying the semantic and historical nuances could be introduced. Let us look at the following entries:

Ghost Dance^Rn. 1. Among certain North American peoples in the late 19th cent.: a shuffling, circular ritual dance, often lasting several days, intended to bring about the restoration of traditional lands and ways of life, the expulsion of white settlers, and the return of the dead; 2. A religious movement among certain North American peoples in the late 19th cent., based on participation in the Ghost Dance (see sense 1) and intended to bring about the restoration of traditional lands and ways of life, the expulsion of white settlers, and the return of the dead.

Note: The first movement developed in 1869, when a Northern Paiute man, Wodziwob, had a vision in which he saw the removal of white settlers from the earth by means of a large earthquake. Usage in this sense was originally and most commonly with reference to the movement inspired by the visions of the Northern Paiute prophet, Wovoka, in the late 1880s and incorporated into ongoing resistance to the U.S. Federal government by peoples of the northern Great Plains, esp. the Lakota Sioux. Official concern at the spread of this movement is typically viewed as an important contributory factor in, e.g., the killing of Sitting Bull and the massacre at Wounded Knee (both in December, 1890).

➣ Addition to the Note:

    Native American spiritual movements, dating back to the 17th century, were a reaction to hardships caused by white people, including the loss of lands, the destruction of the ecosystem, and diseases devastating the population.

grass dance^Rn. A ceremonial dance of some Indigenous peoples of the Great Plains, typically performed in clothing featuring long, brightly coloured fringes that sway with the dancers’ movements.

➣ grass dancen. A ceremonial dance of some Indigenous peoples of the Great Plains, typically performed in clothing featuring long, brightly coloured fringes symbolising either enemy scalps or grass.

Note: The dance took its name from the custom of wearing braided grass to symbolise the scalps of enemies; originally only experienced warriors belonged to the Grass Dance Society. In other tribes (e.g., Sioux), each member had a bunch of grass attached to the waist as an emblem of abundance, since, on the Great Plains, the grass nourished both domestic and wild animals.

sun dance^Rn. In the cultures of certain Indigenous peoples of the Great Plains: an annual religious ceremony lasting several days at which individuals pray, fast, and offer personal sacrifices on behalf of their family, community, etc., and at which young men typically participate in self-mortifying dances which require them to endure pain and exhaustion; sometimes also more fully sun dance ceremony. Also spec.: the self-mortifying dance in which young men typically participate at this ceremony.

Note: The misconception among outsiders that the purpose of such dances is to honour and worship the sun is due to the misinterpretation of the Lakota name for the ceremony by the missionary M. H. Eastman; see note in etymology and quot. 1849. In some ceremonies the dancers dance around a pole to which each is attached by a strip of rawhide that is pierced through the skin of his chest or (less commonly) back; cf. quot. 1894.

➣ Addition to the Note:

    In Native American cultures, the sun dance served important religious and social functions. From 1881 onwards, misinterpreting it as a heathenish and seditious ritual, the federal government issued a series of edicts that outlawed it. The ban was lifted in 1934, but even then the government insisted that it be modified to suit white sensibilities.

In my view, the history of Ghost Dance needs to be foregrounded, explaining the reasons that contributed to Native American mass movements. The entries for grass dance and sun dance, by contrast, would benefit from details extracted from specialist sources available from Google Books. The former expedites the symbol of grass, whilst the latter elaborates on the plight of the dance in its sociohistorical setting. By providing such information, partly encyclopedic in form, OED3 may indeed be considered ‘more than a historical dictionary’.³⁸

3.4. Illustrative examples

As mentioned above, the definition is not the only source of information the dictionary users may consult; chronologically ordered citations are another. In theory, an analysis of the contextual uses should provide sufficient information on the meaning, but, at least in this case, they tend to be infrequent and unhelpful.³⁹ The undefined compound medicine dance^R, for example, is illustrated with three citations, none of which is particularly illuminating:

medicine dance^Rn.

1805 We sent a man to this Medisan Dance last night.

1855 Then they..Danced their medicine-dance around him.

1991 It [sc. a textbook] reduces our sun dance and medicine dance to a toilet-paper project.

Headwords from the unrevised portions of the dictionary material are documented even more scarcely. Clog-dance^U and shimmy dance^U are followed by a single citation each, neither of which points to any characteristic features, leaving the dictionary users in the dark: ‘A grand international clog-dance’ and ‘Shimmy dance is banned in greater New York’.

It is a simple truth in historical lexicography that citations illustrate headwords to different degrees. Many are too short accurately to explicate the meaning of an undefined compound. While space used to be a vital element in paper dictionaries,⁴⁰ it is no longer a real consideration in an online edition. It is, therefore, suggested that previously abridged illustrations be quoted in their entirety. This may be illustrated by the entry for nosegay dance, supported by quotations from two editions of a seventeenth-century French-English dictionary (Cotgrave, 1611, 1673):

nosegay dance^Rn.

1611 Bransle du bouquet, the nosegay daunce.

    ➣ 1611Bransle du bouquet, the nosegay daunce, or kissing daunce.

    ➣ 1673 The nosegay dance, or a marriage dance (for there is much kissing in it).

Part of my research consisted in verifying usage of the compounds and tracing their sense development. In doing so, I kept in mind that Simpson (2008, p. 126) called the search for the perfect selection of citations ‘an illusory quest’, suggesting that the users be given a digest of data in an easily interpretable way. For the examples below, I would propose adding illustrative examples which might better flesh out the nuances of the compound’s meaning. Three of them predate the first recorded occurrences either as square-bracketed citations or citations proper, whilst two come from sources published in the same year as those in OED3.

character dance^Rn.

1757 All these love-scenes, the girls execute in character-dances.

  ➣ [1712 Orchesography, or the Art of writing down Dances in Characters.] (Weaver, 1712, p. 171)

fan dance^Un.

1879 The most interesting dances were a Club Dance and a Fan Dance.

  ➣ 1872 There is a fan-dance, in which the omnipresent toy is put to more coquettish uses than ever a Rosina dreamed of. (House, 1872, p. 269)

freak dance^Rn. 1

1893 (April) Next Week—theFreak Dance.

1893 I have already copyrighted my dance. It is called the ‘Freak Dance’, and I have also patented the apparatus.

          ➣ 1893 I have already copyrighted my dance. It is called the ‘Freak Dance’, and I have also patented the apparatus. Miss Robson’s leg is fastened here,’ and Mr. Cline placed one hand on his abdomen, ‘but in my dance the two legs are placed over the hips. (Anonymous, 1893, p. 8)

pipe dance^Rn. 1

1778 The Indians have several kinds of dances..as the Pipe or Calumet Dance, the War Dance, the Marriage Dance.

          ➣ 1778 The Pipe Dance is the principal, and the most pleasing to a spectator of any of them, being the least frantic, and the movements of it the most graceful. (Carver, 1778, p. 268)

rabbit dance^Rn.

1892 The moose-dance, rabbit-dance, and duck-dance were kept up till the small hours.

   ➣ 1878 The dancers of the rabbit dance stood arranged in two lines. All sang together a French song, which they did not understand, that expressed the chase after rabbit. (Rey, 1878, p. 541)

   ➣ 1886 Now, yo’ all ready fo’ de rabbit dance,’ cried Uncle Mingo, and the dancers gathered in a circle, one in the center personating ‘Brer Rabbit,’ and with heads, hands, and feet, they kept time to the music. (Folsom, 1886, p. 486)

My hypothesis that space in an online dictionary is no serious limitation may be elaborated further: if the citations selected for the dictionary entry are enigmatic, a longer passage clarifying the details would be a welcome addition. A huge number of the illustrative examples incorporated in the dictionary, while retaining their authenticity, pose interpretative problems because they have been taken out of context. The first sense of freak dance ‘† A novelty dance, briefly popular in the late 19th cent., reportedly involving the use of one or more artificial limbs by the dancer’, for instance, has been supported with the 1893 issue of Music & Drama, but a fuller excerpt would better depict the peculiarities of the dance. In fact, some citations in the revised entries (e.g., Ghost Dance, line dance, and trance dance) are longer than one sentence.

Each instance needs to be evaluated carefully in terms of semantic adequacy. Wine dance^U ‘A dance performed in celebration of wine’, for example, could be both antedated and postdated, but the compound is displayed in different cultural contexts and, consequently, in different shades of meaning. Let us compare the only citation in OED3 with usages extracted from three sources:

wine dance^Un.

1920 They begin to sing, dancing meanwhile, in a free little ballet-manner, a wine-dance, dancing separate and then together.

  ➣ 1864 As in the most primitive ages also, the practice prevails of treading out the grapes with the human foot. This is the ‘wine dance,’ in which strong active young men are employed. (Milner, 1864, p. 298)

  ➣ 1901 There are also ballet dances, the Kermess, the Harlequin, the ancient dance of Brittany, the old Wine dance of pagan origin. (Staunton Moore, 1901, p. 188)

  ➣ 1980 This practice of carrying things on top of the head is also seen in Philippine dances like the Pandanggo sa Ilaw (fandango of light) where candles inside glasses are balanced atop the dancers’ heads and the Binasuan (wine dance) where a glass full of wine is balanced on the head. (Lopez, 1980, p. 427)

The 1864 citation refers to France and its former wine-producing method of treading grapes with bare feet, largely the same as in Spain and Italy. The second concentrates on the amusements of the Paris World Exhibition of 1900 or, more precisely, its palace of the dance. The last, on the contrary, describes the tradition of the Filipinos balancing a glass of wine on their heads; a similar dance has been performed in East Asia (e.g., Korea, China, and Mongolia). One might wonder whether these examples should be accepted and, if so, on what grounds. Frequency would normally be the main criterion, but in the case of what might be termed cultural dispersion, sometimes more fine-grained criteria would have to come into play (cf. Gypsy wine dance, cheese and wine dance). The origins of wine dance are reportedly traced back to pagan antiquity (Pearson, 1974, p. 260).

It is worth reiterating that, in the case of defined headwords, illustrative material plays a supplementary role, whereas it is the main or only source of semantic information for undefined headwords. The choice of citations in terms of their explanatory potential remains a crucial issue in historical lexicography.

3.3. Antedatings

As mentioned, OED3 editors have at their disposal massive sets of data to document headwords, so I took it for granted that antedatings for fully updated entries would be few and far between. Google Books will definitely help to predate headwords awaiting revision, especially if they go back to the eighteenth and nineteenth centuries, as sources dated to the sixteenth or seventeenth centuries are represented rather sparsely (cf. Nevalainen, 2013, p. 41). For fifty-two unrevised compounds in my sample, I found thirty-eight citations (73%) antedating the dance compounds or their senses by a couple of years to more than a century. Kissing dance is an exception, setting the record by being predated by 288 years. A few examples are given below.

buffalo dance^Un.

1807−8

  ➣ 1760 In the Western parts there is a different sort of dance which they call the Buffalo dance. (Jefferys, 1760, p. 75)

clog dance^Un.

1881

  ➣ 1823 A Clog Dance in Fetters. (Anonymous, 1823, p.128)

duty dance^Un.

1850

  ➣ 1820 In the first place, duty reading (like duty dances, when we are linked to the person we least like in the room, and that too by our own imperious sense of decorum), is, to say the least of it, very dismal. (Anonymous, 1820, p. 537)

kissing dance^Un.

1899

  ➣ 1611Bransle du bouquet, the nosegay daunce, or kissing daunce. (Cotgrave, 1611)

  ➣ 1739 I must confess I am afraid that my Correspondent had too much Reason to be a little out of Humour at the Treatment of his Daughter, but I conclude that he would have been much more so, had he seen one of the kissing Dances in which… they are obliged to dwell almost a Minute on the Fair One’s Lips, or they will be too quick for the Musick, and dance quite out of Time. (Anonymous, 1739, p. 263)

ladder dance^Un.

1801

  ➣ 1701 Let her be so by well conducted Scenes, Nor be secur’d, intrench’d behind Machines; Less by a Name from Italy or France, By Clinch of Barnet, or the Ladder-Dance. (Trotter, 1701)

scalp-dance^Un.

1791

  ➣  1755 Mohawks − his warriors dance the scalp dance. (Anonymous, 1755, p. 604)⁴¹

skirt dance^Un.

1894

  ➣ 1850 New Skirt Dance. (Martin, 1850, p. 6)

  ➣ 1892 She was the originator of what is called the ‘skirt dance’. (Anonymous, 1892, p. 297)

tea dance^Un.

1885

  ➣ 1842 But I knew all this before, and the ascendency of the merchants, and their coldness towards poetry, and philosophy, and the arts; and the want of elevation in the women, that always keep pace with the others, and on whose heads there are rarely faces such as one meets in the Wendelschentea dance by the dozen. ([Richter & Lee], 1842, pp. 280−281)

victory dance^Un.

1921

  ➣ 1857 Victory dance. (Schoolcraft, 1857, p. xxvii)

  ➣ 1858 On her return another noisy meeting was called, and they spent the night in one of their victory dances. (Stratton, 1858, p. 227)

The revised and newly introduced entries are of considerable interest. Contrary to my anticipations, I collected 39 antedatings (46%) in this category.⁴² This should not be surprising, as access to new data is likely to produce new findings. Google Books also provided first attestations for two undated compounds, dew dance and torch dance,⁴³ and a handful to be regarded as square-bracketed citations. Here is a selection of examples, sometimes with more than one antedating per compound:

ballet dance^Rn.

1760

  ➣ 1739 That the Mascherata dell’ Ingrate, an interlude or ballet-dance, at the marriage of the Prince of Mantua, was printed at Mantua in the year 1608, in 4to. ([Bayle et al.], 1739, p. 735)

beast dance^Rn.

1900

  ➣ [1866 Far away there we must now leave him a little longer, seeing fox-dances, buffalo-dances, and all sorts of wild-beast dances, war-parties, hunting-parties, gambling-parties, and all sorts of savage parties … retired in these wild solitudes, with their uncouth but not ungrateful children, sporting with them, teaching them, and protecting them, revered by them in his life, and seeing their dark weeping faces around him in his death.] (Foxton, 1866, p. 199)

  ➣ 1897 But I here confess that while beast-dances and wearing of skins of sacred beats are common, to prove these sacred beast to be totems is another matter. (Lang, 1897, p. 142)

  ➣ 1897 In the eleventh there is an important discussion of the Arcadian question, and the relation of Artemis to beast dances and similar observances. (Macalister, 1897, p. 121)

dance macabre^Rn.

1841

  ➣ 1817Douce, Mr. Francis … copy of the Dance Macabre. (Dibdin, 1817, p. 511)

  ➣ 1821 The Troyes edition of 1723… consists of 38 leaves, having wood-cut head-pieces to almost every page: each cut of the Danse Macabre contains four figures, viz. two of Death and two of the Personages he is addressing. (Davis, 1821, p. 22)

  ➣ 1836 This edition has sometimes the Dance Macabre annexed to it, which is attached the date of 1554. (Anonymous, 1836, p. 26)

devil dance^Rn.

1849

  ➣ 1585 Thus all thinges sette in order, then haue they their hobby horses, Drago[n]s and other Antiques, together with their baudie Pipers, and thundering Drommers, to strike vp the Deuilles Daunce withal, them marche these Deathen companie towards the Churche… to see these goodly pageauntes, solemnized in this sort. (Stubbs, 1585, p. 92−93)

  ➣ 1683 For Lust, and Pride is their Inheritance, So they run on within the Devil’s dance… In meat, and drinks, in cloaths, and dresses brave, which many damn, but never any save. ([Keach & Mason], 1683, p. 111)

  ➣ 1815 Most of the native Christians are Buddhites in their hearts, and frequently attend idol worship, and devil-dances; and the Roman Catholics are scarcely a remove in Christian knowledge or practice from them. ([Pratt], 1815, p. 131)

fairy dance^Rn.

1676

  ➣ 1584 Item, Bodin reporteth, that one confessed that he went out, or rather up into the aire, and was transported manie miles to the fairies danse, onelie bicause he would spie unto what place his wife went to hagging, and how she behaves hir selfe. (Scot, 1584, p. 25)

  ➣ 1640 And others said that it is generated in the cloud, and falleth downe the field, in a circle, on those round circles, which are seen in many fields, that ignorant people affirme to be the rings of the Fairies dances. (Fulke, 1640, p. 68)

  ➣ 1673 When my bless’d Spirit took me in my Trance, And led me a fantastick Fairy Dance: And through new paths, and through new ways me bore, Till she had set me down there where I was before. (Anonymous, 1673, p. 18)

line dance^Rn.

1961

  ➣ [1882 Turn! Turn back, thou perfect Jerusalemites! turn back, turn back… what shall your eyes, admiring, lay open? in respect to the Jerusalemites; [they respond, saying − in pursuance of the dual-line dance of young men and maidens.] (Garstang, 1882, p. 25)

  ➣ 1908 The line dance was used to express some degree of contest. (Crawford, 1908, p. ix)

  ➣ 1929−1930 A mixed dance, a line dance in which all the participants are differently costumed and masked, may be performed by any kiva during the winter or summer series. (Parson, 1929−1930, p. 1025)

minuet dance^Rn.

1765

  ➣ 1708 Shou’d Roger advance, to a Minuet Dance, and with an Air gracefully Bow, If you ask him e’er after, He’d answer with Laughter, To Handle the Harrow and Plow. (Anonymous, 1708)

  ➣ 1735 During this he hands or introduces the Lady into the Dance in the most agreeable manner he possibly can… concluding the first Division or Part of the Minuet Dance in the Hat’s being put on in a graceful Manner. (Tomlinson, 1735, pp. 125−126)

pipe dance^Rn. 1

1778

  ➣ 1698 They made a publick Feast where the Calumet or Pipe dance was danced, during four and twenty hours, with Songs fitted to that purpose. Which the Chief man amongst them tuned with all his strength. (Hennepin, 1698, p. 43)

school dance^Rn.

1823

  ➣ 1751 The gentleman-commoner took miss Betsy’s hand, and led her some steps to a minuet, then fell into a rigadoon, then into the louvre, and so ran through all the school-dances, without regularly beginning or ending any one of them… so that between singing, dancing, and laughing, they all grew extremely warm. ([Haywood],1751, p. 85)

  ➣ 1805 But there are some pointed traits in a just and elegant performance of the Minuet, as well as in all other school dances, which ought to be attended to precision… the good position, and proportionate distance of the Feet, in all their motions; a free, yet nervous, play of the Instep. (Peacock, 1805, p. 75)

sun dance^Rn.

1849

  ➣ 1806 4. Mode of making peace, smoking the great pipe, and dancing in different ways, such as the calumet-dance, the sun-dance, and the bull-dance. (Mitchill & Miller, 1806, p. 315)

➣ 1807 The savages have different sorts of dances… The principal ones are, the scalp dance, of which I have before spoken, the calumet-dance, the sun-dance, and the ox-dance. (Perrin du Lac, 1807, p. 75)

touchdown dance^Rn.

1971

  ➣ 1937 Coaches and Captains of Senior Rugby Teams to Be Guests of Pep Club at ‘Touchdown’ Dance. (Anonymous, 1937, p. 1)

➣ 1953 They charged to a quick touchdown which Gutowski scored on a short plunge, the ‘Dutch’ Clark broke away on a touchdown dance. (Roberts, 1953, p. 129)

➣ 1966 After a few regular dances announce a Touchdown Dance. At regular intervals throughout the dance someone will call out, ‘Touchdown!’ and the dancers all cheer and change partners with the couple nearest them. (Harris, 1966, p. 159)

witch dance^Rn.

1824

  ➣ 1753 Grim Pluto, you know, the black-peruked Monarch, must bellow in bass, and the attendant devils capers in flame-coloured stockings… after Hercules has resigned his club, to celebrate her triumph I shall introduce a grand dance of distaffs, in emulation of the Witches dance of broomsticks. ([Hawkesworth], 1753, p. 18)

  ➣ 1818 The second strain in the ‘witches dance’ is amongst the most singularly fine adaptations of ideal expression that we know of. (Anonymous, 1818, p. 196)

  ➣ 1821 That he was placed in the rear at the witch dances, and beat up them that were too slow. (Aiton, 1821, p. 79)

Most of the antedatings push back the first recorded use by a few decades, but there are also truly amazing discoveries. Two such discoveries would be the citations for devil dance,⁴⁴ predating the earliest known occurrence by as many as 266 years, and fairy dance, predating it by ninety-two. These findings justify queries for older versions of both the modifiers and the heads. One should note that the two cases, and witch dance as a third, are peculiar in that they are, in fact, possessive compounds. The missing apostrophe is not treated as a mistake, because it came into wider use only in the late seventeenth century (McArthur et al., 2018, p. 114). Taylor (2000, p. 302) posits that ‘there would seem to be every reason’ to treat possessive compounds as a subcategory of nominal compounds on both structural and semantic grounds, but the motivations for the existence of competitive forms are not entirely clear. The sample given above, however modest, might be indicative of the same path of development for each of the three patterns, that is, from a possessive to a non-possessive compound, but this may be an oversimplification. According to OED3, two genitive variants, witches’ dance and Dance of Death, are acknowledged to have operated alongside witch dance and death dance respectively. Further research is, therefore, necessary to investigate the factors that triggered lexicalization processes from the diachronic perspective (see, e.g., Juvonen, 2021).

A preliminary search in British and American newspaper archives for the compounds I was unable to antedate with Google Books shows that they, too, are worthwhile considering in the revision process. Shimmy dance (1919) has, for instance, been predated to 1918; slam dance (1981) to 1976; spot dance (1911) to 1829; and supper dance (1820) to 1814.⁴⁵

4. Conclusions

The study indicates that the treatment of dance compounds and, one might presume, other types of headword in OED3 has been unbalanced. This is hardly surprising, inasmuch as a situation in which half of the dictionary is fully revised and the other half remains steeped in nineteenth-century lexicographical tradition presupposes palpable asymmetry. Brewer (2009b, p. 210) observes that it was ‘impossible, over so many years of composition, by different editors and under often difficult conditions, to maintain consistent scholarly standards’. Almost a hundred years after the completion of OED1, her words ring equally true. Be that as it may, James Murray collaborated with three subeditors, whereas the editorial staff of OED3 in 2014 consisted of seventy-five editors, based mostly in Oxford, but also in New York and ‘elsewhere in the world’ (Simpson, 2014, p. 26). Their number alone is a factor affecting the uniformity of editorial practices. To put it bluntly, at this stage, the dictionary is not perfect, but certainly perfectible.

The objective of my study, like Schäfer’s (1984, p. 145) four decades ago, was not to discredit the OED’s documentation, but to ‘identify and locate its specific deficiencies’. The revised entries in OED3 are more comprehensive than those in OED2, but some components, including the choice of the quotations, the wording of the definitions, and the dating of the headwords or senses, remain to be refined and brought up to date. The editors, moreover, need to make sure that the dictionary system is efficient and easy to navigate, supplying the users with the kind of information they are seeking. Since the old and new entries have not been integrated as seamlessly as expected, there is room for further progress in this respect. The major disturbing feature of the revised and newly added compounds is that a proportion of them is left undefined, even though their meanings are not always transparent.

Despite the fact that updates to the dictionary, including new and revised words from across the alphabet, appear regularly on OED3 website, what the compilation consists in is rarely shared with the general public. One source reveals that the lexicographers ‘systematically search a range of historical text resources, narrowing down the examples of polysemous words by context, as well as manually comparing example sentences… relying a great deal on researchers’ own intuitions’.⁴⁶ What has been offered here might look very similar, but the major difference lies in the size: Google Books is big data. As emphasised by Kilgarriff (2007, p. 151), enormous data sets provide researchers with better results, which is but a single reason why their investigations are valuable in the long run. One has to admit that Mark Davies’ Google Books corpora have a user-friendly interface, but, in terms of historical detail, the entire Google Books archive is far more extensive.

Regarding the priorities adopted by the editors, words for revision or addition are selected on the basis of several factors, including frequency; topical coverage, such as sexuality and gender identities (see, e.g., Dent, 2018); and special initiatives, such as neologisms (see, e.g., Connor Martin 2019) and world Englishes (see, e.g., Salazar, 2021). This inevitably results in a mixture of lexical and semantic categories from various areas of the English lexicon, perhaps those most prominent in online sources. A systematic cross-check of the diachronic material of Google Books and the synchronic material of contemporary corpora would be recommended for a fully holistic treatment. Dance names, for instance, have been covered fairly well in OED3, but there are more high-frequency terms (e.g., aerial dance, bachata, charanga, concert dance, cumbia, dragon dance, Eagle Dance, ethnic dance, harlem shake, highland dance, hip-hop dance, hoop dance, house dance, and urban dance), both in Google Books and enTenTen21, to be considered for inclusion.

A handful of compounds for Native American dances (e.g., bear dance, Ghost Dance, and medicine dance) are related to the special initiative focusing on expanding the coverage of world Englishes. The dance terms are not borrowings, the initiative’s primary concern, but they call for a re-examination of the cultural heritage of Indigenous peoples, originally presented ‘entirely through the lens of Victorian Britain’ (Salazar, 2021, p. 273) and, surely, in line with white colonial supremacy. This would entail identifying the full repertoire of tribal dance forms which either originated from, or are an expression of, Native American culture (e.g., bull dance, fancy dance, and war dance); revising the definitions (e.g., kachina dance); excluding disrespectful citations (e.g., ‘A wolf-dance [by] painted naked savages’); and adding sociohistorical notes such as the one recommended for sun dance.

Taken together, Google Books is by no means an ideal diachronic resource, but it is unparalleled in terms of size, topicality, and variety of genres, of which only newspapers are represented more sparsely. There are reasons for claiming that the millions of texts it contains, particularly out-of-print and obscure resources that had languished for decades in library stockrooms (Dinscore, 2015, p. 181), have been under-explored, so mining them for carefully selected content is likely to fill out the gaps in the dictionary’s patchy evidence. Drawing on my findings, I provide some suggestions as to how the treatment of OED3 headwords could be improved. After all, before the revision of OED3 eventually reaches completion, ‘there are more words to document, more still to expand and update and antedate’ (Williams, 2021, p. 59).

One might wonder to what extent technological innovations will influence the workflow in historical lexicography. Enhancing the efficacy of existing resources with computational techniques is now being implemented and there is no doubt that artificial intelligence (AI) tools will soon drive the field forward. It seems highly unlikely, however, that the compilation process will ever become fully automated to the extent that neglects the editors’ judgement and expertise.

References

Footnotes

Appendix 1: Dance compounds with Google Books frequencies