2. Garner's lexicographic methodology

The most significant innovation in lexicographic methodology in Garner4 is the use of Google Books in combination with Google's ngrams application.Footnote ¹ This deserves some attention since Garner dedicates a significant portion of the preface to discussing the innovation. The information that is obtainable through Google Books Ngram Viewer can be highly useful, but since it does not come without its complications, applying Google ngrams to quantitative lexicography should be done with caution. Although it can be used to calculate ratios of two or more variant uses, as Garner does, these ratios cannot always be taken at face value (as he seems to do).

Calculation of the ratios is affected when there is polysemy in one of the variant uses: the occurrence of words or forms with different meanings cannot be taken as simple variants of the usage problem under investigation. This has to be taken into account, and contextual information is needed to overcome this complication. However, both the required co-text and context are available only in a limited way, since Google Books was not designed for linguistic inquiry at this level of detail.

Another complication is that, although the Ngram Viewer should ostensibly make it ‘possible to calculate ratios on word frequency and phrase frequency in […] the two major sub-types: American and British English’ (Garner, Reference Garner2016: ix), we do not actually know what criteria Google employs to categorise the language variety of a book as American or British English (or other varieties of English for that matter).Footnote ² And as mentioned, potential polysemy in one variety of English but not in the other renders the numbers unreliable. This obviously needs to be taken into account when commenting on usages across varieties of English.

First introduced as a methodological innovation in the third edition, Garner's Modern American Usage (Garner, Reference Garner2009, hereafter Garner3), the Language Change Index (LCI) (see also Smits, Reference Smits, van Ostade and Percy2017, and Peters, Reference Peters and van Ostade2018) is also present in Garner4. This index describes five stages of language change and the spread of new forms, as represented in Table 1.

In order to compare the general level of acceptability of usage items between Garner3 and Garner4, I manually counted the number of times each stage occurs in the printed versions. The distribution of each of the five stages of the LCI in Garner3 and Garner4 is given in Figure 1.

Figure 1. Relative frequencies of the LCI stages in Garner3 and Garner4 as percentages of total number of LCI attributions

Figure 1 shows two things: the occurrence of Stage 1 seems disproportionally high compared to the other four stages, and Garner's evaluation has not significantly changed between Garner3 and Garner4. Stage 1 means that the usage in question is ‘widely rejected’ (Garner, Reference Garner2009: lv; Reference Garner2016: li). The five stages are presented with analogies with several other domains, and one of these is moral behaviour, with Stage 1 representing a ‘mortal sin’ (Garner, Reference Garner2016: xxxi). This seems to confirm Peters’ assessment that the LCIs ‘represent degrees of social acceptance […] rather than the stage of assimilation which a new usage has attained’ (Peters, Reference Peters and van Ostade2018: 38). The disproportionally high relative frequency of occurrence of Stage 1 could thus suggest that, although Garner's ostensible purpose was to be descriptive, i.e. ‘to measure how widely accepted various linguistic innovations have become’ (Garner, Reference Garner2009: lv), the attribution of Stage 1 to a particular usage seems to be a de facto proscriptive device.

3. Attitudes towards prescriptivism, descriptivism and linguistics

Since Garner generally appears to be reasonably accepting of the application of linguistic ideas to lexicography, it is regrettable that the information he uses to support his points of view is at times considerably out of date. Just like its two predecessors, Garner4 contains two essays on what he calls ‘language wars’. Both essays were written in response to linguists’ reactions to earlier editions, and ostensibly meant to bridge the philosophical and ideological gap between prescriptivists (like Garner himself) and linguists. In essence, these two essays are defences of prescriptivism, and since they are informative of Garner's attitude as a prescriptivist writer, it is worth exploring them in a little more detail.

The essay ‘Making peace in the language wars’ first appeared in the second edition, Garner's Modern American Usage (Garner, Reference Garner2003, hereafter Garner2), and was based on an article Garner published in Reference Garner2001 as part of a brief polemic he was engaged in with the linguist Tom McArthur. Their argument centred around differing views on the actions and responsibilities of the two groups of people that Garner calls ‘prescribers’ and ‘describers’. Garner puts the difference between descriptive linguists and prescriptivists as follows: ‘[e]ssentially, descriptivists and prescriptivists are approaching different problems. Descriptivists want to record the language as it's actually used’, while ‘[p]rescriptivists […] want to figure out the most effective uses of language, both grammatically and rhetorically’ (Garner, Reference Garner2016: xiv–xv). ‘I am a prescriber who uses descriptivist methods – in effect, a descriptive prescriber,’ Garner claims (Reference Garner2016: xliv).

Differences between these two groups still exist, and in some ways they are as great as they have ever been (see for instance Beal, Reference Beal, van Ostade and van der Wurff2009 on the rise of the ‘new prescriptivism’). There has, however, been a considerable change in ideas about and attitudes towards prescriptivism among members of both groups since Garner's essay was first published, and a new edition of a usage guide should reflect this. Instead, describers’ attitudes are represented by linguists’ remarks from the last century. Citing Robert Hall's out-dated remarks from 1950 that there is no such thing as ungrammaticality as evidence of the position of present-day linguists can be seen as negligent, misleading and even irresponsible. At the very least, current, critical attitudes of linguists to important issues like this ought to be included.

There are other instances in which Garner does not take recent developments into account. One example is the suggestion that linguists do not recognise and accept the value of knowing and teaching Standard English (Garner, Reference Garner2016: xxxvi). This is no longer the case (see for instance Carter, Reference Carter, Bex and Watts1999: 163), and suggestions as to how to responsibly approach the teaching of Standard English had already been offered previously (e.g. Curzan, Reference Curzan2002). However, Garner writes that ‘[t]oday the teaching of standard English is being labeled discriminatory’ (Garner, Reference Garner2016: xliv), referring to James Milroy's chapter in Language Myths (Reference Milroy, Bauer and Trudgill1998) in which it is claimed that a person's use of language is one of the last remaining socially acceptable forms of overt discrimination. It is, of course, not so much the teaching of standard English itself that is considered discriminatory by linguists, as the teaching of the ideology that the standard is linguistically superior to other regional and social varieties of English. Garner must at some level be convinced of this superiority, since it is clear from his LCI that he believes that some forms of English are better than others, and that some users are better than others as well, which is by definition discriminatory.

Another example is the charge that linguists have neglected to include prescriptivism in their research. Garner writes that he does not ‘doubt the value of descriptive linguistics – up to the point at which describers dogmatically refuse to acknowledge the value of prescriptivism’ (Garner, Reference Garner2016: xliv). While this has been the case in the past, more recent studies have investigated the level of success of prescriptive efforts, thereby implicitly accepting that prescriptive efforts can actually be successful. Examples are Anderwald (Reference Anderwald2014), Curzan (Reference Curzan2014) and Hinrichs, Szmrecsanyi and Bohmann (Reference Hinrichs, Szmrecsanyi and Bohmann2015), which answer Garner's wish that if ‘the prescribers have moderate success, then the describers should simply describe those successes’ (Garner, Reference Garner2016: xliv). In addition, there has been an increasing amount of research that describes prescriptivism as part of linguistic reality and therefore worthy of study.Footnote ³

As a third example Garner mentions linguists’ resistance to prescription exemplified by their apparent aversion to spell-checking applications in word-processing software. This notion is based on an article that allegedly argues that these applications are bad because they stifle experimentation in spelling, but the article about spell-checkers referred to was published in the late 1980s. Reference to an article from that period may have been acceptable at the time of Garner1, but attitudes have evolved since then. The main criticism that sociolinguists aim at spell-checking applications these days is that their prescriptivism is ubiquitous, rarely perceived as optional by the users of the application, and that the norms from which the prescriptions originate are hidden behind the technology (see Curzan, Reference Curzan2014: 64–92).

‘Making peace in the language wars’ was first published 15 years before the most current edition, and its sources have not been updated for any of the editions after those in which they were originally published, including Garner4. Garner is a lexicographer, not a linguist, and therefore may not be fully up to date on the developments discussed. But the developments concerned are important and reflect broad changes in attitudes in an area that he is interested in and about which he obviously cares. So, especially for a descriptive prescriber it would be useful to at least refer to more recent developments in the field.

Originally called ‘The ongoing struggles of garlic-hangers’ in Garner3,Footnote ⁴ the essay ‘The ongoing tumult in English usage’ seems to have been a reply to an article entitled ‘Language wars truce accepted (with conditions)’ in a legal journal (Tiersma, Reference Tiersma2005). It is worth examining Tiersma's essay before discussing Garner's views in more detail.

Tiersma's central point is that ‘it matters if Garner is a describer or a prescriber because he has become a force to be reckoned with in […] what we might call “the style business”’ (Reference Tiersma2005: 139), and that the criticism that linguists aim at ‘typical prescribers is that, instead of acknowledging the existence of variation, they seem to have an almost irresistible urge to label one of the two variants as wrong or improper’ (Reference Tiersma2005: 142). Tiersma argues that ‘Garner is free to argue in favor of proved or strove. He can even prescribe it to those who seek his advice. He should do so, however, in light of solid descriptive evidence’ (Reference Tiersma2005: 143). He concludes that ‘what it boils down to is this, can we trust Brian [sic] Garner and other prescribers to decide for us what is correct English and what is not?’ (Reference Tiersma2005: 141).

So it is a question of authority. Smits (Reference Smits, van Ostade and Percy2017: 224) notes that ‘Garner's appeals to authority’, that is ‘the sheer enumeration of […] different “authoritative devices”, are meant to convince (perhaps overwhelm) the reader to believe that his judgement can be trusted’, adding, ‘[i]n that sense, Garner is particularly knowledgeable about the developments in the usage guide tradition’. To this I would add that Garner is also very much aware of his own place in that tradition as it is currently being shaped, which I will discuss in Section 5.

4. A discussion of some of Garner's usage problems

Some of the criticism of Garner's methodology may be illustrated by a number of well-known usage problems: disinterested, the ‘new’ like, snuck, singular they, and unique. In Garner's discussion of these usage problems, either the methodology on which his prescriptions are based is insufficient, or the information on actual, current usage is not up to date. However, these instances are also illustrative of how descriptive Garner can be.

Garner's prescription for the use of disinterested in the sense of uninterested has not changed since Garner1: ‘the distinction is still best recognized and followed because disinterested captures a nuance that no other word quite does’ (Garner, Reference Garner1998: 213). Garner3 added the LCI at Stage 4 (Garner, Reference Garner2009: 266), and Garner4 adds: ‘Current ratio (uninterested children vs. *disinterested children): 1:1’ (Garner, Reference Garner2016: 290), indicating how frequent this alleged misuse is. Fowler4, also based on corpus data (Butterfield, Reference Butterfield2015: viii), says that disinterested occurs slightly more often than uninterested, but that ‘there is no reason whatsoever that [disinterested] should be prevented from having two different meanings; most words do’ (Butterfield, Reference Butterfield2015: 224). This polysemy of disinterested was already put forward as a linguistic fait-accompli in the corpus-based Cambridge Guide to English Usage: ‘Given that disinterested carries several meanings, we effectively rely on the context to show which is intended - as is true of many words’ (Peters, Reference Peters2004: 158). The overlap in meaning noted by Butterfield and Peters is ignored when Garner represents usage as simply being a ratio of frequencies of occurrence. Also, by merely citing this ratio, Garner seems to implicitly comment on the ‘erroneous’ use of uninterested to mean ‘impartial’, even though this is a mistake that no one would make (which Garner does not mention, nor do most usage guides).

Garner discusses the various ‘new’ uses of like under like as a ‘Vogue word and verbal tic’ or ‘space-filler’ (LCI Stage 2), as he calls them (Garner, Reference Garner2016: 566). This instance conflates various uses of like as a discourse marker or a discourse particle, which, rather than being random insertions into discourse, have been shown to have specific functions. One function of what is sometimes referred to as the ‘new like’ is its use as a quotative complementiser in the construction be like (D'Arcy, Reference D'Arcy2007: 392). Although relatively new, the use of quotative like was already noted thirty-five years ago (Butters, Reference Butters1982: 149) and has been studied fairly extensively since.Footnote ⁵ Garner acknowledges the history of this use of like: ‘[s]ince the 1980s, be like has also been a low casualism equivalent to said in relating a conversation, especially among juveniles. […] In teenagers, this usage is all but ubiquitous. In adults, it shows arrested development’ (Garner, Reference Garner2016: 566). However, he does not mention the word's actual function, or how it is used differently from said, since in its quotative function, according to D'Arcy, be like ‘performs the specialized role of introducing reported speech, thought and nonlexicalized sounds’ (Reference D'Arcy2007: 392). Moreover, its use is not limited to young people, since ‘adults of all ages use them to some extent or another’, D'Arcy claims, although ‘adolescents use the vernacular forms more frequently than adult cohorts’ (Reference D'Arcy2007: 411). Thus, the information Garner uses is not very up to date.

In Garner3, the use of snuck as an alternative form for the past tense and past participle of the verb to sneak (sneaked) had an LCI of Stage 3. Fowler4 reads that snuck ‘is recognized as standard in AmE’ (Butterfield, Reference Butterfield2015: 756), though I have not been able to track down what that statement is based on other than its frequency in the Oxford English Corpus, which provides Butterfield with his data. In the fourth edition, Garner acknowledges the form's widespread currency and puts its LCI at Stage 4 (Garner, Reference Garner2016: 841). This indicates that he has noted the increased use of snuck, which, as Smits shows, has overtaken the use of sneaked in American English from 2005 onwards (Reference Smits, van Ostade and Percy2017: 232). Despite this, however, and contrary to Fowler4, Garner still considers snuck ‘a nonstandard past tense and past participle of sneak common in American speech and writing’ (Garner, Reference Garner2016: 840).

The discussion of they (and them) as a common-sex singular pronoun is given a descriptive approach in Garner4. It is discussed in the lemmas ‘concord’, ‘pronouns’ and ‘sexism’. Under ‘concord’ Garner notes, as he already did in the first edition, that singular they ‘is especially common in BrE’, that it has become so common because it is ‘the most convenient solution to the single biggest problem in sexist language – the generic masculine pronoun’, and that ‘[i]n all varieties of World English, resistance to the singular they is fast receding’ (Garner, Reference Garner2016: 195–196). Under ‘pronouns’ Garner observes that many indefinite pronouns often combine with they despite being traditionally considered singular ‘because they has increasingly moved towards singular senses’ (Garner, Reference Garner2016: 736). And under ‘sexism’, he notes that, although ‘the masculine singular pronoun may survive awhile longer as a generic term, it will probably be ultimately displaced by they, which is coming to be used alternatively as singular or plural. […] Speakers of AmE resist this development more than speakers of BrE’, which is ‘an unfortunate obstacle to what promises to be the ultimate solution to the problem’ (Garner, Reference Garner2016: 822).

Fowler4 reads that ‘[o]ver the centuries, writers of standing have used they, their and them with anaphoric reference to a singular pronoun or noun […] to the point that […] such constructions are hardly noticed any more or are not widely felt to lie in a prohibited zone’, adding that ‘[t]he process now seems irreversible’ (Butterfield, Reference Butterfield2015: 814). Garner's assessment of the British situation thus seems correct, and his comments show that he describes singular they using up-to-date information with regard to its use and the attitudes towards its use.

The treatment of unique highlights the problem of basing usage arguments on logic. Garner discusses the use of modifiers with unique and other absolute adjectives in the essay entry ‘adjectives’,Footnote ⁶ as well as under the lemma ‘unique’. The argument is based on logic (there is a cross-reference to the essay entry ‘illogic’) and on the unambiguous meaning of unique. He argues that ‘uniqueness […] is an absolute quality, something cannot logically be more or less unique than something else’, and that ‘[u]nless the thing is the only one of its kind, rarity does not make it unique’ (Garner, Reference Garner2016: 930); ‘something is either unique or not unique, there can be no degrees of uniqueness. Hence *more unique and *very unique are incorrect’ (Garner, Reference Garner2016: 20). If we accept this as logically sound, then the application of this logic to an adjective such as perfect would also disallow the use of more perfect. However, Garner does allow more perfect and even considers what is arguably a more logical phrase, more nearly perfect, to be pedantic, adding: ‘[a]lthough the Constitution [in which the phrase occurs] is not without stylistic blemishes, this probably isn't one of them’ (Garner, Reference Garner2016: 20). The reason for this may be the historical importance of the document mentioned.

Chapman, among others, argues that appealing to logic is not always the best approach to linguistic problems since ‘language works on principles extending beyond logic’ (Chapman, Reference Chapman, Cloutier, Hamilton–Brehm and Kretzschmar2010: 149). Garner is aware of this as well, since under ‘illogic’, he writes that ‘grammatical distinctions’ do not ‘necessarily reflect logical ones’, and that ‘[o]ur language is full of idioms that defy logic, many of them literary and many colloquial’ (Garner, Reference Garner2016: 484). Thus, with regard to unique as a gradable adjective, Garner should probably have heeded his own advice more.

5. An American Fowler

Bryan Garner is a self-identified Fowlerian. As early as the ‘Autobiographical Note’ in the preface to Garner1, he wrote about how he became infatuated with A Dictionary of Modern English Usage (Fowler, Reference Fowler1926).

Garner's admiration of Fowler led him to found the H.W. Fowler Society in the United States in 1980. His identification with Fowler seems to have started when the linguist Tom McArthur opened his review of Garner1 with the following words:

McArthur made the comparison to Fowler because he found that Garner1 was ‘quite astonishingly close in content, spirit, and layout to the original, published in 1926’ (McArthur, Reference McArthur1999: 59). Garner replied to this by writing ‘how pleasing it was to see the opening words of Tom admitting McArthur's review’ (Reference Garner2000: 3), though acknowledging that ‘McArthur didn't intend much, if any, praise in his comment’ (Garner, Reference Garner2000: 3). Nevertheless, Garner seems to have taken up Fowler's mantle eagerly, and appears to have been actively branding himself as the American version of Fowler since: ‘The Fowlerian echo is purposeful […] It's just that I'm writing from an American perspective’ (Garner, Reference Garner2000: 4). That perspective, however, seems to have changed in the most recent edition of his work.

Garner's pursuit of Fowler is particularly evident in the original title of his usage guide and in the ways in which it has changed over subsequent editions.Footnote ⁷ The title of the original 1998 edition, A Dictionary of Modern American Usage, clearly referenced Fowler's A Dictionary of Modern English Usage (Reference Fowler1926). From Garner2 onwards, Garner's name was included in the title of his book, thereby echoing the titles of the later editions of Fowler.Footnote ⁸ From Garner1 to Garner3, the author and his book firmly remained an authority on specifically American usage. However, judging by the change in title for Garner4, the book is now officially about English, rather than merely American usage. Perhaps to pre-empt any ad-hominem attacks, Garner strategically mentions in the first sentence of his Preface to this edition that this change was the publisher's choice:

It has not been possible to comprehensively analyse the book for references to other varieties of English, but it will be interesting to see to what extent norms of World English – insofar as these can be identified – are indeed included, or whether Garner's Modern English Usage still retains much of the existing British and American norms for correct usage of Fowler and Garner.

The change in Garner4 can, however, be illustrated by counting the number of references to BrE in the various editions of Garner's usage guide, as is shown in Figure 2. The graph clearly shows that Garner pays much more attention to BrE usage in the latest edition compared to the previous ones, as was to be expected by the change in the title.

Figure 2. Number of references to BrE usage per page (of main text) in all four editions of Garner's usage dictionary

I have argued elsewhere that the development of Fowler's usage guide has both mirrored and directed the historical development of the usage guide as a genre during the twentieth century (Straaijer, Reference Straaijer2016, Reference Straaijer and van Ostade2018). The evolution of Garner's book follows that of Fowler, though edited by three different writers in the course of its history. After the first edition, the second edition (Gowers Reference Gowers1965) remained largely unaltered (Busse & Schröder, Reference Busse and Schröder2010: 47), as in the case of Garner2; the third edition introduces a new method of gathering information or evaluation, such as a purpose-built corpus in the case of The New Fowler's Modern English Usage (Burchfield, Reference Burchfield1996) and the Language Change Index in Garner3; and the fourth edition then sees an increase in sophistication of this method, as in the use of the Oxford English Corpus in the case of Fowler4 and Google Books in Garner4.

The timing of the publication of Garner4 is interesting. Considering that Garner has explicitly identified himself with Fowler since Garner1, that the two works are published by the same publisher, and that they now also ostensibly deal with the same variety of English, the appearance of Garner4 just a year after the latest edition of Fowler is unlikely to be a coincidence. It seems that ‘Garner’ is now so established in America as a brand that the book no longer needs to have ‘American’ in the title. This close following of Fowler is perhaps a reason why the four editions of Garner's dictionary have appeared in such a relatively short amount of time: while 90 years separate Fowler1 (Reference Fowler1926) from Fowler4 (2015), there are only 28 years between the publication of Garner1 and Garner4.