African Romance

The Fossa regia (in pink) marked the approximate border between the province of Africa and Numidia.

The Roman province of Africa was organized in 146 BC following the defeat of Carthage in the Third Punic War. The city of Carthage, destroyed following the war, was rebuilt during the dictatorship of Julius Caesar as a Roman colony, and by the 1st century, it had grown to be the fourth largest city of the empire, with a population in excess of 100,000 people.[upper-alpha 1] The Fossa regia was an important boundary in North Africa, originally separating the Roman occupied Carthaginian territory from Numidia,[5] and may have served as a cultural boundary indicating Romanization.[6]

In the time of the Roman Empire, the province had become populous and prosperous and Carthage was the second-largest Latin-speaking city in the Empire. Latin was, however, largely an urban and coastal speech. Carthaginian Punic continued to be spoken in inland and rural areas as late as the mid-5th century, but also in the cities.[7] It is probable that Berber languages were spoken in some areas as well.

Funerary stelae chronicle the Romanization of art and religion in North Africa.[8] Notable differences, however, existed in the penetration and survival of the Latin, Punic and Berber languages.[9] These indicated regional differences: Neo-Punic had a revival in Tripolitania, around Hippo Regius there is a cluster of Libyan inscriptions, while in the mountainous regions of Kabylie and Aures, Latin was scarcer, though not absent.[9]

Africa was occupied by the Germanic Vandal tribe for over a century, between 429 and 534 AD, when the province was reconquered by the Byzantine Emperor Justinian I. The changes that occurred in spoken Latin during that time are unknown. Literary Latin, however, was maintained at a high standard, as seen in the Latin poetry of the African writer Corippus. The area around Carthage remained fully Latin-speaking until the arrival of the Arabs.

An inscription from one of the gates to the theatre at Leptis Magna, indicating that Latin and Punic co-existed in Northern Africa for centuries.

Like all Romance languages, African Romance descended from Vulgar Latin, the non-standard (in contrast to Classical Latin) form of the Latin language, which was spoken by soldiers and merchants throughout the Roman Empire. With the expansion of the empire, Vulgar Latin came to be spoken by inhabitants of the various Roman-controlled territories in North Africa. Latin and its descendants were spoken in the Province of Africa following the Punic Wars, when the Romans conquered the territory. Spoken Latin, and Latin inscriptions developed while Punic was still being used.[10] Bilingual inscriptions were engraved, some of which reflect the introduction of Roman institutions into Africa, using new Punic expressions.[10]

Latin, and then some Romance variant of it, was spoken by generations of speakers, for about fifteen centuries.[2] This was demonstrated by African-born speakers of African Romance who continued to create Latin inscriptions until the first half of the 11th century.[2] Evidence for a spoken Romance variety which developed locally out of Latin persisted in rural areas of Tunisia – possibly as late as the last two decades of the 15th century in some sources.[11]

By the late 19th century and early 20th century, the possible existence of African Latin was controversial,[12] with debates on the existence of Africitas as a putative African dialect of Latin. In 1882, the German scholar Karl Sittl used unconvincing material to adduce features particular to Latin in Africa.[12] This unconvincing evidence was attacked by Wilhelm Kroll in 1897,[13] and again by Madeline D. Brock in 1911.[14] Brock went so far as to assert that "African Latin was free from provincialism",[15] and that African Latin was "the Latin of an epoch rather than that of a country".[16] This view shifted in recent decades, with modern philologists going so far as to say that African Latin "was not free from provincialism"[17] and that, given the remoteness of parts of Africa, there were "probably a plurality of varieties of Latin, rather than a single African Latin".[17] Other researchers believe that features peculiar to African Latin existed, but are "not to be found where Sittl looked for it".[17]

Fifth century AD inscription at the Leptis Magna forum, Libya.

While as a language African Romance is extinct, there is some evidence of regional varieties in African Latin that helps reconstruct some of its features.[18] Some historical evidence on the phonetic and lexical features of the Afri were already observed in ancient times. Pliny observes how walls in Africa and Spain are called formacei, or "framed walls, because they are made by packing in a frame enclosed between two boards, one on each side".[19] Nonius Marcellus, a Roman grammarian, provides further, if uncertain, evidence regarding vocabulary and possible "Africanisms".[20][upper-alpha 2] In the Historia Augusta, the North African Roman Emperor Septimius Severus is said to have retained an African accent until old age.[upper-alpha 3] More recent analysis focuses on a body of literary texts, being literary pieces written by African and non-African writers.[23] These show the existence of an African pronunciation of Latin, then moving on to a further study of lexical material drawn from sub-literary sources, such as practical texts and ostraca, from multiple African communities, that is military writers, landholders and doctors.[23]

The Romance philologist James Noel Adams lists a number of possible Africanisms found in this wider Latin literary corpus.[24] Only two refer to constructions found in Sittl,[25] with the other examples deriving from medical texts,[26] various ostraca and other non-traditional sources. Two sorts of regional features can be observed. The first are loanwords from a substrate language, such is the case with Britain. In African Latin, this substrate was Punic. African Romance included words such as ginga for "henbane", boba for "mallow," girba for "mortar" and gelela for the inner flesh of a gourd.[27] The second refers to use of Latin words with particular meanings not found elsewhere, or in limited contexts. Of particular note is the African Romance use of the word rostrum for "mouth" instead of the original meaning in Latin, which is "beak",[28] and baiae for "baths" being a late Latin and particularly African generalisation from the place-name Baiae.[29] Pullus meaning "cock" or "rooster", was probably borrowed by Berber dialects from African Romance, for use instead of the Latin gallus.[30] The originally abstract word dulcor is seen applied as a probable medical African specialisation relating to sweet wine instead of the Latin passum or mustum.[31] The Latin for grape, traditionally indeterminate (acinis), male (acinus) or neuter (acinum), in various African Latin sources changes to the feminine acina.[32] Other examples include the use of pala as a metaphor for the shoulder blade; centenarium, which only occurs in the Albertini Tablets and may have meant "granary";[33] and infantilisms such as dida, which apparently meant "breast/nipple" or "wet nurse".[34]'

Both Africans, such as Augustine of Hippo and the grammarian Pompeius, as well as non-Africans, such as Consentius and Jerome, wrote on African features, some in very specific terms.[35] Indeed in his De Ordine, dated to late 386, Augustine remarks how he was still criticised by the Italians for his pronunciation, while he himself often found fault with theirs.[36] While modern scholars may express doubts on the interpretation or accuracy of some of these writings, they contend that African Latin must have been distinctive enough to inspire so much discussion.[37]

The Exarchate of Africa within the Byzantine Empire after the reconquest of Justinian.

Prior to the Arab conquest in 696–705 AD, a Romance language was probably spoken alongside Berber languages in the region.[38] Loanwords from Northwest African Romance to Berber are attested, usually in the accusative form: examples include atmun ("plough-beam") from temonem.[38]

Following the conquest, it becomes difficult to trace the fate of African Romance. The Umayyad administration did at first utilize the local Latin language in coinage from Carthage and Kairouan in the early 7th century, displaying Latin inscriptions of Islamic phrases such as D[e]us tu[us] D[e]us et a[li]us non e[st] ("God is your God and there is no other"), a variation of the shahada, or Muslim declaration of faith.[39] African Latin was soon replaced by Arabic as the primary administrative language, but it existed at least until the arrival of the Banu Hilal Arabs in the 11th century and probably until the beginning of the 14th century.[40] It likely continued to be widely spoken in various parts of the littoral of Africa into the 12th century,[1] exerting a significant influence on Northwest African Arabic, particularly the language of northwestern Morocco.[38]

Map highlighting in black the "Romania submersa", being those Roman, or formerly Roman regions where forms of neo-Latin disappeared after some centuries, including Northern Africa.

Amongst the Berbers of Ifriqiya, African Romance was linked to Christianity, which survived in North Africa (outside of Egypt) until the 14th century.[3] Spoken Latin or Romance is attested in Gabès by Ibn Khordadbeh; in Béja, Biskra, Tlemcen, and Niffis by al-Bakri; and in Gafsa and Monastir by al-Idrisi,[1] who observes that the people in Gafsa "are Berberised, and most of them speak the African Latin tongue."[1][41][upper-alpha 4] There is also a possible reference to spoken Latin or African Romance in the 11th century, when the Rustamid governor Abu Ubayda Abd al-Hamid al-Jannawni was said to have sworn his oath of office in Arabic, Berber and in an unspecified "town language", which might be interpreted as a Romance variety; in the oath, the Arabic-rendered phrase bar diyyu could represent some variation of Latin per Deu(m) ("by God".)[42] In their quest to conquer the Kingdom of Africa in the 12th century, the Normans were aided by the remaining Christian population of Tunisia, who some linguists, among them Vermondo Brugnatelli, argue had been speaking a Romance language for centuries.[43]

The final attestations of African Romance come from the Renaissance period. The 15th century Italian humanist Paolo Pompilio makes the most significant remarks on the language and its features, reporting that a Catalan merchant named Riaria who had lived in North Africa for thirty years told him that the villagers in the Aurès mountain region "speak an almost intact Latin and, when Latin words are corrupted, then they pass to the sound and habits of the Sardinian language".[44] The 16th century geographer and diplomat Leo Africanus, who was born into a Muslim family in Granada and fled the Reconquista to Morocco, also says that the North Africans retained their own language after the Islamic conquest which he calls "Italian", which must refer to Romance.[45] A statement by Mawlâ Aḥmad is sometimes interpreted as implying the survival of a Christian community in Tozeur into the eighteenth century, but this is unlikely; Prevost estimates that Christianity disappeared around the middle of the thirteenth century in southern Tunisia.[3]

The earliest known portrait of Saint Augustine in a 6th-century fresco, Lateran, Rome.

Starting from African Romance's similarity with Sardinian, scholars theorise that the similarity may be pinned down to specific phonological properties.[11] Sardinian lacks palatization of velar stops before front vowels, and features the pairwise merger of short and long non-low vowels.[2] Evidence is found that both isoglosses were present in African Latin (at least in the central province of Africa Proconsularis):

• Velar stops also remain unaffected in Latin loanwords in Berber.[2] For example, tkilsit ("mulberry tree") < (morus) celsa in Latin,[75] and i-kīkər ("chickpea") < cicer in Latin,[76] or ig(e)r , ("field") < ager in Latin.[77]
• Inscriptions from Tripolitania, written as late as the 10th or 11th century are written with a ⟨k⟩, diverging from contemporary European Latin uses.[78] Thus, there are forms such as dikite ("say", 2nd person plural imperative), iaket ("he/she lies down"), dekember ("December") and pake ("peace") with ⟨k⟩ such as dilektus ("beloved"), karus ("dear") and Afrikana ("African"), found in an inscriptional corpus.[79]
• Some evidence that Latin words with a "v" are often written with a "b" in African Romance, as reported by Isidore of Seville: birtus ("virtue") < virtus in Latin, boluntas ("will") < voluntas, and bita ("life") < vita.[upper-alpha 10] Adams and Adamik both observe frequent b/v confusions in inscriptions and non-literary documents from Africa Proconsularis.[81][82] On the other hand, according to Adamik Mauretania Caesariensis shows a much lower rate of betacism, comparable to Hispania or Gaul.[83]
• The 5th century Albertini Tablets suggest high levels of phonetic errors and an uncertainty in the use of Latin cases.[84]
• In a study of errors on stressed vowels in a corpus of 279 inscriptions, scholars noted how African inscriptions confused between over-stressed and under-stressed vowels between the 1st and 4th century AD, with Rome reaching comparable error rates only by the late 4th to 6th centuries.[85]
• As aforementioned, there is strong evidence that the Latin/Romance of Africa Proconsularis shared an identical five-vowel system (probably /a, ɛ, i, ɔ, u/) with Sardinian.[78] Augustine of Hippo's testimony on how ōs ("mouth") in Latin was to African ears indistinguishable from ŏs ("bone") indicates the merger of vowels and the loss of the original allophonic quality distinction in vowels.[86][upper-alpha 11]
• Analysis of inscriptions and non-literary documents, including the 3rd century Bu Njem ostraca and Albertini Tablets demonstrate that, despite the presence of other common Vulgar Latin sound changes (such as loss of aspirate h, monophthongization of ae and loss of final m), confusion between ē and ǐ and ō and ŭ is almost nonexistent. Adams finds that inscriptions display a rate of ē/ǐ and ō/ŭ mistakes of only 0.7%, while in the Bu Njem ostraca ē/ǐ and ō/ŭ confusion is totally absent and in the Albertini Tablets there are fewer than two of each error.[88] In contrast, by the 6th-7th c. inscriptions from Gaul show that misspellings of ē/ǐ had come to surpass instances of correct spellings, with an error rate of 51%[89] Data by Adamik shows very slightly higher rates of confusion in Africa Proconsularis (5.3%), but still far lower than in Gaul (63.5%) or Hispania (21.3%), and he similarly concludes that "the later Latin of Africa Proconsularis undoubtedly belonged to the Sardinian Romance type of vocalism."[90]
• In Latin loanwords in Berber, Latin short ĭ, ŭ also result in i, u (instead of e, o) as in Sardinian. For example, pullus ("chicken") > afullus, cicer ("chickpea") > i-kīkər, pirus ("pear tree") > ti-firest.[91] However, as Adams points out, assumptions about African Romance vocalism based solely on loanwords should be taken with caution due to the lack of vowels /e, o/ in Berber languages.[92][93]
• On the contrary, Adamik states that data from inscriptions show that the vowel system of the western province of Mauretania Caesariensis seemed to be developing differently from that of Africa Proconsularis/Sardinia. Due to a higher number of ē/ǐ confusions found than ō/ŭ confusions (at 4.6% vs. 1.3%), it is suggested that the vocalic system of Mauretanian Latin might possibly "have started to develop toward the eastern or Balkan (more exactly Rumanian) type of vocalism": e.g., merging ē with ĭ, ĭ with ī and ŭ with ū, potentially resulting in the six-vowel system /a, ɛ, e, i, o, u/. However, spelling error rates are still too low for a definite conclusion to be made on the classification of the dialect of Mauretania Caesariensis.[65]
• There is additional evidence both from commentary and metrical inscriptions for confusion over syllable length and stress resulting from the collapse of vowel length distinctions. In another passage, Augustine writes that a sound change which was criticized by grammarians was to lengthen formerly short stressed vowels in words like cano ("to sing'), e.g. as [ˈkaːnɔ] instead of [ˈkanoː].[upper-alpha 12] Consentius makes a similar remark that Africans tended to mispronounce piper ("pepper") with the formerly short stressed syllable lengthened, e.g. [ˈpiːpɛr] instead of [ˈpɪpɛr],[upper-alpha 13] while also shortening formerly long unstressed syllables, as in orator with short o, e.g. [ɔˈraːtɔr] instead of [oːraːtɔr].[upper-alpha 14]
• Herman also finds evidence metrical inscriptions from the 1st-4th centuries AD for the lengthening of short stressed vowels and shortening of long unstressed vowels, which could point to an earlier loss of contrastive vowel length than in Rome.[96] Adams' analysis of 3rd century poems from Bu Njem written by Italian and African soldiers seem to reflect a dialectal contrast in the vowel systems of the two regions, with the Italian Avidius writer preserving Classical prosody and the African Iasucthan displaying vowel length distinction loss (given that Iasucthan is assumed to have been a native Latin speaker.)[97]

• Africitas, a purported "style" of African Latin.
• Southern Romance, a proposed hypothetical Romance classification, including Sardinian and African Romance
• Lausberg area, a region of southern Italy covering Basilicata where the local Neapolitan dialects display variation between Sardinian, Sicilian and Romanian-like vowel systems
• Sardinian language, as stated above, theorized to be the closest surviving language to African Romance.
• British Latin, another extinct dialect of Latin.
• Moselle Romance, another extinct dialect of Latin.
• Pannonian Romance, another extinct dialect of Latin.

• Sayahi, Lotfi (2014). Diglossia and Language Contact: Language Variation and Change in North Africa. Cambridge University Press. ISBN 9780521119368.