Abstract
In this paper, a quantitative and qualitative analysis is conducted on the <b>/<v> alternations found in the epigraphic texts from three representative subsets of Roman African inscriptions from both urban centres and more peripheral areas (1st century BCE – 7th century CE). The distribution of the confusion has been related to the dating and provenance place of the inscriptions and the level of literacy of those involved in their crafting. The results show a difference in the distribution of <b>/<v> confusions in the three areas examined, with a higher incidence in later inscriptions from Sabratha. Thus, it is discussed whether the different distribution of the <b>/<v> confusions observed in the different regions might be a cue for internal diatopic variation.
1 Introduction
Among the graphemic alternations attested in the epigraphic texts of the Roman Empire, the confusion between <b> and <v> is one of the most widespread phenomena. Occurrences of alternations such as beni for ueni or uene for bene are found in inscriptions, at least from the first century CE,1 as well as in non-literary sources such as the Bu Njem ostraca and the Albertini tablets.2 Such confusions are traditionally explained as being due to a fricativization of /w/ and, in general, to the partial merger of the Classical Latin bilabial voiced stop /b/ and the labiovelar semivowel /w/ into a bilabial fricative [β], which are represented in Latin script either with <b> or <v>.3 This hypothesis seems to be confirmed also by the testimonies of ancient grammarians: see e.g. Velius Longus (Keil Gramm. Lat. VII 58. 17–19), speaking of the pronunciation of /w/ cum aliqua adspiratione, and later Priscianus (Keil Gramm. Lat. V 23. 9–10: Apud nos quoque est inuenire, quod pro u consonante b ponitur). Confusions between /b/ and /w/ are also attested in the differentiae uerborum corpus, i.e. lists of Latin lexical couples including homophones, where couples such as labat/lauat, iubat/iuuat (see e.g. the Appendix Probi) are found.4 The confusion is the subject of Martyrius' treatise on orthography De B muta et V uocali (Keil Gramm. Lat. VII 165–199), mentioning couples such as uis/bis, bos/uos. Although the treaty refers primarily to orthographic issues, this testimony might hint at a confusion of /b/ and /w/.5
Focusing on epigraphic attestations, the analyses carried out on the topic show an unequal diffusion of the phenomenon through the Empire.6 Interestingly, Africa seems to be especially targeted, as was shown by Joseph Louis Barbarino7 in his examination of the Latin Christian inscriptions dating back mainly to the 3rd–7th century CE from Northern Africa, Britain, the Balkans, Dalmatia, Spain, Gaul, Rome, and Italy. In this study, the relative frequency of the confusions between <b> and <v> is calculated over the corresponding Classical spellings, i.e. the number of instances of <b> for Classical Latin /b/ and <v> for Classical Latin /w/ for each area. The results show a higher incidence of the merger in Rome (27%), Southern Italy (23%) and Africa (16%), with higher percentages than Gallia Narbonensis, Baetica, Britain and the Balkans, which display an error rate of less than 5%. Interestingly, these data seem also to be confirmed by non-literary sources from Africa: in both the Bu Njem ostraca and the Albertini tablets a high incidence of <b>/<v> confusions is attested.8
The causes of the higher incidence of these alternations in Roman Africa, however, cannot be determined at the time of writing. The main explanations provided in the literature for the different incidences of <b>/<v> confusions throughout the Empire, e.g. language contact, do not hold for Africa. Greek influence9 cannot be assumed, since Africa was not a predominantly Greek-speaking area, to an extent that would justify the transfer of a phonetic feature of Greek into African Latin.10 Also, the hypothesis of a ‘Libyan’ influence mentioned by Acquati11 to explain the higher frequency of confusion in Africa is problematic. Indeed, Libyan is attested by inscriptional evidence (see the texts included in the Recueil des Inscriptions libyques by Chabot).12 However, these sources do not allow for a reconstructing of its phonology, since they are either difficult to interpret or consist mainly of proper names and formulae.13
Acknowledging this problematic issue, however, we believe that the high number of <b>/<v> confusions found in North African sources still deserves special attention. However, a fine-grained analysis of the confusion found in the inscriptions from this province is not available at the time of writing. Barbarino's examination does examine the African area in detail, although variables such as internal diatopic variation within Roman Africa and literacy level are not assessed. Furthermore, Africa was not examined in the other investigations on the topic, i.e. Herman's analysis and Adamik's (since the province was not yet included in the LLDB database at that time) quoted above. For this reason, in the next section, we will examine three sets of representative inscriptions from Africa Proconsularis, and consider linguistic and extra-linguistic variables such as the lemmas involved, the dating, provenance place and text type of the inscriptions in our analysis.
2 The corpus
Our corpus contains 835 dated inscriptions from Africa Proconsularis, for a total number of 15,945 tokens. The texts have been selected through an examination of the inscriptions available in the Epigraphik Datenbank Clauss-Slaby database (https://db.edcs.eu/epigr/hinweise/hinweis-it.html), which is the most comprehensive digital source of Latin inscriptions available at the time of writing. To perform a more accurate linguistic analysis, only dated inscriptions were selected: the dating was determined by consulting the available editions of the texts and the information included in other online databases, such as the Epigraphic Database Roma (http://www.edr-edr.it/default/index.php).14 Several areas are included in our corpus. Firstly, we selected Lepcis Magna and Theueste, since they display a high epigraphic density and a comparable number of dated texts (ca. 250 inscriptions);15 the colony of Sabratha was also included given its prominence in the Roman Empire,16 although it displays a lower number of texts (88; see below). Finally, to compare these data with those from cities with an inferior epigraphic density, we decided to select a set of close areas with at least 10 inscriptions available (Fig. 1). This group, henceforth labelled as ‘other areas’, comprises the following locations: Uchi Maius, Mustis, Thuburbo Maius, Mactaris, Sufetula, Cillium, Talah, Aradi, Limisa and Mascilianae.
Providing more detail, the subset from Lepcis Magna contains 250 inscriptions (5,836 tokens). The texts date mainly from the 1st to the 7th century CE. However, the majority of them date back to the 1st–3rd century (187 texts, i.e. 75%). Several text types are available: funerary inscriptions, texts on public monuments, sacred texts and a small number of instrumenta domestica. However, most of them are tituli honorarii, i.e., inscriptions carved on public monuments (80% of the tokens belong to this category; see Table 1).
Lepcis Magna: text types
Text type | Tokens | Texts | ||
N. | % | N. | % | |
Tituli sepulcrales | 287 | 5% | 18 | 7% |
Tituli honorarii | 4,698 | 80% | 189 | 76% |
Instr. dom. | 5 | 0% | 1 | 0% |
Tituli sacri | 846 | 15% | 42 | 17% |
Total | 5,836 | 100% | 250 | 100% |
The 88 inscriptions from Sabratha (1,205 tokens) are more balanced in terms of dating (52% of the texts, i.e. 46, date back to the 1st–3rd century CE); like those in Lepcis Magna, public inscriptions are more represented than other text types (Table 2).
Sabratha: text types
Text type | Tokens | Texts | ||
N. | % | N. | % | |
Tituli sepulcrales | 498 | 41% | 37 | 42% |
Tituli honorarii | 675 | 56% | 50 | 57% |
Tituli sacri | 32 | 3% | 1 | 1% |
Total | 1,205 | 100% | 88 | 100% |
The subset from Theueste contains 271 inscriptions (4,324 tokens), mainly from the 1st–3rd century (259 texts, i.e. 96%). Unlike Lepcis Magna and Sabratha, funerary inscriptions are by far the most frequent among the attested text types in this area (Table 3).
Theueste: text types
Text type | Tokens | Texts | ||
N. | % | N. | % | |
Tituli sepulcrales | 4,048 | 94% | 258 | 95% |
Tituli honorarii | 190 | 4% | 9 | 3% |
Tituli sacri | 86 | 2% | 4 | 2% |
Total | 4,324 | 100% | 271 | 100% |
Finally, the subset of inscriptions displaying an inferior epigraphic density contains 226 texts in total (4,570 tokens). Again, public inscriptions of the first centuries of the Empire are more represented (77% of the texts, i.e., 174, date back to the 1st–3rd century CE; see Table 4 for the distribution of text types).
Other areas: text type
Text type | Tokens | Texts | ||
N. | % | N. | % | |
Tituli sepulcrales | 1,445 | 32% | 113 | 50% |
Tituli honorarii | 2,651 | 58% | 93 | 41% |
Tituli sacri | 474 | 10% | 20 | 9% |
Total | 4,570 | 100% | 226 | 100% |
One further remark is needed. The picture outlined above clearly shows that our data are not balanced in terms of dating and text type for the different areas. In addition, the inscriptions from Sabratha are less numerous than those from the other regions. This issue relates to the unbalancing in the documentation available for epigraphic texts, a well-known shortcoming of epigraphic studies that encompasses all the regions of the Empire. We will try to partially overcome this issue in two ways: (i) text type and dating will be taken into account as variables in our linguistic analysis, and a fine-grained examination of a subset of comparable texts of the same dating and type will be performed (§4); (ii) the error rate will be calculated as a relative percentage against the total number of possible contexts/other divergent spellings of both each subset and of the whole corpus so that the percentages can be ‘weighed’ so as to take into account the different sizes of the corpora (§§3–4).
3 Error rate
3.1 <b>/<v> alternations in the corpus
As a first parameter, we examined the relative frequency of <b>/<v> alternations over the total number of possible contexts, i.e., of spellings displaying <b> and <v> according to the Classical norms. The frequency of the occurrences of <b> for <v> (e.g., bixit for uixit) and <v> for <b> (e.g. nouilissimo for nobilissimo) was calculated over the number of <b> and <v> ‘Classical’ spellings (e.g. uixit, bene, and the like). The results are represented in Tables 5 and 6.
Occurrences of <b> for <v> (e.g., bixit for uixit)
Lepcis Magna | Theueste | Sabratha | Other areas | |||||
N. | % | N. | % | N. | % | N. | % | |
<b> | 16 | 3% | 2 | 1% | 35 | 33% | 8 | 3% |
<v> | 551 | 97% | 271 | 99% | 70 | 67% | 247 | 97% |
Total | 567 | 100% | 273 | 100% | 105 | 100% | 255 | 100% |
Occurrences of <v> for <b> (e.g., uene for bene)
Lepcis Magna | Theueste | Sabratha | Other areas | |||||
N. | % | N. | % | N. | % | N. | % | |
<v> | 1 | 0.2% | 1 | 1% | 0 | 0% | 0 | 0% |
<b> | 410 | 99.8% | 128 | 99% | 100 | 100% | 247 | 100% |
Total | 411 | 100% | 129 | 100% | 100 | 100% | 247 | 100% |
The Tables show some discrepancies in the type of alternations. Firstly, the occurrences of <b> for <v> are far more attested than <v> for <b> in all areas. Only two instances of <v> for <b> are found, in inscriptions dating from the end of the 3rd century onwards. One occurrence is attested in Lepcis Magna (laudauilis for laudabilis, IRT 522, 291–295 CE) and one in Theueste (nouilissimo for nobilissimo, 337–240 CE), whereas the phenomenon is not attested in the other subsets, although possible contexts of occurrence for the phenomenon were available. The frequency of <v> for <b> is therefore very low in all areas (i.e. equal or below 1%) coherently with the trend observed for other regions of the Empire.17 For this reason, from now on we will focus on the instances of <b> for <v>.
Secondly, a difference in the incidence of the use of <b> for <v> is found in the examined areas (Table 5). A very high relative frequency is found in Sabratha, where nearly one-third of the instances display <b> for <v> (33%). Contrarily, in the other subsets the percentages are far lower, i.e. equal to or below 3%, with the smallest error rate found in Theueste (1%). Although the higher frequency found in Sabratha might be overestimated due to the quantitative unbalancing of the corpora, the difference is statistically significant (χ2 (3) 332.23, p-value < 2.2e-16). This is confirmed also by the results obtained by calculating the relative frequency of the alternations over the total number of contexts available in the whole corpus (Table 7).
<b> for <v> in the corpus: diatopic variation
<b> for <v> | Total contexts | ||
N. | % | ||
Lepcis Magna | 16 | 1.2% | 1,287 |
Theueste | 2 | 0.2% | |
Sabratha | 35 | 2.7% | |
Other areas | 8 | 0.6% |
Although the discrepancy is less marked, it is still evident that more alternations are found in Sabratha, where the frequency of <b> for <v> spellings is more than double Lepcis Magna (2.7% vs 1.2%) and far higher than the other areas.
3.2 <b> for <v>: dating and text type
Regarding dating, the occurrences of <b>/<v> alternations are found almost exclusively in later inscriptions dating from the 4th century onwards. This holds for all areas, despite the presence of possible contexts for the phenomenon available in both earlier and later inscriptions, as is shown in Tables 8 and 9.
Incidence of <b> for <v> before 300 CE
Before 300 CE | ||||||||
Lepcis Magna | Theueste | Sabratha | Other areas | |||||
N. | % | N. | % | N. | % | N. | % | |
<b> for <v> | 1 | 0% | 1 | 0% | 0 | 0% | 0 | 0% |
<v> | 350 | 100% | 258 | 100% | 26 | 100% | 191 | 100% |
Total | 351 | 100% | 259 | 100% | 26 | 100% | 191 | 100% |
Incidence of <b> for <v> after 300 CE
After 300 CE | ||||||||
Lepcis Magna | Theueste | Sabratha | Other areas | |||||
N. | % | N. | % | N. | % | N. | % | |
<b> for <v> | 15 | 7% | 1 | 7% | 35 | 44% | 8 | 12% |
<v> | 201 | 93% | 13 | 93% | 44 | 56% | 56 | 88% |
Total | 216 | 100% | 14 | 100% | 79 | 100% | 64 | 100% |
Focusing on later texts, the discrepancy in frequency outlined above can still be observed: in Sabratha the relative frequency of the confusions is far higher than in the other regions (44% vs ≤12%; Table 9).
As far as text type is concerned, the phenomenon is particularly attested in funerary and public inscriptions (Tables 10–13). This is due to the unbalancing in the text types represented in the corpus since most of the available inscriptions are public or funerary (§2). Since tituli sacri and instrumenta domestica are less represented, not enough contexts of occurrences of the phenomenon are available to make valid observations on these epigraphic genres. As for earlier tituli sacri from Lepcis Magna and the ‘other areas’ subset, where a higher number of contexts are available, no confusion is attested: this is coherent with the data from the earlier funerary and public inscriptions.
<b> for <v> in Lepcis Magna: text type
Lepcis Magna | Before 300 | After 300 | |||||||
Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | ||
<b> for <v> | N. | 0 | 1 | 0 | 0 | 13 | 2 | 0 | 0 |
% | 0% | 0% | 0% | 0% | 62% | 1% | – | – | |
<v> | N. | 7 | 268 | 1 | 74 | 8 | 193 | 0 | 0 |
% | 100% | 100% | 100% | 100% | 38% | 99% | – | – |
<b> for <v> in Sabratha: text type
Sabratha | Before 300 | After 300 | |||||||
Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | ||
<b> for <v> | N. | 0 | 0 | 0 | 0 | 35 | 0 | 0 | 0 |
% | 0% | 0% | – | 0% | 70% | 0% | – | – | |
<v> | N. | 6 | 14 | 0 | 6 | 15 | 29 | 0 | 0 |
% | 100% | 100% | – | 100% | 30% | 100% | – | – |
<b> for <v> in Theueste: text type
Theueste | Before 300 | After 300 | |||||||
Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | ||
<b> for <v> | N. | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
% | 0% | 0% | – | 0% | 17% | 0% | – | – | |
<v> | N. | 235 | 16 | 0 | 7 | 5 | 8 | 0 | 0 |
% | 100% | 100% | – | 100% | 83% | 100% | – | – |
<b> for <v> in other areas: text type
Other areas | Before 300 | After 300 | |||||||
Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | Tit. Sep. | Tit. Hon. | Instr | Tit. Sacr. | ||
<b> for <v> | N. | 0 | 0 | 0 | 0 | 8 | 0 | 0 | 0 |
% | 0% | 0% | 0% | 0% | 24% | 0% | – | 0% | |
<v> | N. | 60 | 109 | 1 | 21 | 25 | 28 | 0 | 3 |
% | 100% | 100% | 100% | 100% | 76% | 100% | – | 100% |
Focusing on later funerary inscriptions, which cover most of the token instances, a discrepancy in the incidence of the phenomenon is still attested. Sabratha displays a far higher frequency of confusion (70%), followed by the ‘other areas’ subset (24%). Although Lepcis Magna seems to display a high incidence of confusion (62%), the low number of token instances available for later funerary inscriptions (21 in total) impedes us from drawing reliable conclusions. Similarly, no valid observations can be made for Theueste because of the low number of occurrences (<15).
3.3 Lemmas involved
The alternations found in our corpus mainly involve the lemmas uiuo (in the form bixit for uixit) and requiesco (in the form requiebit for requieuit): this peculiarity seems to be due to the high number of funerary texts available in the corpus (§2) and to the limited number of lexemes found in the inscriptions. The complete list of the lemmas involved for each area is reported below:
- (i)Lepcis Magna. <b> for <v>: lemma uiuo (in the forms bixit and biuerunt), 7 occurrences; lemma seruator (in the form serbator), 3 occurrences; lemmas octauus (in the forms octabae, oktabae) and Lauinatium (Labinatium), 2 occurrences; 1 occurrence of the lemmas beneuolus (beniboli) and uir (bir). As for <v> for <b>, one occurrence of the lemma laudabilis (in the form laudauilis) is attested.
- (ii)Sabratha. <b> for <v>: lemma uiuo (in the form bixit), 26 occurrences; lemma requiesco (in the form requiebit), 6 occurrences; one occurrence of the lemmas Nouember (Nobembris), uox (bocem) and uniuersus (unibersa). No instances of <v> for <b> are recorded.
- (iii)Theueste. <b> for <v>: one occurrence of the lemmas uiuo (in the form bixit) and Nouember (Nobembres). As for <v> for <b>, one occurrence of the lemma nobilis (nouilissimo) is attested.
- (iv)Other areas. <b> for <v>: lemma uiuo (in the form bixit), 6 occurrences; one occurrence of the lemmas requiesco (requiebit) and conuoco (conbocare). No instances of <v> for <b> are recorded.
3.4 Context
Another important variable in the study of <b>/<v> alternations is their context of occurrence, i.e. the position in the word of the phoneme showing the confusion. This examination is particularly interesting given the difference in the target environment throughout the Empire evidenced by the literature.18 For example, according to the data by Adamik,19 some areas (e.g. Apulia et Calabria and Moesia Inferior) show a higher frequency of confusion in word-initial and post-consonantal position,20 whereas others display a preference for the intervocalic context (e.g. Germania Superior and Venetia et Histria). The picture displayed by the African areas examined here, however, has not yet been examined in detail.
For this parameter, we focused on the instances of <b> for <v>, since the low number of attestations displaying <v> for <b> does not allow us to perform such an analysis. The data have been classified through the following labels: ‘initial’, when the confusion is found in an absolute initial position; ‘intervocalic’, when the confusion is preceded by a vowel (including the occurrences where it is preceded by a word ending in a vowel, e.g., qui bixit for qui uixit); and post-consonantal, when the confusion is preceded by a consonant. According to our data, in almost all areas, the initial position seems to be more frequently targeted by the phenomenon.21 However, when phonosyntax is taken into account (i.e. occurrences such as qui bixit are treated as intervocalic rather than in the initial position),22 the phenomenon is mostly attested in the intervocalic context (Tables 14 and 15). A partial exception seems to be found in the ‘other areas’ subset, where confusions in intervocalic position are only half of the total; however, the low number of data forms from this area (<10 tokens) impedes us from making reliable observations.
Context for <b> for <v>
Context | Sabratha | Lepcis Magna | Other areas | Theueste | ||||
N. | % | N. | % | N. | % | N. | % | |
Initial | 27 | 77% | 8 | 47% | 6 | 74% | 1 | 33% |
Intervocalic | 8 | 23% | 6 | 35% | 1 | 13% | 2 | 67% |
Post-consonantal | 0 | 0% | 3 | 18% | 1 | 13% | 0 | 0% |
Total | 35 | 100% | 17 | 100% | 8 | 100% | 3 | 100% |
Context for <b> for <v> in phonosyntax
Context (phonosyntax) | Sabratha | Lepcis Magna | Other areas | Theueste | ||||
N. | % | N. | % | N. | % | N. | % | |
Initial | 0 | 0% | 1 | 6% | 3 | 37% | 0 | 0% |
Intervocalic | 35 | 100% | 13 | 76% | 4 | 50% | 3 | 100% |
Post-consonantal | 0 | 0% | 3 | 18% | 1 | 13% | 0 | 0% |
Total | 35 | 100% | 17 | 100% | 8 | 100% | 3 | 100% |
Our results are thus coherent with the general trend evidenced by the Latin language23 and Latin inscriptions from other areas of the Empire:24 the merger of CLat. /b/ and /w/ is attested in intervocalic position, possibly also across word boundaries. No internal diatopic difference is found in our corpus.
4 <b> for <v> in later inscriptions: different levels of literacy?
The diatopic difference in the incidence of the phenomenon evidenced in §3.1 deserves further attention. While considering the total contexts available as a variable, the methodology applied in the previous sections does not consider the literacy level of those involved in the crafting of the inscriptions. However, the available literature on epigraphic data shows that this variable cannot be disregarded in linguistic analysis. Theoretically, if the level of literacy of the author(s) was high, a lower frequency of divergent spellings found in a subset of inscriptions could be due to better knowledge of the Classical norms.25
For this reason, we accounted for the level of literacy following the methodology proposed by Herman.26 We examined the frequency of the confusion between <ae> and <e> for the later texts, where most of the <b>/<v> alternations are attested. According to Herman, this datum could provide a sort of ‘cultural index’ for each region. Since the monophthongization ae > e was accomplished centuries before the end of the Empire, <ae>/<e> confusions in later texts might depend on the low orthographic competence of the author(s).27
In other words, following this hypothesis, a high frequency of <ae>/<e> confusions (calculated over the total number of other divergent spellings) might hint at a low level of literacy of those involved in the crafting of the inscriptions and thus help with the interpreting of our data. The results are displayed in Table 16 below and are calculated for later public and funerary inscriptions (after 300 CE), which display most of the <b>/<v> alternations examined here.28
<ae>/<e> confusions in public and funerary inscriptions (after 300 CE)
Sabratha | Other areas | Lepcis Magna | Theueste | |||||
N. | % | N. | % | N. | % | N. | % | |
<ae>/<e> | 36 | 35% | 7 | 17% | 8 | 14% | 1 | 6% |
Other divergent spellings | 68 | 65% | 34 | 83% | 51 | 86% | 15 | 94% |
Total | 104 | 100% | 41 | 100% | 59 | 100% | 16 | 100% |
A different incidence of <ae>/<e> confusion can be spotted in the examined areas. Sabratha displays the highest frequency of confusion (35%), whereas the ‘other areas’ subset and Lepcis Magna display a lower incidence of confusion (17% and 14%, respectively). Not enough data are available for Theueste to make valid observations. Of course, the data sparseness forces us to proceed with caution. However, following Herman's hypothesis, Sabratha seems to display a lower level of literacy than the other areas examined – a variable that might affect our data.
Given these results, it is important to consider literacy as a variable. To do so, we adopted the methodology proposed by Herman and more recently by Béla Adamik,29 i.e. the relative frequency of the <b>/<v> confusions was calculated over the total number of other consonantal divergent spellings found in each area.30 Since most of the data involve the use of <b> for <v> in public and funerary inscriptions dating from 300 CE onwards (§3.2), we limited our analysis to these occurrences. The results are reported in Table 17 below.31
<b> for <v> in public and funerary inscriptions (after 300 CE) and other consonantal divergent spellings
Sabratha | Lepcis Magna | Other areas | Theueste | |||||
N. | % | N. | % | N. | % | N. | % | |
<b> for <v> | 35 | 65% | 15 | 38% | 8 | 24% | 1 | 10% |
Other C | 19 | 35% | 24 | 62% | 25 | 76% | 9 | 90% |
Total | 54 | 100% | 39 | 100% | 33 | 100% | 10 | 100% |
These results allow us to make two important observations. Firstly, the validity of the <ae>/<e> confusions as a ‘cultural index’ seems to be confirmed by the higher number of total consonantal divergent spellings found in Sabratha in comparison with the other areas (54 vs. <40), although a smaller number of texts is available for this subset (88 vs ca. 250; §2).
Secondly, although <b>/<v> confusions are attested in all areas, the application of this methodology on a subset of synchronic inscriptions seems to confirm the higher frequency of the phenomenon found in Sabratha (65%), which almost doubles the one found in Lepcis Magna (38%) and in the other areas (24%). For Theueste the data are too scanty to make valid observations (10 total tokens). The causes of this internal diatopic difference are not clear: as was already mentioned for the whole of Roman Africa, the traditional explanations involving e.g. contact with Greek or ‘Libyan’ do not apply to these areas. There is no evidence of a strong Greek influence in the examined regions that would justify the transfer of a phonetic feature of Greek into African Latin, and little information is available on the phonology of the local languages in contact with Latin (except for Punic) that were part of the complex linguistic background of Roman Africa.32 Further research is needed to address this issue.
5 Some conclusions and methodological remarks
The fine-grained examination of a subset of inscriptions from North Africa illustrated so far allows us to add new data regarding the incidence of <b>/<v> confusions and the literacy level in the examined areas, on the one hand, and to make some methodological observations, on the other.
Firstly, most of the alternations involve the use of <b> for <v> and are found in later texts (after 300 CE) in intervocalic position, also when phonosyntax is considered, coherently with the trend evidenced for Latin inscriptions from other areas of the Empire.33 Furthermore, a diatopic differentiation seems to be found, also when dating, text type, available contexts and literacy level are considered: Sabratha seems to display a higher incidence of confusion than the other areas examined. The causes of this internal diatopic difference are unclear, and we intend to tackle this issue in the near future.
Secondly, different levels of literacy seem to be attested in the several areas examined, with Sabratha displaying the lowest literacy level among the examined regions. In this respect, the frequency of <ae>/<e> confusions seems to be a valid ‘cultural index’ to be considered when carrying out linguistic analysis.
In addition, a few methodological remarks are needed. Firstly, caution is needed due to the low number of available data, which is an inescapable shortcoming of epigraphic texts. The reliability of our observations might also be compromised by the imbalance in the size of the corpora. However, we believe that these issues might be at least partially overcome by calculating the relative frequency of the examined spellings and by taking into account extralinguistic variables such as dating and text type. Finally, we believe it is vital to consider the level of literacy of the writers, in addition to the total number of available contexts of occurrence of the phenomenon: only in this way are we able to make more reliable observations, despite the imbalance in the available data.
Aknowledgements/Funding information
The present paper was prepared within the framework of the PRIN Project ‘Ancient languages and writing systems in contact: a touchstone for language change’ (2017JBFP9H) and the HORIZON-ERC-2022-ADG project no. 101098102 ‘Digital Latin Dialectology (DiLaDi): Tracing Linguistic Variation in the Light of Ancient and Early Medieval Sources’ (see http://lldb.elte.hu/). I wish to thank professors Giovanna Marotta, Francesco Rovai and Béla Adamik for their insightful comments on a preliminary version of this work.
Bibliography
Acquati, A. (1974). Il consonatismo latino-volgare nelle iscrizioni africane. Annali della Facoltà di lettere e filosofia dell'Università degli studi di Milano, 27(1): 21–56.
Adamik, B. (2012). In search of the regional diversification of Latin: some methodological considerations in employing the inscriptional evidence. In: Biville, F.. (ed.), Latin vulgaire – latin tardif IX: Actes du IXe colloque international sur le latin vulgaire et tardif, Lyon, 6–9 septembre 2009. Maison de l’Orient e la Méditerranée, Lyon: pp. 123–139.
Adamik, B. (2017a). On the Vulgar Latin merger of /b/ and /w/ and its correlation with the loss of intervocalic /w/: Dialectological evidence from inscriptions. PALLAS, Revue d’études antiques, 103: 25–36.
Adamik, B. (2017b). Potential Greek influence on the Vulgar Latin sound change [b]>[β]: Dialectological evidence from inscriptions. Acta Antiqua Academiae Scientiarum Hungaricae, 57: 11–33.
Adams, J.N. (2007). The Regional Diversification of Latin, 200 BC-AD 600. Cambridge University Press, Cambridge.
Adams, J.N. (2013). Social Variation and the Latin Language. Cambridge University Press, Cambridge.
Allen, W.S. (1965). Vox Latina: A Guide to the Pronunciation of Classical Latin. 2nd ed. Cambridge University Press, Cambridge.
Barbarino, J.L. (1978). The Evolution of the Latin /b/–/ṷ/ Merger: A Quantitative and Comparative Analysis of the B–V Alternation in Latin Inscriptions. U.N.C. Dept. of Romance Languages, Chapel Hill.
Chabot, J.B. (1940–1941). Recueil des Inscriprions libyques. Imprimerie nationale, Paris.
Clackson, J. – Horrocks, G. (2007). The Blackwell History of the Latin Language. Blackwell Publishing, Malden, Mass.
De Paolis, P. (2010). Problemi di grafia e pronunzia del latino nella trattatistica ortografica tardoantica. In: Anreiter, P. – Kienpointner, M. (eds.), Latin Linguistics Today. Akten des 15. Internationalen Kolloquiums zur Lateinischen Linguistik, Innsbruck, 4.-9. April 2009. Institut für Sprachen und Literaturen der Universität, Innsbruck: pp. 57–74.
Herman, J. (1965). Aspects de la différenciation territoriale du latin sous l’Empire. Bulletin de la Société de Linguistique de Paris, 60: 53–70 = Herman, J. (1990). Du latin aux langues romanes. Études de linguistique historique. Niemeyer, Tübingen, pp. 10–28.
Herman, J. (2000). Differenze territoriali nel latino parlato dell’italia: un contributo preliminare. In: Herman, J. – Marinetti, A. (eds.), La preistoria dell’italiano. Atti della Tavola Rotonda di Linguistica Storica. Università Ca’ Foscari di Venezia 11–13 giugno 1998. Niemeyer, Tübingen: pp. 123–135.
Lancel, S. (1981). La fin et la survie de la latinité en Afrique du Nord: état des questions. Revue des études latines, 59: 269–97.
Lupinu, G. (2000). Latino epigrafico della Sardegna. Aspetti fonetici. Ilisso, Nuoro.
Mancini, M. (2005). La formazione del neostandard latino: il caso delle differentiae uerborum. In: Kiss, S. – Mondin, L. – Salvi, G. (eds.), Latin et langues romanes. Études linguistiques offertes à József Herman à l’occasion de son 80ème anniversaire. Niemeyer, Tübingen: pp. 137–155.
Matthews, K.D. (1957). Cities in the Sand. Lepcis Magna and Sabratha in Roman Africa. University of Pennsylvania Press, Philadelphia.
Millar, F. (1968). Local Cultures in the Roman Empire: Libyan, Punic and Latin in Roman Africa. The Journal of Roman Studies, 58(1–2): 126–134.
Niedermann, M. (1991). Précis de phonétique historique du latin. 5th ed. Klincksieck, Paris.
Pugliarello, M. (2006). Lingua scritta e lingua parlata nel trattato di Martirio De b muta et v vocali. In: Taifacos, I. (ed.), The Origins of European Scholarship: The Cyprus Millennium International Conference, April 6–8 2000, Nicosia. Steiner Verlag, Berlin: pp. 121–129.
Pugliarello, P. (2011). Martirio orthographus: lezioni di corretta scrittura. Koinonia, 35: 123–143.
Sturtevant, E.H. (1920). The Pronunciation of Greek and Latin. University of Chicago Press, Chicago.
Tamponi, L. (2019). The confusion between and <v> in Latin inscriptions from Sardinia. Acta Classica Universitatis Scientiarum Debreceniensis, 55: 127–146.
Tamponi, L. (2022). Variation and Change in Sardinian Latin. The Epigraphic Evidence. Pisa University Press, Pisa.
Väänänen, V. (1965). Étude sur le texte et la langue des Tablettes Albertini. Suomalainen Tiedeakademia, Helsinki.
Väänänen, V. (1963). Introduction au latin vulgaire. Klincksieck, Paris.
Sturtevant (1920) 43; Adams (2013) 183.
Adams (2007) 644 ff.
Allen (1965) 41; Herman (2000) 38–39; Adams (2013) 183–185; (Adamik 2017a) 15.
Mancini (2005) 148.
Herman (1965); Barbarino (1978); Adamik (2017a, 2017b); Herman (2000); Lupinu (2000); Tamponi (2019, 2022).
See Väänänen (1965) and the figures reported by Adams (2007) 643 ff.
See recently Adamik (2017b).
See Adams (2007) 665; Adamik (2017b).
Acquati (1974); see also Lancel (1981) 281.
Millar (1968); Adams (2007) 685.
The datings provided on the basis of linguistic criteria were excluded to avoid the danger of circular argumentation in linguistic analysis.
Carthago, Thugga and Ammaedara were not included in our corpus: since they feature a considerably higher number of texts (more than 400 for Carthago and Ammaedara and almost 300 for Thugga), their inclusion would have caused data imbalance as for the provenance place of the texts.
See e.g. Matthews (1957).
This datum led Adamik (2017b) to propose a Greek influence for the alternations, since in the Greek dialects the sound change [b] > [β] > [v] is attested in all positions, at least by the 1st century CE, and – unlike Latin – affected syllable-final position as well.
In Theueste, the low number of data (only 2 occurrences) leads to an overestimation of the occurrences in intervocalic position, since one of the two instances of the phenomenon is found in Nobembres for Novembres, i.e. in intervocalic position. However, this datum is coherent with the general trend shown by the other areas when phonosyntax is examined (see infra).
For this methodology, we followed Adamik (2017b).
Niedermann (1991) 87–88; Väänänen (1963) 50.
Among Herman's and Adamik's studies, see e.g. Herman (2000) and Adamik (2012).
Herman (2000); see also Adamik (2012) 135.
Herman (2000) 126, n. 7: “Segnaliamo un punto di particolare importanza: dato che alla fine dell'impero la monottongazione ae > e era un processo compiuto da secoli, le frequentissime fluttuazioni e incertezze fra AE e E sono semplicemente ortografiche, senza incidenza fonetica. La loro frequenza dipende dal livello culturale, di competenza ortografica, e non dalla pronuncia.”
If we examine separately funerary and public inscriptions, the data are too scanty to make valid observations. In particular, in Sabratha and Theueste, almost all the data belong to funerary texts (only two other divergent spellings are found in tituli honorarii from Sabratha and four in those from Theueste); the number of tokens for public inscriptions from the ‘other areas’ subset is also low (only 10 instances are available). However, in Lepcis Magna, where more data are available, the results do not change significantly. The frequency of <ae>/<e> confusions is 14% in funerary inscriptions and 13% in public texts (vs 13% reported in Table 16).
See e.g. Herman (1965, 2000); Adamik (2012, 2017a, 2017b).
The other consonantal divergent spellings taken into account are the following: omission of consonants (final -s, -m, -t), insertion of consonants, consonant doubling, occurrences of single for double consonants, confusion between voiced and voiceless stops.
Public inscriptions were included since occurrences of <b> for <v> are found in tituli honorarii from Lepcis Magna (2 occurrences vs 17 other consonantal divergent spellings); in the other areas, no confusions are attested in this text type, although other divergent spellings are found (2 in Sabratha, 4 in the other areas, 3 in Theueste). However, even excluding public inscriptions from the analysis, the results still indicate a higher frequency of the confusions in Sabratha, although the difference with Lepcis Magna is less marked: the instances of <b> for <v> cover 67% of the consonantal divergent spellings in Sabratha (35 occurrences vs 17 other spellings) and 65% in Lepcis Magna (<b> for <v>: 13 occurrences; other consonantal divergent spellings: 17); their frequency is lower in the other areas (28%, i.e. <b> for <v>: 8; other divergent spellings: 21) and in Theueste (14%, i.e. <b> for <v>: 1; other divergent spellings: 6).
Se e.g. Millar (1968) and Clackson–Horrocks (2007) 86–87 on Libyan.