2. Background

While the -ize form is a Greek verbal ending transmitted through Latin and Old French and initially associated with Greek verbs (i.e. baptize), the -ise form is etymologically connected with French verbs ending in -iser (i.e. realise).Footnote ³ As a result of the increase of Latinate and Greek vocabulary in English in the Renaissance and of the familiarity of literate men with Latin, ‘English spelling became as affected by the etymologising process as French had earlier been’ (Scragg, Reference Scragg1974: 53–54). Many grammars and usage guides have since then adopted a purist approach proposing to spell words with a Greek etymon with -ize and Latin ones with -ise.Footnote ⁴ This etymological rationale is, in many cases, opaque to the modern reader and there has been some pressure, at least in British English, to standardise the <s> spelling on the basis of the French form (Carney, Reference Carney1994: 433), while other sources, such as the OED, The Times or The Encyclopaedia Britannica opted to standardise the <z> spelling on the grounds of its pronunciation /z/ (Pink, Reference Pink1932: 88).Footnote ⁵

The origin of this orthographic variation dates back to the latter part of Middle English. Even though the competition was initially solved in favour of the Greek form -ize, this practice was not systematic because of i) the increasing variation of s/z in Old French and Middle English; ii) the uniform spelling of some verbs (comprise, surprise, etc.); and iii) the French decision to deplore the z-spelling in favour of the standardisation of the -ise form in these contexts (Upward & Davidson, Reference Upward and Davidson2011: 171).

It was in the 16^th century when these forms began to proliferate regardless of their origin, whether Greek or Latin/French. Early Modern English writers could spuriously use either form without considering the origin of the word itself, as in metamorphize/metamorphise or exercise/exercize. Interestingly enough, this orthographic fluctuation also spread, albeit sporadically, to other native words which had never been written with the -ize spelling, as in arise/arize.

The dilemma, however, was not resolved with the arrival of prescriptivism, which in many cases proposed to leave it to the user's choice. Even though 18^th-century prescriptive grammarians almost systematically ignored the phenomenon, the general tendency was to avoid any reference to the existence of the <z> spelling. In his Elements of the Grammar, Coote deliberately ignored any reference to Greek-etymon verbs in the following terms:

It was in the 19^th century when grammarians expressed their concern for the unresolved representation of ‘eyes’. In Alford's The Queen's English (Reference Alford1864), for instance, he stated that ‘it does not seem easy to come to a decision’ in view of the ‘present state of our English usage’, recognizing that the question was still open (Alford, Reference Alford1864: 36). 20^th-century usage guides brought some sort of fresh air to the topic in the attempt to disregard the etymological distinction assuming the speakers were not necessarily aware of the origin of particular words. Pink proposed that ‘in the absence of general agreement on the matter, the reader may please himself’ (Reference Pink1932: 88), noting that the ordinary users with limited knowledge on etymology are cutting ‘the Gordian knot by spelling all such words with -ise, and English printers very commonly adopt the same practice’ (Reference Pink1932: 88; Vallins, Reference Vallins1951: 196–197; Reference Vallins1955: 120).Footnote ⁷

Today, the choice between -ise and -ize is generally considered to depend on geographical preferences, the latter ascribed to American English (henceforth AmE) while British English (henceforth BrE) allows either spelling (Carney, Reference Carney1997: 65; Peck & Coyle, Reference Peck and Coyle1999: 48; Upward & Davidson, Reference Upward and Davidson2011: 171). This is a simplistic description of the issue in view of the number of verbs which can adopt either spelling. There is, however, a growing tendency today for the use of -ize, in academic prose in particular, both with high frequency and with rare coinages (Biber et al., Reference Biber, Johansson, Leech, Conrad and Finegan1999: 402; Todd & Hancock, Reference Todd and Hancock1986: 293).Footnote ⁸

Variation increases when it comes to the different varieties of English worldwide. Among the inner circle varieties, ‘both American and Canadian publishers restrict themselves to -ize while Australian and New Zealand publishers tend to use -ise more consistently than their British counterparts, with <z> spellings usually being a sign of learned or scientific writing in those varieties’ (Bauer, Reference Bauer2002: 62; Huddleston & Pullum, Reference Huddleston and Pullum2002: 1715). Notwithstanding this, the variation is not merely a matter of American versus non-American usage; informal written International English, for instance, is characterised by a mixture of British (Commonwealth) and American spellings, ‘with the latter presumably dominating’ (Bauer, Reference Bauer1994: 134; Melchers & Shaw, Reference Melchers and Shaw2003: 187).

Existing accounts are mostly descriptions of the general tendencies across varieties based on assumptions derived from the dominant publishing policy in a particular area, which cannot be taken as the general practice of a community of speakers. The present paper therefore delves into the origin, development and distribution of these competing forms in English. The first part approaches the phenomenon historically, describing the quantitative dimension of the forms from the year 1500, when their actual competition arose. The second provides a distribution of the varying forms in some present-day varieties of English worldwide, both inner and outer circle varieties. The study follows the line initiated by other recent works examining the spelling of relevant English words, such as Achiri–Taboh's account of the dual representation of ‘shun’-words and the choice to spell them as -tion or -sion (Reference Achiri–Taboh2018: 36–42) along with the subsequent discussion on their provenance and historical development in English (Achiri–Taboh, Reference Achiri–Taboh2020; Bulley, Reference Bulley2020).

The study relies on different corpora in view of the diachronic and synchronic dimension of the analysis. The historical analysis covers a time-span of approximately 500 years (the period 1500–1993) and, for convenience, is based on the evidence found in the Early English Books Online Corpus (EEBOC), the Corpus of Late Modern English Texts (CLMET), the Corpus of Historical American English (COHAE) and the British National Corpus (BNC), which contain sizeable textual evidence from Early Modern to present-day English.Footnote ⁹ The synchronic analysis of the varieties of English, on the other hand, is based on the Corpus of Global Web-based English (GloWbE).Footnote ¹⁰

A holistic treatment of the phenomenon is not feasible in view of the different etymological provenance of the words. For convenience, an etymological perspective has been adopted considering the evidence provided by the most frequent words of French/Latin and Greek origin with dual forms. Mark Davies’ tagged versions of EEBOC, COHAE, BNC and GloWbE have been designed using the CLAWS7 POS-tagger, therefore facilitating introspection not only by lemma but also by part-of-speech.Footnote ¹¹ The instances were automatically retrieved by searching for the verbal uses of -ise and -ize with the notation *ise*_v* and *ize*_v*, so that both the present and the past tense forms could be automatically generated, third person inflection included, irrespective of any orthographic variation in the rendering of the word (such as advertise, aduertise, advertised, advertises). Additional searches were needed, however, for the -ing forms (-ising* and *-izing*, respectively). CLMET is unfortunately offered as a plain text version, a fact which complicated the automatic generation of the instances. Irrespective of the corpus, the results required manual disambiguation to eliminate the instances beyond the scope of the present research.

3. Variation in the spelling of -eyes in the history of English

The diachronic study is based on the occurrences provided by the topmost frequent French/Latin and Greek words allowing for the forms -ise and -ize. The words of French/Latin provenance include advertise, authorize, baptize, circumcise, criticise, exercise and recognize and the Greek ones include anathematize, anatomize, epitomize, metamorphize, phlebotomize, philosophize, synchronize.Footnote ¹² As shown, -ise outnumbers -ize in Early Modern English, with 64.6 and 0.5 occurrences, respectively (normalised to 100,000 words). In themselves, these figures cannot be taken as a typical case of spelling forms in competition, at least at that early stage of development, but two etymologically distinct forms with apparently different uses. The use of -ise and -ize in the period 1500–1700 depends on the etymology, -ise systematically preferred with Latinate and -ize more strongly associated with Greek words, as shown in Figure 1. The rise of -ise with Latinate words in the period 1500s–1540s is surely corpus-biased as these decades display a smaller number of words than the other decades, and this was precisely the period when some French borrowings were still incipient in English. After the 1540s, however, -ise is found to have a balanced distribution over the period, with a slight decrease towards the second half of the 17^th century. The -ize form is negligible with less than one occurrence every 100,000 words throughout the whole period. With Greek words, on the other hand, -ize became the preferred form immediately after the borrowing of these terms into English in the second half of the 16th century, rising from 0.26 to 0.51 occurrences in the 1600s and 1690s.Footnote ¹³ Interestingly enough, -ise becomes sporadic throughout the whole period.

Figure 1. The forms -ise and -ize with Latinate and Greek words (n.f.)

The period 1700–1900 is characterised by the actual competition of these variant forms until the eventual adoption of -ise in BrE and -ize in AmE. Figure 2 presents their distribution with both types of words in CLMET in view of the three 70-year sub-periods of the corpus (1710–1780, 1780–1850 and 1850–1920). These forms were in an on-going process of competition until the second half of the 19^th century when -ise definitely outnumbered -ize with Latinate verbs, to such an extent that the former almost doubled the occurrence of -ize in the period 1850–1920, after -ize had been found to be slightly more frequent than -ise in the first two periods.

Figure 2. The forms -ise and -ize with Latinate and Greek words in CLMET (n.f.)

In combination with Greek words, the period 1710–1780 is a follow-up of the Early Modern English practice with an outstanding use of -ize (with 6.11 and 0.19 occurrences, respectively). The period 1780–1850 shows a significant decline of -ize (with 2.13 and 1.24 occurrences, respectively). Even though -ize slightly outnumbers -ise, they present a parallel distribution in the period 1850–1920 and, unlike words of Latinate origin, there is no clear preference for either of them until the first quarter of the 20^th century.

The period 1960–1993, in turn, confirms the definite adoption of -ise in BrE, irrespective of the provenance of the word. As shown in Figure 3, in the first sub-period (1960–1974) -ize triples the occurrence of-ise (357.8 and 116.3 occurrences, respectively). This is an unexpected occurrence considering the tendency observed in the last period of CLMET where -ise already outnumbered -ize at the beginning of the 20^th century. The drastic rise of -ize in the first half of the 20^th century is surely associated with the impetus received at that time from reputed sources such as the OED, The Times or The Encyclopaedia Britannica, which opted to standardise the <z> spelling on the grounds of its pronunciation /z/. Notwithstanding this, -ise not only resisted in the period 1960–1974, but also managed to rise again in the period 1975–1984, when it outnumbered -ize with Latinate words (with 274.5 and 185.5 occurrences, respectively). This preference remained in the last sub-period of the corpus where both forms occur with a rate of 63.8% and 36.2% towards the end of the century. Greek ones, on the other hand, confirm the progressive standardisation of -ise in BrE. The -ize form is still preferred over -ise in the first sub-periods of the corpus and the latter proliferated in the last sub-period along with the concomitant decline of -ize.

Figure 3. The forms -ise and -ize with Latinate and Greek words in BNC (n.f.)

Unlike BrE where the competition stretched until the last quarter of the 20^th century, these orthographic variants never competed with each other in AmE. As shown in Figure 4, -ize is already the preferred form in AmE since the beginning of the 19^th century and, more importantly, it is found to be on the rise throughout that same century until reaching its climax at the beginning of the following century, coinciding with the progressive demise of -ise which becomes sporadic since then. The standardisation of -ize in AmE took place regardless of the Latinate or Greek origin of the word.

Figure 4. The forms -ise and -ize with Latinate words in COHAE (n.f.)

4. Variation in the spelling of -eyes across varieties of English

This section analyses the distribution of these spelling forms in different varieties of English worldwide. The study first discusses the alternation in the inner circle varieties to proceed with the outer circle, both in Asian Englishes and African Englishes. Jamaican English (JamE) is also offered as an example of a transatlantic variety of English.

The analysis of spelling variation from a diatopic perspective must necessarily begin with the study of the item in the inner circle to evaluate its diffusion in the other less influential varieties. Figure 5 collapses the distribution of -ise and -ize with words of Latinate provenance. In the inner circle, as expected, two major tendencies stand out. The -ize form, on the one hand, systematically predominates in AmE and CanE, while -ise is sporadic. The preference for -ize in CanE is logically the result of the direct contact with the American language and culture. CanE is described as a blend of British and American features where pronunciation is American and spelling of British influence (Schneider, Reference Schneider, Kachru, Kachru and Nelson2006: 66), albeit the dilemma in this case is clearly resolved in favour of the American spelling.Footnote ¹⁴ The -ise form, on the other hand, is preferred in BrE, AusE and NZE, although -ize finds considerable room in these varieties. The diffusion of -ize in BrE is explained as a result of the decision of some reputed sources to adopt this spelling (the OED, The Times, among others). AusE and NZE, the former in particular, have consolidated as distinctive varieties of English with their own norms, mostly relying on BrE as a result of a national hostility to the American usage (Burridge, Reference Burridge and Kirkpatrick2010: 148), a fact which may have contributed to the diffusion of the BrE spelling. There is, however, a growing process of Americanisation driven by the younger speakers of the language (Peters, Reference Peters and Moore2001; Taylor, Reference Taylor, Collins and Blair1989), which is now spreading the use of -ize in these varieties.

Figure 5. The forms -ise and -ize with Latinate words (n.f.)

In view of these tendencies, the moot point is to discern whether the outer circle varieties have adopted the American- or the British-based model for the rendering of eyes-words. In the Asian varieties, the dilemma is clearly resolved for the adoption of the American rendering, although the phenomenon evolves at a different speed in each case. PhilE, as an American-based variety, is clearly pioneering the diffusion of -ize, followed by HKE in spite of the relatively recent presence of the British in Hong Kong, who returned to China in 1997. Finally, IndE and SingE present a higher distribution of -ise. IndE is elsewhere referred to as a conservative variety of English, often consistent to the BrE practice and impervious to many morpho-syntactic innovations (Calle–Martín & Romero–Barranco, Reference Calle–Martín and Romero–Barranco2014: 136; Seoane–Posse & Suárez–Gómez, Reference Seoane and Suárez–Gómez2013: 12). India was under the rule of the British Empire from 1765 until independence in 1947, a nearly 200-year period which resulted in its status as an associate official language in the country together with Hindi (Gargesh, Reference Gargesh, Kachru, Kachru and Nelson2006: 94). In contrast with the other Asian varieties, ‘the syntax of Indian English, as opposed to phonology and lexis, is said to conform most to standard British English’ (Saijala, Reference Saijala2009: 39), a fact which may also explain the high distribution of -ise. SingE stands out as a variety with a certain proneness to morpho-syntactic innovations (Seoane–Posse, Reference Seoane–Posse2017: 118), although with some spelling ties to the BrE standard associated with the government attention to language issues, which have encouraged teachers ‘to promote standard English in the classroom and ensure that their pupils develop the ability to communicate reasonably proficiently in this more formal variety’ (Deterding, Reference Deterding2007: 90). The promotion of the standard here has surely had greater impact on spelling than in the other levels of language.

In the African varieties, two tendencies are observed. SAE, on the one hand, is more prone to retain the BrE spelling (57.3%), although -ize closely follows (47.7%). The BrE imprint here is surely the outcome of the policy of Anglicisation by the British when they retook control of the territory from Dutch hands in 1806 lasting until 1910 when the Union of South Africa was formed giving English and Dutch the status of co-official languages (Kamwangamalu, Reference Kamwangamalu, Kachru, Kachru and Nelson2006: 159–160). As in the case of IndE, this 100-year period of Anglicisation has undoubtedly left a stronger imprint of the British standard than in the other African varieties. NigE and KenE, on the other hand, are more prone to the adoption of -ize, KenE in particular. Unlike South Africa, the European languages arrived relatively late both in West and East Africa as a consequence of the colonialist expansion, and today the English language is used there almost exclusively in a socio-educational context. English in East Africa is mostly based on the characteristic features of New Englishes, not directly transmitted through native-speaker settlers and, consequently, ‘the theoretical BrE norm in grammar is still upheld in books but rarely used or experienced in use in present-day East Africa’ (Schmied, Reference Schmied, Kachru, Kachru and Nelson2006: 191). This casts light on a freer attitude towards the British standard in these varieties, the incidence of -ize surely interpreted as the effect of the growing interest in American culture.

The arrival of English in Jamaica was the result of the British colonialist expansion throughout the West Caribbean in the 17^th century. Consequently, this variety of English is ‘more British-oriented, at least in [its] phonology, though in the last century American and Canadian influence can be expected and documented’ (Aceto, Reference Aceto, Kachru, Kachru and Nelson2006: 211). The relevance of -ise is negligible in JamE, representing approximately one third of the instances. The - ize form is favoured here in spite of the long-standing influence of BrE, surely as a result of the American-Canadian influence throughout the last century.

The incidence of -ize in these varieties is not exclusively explained in terms of their colonial history and current legislation. More decisive is the contribution of American imperialism and growth of pop-culture after World War II, raising ‘America to the height of political, economic, commercial, technological strength which saw the transformation of English from being a reserve of the British Isles and their queen, to a code of international linguistic transaction’ (Anchimbe, Reference Anchimbe2006: 3; Graddol, Reference Graddol1997: 9). In itself, this Americanisation of the English language ‘presupposes a greater or less submergence of regional and national varieties and cultures into a far greater and more powerful American-determined variety’ (Anchimbe, Reference Anchimbe2006: 9). This idea of submergence is what comes to light with the dissemination of -ize as a form on the rise even in the non-American-determined varieties.

Figure 6 presents the occurrence of -ise and -ize with words of Greek provenance in the 13 varieties analysed, where the figures have also been normalised for comparison. The etymological rationale is rejected in these varieties as they practically follow the tendencies found with Latinate words. In the inner circle, both AmE and CanE again show a natural predisposition to -ize while BrE, AusE and NZE are prone to the use of -ise. Still, -ize is not a residual form in the latter varieties insofar as it represents one fourth of the total of instances in BrE and AusE (5.1 and 4.9 occurrences, respectively) and one third in NZE (5.1 occurrences).

Figure 6. The forms -ise and -ize with Greek words (n.f.)

The Asian varieties present a similar state of affairs with a greater preference for -ize. As an American-based variety, PhilE again shows the highest distribution of this form (91.2%), followed by HKE (76.2%), IndE (65.5%) and SingE (62.3%). In the African varieties, SAE stands out as the only variety consistently following the British form with -ise while both NigE and KenE are more prone to the adoption of -ize. Finally, JamE also adopts the spelling -ize with words of Greek origin.

Figure 7 presents the frequency of -ise and -ize across the varieties and text types. With the necessary caveats, web pages are taken to represent more formal writing for general purposes, in most cases for commercial purposes, requiring a more elaborated and complex type of prose. Blogs, in turn, display a less formal instantaneous writing, often consisting of individual or group discussions on a particular topic, which therefore implies a less elaborated and simpler type of prose. The results show a slightly greater frequency of -ize in blogs than in web pages, at least in non-American-based varieties such as BrE, AusE and NZE. In the Asian varieties, this difference can be taken as an indication of the productivity and growing level of diffusion of this form. IndE, on the one hand, presents a higher incidence of -ize in the less formal type of writing, which confirms that it is well advanced in the process of adoption of the American form. SingE, on the other hand, exhibits a balanced distribution of -ize in both text types, showing a more constrained diffusion. In the light of this, the -ize form is now spreading more widely in IndE HKE than in SingE.

Figure 7. The forms -ise and -ize across text types (n.f.)

The African varieties show a similar state of affairs. While NigE and KenE present a more widespread use of -ize in blog material, SAE is again faithful to the BrE spelling in both types of texts. In itself, SAE stands out as the most conservative variety in the adoption of -ize and there are not clear symptoms that the American spelling will eventually displace the British form, at least in view of the positive attitude towards the latter. Finally, JamE follows the same trend with a higher frequency of -ize in blogs than in webpages.

As mentioned above, the diffusion of -ize is again confirmed as the result of an ongoing process of Americanisation in many of these varieties, which have progressively lost their original ties with Britain to develop a growing interest in American culture together with the adoption of the American linguistic variety. Different studies have investigated the level of Americanisation of some varieties of English worldwide, especially on the grounds of spelling and vocabulary (Awonusi, Reference Awonusi1994; Collins, Reference Collins, Peters, Collins and Smith2009; Fuchs, Reference Fuchs2017; Modiano, Reference Modiano1996; etc.). In the particular case at hand, it goes without saying that the BrE spelling has lost substantial ground in favour of the American form, BrE varieties included. The outstanding use of this spelling in blogs may be taken as a reliable evidence of its success worldwide. This material has been highlighted ‘as a text-type where American forms are preferred’ (Gonçalves et al., Reference Gonçalves, Loureiro–Porto, Ramasco and Sánchez2018: 2) and its online status stands out as a convincing argument in favour of the Americanisation of -ize in English worldwide.