1.1 Change affecting Latin in general

It is not being unfair to say that scholars have sometimes blindly assigned phenomena to Vulgar Latin that were well established in the literary language (as well as in lower varieties), for no better reason than that these phenomena influenced the Romance languages.

A typical example of a development affecting all levels of Latin that has suffered this fate was the loss by the ablative of the gerund of its instrumental force, such that it encroached on the nominative present participle (xxvii). This was a usage favoured by high stylists such as Livy and Tacitus and found also in e.g. Cicero and Virgil, as well as in the substandard letters of Terentianus. Many of the examples are ambiguous between an instrumental and a weakened sense, and that suggests that the participial (i.e. non-instrumental) function was developing in ambiguous contexts: change was in progress in the educated language, and it was not a matter of a stigmatised usage of lower sociolects borrowed from time to time by the educated.

Another such case is the anticausative/passive use of reflexive verbs, the distinction between which tends not to be satisfactorily made (xxvi). Some scholars have concentrated on low-register texts such as the Mulomedicina Chironis and constructed a narrative whereby supposed proto-Romance uses of the reflexive had their origin in the Vulgar Latin of late antiquity. But the same uses are found in the classical period and even earlier. We showed that there is a close parallelism between the Mulomedicina Chironis, of the fourth century, and the classiciser Celsus, three centuries earlier. Anticausative reflexives are as old as extant Latin itself, and found in literary sources. As for reflexives that are open to a passive interpretation, these turn up occasionally from an early period onwards and are not confined to low-register sources (xxvi.12).

The reflexive dative with transitive verbs (e.g. uolo, sumo, habeo) has tended to be marginalised and treated as colloquial or the like (xvi.2), but such uses were well established in literary language as well, often conveying a clear nuance (xvi.7).

The factor lying behind suppletion of the verb ‘go’ shows up in Latin in general and not merely in the types of low-register texts that might be taken to exemplify lower social dialects (xxxi). The loss of the ‘weak’ monosyllabic forms of ire occurred in the republican literary language at some time between Terence and Cicero. It is only if we turn a blind eye to Cicero, Livy and many other literary texts that we could possibly conclude that Vulgar Latin played a vital role in establishing the conditions that were to lead to suppletion.

The construction inf. + habeo with future meaning might qualify for inclusion in the category ‘change from above’, were it not for a stray example (without context) on an ostracon from Egypt (xxv.1.2). The construction can be observed extending its semantic range towards the expression of futurity in the educated language from the classical period through the second century to the fourth century and beyond. In the later period, when it often is open to a future meaning, it tends to be in learned texts of theologians and grammarians, and is frequently in conditional clauses in logical argument. Such contexts would not favour the idea that it was a stigmatised usage that had started life well down the social scale and was admitted from time to time by the educated. Either it developed first in educated varieties of the language, or, if we attach weight to the ostracon, it was spread across the social/educational spectrum.

The present indicative referring to the future (xxv.3), a usage with structural and semantic limitations early on but of wider range in late Latin, occurs not only in Plautus, low-register texts and non-literary documents on papyrus and ostraca, but also in a speech in Caesar. Later grammarians attempted to resist it, but one acknowledged that there was no more common ‘solecism’ even among the learned, and another let slip that it had its (learned) defenders. The profile of the construction is suggestive of a usage that had a place in casual speech right across the social spectrum.

Ever since Cooper's (Reference Cooper1895) book dealing supposedly with word formation in the sermo plebeius there has been a view that numerous adjectival suffixes were vulgar in some sense. One of these, -osus, has attracted the attention of literary scholars and prompted speculation about the motives of certain poets in admitting such words (xxii.11). It has been argued here that, far from being domiciled down the social scale, -osus was a resource of the language in general. Just one oddity was identified (bibosus) (xxii.11.1.4), but that was probably an ad hoc coinage. Many other suffixes and suffixal uses were shown in xxii to be general to the language, or to have been motivated in particular texts not by any social status that they might have had but by their functional properties. One or two low-level suffixes or usages (such as the diminutive formation -inus: xxii.7) were noted, but in general suffixation is functionally not socially determined, and most suffixes are likely to turn up in a wide variety of genres.

Later grammarians stigmatised compound adverbs (prepositions), particularly with separative prefixes, even though the type of formation was old and had long been acceptable in the literary language (xxiii.1). What was new was the proliferation of words of this formation in later Latin. There were probably mixed attitudes in the late period. Some purists were resistant to such formations (xxiii.5 (1), 7), which are common in low-register texts (xxiii.6). However, if we read between the lines we find that grammarians were well aware that they were used by the educated class as well, and even defended (xxiii.5 (2)). The formation had probably flooded the later language in general and not just lower social varieties, though some may have excluded it from writing.

Various phonological developments were shown to have taken place in the language in general. Since misspellings reflecting these developments are almost by definition found only in badly written non-literary documents, there has been a tendency among scholars to speak of vulgarisms, when it is only the spellings and not the pronunciations behind them that might be so described. Phenomena in this class include the omission of final m (viii.1), assimilation of the final nasal in a monosyllable or ‘grammatical’ word to a following consonant (viii.1.2), loss of t at the end of a word or syllable when it was followed by a consonant (e.g. pos paucos) (viii.3.4.1), the conflation of ad and at in speech (viii.3.5.1), assimilation in the consonant cluster ns (ix.7), the pronunciation lying behind the writing of B for V (x), and assibilation and palatalisation in words such as hodie and diebus (vi.7.2; see further xi.1–2). Syncope was so widespread in the language in general (see v.1) that we sometimes find educated purists castigating the original full form (so caldus and audacter were advocated over calidus and audaciter) (v.2). It was remarked at xi.3 that we had found in Part 1 hardly any phonological phenomena that could be assigned exclusively to low social dialects.

Grammarians’ pronouncements may at first lead us to think that a development was resisted among the educated classes (and therefore by implication confined to the uneducated classes), but the data may reveal rather that the change was spread across the social classes (see also above on the present for future, and on grammarians and their motives see further below, 5). That is particularly so of the loss of distinctions of vowel quantity (iii.6). Grammarians made some effort to preserve oppositions of vowel length, but that does not mean that in about the fourth century there were genuinely two vowel systems current, the one among the educated classes and the other among the uneducated. The fact is that grammarians were prone to the very errors of vowel length that they were criticising, and we may surmise that there was merely an artificial (and no doubt inconsistent) attempt to maintain quantitative distinctions mainly for the reading of classical verse. Another sphere in which the ideals of grammarians were probably unrealistic lies in their resistance (as revealed by Augustine) to the insertion of glides between vowels of different quality in hiatus (vi.6). Augustine makes it obvious that this resistance was ridiculous, and the implication is that most speakers of his own class would have made such insertions. Again, we find a grammarian finding fault with the indicative in indirect questions, but another grammarian of much the same time himself using the construction extensively (xxix.1.8).

The expansion of prepositional expressions to convey case functions cannot be located merely in lower sociolects. An important general point is that a prepositional usage resembling a use of a case inflection (such as ad alongside the dative of the indirect object) may have a nuance that distinguishes it from the plain case. A failure to notice the distinction may cause the prepositional expression to be wrongly classified as vulgar. The use of ad with nuntiare in Plautus has been interpreted as equivalent to the dative, and the expression treated as a proto-Romance vulgarism already found in early Latin but thereafter largely submerged. In fact with this verb the preposition implies movement, and there are parallels in the literary language, e.g. of Cicero (dare/scribere ad of the dispatch of letters over a distance) (xiii.5.1). Literary stylists such as Sallust and Tacitus experimented with prepositions as substitutes for unaccompanied cases, and were perhaps greater innovators in this respect than lower sociolects as represented in low-register documents (xiii.3). The gradual encroachment of some prepositions on case uses (as for example de, with its separative force intact, rivalling the partitive genitive but sometimes distinguishable from it: xiii.4.2–3) can be observed at different levels of the language. The use of ab to mark departure from a town (as in ab Roma with verbs of motion) is already the norm in the historian Livy in the Augustan period (xv.2). But the status of prepositional innovations may vary from instance to instance, and one should avoid over-generalising. On the one hand the adnominal use of ad to express possession can only safely be assigned to lower social dialects (xiii.5.5). On the other hand the instrumental use of ad, far from being stigmatised, is more common in the classiciser Vegetius than in his substandard source the Mulomedicina Chironis (xiii.6.4.4).

Ecce, which along with the form eccum coalesced with certain demonstratives in some Romance languages, was a word used in abundance in all varieties of the language, including high literature. We saw it juxtaposed with demonstratives (a precondition for compounding) at a low social level in the direct speeches in the Peregrinatio Aetheriae, but at the other end of the scale in a grammarian and other learned writers (xx.5.4.1).

It was suggested (xxxii.9) that a predominating word order aux. + inf. (with the associated VO) may well have been a feature of speech, as distinct from writing, and indeed of speech well up the educational scale and not only of the poorly educated.

1.2 Change from below

The key question addressed in this book is whether we can ever securely talk of change emanating from lower social dialects, and perhaps later influencing the Romance languages. Despite the traditional misclassifications referred to in the preceding section, there has emerged here a mass of incontrovertible evidence for usages either found in low-register sources but avoided completely by the educated, or explicitly stigmatised by the educated. These constantly survive in Romance. Sometimes we have evidence of more than one of the types listed at i.7 bearing on the social level of a phenomenon. Sometimes a usage from lower sociolects was at first rejected by the educated but in time became acceptable.

We can only guess why it was that stigmatised forms of expression should so readily have imposed themselves on the educated, and then have survived in whole areas. One factor might have been the high proportion of uneducated speakers in both urban and rural populations. In the modern Western world there is universal education, and, although educational levels vary widely, there is mass literacy, which creates a sense in speakers that some linguistic features are considered proper and others are not. In such societies change from above is commonplace, as speakers take up what they perceive to be prestige usages from higher educational classes. In the Roman world by contrast the highly educated elite must have been a tiny group, and their social dialect was perhaps more easily swamped than might have been the case had they been more numerous. The upper classes communicated regularly with slaves and freedmen, and there was a feeling among them that less formal speech was appropriate for such communication (Quint. 12.10.40). The same passage of Quintilian envisages the same level of language being used by the educated with their friends, and Cicero (Fam. 9.21.1) refers to the ‘plebeian language’ that he adopts with one of his friends (see i.7 (vi)).

A particularly interesting passage is at Quint. 1.6.44–5:

In the second sentence Quintilian alludes (with disapproval) to the language of the uneducated (imperiti, here alongside uulgo of their general speech habits), and then adds that whole theatres and circus crowds shout out in barbarous language. The point can only be that it is not only the uneducated who make up these crowds but also the educated, and the latter too use the barbarisms of the uneducated masses. They accommodate their speech to those around them, or in other words are influenced by lower social dialects. In the first sentence Quintilian makes it clear that ‘bad habits’ (note uitiose) may become ingrained in many people (multis), and states that these should not be accepted as correct simply because they represent the usage of the ‘many’; it is the usage of the educated (eruditorum) that should be followed, and these by implication are not many. This contrast between the educated and the many shows that the many are the uneducated, and it becomes clear that the final verb phrase of the first sentence is in effect a call to resist change from below, which by implication was hard to do.

We saw that a problem of communication across the classes was perceived particularly by later Christian Fathers (see i.7 (ii), and below, 2). Augustine appears to have taken pride in employing usages castigated by grammarians so that he could be understood by the ordinary people (see i.7 (ii); also vii). This attitude was widespread (see Herman Reference Herman1991), and was conducive to change from below. There was also perhaps a feeling among practically educated men such as architects and doctors that the linguistic practices of those trained by grammarians and rhetors were so rarefied as to be unattainable. The architect Vitruvius makes this point early in his first book (1.1.18), and he happily himself admitted usages rejected by the educated elite (see e.g. E. Löfstedt Reference Löfstedt1956: ii.92 on si qui for si quis; also below, 1.4). Practical men were probably more willing to associate themselves linguistically with those they worked with (builders, craftsmen, midwives, and so on) than with the pupils of rhetoricians, and that accommodation would have favoured change from below. When the medical writer Soranus Lat. (Mustio) said in his preface (p. 3 Rose) that he would speak ‘simply’ to the illiterate midwives he was addressing and use women's words (see also above, i.7 (ii)), he was promising precisely this form of accommodation, and the attitude conveyed if fairly widespread was bound to cause some usages to move up the social/educational scale.

We offer a summary of the usages discussed in this book that were definitely domiciled in lower social dialects, and rejected by the educated either entirely or for a time.

The case system was one sphere in which there is particularly good evidence for stigmatised usages that were rigorously resisted by the educated over a long period. At xviii.2 there is a list of eleven such phenomena, which need not be repeated here, though some are worth recalling.

First, the point was made at xviii.2, p. 377 that the most striking substandard phenomena down to about the fourth century are all local expressions, static, directional or separative. The evidence that we have tends to come from different types of sources. For example, the directional use of the locative Capuae is put into the mouth of a freedman by Petronius, and a little later the same usage (Alexandrie) is used several times in the substandard letters of Terentianus from Egypt (xv.4). The two sources, one conveying an attitude by an educated speaker, complement each other and establish that there was a submerged method of referring to motion to a (well-known) town. Both writers have the locative with the same verb, exeo, which is suggestive of an idiom.

Second, there is also information from a variety of sources about the use of in + acc. to denote motion to a town. This usage was castigated by Quintilian and it turns up in low-register subliterary documents from Egypt (xv.2). It is also found in Plautus, whose diverse language incorporates elements from lower sociolects considered appropriate to stage performances before a popular audience. Since the emperor Augustus, we are told, used prepositions with names of towns, it is a reasonable guess that this stigmatised construction had some place in the casual speech of the educated. Augustus’ motive, according to Suetonius, was to achieve clarity, and that aim reminds one of the readiness of Church Fathers to accommodate their speech to that of their audience so that they could be understood, an accommodation, as we noted, that represents, or might generate, change from below.

Third, fossilised place names were commented on by a grammarian (xv.4). Those with the ending -as, used with several case functions, are largely restricted to inscriptions and a mundane bureaucratic text, the Notitia dignitatum (xv.5). It is only in very late (early medieval) Latin that the nominative use of the -as ending in general (i.e. in common nouns as well as place names) became acceptable.

Fourth, the use of the accusative as subject of passive or inactive intransitive verbs shows up exclusively in late Latin in low-register texts (xii.6.3, 6.4).

Fifth, the nominative of personal names dependent on prepositions occurs only in substandard texts, mainly late (xii.3.1). Lower social dialects also seem to have led the way in the fossilising of the nominative in certain naming constructions (xii.4).

Finally, the reflexive dative with intransitive verbs, particularly with the nuance ‘spontaneously’ and the like, seems to be domiciled largely in low-register texts (xvi.5, 7). It was eliminated by Vegetius when he found it in his source, and was avoided by Jerome in the Old Testament.

A morphological feature that remained all but completely submerged was the possession by various pronouns of distinct feminine dative and genitive singular forms (xx.3). These are restricted to low-register writing tablets and inscriptions, where they have not always been recognised for what they are, and must for centuries have been rejected by the literary classes. They partly survive in Romance languages.

Gender in Latin was always somewhat variable, with many nouns showing more than one gender or changing gender ad hoc in response to a variety of factors. However, a systematic account of aberrant genders in Plautus on the one hand and Petronius on the other reveals a diachronic development in the use of the neuter (xix). In Plautus variations of gender affect nouns of all three genders. In Petronius there are two distinctive features of gender variation: all examples are in speeches by freedmen, and all are related to the neuter. The neuter had not been lost, but neuter nouns were tending to switch into the masculine under restricted conditions. Since Petronius only allows this feature in freedmen's speeches, he associated it with the middle to lower end of the social spectrum. One other suggestive item was the appearance, in British writing tablets of military provenance and in the graffiti of La Graufesenque, of a group of neuter plurals of nouns otherwise masculine (e.g. gladia, carra, radia, modiola) (xix.11; also below, 1.4). Do we have here, at relatively low social levels, an anticipation of the neuter plural forms that were to survive in some Romance languages (as feminine plurals) with the function of designating groups of weakly differentiated entities?

The history of the indicative in indirect questions is complicated. In the first century ad it does seem to have been socially stigmatised, but by about the fifth century it could be used freely even by a grammarian. This pattern suggests that in the intervening period it had spread from below (xxix.1.9). The infinitive for subjunctive in indirect deliberative questions and in potential/generic relative clauses definitely started in lower sociolects and even in late Latin is largely restricted to low-register texts (xxix.2).

Many misspellings with phonological motivations merely reveal pronunciations that were not socially restricted (see above, 1.1). However, the four structurally parallel types of contact assimilation discussed in Chapter ix (such as otto < octo) do seem to have been restricted. These had had no influence on Latin as it appears at the start of the historical period, which is remarkable, because the effects of other types of contact assimilation are widespread. Many of these other types had taken place before Latin came to be written down, as the original unassimilated forms are themselves often not attested. The first secure attestations of the four assimilations are in Pompeian graffiti, an early source for lower-class Latin. Thereafter they are rare and very limited in distribution, to a few low-register inscriptions and writing tablets; they also have some parallels in Oscan and Umbrian, which some have taken as suggestive of Italic influence on (lower-class?) Latin. We cannot prove that the underlying pronunciations were not socially widespread in speech, but the history of the four types and the other, structurally different, types in the written language is so different that the educated are likely to have been resisting the former in speech for a long time (see further ii.2.2 (ii)).

The aspirate particularly in initial position has an interesting history in Latin (vii). There are signs that it was unstable in country districts from an early period (note for example anser ‘goose’, which etymologically ought to have an initial h but is never attested in that form). In the late Republic dropping of the aspirate and its false addition by hypercorrection were clearly stigmatised by the educated (there is a relevant fragment of Nigidius Figulus, and Catull. 84). At the time of Augustine grammarians were still trying to preserve the aspirate but were mocked for this by Augustine, who accuses them of anachronism. Here then is a feature that started from below, probably both in regional and lower social dialects, and then spread up the social scale, despite the resistance of some grammarians. We have seen other clear-cut pieces of evidence for a rise in acceptability of a once stigmatised usage, perhaps most noticeably in the case of some compounded adverbs/prepositions with the prefix de- (desursum, deintus etc.) (xxiii.8 with 5 ((3)) and of the nominative plural use of the feminine ending -as (xii.7). See also iv.2.6 on the spread of the monophthong e for ae.

One type of syncope that seems to have started in lower sociolects is that in perfect verb forms such as donaut, exiut (v.2.1).

We noted some slight evidence of very late date that lower social dialects might have led the way in developing articloid uses of ille (see xxi.9, p. 526). Ipse had a far more complicated history.

One whole area of the language in which usages were often established in lower sociolects but avoided higher up the social/educational scale was that of the lexicon (but see below, 2 on the problem of submerged Latin). Particularly revealing are terms that survive widely in Romance languages and are attested in Latin just a few times in low-register texts (e.g. ebriacus: see further xxi.6, p. 565; see in general xxx).

1.3 Change from above

As was noted, at least one of the phenomena discussed at 1.1 above might arguably belong in this category, namely the future periphrasis inf. + habeo. The stray example in a substandard text is in a fragmentary context and its interpretation is uncertain. The construction is overwhelmingly in learned texts.

The dico quod-construction is absent from low-register documents such as writing tablets, in which speech is often quoted either indirectly by means of the acc. + inf. or in direct form, but can be seen emerging in the literary language (xxviii). The acc. + inf. is also overwhelmingly preferred by the freedmen in Petronius’ Cena Trimalchionis, where there is just one instance of the quod-construction. That however is matched by an example in another part of the novel, a distribution which at least shows that Petronius did not perceive the new construction to be distinctively lower-class. There are two examples of the dico quia-construction in speeches by freedmen, the first examples of this usage extant, but it is not impossible to give quia a remnant of its causal meaning in both places. In later Latin dico + quod, quia and quoniam had a firm place in literary language, and it is not possible to assign the construction predominantly to low-register texts.

The comparative construction with ab was a prepositional equivalent to the ablative of comparison, alongside which it is discussed by grammarians (xvii.2), and the ablative of comparison was a literary construction. If we see the ab-construction as a prepositional development out of the ablative, then that development is likely to have taken place at higher levels of the language (xvii.4).

Habeo + past participle finds its way into handbooks of Vulgar Latin (see xxiv.5), predictably, since it has Romance outcomes, but without justification. Several different constructions underlie this combination, depending on the meaning of habeo, and they are all in literary and legal prose. The type which seems to function as a perfect periphrasis is most securely attested in the classical period in Cicero, with participles from verbs denoting mental acquisition or cognition. Otherwise possible perfect equivalents are scattered about in different types of literary texts and do not seem to occur in substandard texts. The argument that there was a new shift towards a genuine perfect periphrasis in the ‘late Vulgar Latin of Gaul’ as evidenced by Gregory of Tours is far-fetched (xxiv.3), as on the one hand Gregory was a learned writer and on the other his usage in this respect cannot clearly be distinguished from that of Cicero. Scholars have seemed unwilling to see the origin of the Romance construction in the literary language, and there have been claims that much must have been going on beneath the surface, in lower sociolects.

The main suppletive of ire was uado, which survives extensively in the paradigm of the verb ‘go’ in Romance languages (xxxi). In late low-register texts it is commonplace. In the classical period, however, it was very much a literary word, at home in poetry and high-style prose, whereas in low-level writing tablets it is non-existent. There are signs particularly in Ovid of a loss of semantic markedness, notably in the form uade (xxxi.4.3), which was to become a replacement for i. Given the frequency of verbs of motion in subliterary writing tablets and the absence from these of uado (xxxi.5), it would be implausible to suggest that, at the time when it was a literary word, it also had an independent existence in the language of the uulgus. Here seems to be a verb that started off in higher social dialects denoting marked forms of motion, underwent semantic weakening in those dialects and sometimes indeed stood in for monosyllabic forms of ire, and then spread down the social scale in this weakened sense and became a suppletive, encroaching on ire in its weak forms.

Final -s was actively discussed by scholars in the late Republic, and its complete restoration in the literary language of the classical period seems to have been the result of a standardisation movement among the educated (viii.2.2). If we were right to argue from low-register non-literary documents of the Empire that the restoration also took place in lower social dialects (viii.2.4), then it is possible that a prestige pronunciation advocated by educated purists had spread down the social scale as well.

It has to be acknowledged, however, that identifying change from above may be problematic. A difficulty is that experiment in the literary language may foreshadow a Romance usage but be independent of later developments at a lower social level (see below, 6.5). Sallust and Tacitus, for example, occasionally use de in contexts in which an objective genitive might have been expected (xiii.3.3). This usage anticipates de as a substitute for the objective genitive in late low-register texts and as its replacement in Romance. Should we say that the objective-genitive equi-valent first emerged in literary Latin and then spread from above? It is far more likely that the isolated phrases in Sallust and Tacitus were ad hoc creations unrelated to extensions of the functions of de at a late date and a different social/educational level. To make a case for change from above one needs continuity of evidence over a period suggesting the direction of change. There is such evidence for uado, and also for a usage not dealt with in this book, the quasi-adverbial use of mente with adjectives (e.g. tota mente) anticipating the Romance role of mente as an adverbial suffix (see Karlsson Reference Karlsson1981, Bauer Reference Bauer2010). Such phrases with mente occur in the classical literary language and then spread later in different types of texts.

1.4 Specific social groups within the uulgus

For reasons set out at the beginning of the book (i.5) we have been forced to speak vaguely about usages of the uulgus without identifying different social groups within the relatively uneducated masses. Imprecise terms such as ‘lower sociolects’, ‘lower social dialects’, ‘uneducated’ versus ‘educated’ usage abound, inevitably, given the deficiencies of the evidence available. Occasionally, however, snippets of information have emerged about the linguistic practices or general linguistic level of specific groups, and it may be worthwhile to collect these here. Grammarians had a habit of attributing linguistic ‘vices’ to different peoples (gentes: gentilia uitia), particularly Africans. These are relevant to regional, not social variation in the strict sense, and have been collected elsewhere (see Adams Reference Adams2007: 806 s.v. ‘“vices”, linguistic of certain peoples’). Similarly ‘rustic’ Latin may have a social dimension, since rustics were often presented as uneducated, but again regional variation is at issue, and we have not dealt systematically with rural Latin here (see Adams Reference Adams2007: 804 s.v. ‘“rustic” Latin and rustics’; see also above, i.7 (iii)).

Within the uulgus a prominent group was the Roman plebs, and usages are ascribed to them by grammarians (Consentius and Fortunatianus: see i.7 (i)), with disparagement. When referring to the ‘barbarisms’ that even the educated (by implication) fell into in the circus (see above, 1.2) Quintilian no doubt had in mind the urban plebs as the driving force. The plebs were not, however, a uniform mass. There were Greeks and other types of outsiders (see e.g. Cic. Fam. 9.15.2, Brut. 258), and there must have been a good deal of linguistic diversity. Remarks by grammarians are likely to have been over-general.

The diminutive use of the suffix -īnus had its roots in a feature of the nomenclature of the plebs (xxii.7), and some terms of this diminutive formation can be ascribed specifically to wet nurses and midwives (xxii.7).

The best evidence concerns freedmen. These were not of uniform educational level but ranged from educated scholars to the uneducated types presented by Petronius (i.7 (iii)). It has been noticed that even among Petronius’ uncultured freedmen there are linguistic variations (see i.7 (iii)). Slaves and freedmen were often of Greek origin, and the speeches in the Cena Trimalchionis, particularly of one character, Hermeros, are full of Greek elements. In Plautus too slaves use Greek terms and code-switching (whereas women in particular do not: see Clackson Reference Clackson2011c: 509 with bibliography). The occasional grecising element became established (under restricted conditions) in the Latin of such speakers, notably the feminine genitive ending -es and its Latinate variant -aes (see i.7 (iv)). We have in the hand of a real freedman from the Bay of Naples (the location of the Cena), C. Novius Eunus, some legal documents of the first century ad which in their spelling display a lack of education (see Adams Reference Adams1990a). Eunus at one point construes the preposition per with the nominative of a personal name, a usage that can be paralleled much later in Latin but is abundant in some corpora of Greek papyri of comparable date (roughly the second century) to Eunus himself (see xii.3.1), and it is possible that he has lapsed into a contemporary Greek practice. His use elsewhere of a neuter plural subject with the singular verb est is open to various explanations (viii.3.4), and need not reflect Greek influence. But while we may allow that the Latin of some freedmen and slaves was subject to different types of Greek influence, which would have rendered it distinctive, it is not so much interference phenomena as internal Latin developments that make the freedmen's speeches in Petronius interesting. Petronius had noticed features of the language of this social group that he put into their speeches but kept out of the rest of the novel, most notably some uses of the masculine for neuter (see xix.5.3, and above, 1.2), thereby portraying a real social variation. Certain uses of the indicative in indirect questions are also given only to the freedmen (xxix.1.7.1). Plautus too assigns some special (Latin) idioms to slaves (i.7 (iii)).

It is worth mentioning here a remark by a freedman which is revealingly discussed by Clackson (Reference Clackson2011c: 507). Echion, a humble centonarius, accuses Agamemnon, a rhetorician, of looking down at his language: 46.1non es nostrae fasciae, et ideo pauperorum uerba derides. scimus te prae litteras fatuum esse (‘you are not one of our bunch, and so you laugh at the speech of the poor. We know that you are crazy for learning’, Clackson). Clackson notes that here we have implied a notion that language may be a mark of group identity, remarking:

The sense that a usage may be appropriate to one's class (see Chambers Reference Chambers2002: 350, cited at i.2 above) and thus a mark of group identity is one of the reasons why speakers maintain stable sociolinguistic variables over a long period, even if they are aware that the variables they use may be stigmatised. In these two sentences and immediately afterwards Echion seems aggressively to be adopting usages that were substandard, the genitive plural pauperorum, prae + acc. (see TLLx.2.379.5ff.), and then persuadeo + acc. for dative. The linguistic difference that he clearly perceives between himself and the rhetorician is echoed in a remark by the architect Vitruvius that has been referred to earlier (i.5, and this chapter, 1.2). Vitruvius (1.1.18) apologises to the emperor for any departures that his work may contain from the ‘rules of the ars grammatica’ (si quid parum ad regulam artis grammaticae fuerit explicatum, ignoscatur), stating that he does not write as a rhetor disertus, but as an architect (sed ut architectus his litteris imbutus haec nisus sum scribere). This is not an unequivocal apology but a statement of identity: his language is that of an architect, who has merely dipped into grammatical rules.

A sense of belonging to a group as motivating the adoption of usages that higher status groups might look down on perhaps lies behind a remarkable morphological item in specimens of soldiers’ Latin, to which category we now turn.

Slang terms and technical terms used by soldiers are attested and have been collected (see Heraeus Reference Heraeus1902b, Reference Heraeus and Hofmann1937: 151–7), but these are usually not the same as sociolinguistic variables, and we have not gone into details of this kind in the book. There are however some more relevant usages in soldiers’ Latin that have come up. One of these was mentioned earlier (i.7 (iv)), the repeated use of debunt for debent by a group of under-officers at Vindolanda in a military outpost where debent was also current (and is attested). Given that at least some of those who wrote debunt must have been familiar with debent, their regular choice of the non-standard variant smacks of the influence of their sense of group membership. A comparable form, habunt (see Adams Reference Adams2003b: 544–5), has turned up in a letter (628) written by a decurion to the commanding officer Cerialis, whose own Latin is very correct. Habunt lies behind Fr. ont. It is interesting to find these -unt forms used by lower officers, and to see that the decurion does not modify his usage when addressing a superior. The cases of debunt and habunt are a reminder that non-standard variables are not only used by the lowest social classes.

Another feature of the Vindolanda documents (and also the comparable ones from nearby Carlisle) is the accumulation of about half a dozen unusual neuter plural forms (see xix.11). We see within a specific group below the elite a tendency to convert some masculine inanimates into the neuter in the plural (for possible motivations see 1.2).

A comparable neuter plural, catilla alongside the singular catillus, was observed within another identifiable non-elite group (potters, craftsmen) (see xix.11).

6.1 Phonological variables

Socially determined phonological variables are represented by spelling variations, and it is surprising to what extent these variables play a part in distinguishing genres or establishing the tone of a particular context or characterising a speaker or referent. A clear case (with regional significance primarily but also social) is the satirist Lucilius’ use of the forms Cecilius and pretor at 1130 to stigmatise the person referred to (iv.2.1). There is evidence too that rustics in Atellan farce were assigned a form with e for ae (iv.2.2), no doubt as a means of characterisation. The monophthongal variant (o) of the au diphthong also occurs in some significant contexts. Plaustrum, for example, is the form preferred in literary Latin, but Cato and Vitruvius in their practical manuals favour plostrum (iv.3.1.6). The diminutive plostellum is also in Horace's Satires, and satire, like farce, is a genre in which disparaged variables have a role (see above and further below). Cicero uses two o-forms in proverbial expressions in letters (see iv.3.1.2 on oricula and iv.3.1.7 on olla), to impart a tone suitable both to the homely phrases and to the genre. Catullus too has both forms in proverbs in shorter poems. More problematic is the interpretation of Celsus’ use (if the manuscripts are to be trusted) of colis ‘penis’ (see iv.3.1.5), a form which appears to make the term down-to-earth and ‘Italian’, and to dissociate it from a similar Greek word.

Syncope has complex motivations, and cannot simply be attributed to lower sociolects. We did see a few interesting literary variations. Syncope in calidus (> caldus) is attested mainly in satire and epigram (Lucilius, Horace and Martial), and in prose in practical texts (v.2.2). The high literary term ualidus resisted syncope, but the colloquial derivative adverb ualde shows it almost invariably (v.2.5). Occasionally orators as a group are held up as users of what is socially acceptable (see Gell. 1.22.2 at i.1), and that is so in the case of audacter (v.2.3), where it is the syncopated form that is recommended, though not by extreme pedants. These cases in different ways show the interaction of social and stylistic determinants. The stigmatised syncopated variable is admitted in casual/informal style, but in some words may become the norm, and its full variant is then treated by some as no longer acceptable.

It has been suggested in the literature that a stylistic factor influences the use of another variable in early Latin, that is the deletion of final -s under certain phonetic conditions (viii.2.2). Deletion has been reported to be more common in spoken than sung metres in Plautus, and more common also in funerary than in sacral inscriptions, and that would suggest a resistance to loss in the most formal styles. We questioned, however, the reliability of the evidence.

There is a particular literary interest to several terms with an open vowel (e) in hiatus where the educated language has i. The phenomenon starts out as regional (e.g. in filea for filia in inscriptions probably from Praeneste, and in Plautus’ conea for ciconia, attributed to Praenestines), but one such term (labeae; also with a change of gender) reached the city, where it was undoubtedly socially stigmatised, as it occurs only in Atellan farce (Pomponius and Novius), Lucilius and in a passage of Plautus where the dramatic illusion is dropped (see xix.4.1 (12) and also vi.2). It is found in several phrases of aggressive content, and there can be little doubt that these various writers were admitting a disparaged form (there were various synonyms available) to suit the genre and context.

Another term with a provincial feature that reached the city, no doubt through movements of population at a low social level, and was admitted to a lower genre was glaucuma (xix.4.2.3 (30)).

6.2 Morphological variables

Three categories of phenomena with relevance to generic variation in literature have come up.

First, various Greek neuters in -μα (e.g. schema, stigma, glaucuma) are used as feminines in genres such as comedy, farce, mime and satire and also a freedman's speech in Petronius (see xix.5 (ii), 4.2.3 [30, 31], 12.2), but in the high literary language are given their original neuter gender. One such term, schema, is neuter in early tragedy but feminine in comedy (xix.12.2). The change of gender reflects a popular or acoustic borrowing, and the popular forms were obviously felt to be suitable to lower genres.

Second, by the late Republic the reinforced pronoun form istaec (neuter plural) is admitted only in informal style (letters, Catullus, a speech in the senate containing a sexual charge) (xx.2.2), particularly in hackneyed phrases. It might also have been current in low social dialects, though there is no evidence of the relevant type at this date.

Third, various hybrids, with Latin base but foreign suffix (Greek, Etruscan or Celtic), are found only in the lower genres mime, comedy and Menippean satire, and in a speech in Petronius (xxii.6 leuenna, ebriacus, hamiota, lupatria), and one of these (ebriacus) is stigmatised by a grammarian.

On the other hand it was argued (xxii.11) that the suffix -osus was without social significance and that it should have no place in attempts to find rustic or plebeian elements in literary genres.

6.3 Syntactic variables

Directional expressions show some variation, with forms considered acceptable by the educated sometimes rivalled by stigmatised forms. We observed (xv.2; also above, i.2) a familiar pattern in the use of the preposition in with names of towns: it is criticised by a grammarian and attested in non-literary documents from Egypt, but also admitted in comedy. Of particular interest is the fact that another castigated usage, intus for intro, sometimes intrudes into high literature (e.g. Lucretius, Ovid, Tacitus: see xv.3, xxiii.3). Here is evidence for a general point that has often been made here, that distinctions between social dialects are not absolute, and that one cannot always see a motive when a socially disparaged usage is admitted in high style. In Latin literature it is usually the case, as has been shown above, that disparaged usages turn up in what we have referred to as lower genres, but that is not always so. There was also some slippage in the distinction between foris and foras, with the former admitted for the latter by Pliny the Elder (xv.3).

There is some evidence that reflexive datives (with various nuances) with intransitive verbs were frowned on by some purists. It is interesting to find one (which seems pleonastic) in Horace's Satires (see xvi.5). This seems to be an anticipation of a type of usage found much later, mainly in low-register texts.

6.4 The lexicon

The lexicon has had only a marginal place in this book, but we have seen a few items, largely submerged though in some cases with Romance reflexes, which turn up in significant texts or contexts. A proto-Romance use of bucca is admitted by Cicero in casual style (letters) (xxx.2.7). Various weakly attested anatomical terms, pantex (xxx.2.15), aqualiculus (xxx.2.15) and rostrum of humans (xxx.2.7), appear in lower genres such as farce and satire, typically in pejorative contexts. More striking is the appearance of pisinnus and manduco, both of which belonged to low sociolects (or in the case of manduco to casual style as well among the educated), in an epic fragment (i.7 (vi), xxii.9). It is not clear whether the writer was striving for a particular effect or showing indifference to normal social restrictions.

6.5 Some conclusions

The variables seen above that had associations with lower sociolects were admitted by the educated into literary compositions under various conditions or for various purposes. Socially disparaged variables might be (1) suited to the general tone of certain lower literary genres (comedy, farce, mime, satire) that presented scenes from mundane life; (2) suited to the micro-context within a work, as for example to a popular proverb or abusive remark; (3) used to characterise a speaker within a work that is otherwise in a higher style; (4) chosen to clash with their surroundings to achieve some sort of effect, as when the epic poet used both pisinnus and manduco in a single sentence in a translation of Homer; (5) used to achieve a casual style in epistolography; or (6) suited to the readership, a factor of which Jerome was aware (xix.12). Jerome's Latin, across a range of genres, is very variable (xvi.5, xxxi.4.1, xxxii.4). To these six categories, which are not completely discrete, there might be added the phenomenon that we referred to as slippage, that is when a stigmatised usage is admitted in high literature either for no apparent reason other than that no one resists disparaged usages all the time, or because different writers may evaluate the same usage differently (see above on intus).

Sometimes a usage that was to have a Romance outcome makes an appearance in literature not as a manifestation of any of the above factors, but because it was an innovation that first occurred at higher social or educational levels of the language and then spread from above (see above, 1.3).

We also saw a variation on this pattern. Sometimes high-style writers in their striving for unusual forms of expression happen to anticipate developments of a much later period that in some cases were destined to have an outcome in Romance. There need be no continuity between the anticipation and the later development. The innovation is ephemeral and is made again centuries later, usually at a lower social/educational level of the language (see also above, 1.3). Into this category fall the use by Sallust and Tacitus of de as equivalent to the objective genitive (xiii.3.3), the use by Tacitus of instrumental per with concrete nouns (xiii.3.1; cf. 6.2), percutio with quasi-instrumental de in Ovid (xiii.6.4.2; cf. xiii.7, p. 319 for percutio de in late low-register texts), ab as a substitute for the ablative of comparison in Ovid (xvii.2), and a use by Propertius of uenio with a perfect participle that comes close to a perfect periphrasis (xxvi.14.2).