What This Book Is About

Flavius Cerialis was the prefect of the Ninth Cohort of Batavians at the fortress of Vindolanda, in northern Britain, in the late 90s and early 100s AD. He was probably a Batavian noble, and necessarily of equestrian rank. We have some texts probably written in his own hand, including a draft of a letter (Tab. Vindol. 225), of which Reference AdamsAdams (1995: 129) has observed that ‘[i]ts orthography is consistently correct, and it has two types of old-fashioned spelling (the etymologically correct -ss- in occassio, twice, and saluom)’. On the basis of this and other evidence, he concludes that Cerialis’ father was probably made a Roman citizen for loyalty to Rome and that Cerialis received a formal education in upper-class Roman literary culture.

But ‘old-fashioned’ spelling is by no means restricted to texts written by the highly educated upper class.Footnote ¹ As Reference AdamsAdams notes (1995: 130–1), examples can also be found in the writing of the scribes of Vindolanda, showing that education received by these professionals, whose spelling is generally highly standard, had apparently included such features. And in fact, even in a text whose spelling is aberrant enough to give ‘support to the idea that [the writer] may have been a civilian trader without access to military scribes’ (Reference AdamsAdams 1995: 130–1, on Tab. Vindol. 343, letter from Octavius), there is evidence that this writer too had been taught to use ‘old-fashioned’ spellings (although not always correctly).

It turns out that Octavius is by no means the only writer who combines substandard and ‘old-fashioned’ spelling: we will see examples from, among other places and times, first century AD Pompeii, second century AD Egypt and fourth century AD Britain. This – along with other types of what I shall call ‘optional’ spelling features – provides a unique, and unexpected, insight into the kind of education that was received by those who did not belong to the highest stratum of society, predominantly in the first to fourth centuries AD.Footnote ² As will be discussed below, our direct access to knowledge about sub-elite education, in the form of information provided by ancient authors, is very limited. Consequently, if we want to find out about this important question in the study of Roman society, we must take indirect approaches.Footnote ³

By ‘optional’ spellings I mean those which were available for writers educated in the standard orthography of the day to use (and were hence not considered incorrect), but whose absence would not have led the educated to consider their writer to be un- or under-educated. In addition, they are non-intuitive, that is they will not be produced by a writer who has simply learned a basic mapping of individual letters to sounds.

In this book I will consider two categories of optional spelling: ‘old-fashioned’ features (on the definition of which see pp. 10–15) and diacritics used to mark vowels and glides in the form of the apex and i-longa. To do this, I use a range of corpora whose writers can be assumed, in the main, to belong to the sub-elite, even though certain of the texts in some of them may have been written by those who belong to the higher echelons of society (e.g. the equestrian prefect Cerialis at Vindolanda); as far as our knowledge allows, I will take the background of the writers into account. In addition to this primary purpose, a secondary, but not unimportant, aim is to contribute to the understanding of the development of Roman orthography – and in some cases also sound change – more generally, in order to be useful for both epigraphists and linguists.

Sub-elite Education in Literacy

The question of the extent and type of literacy in the ancient world is a perennial one and is difficult to answer. Reference HarrisHarris (1989: 259–73) estimates levels of literacy under the Roman empire to be no greater than 15% in Italy and 5–10% of the population in the Western provinces. These figures are problematic in a number of ways and are really only ‘guesstimates’. More important is his emphasis on the great variation in literacy across the empire, which was affected by a large number of factors, including social class (including slave vs free), wealth, occupation, gender, geography (e.g. location in the empire, rural vs urban, local infrastructure), linguistic background and many others.Footnote ⁴

There is also the issue of how to define literacy, which is hard enough to establish in the modern day: clearly most male members of the elite had received an education which rendered them capable of reading and writing highly complex literary works, but on the evidence available to us it is often difficult to know whether, for example, a craftsman who could write his name could do only this or much more. However, what is clear is that literacy, while not wide, could be deep, in the sense that certain members of the sub-elite were often literate and could read and write to a fairly high level. We have plenty of evidence for slaves of the elite acting as secretaries and reading-machines for their masters, for instance, but there are many other occupations, both among slave and (sub-elite) free, where literacy is attested or implied, and the written word was pervasive, even if it was not a great impediment to be illiterate (Reference HarrisHarris 1989: 196–233; Reference WilliWilli 2021: 14–19). For example, the majority of those carrying out business with the financier family of the Sulpicii in Puteoli, and in the similar tablets from Herculaneum (on which, see pp. 28–31), were literate, since they were able to write out a contract in their own hand (about a fifth of these chirographa were written by someone else; Reference CamodecaCamodeca 2017b: 24). Both they and the scribes who wrote the rest of these documents achieved a largely standard orthography (for some exceptions, see p. 262).

However, what is lacking is much evidence for the educational system by which those in the sub-elite learnt to read and write. Reference Bloomer, Grubbs and ParkinBloomer (2013: 451), for example, tells us:

What Bloomer does not specify is that this describes the educational career of a child who was a member of the elite. Works by writers like Quintilian, on whom Bloomer is leaning here, were written by the elite for the elite; they were not interested in describing the education of the sub-elite: as Reference Sigismund-Nielsen, Grubbs and ParkinSigismund-Nielsen (2013: 289) says, ‘[w]e meet freeborn children from the lower classes very infrequently in our sources. They were simply not interesting enough’.Footnote ⁵

Nonetheless, as we have established, it is clear that literate education could be available to the sub-elite (see also Reference Mullen and BowmanMullen and Bowman 2021: 61). Slaves could be taught in a paedagogium in their owner’s villa; slaves of the imperial household were taught in the Paedagogium on the Palatine – presumably often to a high level (Reference Sigismund-Nielsen, Grubbs and ParkinSigismund-Nielsen 2013: 296; although note the scepticism of Reference HarrisHarris 1989: 247–8). Scribes, whose work will form much of the data used in this study, clearly were educated in some fashion – and, as I shall show, in a fashion that in some respects at least was different from that of non-scribes – but we know very little about how they were trained (Reference MorganMorgan 1998: 32).

Likewise, there appears to have been some literacy education that took place in the army, perhaps for scribal purposes (see pp. 273–6), perhaps for soldiers more generally;Footnote ⁶ the tendency for letters written from and to Vindolanda to end with a greeting in a different hand from that which writes most of the letter suggests that some level of literacy among non-scribes was not uncommon. Reference HarrisHarris (1989: 253–5) suggests that literacy was much higher among legionaries than auxiliaries, but at least some auxiliaries could write, as demonstrated by the letters of Chrauttius from Vindolanda (Tab. Vindol. 264 and 310). Chrauttius will have been a Batavian or Tungrian auxiliary, and probably learnt a non-standard version of Latin in the army, perhaps showing some influence from his first language (Reference AdamsAdams 1995: 129–30), but is capable of writing a greeting formula.

Similarly, Reference Bowman and ThomasBowman and Thomas (1994: 74) suggest that the military reports with the heading ‘renuntium’ were written by the optiones making the report themselves, on the basis of the different hands of the writers. Reference AdamsAdams (1995: 102–3) has argued that the appearance of debunt ‘they ought’ in place of standard debent suggests that the exemplar on which these reports are based was also written by a non-scribe (perhaps also one of the optiones). He observes that this provides evidence of different degrees of education among the writers at Vindolanda: ‘[t]he renuntia thus give us an intriguing glimpse of a social class (probably that of the optiones) who regularly used the substandard form debunt, yet were literate’ (Reference AdamsAdams 1995: 131; emphasis in the original).

Further evidence of education at Vindolanda comes from lines of Virgil, possibly Catullus, and pseudo-Virgil (Tab. Vindol. 118, 119, 854 and 856), which were presumably produced for writing practice, although the ‘literary’ hands used are different from the usual scribal scripts. The editors suggest that 118 may have been the output of children of the prefect Flavius Cerialis, although there is no evidence of such a connection for 854 and 856.Footnote ⁷

It might be assumed that the education undergone by sub-elite members of society largely followed the same pattern as that described by Bloomer above, except that education stopped at some earlier point in the process – exactly at which stage might depend on the resources and aims of the child’s parents, on what teaching was available or other factors. To some extent, this is probably true; in the context of learning Greek in Hellenistic and Roman Egypt, Reference MorganMorgan (1998: 56–7) finds that papyri and other writing materials containing learning exercises and school texts have a different geographical distribution, with letters and alphabets widely scattered, including in villages, as are wordlists and literature, while scholia, rhetorical exercises and grammatical texts are far more restricted, especially to more urban areas:Footnote ⁸ she concludes, ‘[i]t looks rather as though the number of people in Upper Egypt whose education progressed as far as learning grammar and rhetoric was a very small proportion of those who acquired some basic literacy and read some literature’ (Reference MorganMorgan 1998: 57).

However, we should be careful of making too many assumptions along these lines: even if we assume the Egyptian situation is representative of learning elsewhere in the empire, we can seldom identify any clues about the social background of those using these materials, so it is possible that they still largely reflect the education of a fairly small elite.Footnote ⁹ Moreover, Reference MorganMorgan (1998: 67–73) has emphasised, again on the basis of the Egyptian Greek material, that the process by which children were educated was less a curriculum whereby everyone studied the same thing, but some people dropped out earlier than others, but rather a system involving certain ‘core’ exercises and texts, and a much wider ‘periphery’ whose contents were heterogeneous and depended on the choice of the teacher (and presumably other factors, such as access to texts). Morgan includes in the ‘core’ the kind of basic literate education that to some extent this book focusses on:

It makes sense that learning to read and write should be at the core for everyone, since very basic literacy is perhaps open to less variation than other types of education.Footnote ¹⁰ But the periphery might have been very different from what Morgan finds in Egypt, depending on the kind of use that literacy was to be put to. For example, shorthand, which is used in a number of texts at Vindolanda (Tab. Vindol. 122–6), was presumably not part of the standard educational system but was a speciality of those who were being educated as scribes or secretaries. And even at the level of the core, some variation existed: as already noted, syllabaries seem to be used less than other learning materials, in Egypt at least. And Quintilian (Institutio oratoria 1.1.26–32) mentions various approaches to learning to read and write of which he approves (ivory letter shapes to play with) or disapproves (learning the names and order of the letters before their shape; putting off the most difficult syllables; haste in moving on to pronouncing words and sentences).

As we shall see, there might also be variation as to what spellings a teacher might favour: they could be conservative or innovatory. The teachers themselves might also be of higher or lower literacy levels, have access to more or fewer resources, or even make greater or lesser effort. It is not unreasonable to suppose that, on the whole, the cheaper the teacher, the less they might have to impart and the less enthusiasm they might have to impart it. The contexts in which literate education took place might also have varied significantly; both ancient and modern discussions assume that it is children who are learning to read and write, but again, the Vindolanda tablets might provide an exception, if we assume that soldiers like Chrauttius learned to write only in the army, and hence presumably as adults. Another example of learning taking place after childhood comes from the tablets of the Sulpicii of Puteoli, where the Claudian letter Ⅎ in place of <u> for /w/ is found in TPSulp. 5 (by a scribe), 27 (non-scribe), 32 (scribe), 48 (both scribe and non-scribe), 77 (non-scribe) and 101 (scribe), mostly in the names of the consuls Vitellius and Vipstanus, but once in uenalium ‘for sale’ (77) and once in uadimonium (5). All of these tablets are from AD 48 (apart from 5, which is undated) and reflect the introduction of the Emperor Claudius’ new letters (on which, see Reference OliverOliver 1949). Clearly, some scribes and non-scribes alike had heard about and adopted this new letter, at least for the formal context of consular dating. Scribes in the army and elsewhere might also have received additional training, on top of whatever literacy skills they arrived with. And much of this assumes some kind of formal education, with a paid teacher: some may have learnt informally, from their parents, friends or peers, in which case the process might have been quite different, and presumably less systematic.

It does seem likely that scribes must have received some kind of education for their role. The fact that the spelling of the scribes in the tablets from Pompeii, Herculaneum and Vindolanda contains so few substandard features in itself implies a certain degree of homogenisation amongst these groups, which might be due to specific education (perhaps in the form of top-up training). However, it is also possible that people who became scribes were more likely to have already received a high-quality education in standard spelling. As we shall see, the enquiries in Parts I and II will reveal other ways in which scribes show homogenisation in spelling that implies a process of education specifically for scribes.

As important as the question of what was taught, and how, in different contexts is how well it is taught (or, since teaching is at least a two-person process, learnt). That is, the difference in educational content when it comes to spelling might not be very great between those who are well taught and those who are badly taught; the major difference might be their ability to use what they have been taught consistently according to the canons of the elite standard. For example, two writers may both have learnt that the digraph <ae> is used to spell certain words which contain the vowel /ɛː/; one of them consistently remembers which words contain <ae>, while the other remembers only some words or only remembers some of the time, and the rest of the time uses <e>, or hypercorrects by using <ae> in words for which the standard spelling requires <e>. In this case, the two writers have received the same educational content (existence and use of <ae>) but not the same quality of education. One could imagine yet another writer whose education has been so basic that they were simply taught the names and values of the individual letters corresponding to the sounds in their idiolect; this writer would therefore never have learnt the existence of <ae> and will always write <e> for /ɛː/.Footnote ¹¹ Here this has been a difference in content as well as quality.

This distinction allows us to be more precise in our examination of whether the content of the orthographic education which was received by elite or sub-elite, or standard and non-standard spellers, was much the same, or not. If it was not, ‘old-fashioned’ or otherwise non-intuitive features such as apices and i-longa will appear only in the writing of elite or standard spellers; if it was, we should expect to find old-fashioned spellings in both elite and sub-elite writings, by both standard and substandard spellers: any differences in the orthography of these categories will then be ones of consistency or correctness, reflecting quality of education rather than content.

Defining ‘Old-fashioned’ Spelling

A good example of the complicated issues involved in identifying ‘old-fashioned’ spellings is the letter of Suneros (CEL 10), from Oxyrhynchus, dated to the Augustan period. We find the following features:

• <ei> for /i/ < /iː/ by iambic shortening in tibei for tibi ‘to you’, and for /i/ in uocareis for uocāris ‘you will have called’ (in error, since the /i/ in the final syllable was never long, but presumably due to confusion with the perfect subjunctive uocārīs).

• <e> for /iː/ (deuom for dīuum ‘of the gods’).

• <xs> for <x>: adduxsit for addūxit ‘(s)he brought’, Oxsyrychitem for Oxyrhynchitem, maxsuma for maxima ‘greatest’.

• <u> for /i/ before a labial: maxsuma for maxima ‘greatest’.

• <q> for /k/ before <u>: qum for cum ‘when’.

• <uo> for /wu/: uolt for uult ‘wants’, deuom for dīuum ‘of the gods’.

The editor Cugusi describes <ei> as a ‘sign of antiquity’ (‘segno di antichità’) and <uo> in uolt as an ‘archaising spelling’ (‘grafia archaizzante’) but for deuom notes that the ending ‘-om continued in use more or less to the end of the Republic’ (‘-om ci porta pressappoco alla fine della Repubblica’), describes <q> as ‘probably already in this period a “scholarly” spelling’ (‘probabilmente già in questo periodo grafia “scolastica”’), does not consider <xs> old-fashioned, and does not comment on <u> in maxsuma. He sees <e> in deuom as due to a confusion between /eː/ and /iː/ found in inscriptions (for slightly more clarity here, see also Reference CugusiCugusi 1973: 667). Reference AdamsAdams (2016: 208–9) says that Suneros ‘uses the old spelling tibei, and deuom is archaising on two counts.Footnote ¹² Vocareis is a false use of orthographic archaism’. However, on the whole he takes a nuanced approach, emphasising that use of <u> continued into the first century AD, and that ‘[s]pelling reform does not take place overnight, and personal preference was a factor’ (Reference AdamsAdams 2016: 205).

As I shall show in the rest of this book, the ‘old-fashioned’ nature of these spellings is overemphasised by Cugusi and even by the more careful Adams. Even leaving aside the question of when in the Augustan period the letter was written (since habits probably changed a fair amount between 31 BC and AD 14), most of these spellings were still in fairly common usage at that time. This is the case for <ei> for /iː/ (Chapter 3); /wɔ/ had become /wu/ probably not much earlier in the century, and <uo> was still the more common spelling, although a move to use <uu> in official orthography does seem to be visible (Chapter 8). The use of <u> for <i> in maxsuma is still found in the first century AD in high-register texts but was perhaps moving out of the standard (Chapter 6). As for <xs> (Chapter 14) and <q> before /u/ (Chapter 12), these were always minority usages in the Republic but were beginning to lose favour in official orthography of the time; their use was perhaps not particularly striking in the context of an informal letter. The use of <e> in deuom may reasonably be considered old-fashioned (Chapter 3). On the whole, Suneros was perhaps a conservative speller (as well as a substandard one),Footnote ¹³ but most of his spellings are not archaic for the time.

In fact, despite the frequency with which terms such as ‘archaising’ or ‘old-fashioned’ are used, as with the letter of Suneros, to describe the spelling of a given document, defining them is surprisingly difficult, not least because there are several ways in which a spelling could be said to fall into such a category. The least useful definition is that which compares the spelling of a text with the kind of idealised, abstract and anachronistic notion of ‘standard’ Latin spelling that appears in editions of Latin literary texts and often also in inscriptional texts (including in the large online corpora).

Where the spelling feature in question represents a phoneme or series of phonemes which have undergone change in the history of Latin, it is possible to define ‘old-fashioned’ against this sound change. Thus, for example, once the diphthong /ɛi/ developed to /eː/ around the middle of the third century BC, the spelling <ei> could be considered ‘old-fashioned’ relative to <e>, and likewise once /eː/ had become /iː/ around the middle of the second century, the use of both <ei> and <e> could be considered old-fashioned relative to <i>.

Such a definition proves rather unhelpful, in a number of ways. In the first place, it ignores the fact that spelling takes time to catch up to phonological change (if it ever does): as we shall see, the use of <ei> and <e> demonstrably continued long after the relevant sound changes took place, but only after a certain time will their use have been considered old-fashioned. Secondly, under this sound-change-based definition of ‘old-fashioned’, it would be necessary to consider the continued use of <ae> to represent /ɛː/, the result of monophthongisation of the diphthong /ae̯/, old-fashioned from the point at which the sound change first took place.Footnote ¹⁴ But this definition fails to identify the difference, at least under the empire, between (correct) use of <ae>, which was simply the standard spelling – and whose absence would mark out the writer as undereducated – and use of <ei>, whose absence would not have the same effect (for more on this, see 50–57). Lastly, many of the features called ‘old-fashioned’ are not the result of sound changes, for example use of <xs> beside <x>, or <k> and <q> beside <c> to represent /k/. Identification of ‘old-fashioned’ spellings only with regard to sound change will therefore not help us with these cases.

So the definition of ‘old-fashioned’ needs to be usage based: spellings are ‘old-fashioned’ when they are no longer part of the core repertoire of standard orthography. This is not to say that they are necessarily substandard; simply that their use is not necessary for a writer’s orthography to be accepted as hewing to the educated standard (on the use of the terms ‘standard’ and ‘substandard’, see pp. 15–18). Depending on the context, and on the status of individual features, their usage may have made the writer seem to readers to be highly educated, conservative, idiosyncratic, fuddy-duddy or any number of other effects which it is difficult or impossible for us to pin down.

Taking these issues into account, I adopt three methods of assessing whether spellings were ‘old-fashioned’. Firstly, I have considered evidence from the writers on language whose works are dated more or less to the period being considered in this book (i.e. from the first to the fourth centuries AD). For more information on these authors, see pp. 37–9. These authors sometimes tell us (relatively) explicitly of their view on the currency of a particular spelling; even when they do not directly provide us with this information, the very fact that they mention something suggests that its existence was relevant to their intended audience.

There are, however, several issues that we must be careful of when dealing with this kind of information. The tradition of writing about language was tralaticious; its audience was also primarily interested in the reading of works of literature often dating from several centuries previously. The effect of these characteristics might be to foster mention of orthography which was very highly archaic by the time of the writer – perhaps not at all or barely used by anyone at the time.

Furthermore, different writers might have different attitudes towards orthography, preferring either more modern or more old-fashioned spellings, which may lead us to misanalyse the ‘old-fashionedness’ of particular orthography. In some cases the authors tell us explicitly about their approach; for example, Quintilian paints himself as (somewhat) in favour of more modern spelling, at least insofar as this reflects contemporary speech:

Lastly, it must be remembered that, while the writers on language had different backgrounds and audiences, they were writing for the (children of) the elite, or their teachers (or at least for those who had the resources to aim for elite status for their children). Consequently, their claims about the status of particular spellings are not necessarily likely to represent the habits of the sub-elite. Indeed, as Reference Biddau, Ferri and ZagoBiddau (2016: 51–2) observes, we cannot be sure that anyone at all shared a writer’s view.

The second method is to investigate the spelling of ‘official’ inscriptions emanating from and concerning the imperial (and perhaps local) elite and legal system, or of other texts known to be written by members to the social elite, on the basis that this represents standard orthography, and that spellings which have fallen out of use here are consequently ‘old-fashioned’. As with looking to the statements of the writers of language, this has the effect of privileging the elite over the sub-elite in defining the standard (although this is not in itself necessarily incorrect, since the development and imposition of a standard orthography is often a top-down process).Footnote ¹⁵ It is also less easy than it sounds, since there was not one genre of ‘official’ inscriptions: it is often claimed, for example, that legal texts continued certain spellings for longer than other kinds of texts. Should we, then, discount legal texts? And what, if any, other distinctions should be made? While these are interesting questions, they are not the focus of the present investigation, nor is there the space to do a thorough investigation of ‘official’ orthography across the whole of the chronological span that I am considering. Consequently, I do not draw distinctions between the different types of ‘official’ inscriptions, and I only treat them separately from the overall picture of Roman epigraphy as a whole mentioned in the following paragraph when it seems particularly relevant.

My final method of identifying ‘old-fashionedness’ in spellings has been to carry out searches among the massive epigraphic corpus collected as the EDCS, which allow me to roughly identify the frequency of a particular orthographic feature. For most features I have only carried out searches for the whole period of the first to fourth centuries AD, but where relevant, I have searched more precisely within that range to give an idea of changing usage over the period.

The information that these three methods provide on the status of a given spelling will not necessarily be consistent – just as a reader or writer’s view of how ‘old-fashioned’ a feature was may have varied depending on their education, social background or even personality, as well as the genre or register of text. This being the case, it is perhaps unsurprising that one result of the research carried out in this book will be, among other things, to cast doubt on the idea that there was a single, easily identifiable category of ‘old-fashioned’ spelling: this is not to say that use of a given orthographic feature was never seen as ‘old-fashioned’: simply that the point at which a spelling became ‘old-fashioned’ (if it ever did), varied across time as well as according to other variables. Nonetheless, I will follow common (modern) practice in continuing to refer to the particular type of optional spelling practice discussed here as ‘old-fashioned’ for now, while highlighting the problems and inconsistencies in using this term – and dropping the scare quotes. I will return to the question of the use of the term ‘old-fashioned’ in the conclusion (258–61).

Standard and Non-standard Spelling

In this book I will often talk about standard and substandard spelling (as well as old-fashioned spelling); indeed, one of my main claims is that old-fashioned spelling can be found in the writing of substandard spellers as well as those whose spelling is otherwise standard. The idea that Latin of the Classical period had undergone a process of language standardisation is widely held (see e.g. Reference RosénRosén 1999; Reference Versteegh, Adams, Janse and SwainVersteegh 2002; Reference AdamsAdams 2007: 13–17; Reference Clackson and HorrocksClackson and Horrocks 2011: 77–288; for an overview of bibliography, Reference NikitinaNikitina 2015: 3–9; and on the teaching of the standardised language in the later empire, Reference FosterFoster 2019).

However, recently scholars have emphasised the ways in which standardisation of spelling was not complete in the first century BC, and that orthographic variation continued to be found even in the writing of the highly educated elite, and in official inscriptions at least into the first century AD; see, for example, Reference ClacksonClackson (2015); Reference NikitinaNikitina (2015); Reference AdamsAdams (in press). Reference ClacksonClackson (2015) rightly emphasises both that the continued discussion of spelling variants in writers on language can be evidence of diversity in orthography and that there is still variation in spelling in official inscriptions of the first centuries BC and AD (based on the findings of Reference Fischer and CallebatFischer 1995). He concludes that

However, the fact that some variation was acceptable in spelling does not necessarily mean that there was no such thing as standard orthography at all. The types of spellings that Clackson and other scholars have focussed on as showing variation in the first centuries BC and AD are precisely those that are usually called ‘archaic’ or ‘old-fashioned’: Clackson mentions the spelling of pecūnia ‘money’ as pequnia, <u> for <i> before labial consonants, particularly in superlatives and ordinals, <uo> for <uu>, and use of <ii> to represent /jj/ in words like maior and eius. These are spellings which are often traced back to older writers, and indeed often associated with particular named great men of literature or politics. To some extent, and with variation depending on exactly how outmoded the older spelling came to seem, either the newer or the older spelling was acceptable: as Quintilian says, ‘on these matters, the school teacher should use his own judgement: for this ought to have the greatest weight’ ([i]udicium autem suum grammaticus interponat his omnibus: nam hoc ualere plurimum debet, Quintilian, Institutio oratoria 7.30). These are separate from ‘barbarisms’, which include ‘that vice known as a barbarism, of which examples are found everywhere, anyone can easily come up with instances of for himself, such as adding to any word a letter or a syllable or taking one away or replacing one with another or putting the right letter in the wrong place’ (illud uitium barbarismi cuius exempla uulgo sunt plurima, sibi etiam quisque fingere potest, ut uerbo, cui libebit, adiciat litteram syllabamue uel detrahat aut aliam pro alia aut eandem alio, quam rectum est, loco ponat, Quintilian, Institutio oratoria 5.10).

This kind of spelling is not acceptable: it is a uitium.Footnote ¹⁶ And it is this kind of spelling I will refer to as substandard: things like <b> for /w/, <e> for /ae̯/ or for short /i/, or absence of <h> in certain lexemes. Quintilian does not explicitly mention features like these at all, because it would be unthinkable for someone teaching pupils of the social level which Quintilian is writing about to introduce them as acceptable.Footnote ¹⁷ But subsequent scholars would do so, sometimes with great enthusiasm (as in the list of corrections known as the Appendix Probi; Reference PowellPowell 2007). These spellings reflect sound changes which have taken place in the Latin of their users (and often quite possibly of even elite speakers), but have not been admitted into the orthography of educated writers, and instead mark out their users as ill- or under-educated. Only very rarely do we find them in official inscriptions or those commissioned by the elite. Nor do we find them very much in some of the sub-elite texts which we will be looking at here, notably the tablets at Vindolanda, Pompeii and Herculaneum. This is particularly true of those texts and sections of texts which are written by scribes, who had presumably received an education which encouraged such lack of variation (for more on this, see pp. 271–6), but even in most texts which seem to have been written by non-scribes, differences in orthography are relatively minor, which makes texts which do diverge significantly from this norm, such as those of Octavius and Florus at Vindolanda (Tab. Vindol. 343, 643 respectively), and C. Novius Eunus in the tablets of the Sulpicii (TPSulp. 51, 52, 67, 68), all the more striking.

Structure of the Book

After the Introduction, the book consists of two parts, followed by a conclusion. The first part, consisting of Chapters 2–17, deals with old-fashioned spellings, starting with spelling involving vowels and moving on to consonants. In each case, I outline the circumstances that led to the spelling under discussion becoming old-fashioned and provide some context for the use of this spelling in the epigraphic record more generally, mostly in the form of investigation carried out by means of searches of the EDCS. I have not taken a uniform approach in this, both because different spellings require different focus on aspects of their use and because the nature of the database means that different types of search were possible for different spellings. Where relevant, I have also discussed what Roman writers on language had to say about a particular usage. I then provide data on, and discussion of, the usage of each particular spelling in the sub-elite corpora.

The second part, consisting of Chapters 18–24, focusses on two types of what might be considered diacritics: apices and i-longa, primarily in the Isola Sacra funerary inscriptions, the Vindolanda tablets and the tablets of the Sulpicii, since in these corpora their use seems to be restricted largely to stonemasons and scribes, and therefore provides insight into specifically professional writing traditions and education.

In the conclusion, I summarise my findings on these optional spelling features, focussing on three areas: what I have learnt about the use of old-fashioned spellings over time and in different social or geographic contexts; what this investigation tells us about education and social level, and in particular what we can learn about the education that scribes received; and what my research has revealed about sound change in Latin.

Methodology

The Sub-elite Corpora

In this book I use corpora as a way of indirectly assessing the type of education received by the members of the sub-elite in the Roman empire. The sub-elite can be defined straightforwardly as all inhabitants of the empire who were not ‘senators, the equestrian class, and the local governing class’ (Reference TonerToner 2009: 3). Such a definition of course obscures many important distinctions among the several tens of millions who occupied this position, including slave vs free (and vs freed), ‘affairistes’ vs ‘average workers’ vs labourers, women and men, urban vs rural, etc. (see e.g. Reference TonerToner 2009: 1–5; Reference Courrier and GrigCourrier 2017; Reference Grig and GrigGrig 2017: 18–21). Unfortunately, we cannot investigate the orthography of the texts in the corpora with as much granularity as we would like, but the corpora do allow some distinctions to be drawn, particularly scribes vs non-scribes (especially in the tablets of the Sulpicii and of Caecilius Jucundus from Pompeii, and at Vindolanda), and military vs non-military writers (in the form of the texts from the army at Vindolanda, Vindonissa, Dura Europos, Bu Njem, and the Claudius Tiberianus letters as opposed to the other texts).

I have chosen corpora in which it is a reasonable assumption that the majority of writers did not belong to the elite of the empire. In most of them we have a fair idea about the social status of the writers; in addition to scribes and soldiers, identifiable writers of the texts include slaves, contractors and praetorian guards, although there is seldom enough evidence to allow strong conclusions to be drawn about the type of education which these categories of person received. The major exception to this is the curse tablets, which often provide very little or no information about the writer of the text (who may or may not be different from the author). It is often implied that these are by definition written by members of the sub-elite: ‘they come to us largely unmediated by external filters; unlike ancient literary texts, they are devoid of the distortions introduced by factors such as education, social class or status, and literary genres and traditions’ (Reference GagerGager 1992: v).Footnote ²⁹ However, as already noted (pp. 19–20), this is not necessarily the case.

We might assume that tablets with multiple substandard features were not written by a member of the elite, but, just as a high level of orthographic education does not imply a high social position, lack of success in achieving a high level of education does not necessarily rule out membership of the elite (at least as far as writing goes). However, there were relatively low barriers to entry in the creation of these texts, which were not, after all, intended to be read by the public. So where the author and writer were identical, there is at any rate no reason to assume that curse tablets were created more by the elite than the sub-elite; and professional writers of curses would have belonged to the sub-elite.

These ‘sub-elite’ corpora, then, predominantly include texts produced by sub-elite writers. But they do not necessarily only include texts written by the sub-elite. The Vindolanda tablets include letters, (parts of) which are written by prefects of the cohort and their wives, who probably belonged to the equestrian ranks; the corpus of letters includes a number which emanate from the upper levels of the civilian or military bureaucracy; and, as noted above, there is no reason why we should rule out the elite as authors or writers of curse tablets, for example. Where we can identify members of the elite within the corpora, their usage provides a useful point of comparison with the conclusions drawn on the basis of the other texts.

My choice of corpora is obviously skewed by what has survived the vicissitudes of history, as well as what was available to me in good editions (somewhat more restricted than usual during the pandemic, due to lack of library access for part of the time). Within these constraints, I have used corpora which are internally coherent in terms of place, time, social circumstance or genre, while providing coverage of the geographical and chronological extent of the Roman empire.

Writers on Language

I give here a very brief overview of the Roman writings on language referred to in this book.Footnote ³⁸ I have looked for material in the major writers of the first to fourth centuries AD. In the discussion of <k> and <q> I have also included some relevant later writers, mentioned by Reference LindsayLindsay (1894: 6–7), without carrying out a thorough search. For more information on all except Quintilian, see the editions cited, and the useful summaries in Reference ZetzelZetzel (2018: 96–8, 231, 279–329). Where a text is quoted from another edition, but is also included in Grammatici Latini (GL; Reference KeilKeil 1855–80), I also provide a reference to GL. Texts are quoted as printed in the editions, except that I have changed v to u throughout, replaced capital letters at the start of sentences with lower case and used double quotes (“ ”) in place of various equivalent national conventions. I have made no attempt to make them consistent in other ways. In the translations I have not used brackets and slashes to distinguish phonemes from graphemes, since these are not concepts kept distinct by the authors. All translations are my own, except where noted.

• L. Caesellius Vindex. Second century AD. His writings are known only from excerpts in Cassiodorus’ sixth century AD De orthographia, of which I use the text of Reference StoppacciStoppacci (2010), under the names Caesellius and L. Caecilius Vindex. To what extent either of these was actually written by Caesellius Vindex (see the contrasting comments of Reference StoppacciStoppacci 2010: clxvi–vii and Reference ZetzelZetzel 2018: 288), and whether they were altered before inclusion by Cassiodorus are uncertain.Footnote ³⁹

• Flavius Sosipater Charisius, Ars grammatica. Early 360s AD. Reference BarwickBarwick (1964).Footnote ⁴⁰

• Cledonius, Ars grammatica. Fifth century AD. GL (5.1–79).

• L. Annaeus Cornutus. AD 20–65. His De enuntiatione uel orthographia is known only from excerpts in Cassiodorus’ sixth century AD De orthographia, of which I use the text of Reference StoppacciStoppacci (2010). These were not necessarily unchanged by Cassiodorus. An English translation is found in Reference Boys-StonesBoys-Stones (2018: 142–55), but he uses Keil’s (GL 7.147–54) text, which is sometimes quite different.

• Curtius Valerianus. His work is known only from excerpts in Cassiodorus’ sixth century AD De orthographia, of which I use the text of Reference StoppacciStoppacci (2010). These were not necessarily unchanged by Cassiodorus. His date is uncertain (after the second century AD according to Reference ZetzelZetzel 2018: 288; fifth century AD according to Reference StoppacciStoppacci 2010: cxxxix).

• Diomedes, Ars grammatica. Late fourth century AD. GL (1.297–529).

• Donatus, Ars grammatica maior. Floruit mid-fourth century AD. Reference HoltzHoltz (1981).

• Dositheus, Ars grammatica. Probably late fourth century AD. Reference BonnetBonnet (2005).

• Pompeius Festus, De significatu uerborum. Festus’ lexicon was produced in the late second century AD, but based on, and abbreviated from, the work of Verrius Flaccus (c. 55 BC–AD 20). The extent to which the text of Festus reflects that of Verrius is debated (see Reference Glinister, Glinister and WoodsGlinister 2007: 11–12). Only a small fragment remains (Fest.), and the work is otherwise known only through an epitome (Paul. Fest.) made by Paulus in the eighth century. Reference LindsayLindsay (1913).

• Ps-Probus, De catholicis. Early fourth century AD, or later. GL (4.1–43).

• Ps-Probus, Instituta artium. Some parts written around AD 305, although not necessarily in the version preserved. GL (4.47–192).

• M. Fabius Quintilianus (Quintilian), Institutio oratoria. Written in the 90s AD; Quintilian lived from c. AD 35 to c. AD 100. Reference AxAx (2011), who largely follows Reference WinterbottomWinterbottom (1970).

• Servius, Commentarius in Artem Donati. Early fifth century AD. GL (4.405–48).

• Terentianus Maurus, De litteris, de syllabis, de metris. Late second or early third century AD. In verse. Reference CignoloCignolo (2002).

• Q. Terentius Scaurus, De orthographia. His floruit was during the reign of Hadrian (AD 117–138). Reference BiddauBiddau (2008).

• Velius Longus, De orthographia. Probably first third of the second century AD. Reference Di NapoliDi Napoli (2011).

• C. Marius Victorinus, Ars grammatica. Mid-fourth century AD. Reference MariottiMariotti (1967).

• Maximus Victorinus, Ars grammatica. Uncertain date: not before the last part of the fourth century AD. GL (6.187–215).

A Sketch of the Latin Vowel System through Time

Latin inherited from Proto-Italic a system consisting of five long and short vowels which are reconstructed as *i, *ī, *e, *ē, *a, *ā, *o, *ō, *u and *ū, as well as the diphthongs *ei̯, *ai̯, *oi̯, *au̯, ou̯ (and at end of word the long diphthongs *ōi̯, *āi̯ and perhaps *ēi̯). In the International Phonetic Alphabet (International Phonetic Association 1999), I take the following vowel phonemes to have existed in early Latin (Figures 1 and 2).Footnote ⁴¹

Figure 1 Early Latin vowels

Figure 2 Early Latin diphthongs

In the fourth century BC, the diphthong /ɔu/ monophthongised to /oː/, as did /ɔi/ in most contexts, briefly giving a three-way contrast among long vowels on the back axis.Footnote ⁴² This /oː/ was raised to /uː/ in the third century, falling together with inherited /uː/.Footnote ⁴³ Around the middle of the third century, the diphthong /ɛi/ underwent monophthongisation to close mid /eː/, briefly giving a three-way contrast among long vowels on the front axis. About a century later, this /eː/ was raised further to /iː/, thus falling together with inherited /iː/.Footnote ⁴⁴ In the second century, the off-glide of the remaining diphthongs /ai/ and /ɔi/ was lowered to /ae̯/ and /ɔe̯/. The effect of these changes was thus to restore the five-vowel long/short system at the cost of the diphthongs.

At some point long /ɛː/ and /ɔː/ underwent raising to /eː/ and /oː/,Footnote ⁴⁵ eventually falling together in most Romance variants with /i/ and /u/ (after the loss of contrastive vowel length in Romance, the date of which is disputed, but probably towards the fourth or fifth century AD).Footnote ⁴⁶ It is often supposed that /i/ was phonetically [ɪ], facilitating the merger with /eː/. From the first century AD onwards, there is already evidence of /i/ being spelt with <e>, originally probably reflecting a phonetic lowering to [e] in certain contexts.Footnote ⁴⁷ From the second century BC the diphthongs /ae̯/ and /oe̯/ were monophthongised, at least in non-Roman Latin; this was widespread across the empire in the first few centuries AD, although the diphthongs were perhaps maintained by elite speakers for longer. The latter developed to /eː/ on the basis of its reflexes in Romance (e.g. Italian pena /pena/ ‘punishment’),Footnote ⁴⁸ while the former developed to /ɛː/, falling together in the Romance languages with /ɛ/ rather than with the /e/ resulting from original /i/ and /ɛː/ < *ē.Footnote ⁴⁹ Presumably, therefore, original /ɛː/ had already been raised to [eː], and these changes led to phonologisation as /eː/, giving a threefold distinction in the front axis once again.

Since no new /ɔː/ was created to parallel new /ɛː/, it is not clear at what point original /ɔː/ was raised to [oː], but its phonologisation as /o(ː)/ must have been fairly late, since no confusion between original /ɔː/ and /u/ appears to have taken place in African Latin, nor in the Romance languages Sardinian or Romanian (in which /ɔ/ and /ɔː/ fell together). It is possible, therefore, that /o/ only arose as the reflex of original /ɔː/ [oː] and /u/ [o] after the loss of contrastive vowel length.Footnote ⁵⁰

New diphthongs had arisen in the course of the last few centuries BC in a very few words, including /ɛu/ created by apocope in forms like *sei̯u̯e > *sēu̯e > sēu > seu ‘either’, /ɛi/ by contraction across a syllable boundary in words like deinde ‘then’ from /dɛ.indɛ/, and /ui/ in cui, huic (Reference WeissWeiss 2020: 71–3).

For the first three or four centuries of the empire, I therefore assume the following vowel system (Figures 3 and 4);Footnote ⁵¹ /eː/ is the result of raised original /ɛː/ (< *ē) and monophthongised /oe̯/, and /ɛː/ is the result of monophthongised /ae̯/. Short /i/ was probably phonetically [ɪ], although it may have become [e] in final syllables (on which, see pp. 59–64); [o] for /u/ was probably somewhat later (and perhaps also originally only in final syllables or unstressed words).

Figure 3 Latin vowels

Figure 4 Latin diphthongs

An additional feature of the vowel system is the tendency to shortening of word-final vowels that occurred between the third century BC and perhaps the first century AD (or possibly even later). All long vowels followed by a single consonant other than /s/ underwent shortening at the beginning of the second century BC if they were in the final syllable of a polysyllabic word (Reference WeissWeiss 2020: 139–40). Iambic shortening also shortened word-final long vowels in words which originally were of iambic shape, such as bene ‘well’ < benē, ego ‘I’ < egō. In the early Classical period this was largely restricted to words which did not clearly form part of a productive paradigm, where long vowels were restored by analogy, and/or function words, which are particularly likely not to receive phrasal stress (Reference StephensStephens 1985; Reference Selkirk, Morgan and DemuthSelkirk 1996; Reference FortsonFortson 2008: 176–258). Even the first singular present verbal ending in /ɔː/ is sometimes found scanning short in first century BC poetry in iambic words, and by the end of the century word-final /ɔː/ seems to have been shortened in non-iambic words as well, being attested in Horace, Ovid, Propertius and later poets (Reference PlatnauerPlatnauer 1951: 50–2; Reference LeumannLeumann 1977: 110; Reference StephensStephens 1986; Reference MeiserMeiser 1998: 76–7; Reference WeissWeiss 2020: 138–9). It is unclear whether this was already a completed sound change or varied by phonetic or sociolinguistic context (or whether it is a poetic licence by analogy with the iambic forms); whether other word-final long vowels had also become shorter by then seems to be unknown.

By the third century AD, in discussing acceptable clausulae in oratory, the grammarian Sacerdos implies that at least among some speakers long vowels in final syllables (and not just absolute final vowels) had become shortened (GL 6. 494.7–12), although also demonstrating that the educated knew which vowels were supposed to be long.Footnote ⁵² I will assume here that isolated forms like ego, mihi and tibi had a short final syllable from the first century BC onwards, but that all other originally long final vowels, even in iambic word forms which are not paradigmatically isolated, were long. There is some evidence that, at least in some words which had undergone iambic shortening, knowledge of the original length of the vowel remained (e.g. scansion as nemō, cauē, mihī, tibī, spellings such as egó, tibei, tibe, tibì).

Apparently also by the first century BC, at least some vowels in some words had been lengthened before r followed by another consonant (Reference LeumannLeumann 1977: 114; Reference WeissWeiss 2020: 195). This is demonstrated by forms like aarmeis (AE 2008.473, first century BC) ‘weapons (abl.)’, and by reflexes of these vowels in Romance or in languages into which words containing this sequence were borrowed, as in Latin ōrdō ‘order’ > Logudorese órdene, borrowed into Welsh as urdd. However, not all instances of this sequence show lengthening (e.g. Italian fermo from firmus ‘firm’, not ^×fīrmus, notwithstanding Fìrmi, CIL 6.1248, AD 38–49). Whether the difference is due to social variation or the phonetics of this sequence, or some other factor, remains unclear.