Non-standard Latin and local influences in divine names: vowel and declension variants in the religious epigraphy of Roman Pannonia

Introduction

The naming of gods is an important aspect of the interaction between individual(s) and deities. The names and epithets of the gods are part of cultic tradition¹ and are usually very conservative, although the Pannonian epigraphic corpus shows various vernacular Latin traits in theonyms and epithets that differ from the standard language. Linguistic information from inscriptions can provide valuable primary evidence of ritual communication. The onomastic sequences of varying complexity have been used to precisely identify the god addressed. In this regard, periodic (sometimes for shorter periods) and geographical patterns of the gods are observed in the votive inscriptions. The provincial capitals of Carnuntum and Aquincum, together with their immediate surroundings, are disproportionately represented in the Pannonian epigraphic corpus – accounting for over 1,500 votive inscriptions – due to their roles as major administrative and military centres, which naturally generated higher levels of epigraphic activity. The epigraphic corpus clearly reflects the predominance of deities tied to the official Roman cult, with Jupiter, the Genii, and Silvanus receiving the vast majority of dedications – far surpassing local or foreign divinities in frequency.

People from other parts of the Empire migrated to Pannonia, mainly by military service. The largest numbers of Syrians and Thracians were found in the Pannonian army, but there were also significant numbers of Africans and people from Asia Minor.² These temporary or permanent inhabitants brought their native gods with them and their theonyms were integrated within a wholly different cultural and linguistic landscape, resulting in inscriptions depicting a wide range of deities. The name of the deity, through interpretation, could be Roman, a hybrid (Latin name and foreign epithet) or even a traditional Celtic or Pannonian name.

Latin was primarily used in the inscriptions of Pannonia, but individual features (e.g. non-Latin suffixes or stems) in the epithets – Jupiter Optimus Maximus Teutanus, Mars Latobius, Mars Marmogius, etc. – can commonly be expected if the god of a province or a narrower area was worshipped. Non-Roman divine names may have derived either from imported cults or from the reinterpretation of indigenous deities – particularly those with Celtic-root theonyms – within a Latin epigraphic framework. In both scenarios, however, they were worshipped in a recognisably Roman manner, having been integrated into the Latin language and the unmistakably Roman social and cultural contexts of Pannonia. These names are somewhat ambiguous for Latin speakers, either because their form and phonetic formation are foreign to Latin (e.g. Dolichenus and Elagabalus), or because the worshippers themselves were not native Latin speakers, and determining how to incorporate the deity’s name into the Latin text would have required a level of linguistic intuition that they lacked.³ These Latin inscriptions, which invoke both Roman and non-Roman divinities, reflect the multilingual character of the Roman provinces. They exhibit linguistic phenomena that deviate from a purely monolingual use of Latin. The native languages spoken by provincial communities (i.e. the languages used in everyday interpersonal communication – Greek, Celtic, etc.) influenced the use of Latin – the official epigraphic language – manifesting occasionally in lexical or morphological interference.

In Pannonia the number of bi- or multilingual inscriptions is extremely limited: only eleven bilingual inscriptions (Greek and Latin) are known, and only a few of these are votive.⁴ The Greek-speaking population – including both Greeks and other Eastern groups – rapidly assimilated after settling in Pannonia, and their inscriptions were written in Latin.⁵ In the larger Eastern communities, such as those at Intercisa, Ulcisia, and around the legionary forts in Carnuntum, Aquincum, and Brigetio, Latin was the language consistently chosen for inscriptions, whether on instrumenta domestica, funerary monuments, or votive dedications. Strikingly, no inscriptions written in Levantine languages have been found in Pannonia,⁶ despite the presence of substantial Syrian, Palmyrene, and Jewish populations in certain parts of the province.

Although relatively limited in number, the epigraphic corpus of inscriptions exhibiting Latin linguistic interference in divine names reflects a range of underlying phenomena. A previous study has shown certain spatial differences in the declension and case confusion of divine names and their epithets,⁷ with Pannonia and the Gallic provinces displaying the highest frequency of such variations. What accounts for this phenomenon? Why do these inscriptions – normally expected to be among the most linguistically conservative – display the greatest degree of linguistic change in Pannonia? Could the influence of a substratum such as the Celtic language, which survived in Pannonia until the 2nd c. AD, be responsible for the confusion of theonyms and epithets?⁸ W. Meid has addressed this issue in relation to personal nomenclature and has demonstrated that the population still understood pre-Roman names, even when their original etymologies may already have been forgotten.⁹

Examining the theonyms, it becomes evident that most of the mixed deity names appear in Latin inscriptions that incorporate indigenous or non-Latin morphological elements. While differing in emphasis from other contributions in this volume, this chapter aims to offer a complementary perspective by analysing the names of the gods in the Pannonian inscriptions primarily from a linguistic standpoint – focusing in particular on their phonetic adaptation – in light of previous research in the field.¹⁰ At the same time, it acknowledges that these theonyms are embedded within specific social contexts shaped by processes of cultural negotiation, as explored in detail in other chapters of this volume.

Methodology

The analysis is based on the “Computerized Historical Linguistic Database of the Latin Inscriptions of the Imperial Age” (LLDB),¹¹ which contains more than 2 070 datasheets from Pannonia Inferior and 2 210 datasheets from Pannonia Superior, recording the linguistic phenomena appearing in the inscriptions. Previous studies have addressed issues of case and declension confusion,¹² as well as certain preliminary problems concerning theonyms and common epithets more generally.¹³ The history of the Latin language in Pannonia has already been studied by B. Fehér, who analysed the “correct” and “incorrect” forms in the province’s inscriptions on a statistical basis.¹⁴ In contrast to B. Fehér’s methodology, this chapter focuses on the relative proportions of “error” types¹⁵ and draws its data from the LLDB database.

The corpus of votive inscriptions, focusing on non-standard forms of divine names and epithets, comprises 191 LLDB datasheets (135 inscriptions)¹⁶ from both Pannoniae. Deviations from the normative forms of these names can have phonological, morphological, or syntactic causes. Therefore, it is preferable to include the entire corpus derived from the 135 inscriptions in the quantitative analysis, rather than isolating individual occurrences based on presumed causes of linguistic variation. The corpus comprises 15 instances of declensional system confusion, 28 instances of case confusion, 84 instances involving phonological changes, 37 instances of consonant alternations, and 15 instances categorized as miscellaneous non-grammatical errors.¹⁷ The challenges related to the analysis of theonyms and epithets were addressed in a previous study,¹⁸ which identified several limitations. First, there are chronological constraints, as the data span from the 1st to the 3rd ^c. AD. The LLDB database examines linguistic changes beginning in the Augustan period, and by the end of the 3rd c. AD, the number of votive inscriptions declines due to changing epigraphic habits. Second, there are spatial limitations, as the cultic practices associated with certain deities exhibit a distinctly localized character. For instance, the worship of Vidasus and Thana is attested only in the southern Pannonia area of Ad Fines, while the goddesses Quadriviae are known exclusively from Carnuntum and its immediate surroundings. Third, there are also linguistic limitations, as divine names usually belong to the first three declensions (first declension: a-stem; second declension: o-stem; and third declension: consonant or i-stem), being “proper names”,¹⁹ while epithets (typically adjectives) generally fall in the same categories. In the inscriptions, the formulae in which the names of deities appear are most often declined in the dative case. Several factors can explain the syntactic errors observed in the inscriptions, including the contamination between formulaic expressions and the scribes’ limited knowledge of Latin, especially in the phonetic rendering of foreign names.²⁰

Vowel Changes

In the Pannonian corpus of divine names and epithets, 82 cases of vowel confusion have been identified. Several of these phenomena are so widespread and universal across the Empire that they occur in multiple regions. The divine names²¹ most commonly exhibiting such quasi-mistakes are Mithras²² and Jupiter Dolichenus.²³ In Pannonia, 15 examples are attested in which a vowel change occurs in the name of Mithras. ²⁴Partly as a consequence of such changes, Jupiter Dolichenus likewise appears in Pannonian inscriptions in a range of different forms.²⁵

Figure 1 illustrates that the highest proportion corresponds to monophthongisation. In his work, B. Fehér posits that the monophthongisation of the /ae/ diphthong was a particularly rapid process in Pannonia from the time of Trajan onwards, whereas the development of other diphthongs remains unclear.²⁶ A total of 42 instances of this type of monophthongisation can be found in divine names, all of which occur in the –ae dative ending.²⁷

In addition to the monophthongisation of the dative ending, other types of vowel change can be observed in our corpus. A more detailed examination of the form Asculepio, known from Aquincum (AE 1937, 182), suggests two possible explanations. Firstly, phonetic changes may have occurred: the monophthongisation of the ae diphthong at the beginning of the name and the alteration of the e~a sound in the –lap– syllable. Secondly, the form may represent a combination of Latin and Greek variants. The use of either the Greek or Latin form could have depended on the dedicator’s intention and language choice, as well as on local traditions and prevailing trends. In Aquincum, both variants were represented,²⁸ with the Latin form being particularly prevalent.

On occasion, as a hypercorrection of these monophthongisation tendencies (5 %), the vowels /a/ or /e/ are written as the diphthong <ae>. The hypercorrect diphthong is attested in the name Vag|daevercus|ti (LLDB-143622; AE 1935, 163), recorded in an inscription from Vetus Salina (Pannonia Inferior), where the dedicator, Marcus Simplicius Quietus, served as tribune of cohors III Batavorum. The tribune came from Germania, the cult centre of Vagdavercustis, who continued to be venerated in Vetus Salina just as she had been in her place of origin in Germania Inferior. The paucity of evidence regarding Vagdavercustis has led to multiple variations of the name, which appears to have been particularly problematic for some Latin speakers. This name has drawn attention to instances of code-switching associated with linguistic retention in religious contexts. Although the hypercorrect “diphthongisation” of /a/ is not a typical feature of Pannonian Latin,²⁹ 3 examples of hypercorrect spelling with <ae> instead of /e/ are known from Pannonia,³⁰ representing approximately 0.2–0.45 % of cases.³¹

The proportion of changes affecting the /e/ vowel accounts for 4 %, in comparison to the other vowel changes. Within this context, the absence of the vowel /e/ before /a/ is uncommon and is attested in the form Dabus;³² similarly, the phenomenon whereby the non-accented /e/ shifts to /i/ before a vowel is observed in the form Di|abus.³³

The second-largest category consists of changes affecting the /i/ sound, which account for 16% of the total. The use of the spelling <e> of an unstressed /i/ in hiatus is typical in the name Diana, as seen in the two attestations of the form Deana.³⁴ In several cases, the absence of the vowel /i/³⁵ does not reflect a phonetic phenomenon but rather a spelling error. The vowel /y/, occurring in words and names borrowed from Greek, alternates between /u/ and /i/. The pronunciation is elucidated by the use of <i> instead of /y/, as exemplified by Igia in the name Hygia, in an inscription from Aquincum (LLDB-10098, TitAq I, 36).

Another altar relevant to our investigation comes from Aquincum (CIL III, 3474),³⁶ where the spelling of the invoked deity is already controversial, as the altar seems to be dedicated to a Menitrae.³⁷ The god was believed to be a pre-Roman (Celtic or Pannonian) deity, known only from this inscription.³⁸ The altar was erected in the mid-3rd c. AD by Aurelius Florianus, a veteran who served as beneficiarius consularis in an unnamed military unit. The deity’s name presents various confusions. In addition to the /e/-/i/ change that follows imperial patterns, the /m/ is reduplicated with an <n>. Nevertheless, the deity may be identified as Mithras, given the frequency with which his name undergoes changes – a phenomenon observed in other dedications. The substitution of the short /i/ of the stem by a long <e> (at the time bearing a similar sound) also occurs in other Mithraic inscriptions,³⁹ with a relatively wide range of errors documented in the first syllable of the god’s name. In Pannonia, even hypercorrect forms are attested, where <y> is written instead of the original /i/, as seen in the form Mythrae,⁴⁰ which may have arisen under Greek influence.

The corpus contains a limited number of inscriptions that belong to the group of /o/ and /a/ sound changes. In an inscription from Aquincum, the /o/ in the -ona suffix is substituted by an /a/, creating the form Seranae (LLDB-10102, 10103, AE 1982, 806).⁴¹ This phonological variation may reflect the Latin adaptation of the Indo-European vowel /ā/ into Latin occasionally resulted in the /a/~/o/ change, which is characteristic of Moesia Inferior and Pannonia.⁴² The theonym Sirona refers to a Gallic deity whose initial consonant occurs in various written forms and who was worshipped primarily in the Celtic part of the Roman Empire. Consequently, the name Sirona is the Latinised form of the theonym. In this case, the dedicator was Titus Julius Mercator, the decurion of the colonia Aquincum. The votive inscription was erected in the sanctified area surrounding the spring associated with the local aqueduct system, indicating a possible water-related cultic function.

The i > V change is attested only in the name Quadriviae, with the variants Quadruviae, Quadribiae, and Quadrubiae presenting both phonetic and vocalic changes.⁴³ Their theonym derives from the Latin quadrivium (“four roads”).⁴⁴ In Pannonia, these divinities were worshipped differently than in the Celtic areas of the West – not for their association with crossroads, but rather for some local characteristic, which may have been closer in function to the Silvanae. Their name is attested in inscriptions of Pannonia Superior, including those from Carnuntum, Vindobona, and Savaria.⁴⁵

Regarding Jupiter Dolichenus, the most common spellings of the name are Dulceno or Dulcheno,⁴⁶ which reflect the contemporary pronunciation.⁴⁷ In addition to the o > V changes in Dolichenus, only one example represents the o > V change in the dative ending. It appears in an altar erected by a soldier of the Hemesian cohort at the present-day village of Előszállás, near Intercisa. The piece is dedicated to Elagabalu instead of Elagabalo, which represents simply a vocalic change (LLDB-151645, RHP 326).

B. Fehér considers syncopation to be typical in only two sound environments, although he argues that these are proportionally significant or reflect local developments. The first is -c-V-l- in the 1st and 2nd c. AD, and the second is -m-V-n-, primarily from the 3rd c. AD.⁴⁸ In our corpus, the names Dominus and Domina occur in post-tonic syncopation,⁴⁹ while the name of Jupiter Dolichenus displays pre-tonic syncopation.⁵⁰ An inscription from Pannonia Inferior, present-day Sárpentele (LLDB-13552–53; CIL III, 3343), was dedicated to the deity Dolichenus, whose name appears in a syncopated form (Dolc[heno]). The dedication was made collectively by all the priests (sacerdotes) of the province of Pannonia, highlighting the widespread institutional veneration of the god. The syncopated form of the theonym may reflect either a phonological reduction common in spoken Latin or a regional orthographic convention. Such abbreviated forms offer valuable insights into local linguistic practices and the adaptation of foreign theonyms within the Latin epigraphic tradition.

Epenthesis is even more uncommon, typical of the consonant groups -cr- and -cl–.⁵¹ In our corpus, two items represent the epenthesis of the vowels /a/ and /o/, which is less typical – usually the vowels /e/, /i/, or /u/ appear between two consonants. Silavano occurs instead of Silvano (LLDB-117326; AE 2012, 1156) and Tranosito instead of Transitu (LLDB-28103; CIL III, 4444), in cases from Carnuntum. The first was dedicated to Magula or Magla alongside Silvanus. The goddess appears to be an originally Celtic deity with a local centre of distribution in Carnuntum, where the couple Silvanus Domesticus and Mag(u)la was worshipped by people of different origins and status. The three identifiable dedicants are Gaius (AE 2012, 1156), Septimius Asmenus (CIL III, 4440a), and Vind(ius?) Karus (AE 2011, 999). The gentilicium Vindius is of Celtic origin, suggesting possible local or regional ancestry. The Greek cognomen Asmenus is relatively rare and, along with the mononymous Gaius, may indicate freedman status for both individuals. This onomastic evidence contributes to the socio-cultural contextualisation of the dedicants and provides insights into patterns of identity construction and integration in the Roman provincial era.

From all the phonological cases, the most interesting form is Doloceno (LLDB-8503; CIL III, 3999; from Pannonia Superior, Aquae Balissae), due to the acoustic assimilation observed in this spelling (commutationes vocalium variae). His dedicator, Secundius Res(ti)tutus, was a centurio of the legio X Gemina. Although this is the only attested example from Pannonia, similar phonological assimilation is known from other inscriptions as well.⁵² Nevertheless, due to the limited number of documented cases, it remains difficult to draw definitive conclusions. It cannot be ruled out that the observed assimilation reflects a non-linguistic factor, such as a scribal habit, regional orthographic convention, or individual variation.

Consonant Changes

The corpus reveals a diverse set of consonantal changes that can be grouped into broader phonological categories, reflecting various processes of assimilation, simplification, and hypercorrection. Although individual examples vary in frequency and distribution, the overall patterns provide insight into regional linguistic tendencies and orthographic conventions in the Latin of the provinces. Several cases reflect phonological assimilation, either complete or partial, often involving voicing or place of articulation. A significant number of changes relate to the misinterpretation or substitution of Greek aspirated consonants in Latin contexts. Other sporadic changes point to phonetic confusion or orthographic variation. Regarding the consonant changes identified by B. Fehér in Pannonia, it is noteworthy that he describes the shortening of geminates as a characteristic feature of the unstressed syllables /ll/ and especially /nn/.⁵³ The only instances of degemination (3%) of double consonants are attested in the name Tere Matri instead of Terrae Matri (LLDB-13524; CIL III, 10374), while only one gemination (3%) is attested in Dolchenno instead of Dolicheno (LLDB-59669; AE 1972, 445).

Among the consonant groups, changes involving nasal sounds account for 8% of the cases. Within the pre-consonant dropout of n, the ns > s change is considered a separate group, constituting more of a literacy measure, since the /n/ of the -ns- consonant cluster was not spelt.⁵⁴ The pronunciation /ns/ > /s/ was introduced by settlers in the 1st ^c. AD. Two inscriptions illustrate this loss of the nasal /n/: an altar from Gorsium dedicated to Trasit[ui] instead of Transitui (LLDB-13534; RIU-6, 1488), and an altar from Brigetio dedicated to the same personification of the divine movement of Mithras, in the form Tras|ito (LLDB-51829; CIL III, 10963).⁵⁵

The confusion of the intervocalic consonants /b/ and /v/ is attested only in the name variants of Quadriviae – Quadribiae and Quadrubiae –⁵⁶ which account for 10% of all consonant changes observed in divine names and epithets in Pannonia. By contrast, the post-consonantal drop of /v/ is documented in an inscription from Aquincum, where the name Silvano occurs in the form Silan|o (LLDB-19315; CIL III, 10453).

Changes affecting dental consonants are relatively infrequent in the corpus, accounting for approximately 3% of the data. Only two inscriptions from Aquincum attest to the substitution of /d/ with <t>, as in the form teo for expected deo. This phonological change may be interpreted as th > T, particularly if the form derives from the Greek equivalent θεός. The first inscription was dedicated by an individual named Hermes, who is identified as being of Greek origin—an attribution that may support the hypothesis of Greek linguistic influence. The second inscription is significantly more fragmentary and incomplete, precluding any definitive interpretation regarding its phonological features. However, in the case of teo instead of deo, it remains possible – similar to the case of Asculepio – that the form was not the result of a phonetic change, but rather of the uncertainty of the dedicator, commissioner, or engraver, with diverse cultural and linguistic backgrounds. In this way, the Greek θεός and the Latin deus, which share the same meaning, may have been confused.

Upon revising the attestations, the most prevalent variations concern the letter /h/ (60%), although its use may more accurately serve as an indicator of literacy. The drop of /h/ was not yet a widespread phenomenon in Pannonia by the 2nd c. AD, but it became increasingly common in the 3rd c. AD. B. Fehér spoke of an “h renaissance”, which may suggest a partial restoration of the pronunciation by the end of the 2nd c. AD. It is possible that the stylistic reintroduction of /h/ was influenced by settlers arriving from the East after the Marcomannic Wars or perhaps by native Semitic speakers.⁵⁷

The inscriptions provide evidence of phonetic changes in interaction with epigraphic Latin. The proportion of /ch/~/h/ errors is particularly informative. The /ch/~/h/ variations before palatal vowels suggest the lack of palatalisation of /ch/.⁵⁸ Among the heavy consonants, this /ch/ warrants special attention. The change ch>C is clearly a regional feature, indicative of a /k/ pronunciation. From the 3rd c. AD onwards, in Greek words – and in parallel with the corresponding Greek sound change – /ch/ may have begun to be pronounced as /χ/, as evidenced by ch>H patterns. This phenomenon is exemplified by the spelling of Jupiter Dolichenus, which appears in three separate occasions.⁵⁹ The divine names in Pannonia show that the writing and spelling of /h/ represent a significant aspect of the consonant system, with 22 occurrences. The loss of /h/ is attested in the theonym Hygia, which appears in the form Igia (LLDB-10099; TitAq I, 36) in an inscription from Aquincum. A comparable phenomenon is observable in the form Mitrhae, where an intrusive aspiration has been added. Although such forms may alternatively be interpreted as orthographic deviations or irregular spellings of the name Mithras, they have been classified here – by analogy and for the sake of systematic categorisation—under the group of deviations involving the grapheme <h>. This categorisation enables a more coherent treatment of both h-dropping and h-insertion phenomena.

From a historical linguistic perspective, the instability of the Latin /h/ is well-documented. In Classical Latin, /h/ functioned as a voiceless glottal fricative, but its phonemic status had already begun to weaken by the late Republic. In Vulgar Latin, the phoneme was often lost entirely in spoken language, especially in provincial registers. Consequently, h-dropping (e.g., Hygia > Igia) and h-insertion (e.g., Mithrae > Mitrhae) are common features in epigraphic sources from the Roman provinces, where varying degrees of literacy, bilingual interference, and regional phonological developments contributed to orthographic inconsistency. The examples cited here are consistent with these broader diachronic tendencies.

Morphology

In the multilingual contexts addressed by this book from the perspective of religion, the preservation of morphological elements (e.g. suffixes) and their deviation can be understood as evidence of cultural contact. In Pannonia, certain theonyms and divine epithets occur with non-Latin suffixes, despite the remainder of the inscription being composed in Latin. The data reveal that most morphosyntactic changes relate to the transformation of the pronominal system, including case confusion (permixtio casuum: 37% = 16 instances), declension confusion (permixtio declinationum: 33% = 14 instances), innovations in case endings (commutatio in formatione casuum: 28% = 12 instances), and, to a lesser extent, issues in the formation of proper names (commutatio in formatione propriorum: 2% = 1 instance).

A corpus of 58 declension confusions and 174 case confusions can be identified in divine names and epithets across the Empire, based on the morphological errors recorded in the LLDB database. Previous studies have shown a significant concentration of such alternative forms in the Gallic (17 instances)⁶⁰ and Pannonian (15 instances)⁶¹ provinces.⁶² A comparison between the total number of declension confusions in these provinces and those occurring specifically in divine names shows that the Gallic region⁶³ – including Germania Inferior –⁶⁴ and the two Pannoniae⁶⁵ exhibit the highest proportion of declension confusions, also with regard to theonyms.

Declension confusions (permixtio declinationum)

In Pannonia, among these instances, three (22 %) involve contamination between the second and fourth declensions (i.e. declension II used instead of IV), and one (7 %) shows the reverse pattern (declension IV instead of II).⁶⁶ In the case of three personifications (Bonus Eventus, Transitus, and Sedatus), we can observe how theonyms constructed from abstract concepts posed interpretative and grammatical challenges for their worshippers. An inscription from Sirmium attests Bonus Eventus in the form Bono Evento instead of the expected dative Bono Eventui (LLDB-117217; AE 1998, 1051).⁶⁷ In contrast, among the 32 inscriptions associated with this cult, the deity is otherwise predominantly invoked as Bono Eventui or in abbreviated forms.⁶⁸

Another relevant example is the appearance of Transito instead of Transitui. In the Mithraic sanctuaries around Poetovio in Pannonia Superior, altars and statues were used to narrate the Transitus dei – the pivotal moment of “transit” in which Mithras is depicted carrying the bull. Reliefs and statues from the Danubian provinces commonly represent this scene, in which the god bears the bull on his shoulders.⁶⁹ While this moment held considerable significance for Mithraic communities more broadly, the distinctive practice of dedicating offerings specifically to the Transitus in the dative is unique to sites associated with the portorium Illyrici.⁷⁰ Two inscriptions are dedicated to this abstract concept, using the form Transito or Trasito: one, from Carnuntum, was dedicated by Caius Cassius Apronianus, custos armorum in a collegium (LLDB-32410; CIL III, 4444), and another from Tömörd near Brigetio (LLDB-32411; CIL III, 10963).⁷¹ These examples suggest that the phenomenon should not be understood as a series of isolated interpretative errors, but rather as manifestations of a broader linguistic process.

The fourth declension in Pannonia was notably conservative, with relatively few instances of alteration recorded.⁷² Among the attested changes, only the singular ablative -u > -o is worth mentioning, as it indicates a tendency for the fourth declension to merge with the second. Likewise, the simplification of the singular dative ending from -ui to -u is already a recognised phenomenon in Classical Latin. According to B. Fehér, the fourth declension “essentially disappeared in the living Pannonian language” by the 3rd c. AD, with most of its nouns being absorbed into the first through third declensions. In inscriptions, this development is reflected by the appearance of a dative ending -o characteristic of the second declension; nevertheless, one inscription from Brigetio presents the form Sedatui, i.e. declined as a fourth-declension noun, rather than the expected Sedato (LLDB-143623; RIU-2, 429, in Pannonia Superior, Brigetio). Although the context of the inscription is problematic,⁷³ it could be interpreted as deorum Sedatui, suggesting a dedication to peace among the gods. The form Sedatui may have been influenced by the abstract noun sedatus, a fourth-declension term meaning “state of peace”. Nonetheless, comparative evidence indicates that Sedato was the normative form, as all known inscriptions related to this deity use the second declension. The example of Sedatus⁷⁴ therefore illustrates that, as with Bonus Eventus and Transitus, we cannot entirely rule out the possibility that these names were heteroclitic – having multiple declensional patterns – or that their second declension forms were equally normative. These examples suggest that when such terms appear as abstract concepts (i.e. not personified), the fourth declension (transitus, eventus) is more appropriate; however, once personified as divine names (e.g. Sedatus), the second declension becomes the standard, as proper nouns are less commonly declined in the fourth.

The majority of declension confusions involve the second and third declensions, which together account for 71 % of all such errors (with 57 % involving declension II instead of III, and 14 % the reverse). A significant proportion of the observed confusion between the second and third declensions 57 % involve the anomalies, whereby theonyms or epithets that are etymologically or morphologically third declension forms are declined as second declension nouns. In Pannonia, eight such instances have been identified, reflecting this declensional interference. A few examples include the names Dis Pater and Adonis, without epithets, being inflected as second declension nouns.⁷⁵ The largest group within this category involves the epithet Silvestris,⁷⁶ which is often declined under the influence of second-declension morphology (e.g. Silvestro instead of Silvestri).⁷⁷ The epithet Silvestris, associated with rural landscapes, held geographical significance in the Danubian provinces, where Silvanus was worshipped as a forest deity under this title.⁷⁸ According to Fehér’s analysis of the epigraphic material, the adjective silvester appears to have been morphologically reanalysed and transferred into the second declension in Pannonian Latin.⁷⁹ However, this phenomenon is so far attested exclusively in Pannonian inscriptions. In contrast, literary sources consistently retain the third declension form of silvester, though not in direct connection with Silvanus. This discrepancy between epigraphic and literary usage may reflect regional linguistic developments, as well as the differing registers and levels of standardisation between spoken provincial Latin and the literary norm.

14 % of the declension confusions involve the use of the third declension instead of the second. When the name of the god appears in the third declension, the epithet is often adjusted accordingly, resulting in shifts across declensions. Notable examples include two altars bearing the forms Herculi | Augusti instead of Herculi Augusto (LLDB-5597; CIL III, 3305) and Soli | Invi|cti instead of Soli Invicto (LLDB-14882; CIL III, 3475). The frequent alternation between second- and third-declension nouns may have contributed to the use of the second-declension singular genitive -i as a substitute for the dative. This confusion is further complicated by the fact that both second- and third-declension nouns and epithets can end in -i, though this suffix corresponds to different cases depending on the declension.⁸⁰

Case confusions (permixtio casuum)⁸¹

Some examples reflect not only syntactic change but also contamination between different dedicatory formulae or the limited Latin proficiency of the engravers, particularly in cases involving foreign names. These deviations are best explained as the result of a combination of factors.⁸² Case-system alterations are almost always rooted in phonetic developments, or at the very least, phonetic change serves as a catalyst. For instance, the erosion of final -m and the shortening of unstressed (especially word-final) vowels have led to the convergence of the nominative, accusative, and ablative singular forms in the first declension (a-stem), although distinctions remained in the plural.

The most frequent phenomenon observed in divine names and epithets is the confusion of non-dative forms with the dative (94 % = 15 instances), a particularly striking pattern given that divine names in votive inscriptions are almost invariably expected in the dative case.⁸³ In Pannonia, three types of case confusion occur in equal proportion: nominative for dative, accusative for dative, and genitive for dative (each accounting for c. 25 %, or 4 instances).

The most common of these –attested throughout the Roman Empire – is the use of the nominative in place of the dative.⁸⁴ This accounts for one quarter of all case confusions in the Pannonian sample (4 instances). An illustrative example is the use of Sol instead of Soli. Although one might argue that this is an unusual abbreviation, the absence of the final -i is more plausibly a genuine case confusion, particularly when spatial constraints on the altar do not justify the omission. The issue becomes more complex when Sol appears together with the epithet Invictus, as in LLDB-143874 (CIL III, 10364). The ambiguity arises in determining whether the final -i represents the dative singular ending (Soli, without an epithet), or if it is an abbreviated form of Invictus (i.e., Sol i[nvictus]), in which case the name is erroneously left in the nominative. In either scenario, the closing formula of the inscription – votum solvit libens merito – indicates unequivocally that the theonym should stand in the dative, thereby favouring the interpretation of a case confusion. One further instance (6 %) involves a form that may represent either the nominative or ablative used in place of the dative, further supporting the interpretation of case confusion rather than deliberate abbreviation.⁸⁵

The second significant variation pattern involves confusion between the genitive and dative cases (gen. instead of dat., 25 %).⁸⁶ Half of the examples in this category involve the use of Mercuri in place of Mercurio.⁸⁷ According to B. Fehér, a dative adnominal structure – common in regions south of Pannonia and instrumental in the fusion of the genitive and dative cases – is also attested in Pannonia.⁸⁸ However, he argues that the genitive remained relatively well preserved in Pannonia and that no significant degree of case fusion occurred.⁸⁹ In contrast, data from the LLDB database suggest that genitive-dative confusion was particularly pronounced in this region. Across the Roman Empire, between the 1st and 4th c. AD, the average frequency of such confusions (considering a broader range of examples beyond divine names) stands at 12 % (77 instances), whereas in Pannonia it rises to 18 % (23 instances). When these figures are examined specifically in relation to errors that indicate the emergence of the three-case system in “Vulgar” Latin,⁹⁰ the trend becomes even more striking: 24 % at the imperial level, but a remarkable 45 % in Pannonia. This elevated ratio strongly suggests that dative-genitive fusion was far more significant in Pannonia than elsewhere in the Empire.⁹¹

This type of dative-genitive contamination most frequently occurs with nouns of the second and third declensions. The high proportion of such errors in Pannonia can be attributed to analogical confusion between the case endings of these declensions. These patterns of contamination may reflect uncertainty in the spoken language during the period of ongoing morphological simplification, in which speakers increasingly treated dative and genitive endings as interchangeable.

The third most frequent type of case confusion, also occurring at a rate of 25 % (four instances), involves the dative and accusative cases.⁹² These four cases correspond to four altars from various parts of Pannonia. One altar from Campona was dedicated to Silvanas instead of Silvanabus (LLDB-22825; CIL III, 3393), while another from Aquincum bears the dedication Nemesim instead of Nemesi (LLDB-23192; CIL III, 10442). One possible explanation for this confusion is Greek linguistic influence. In Greek inscriptions, it is common to refer to the honoured person in the accusative when a statue is dedicated, particularly in texts inscribed on statue bases. A similar pattern is occasionally found in Latin inscriptions, particularly those involving Genii⁹³ in the western provinces – for example: Genium c(oloniae) [C(laudiae) Savariae] / cum suo tem[plo] from Savaria (CIL III, 4153).⁹⁴

A smaller percentage of case confusion – 13 % (two instances) – involves the substitution of the ablative for the dative.⁹⁵ Both examples, found in inscriptions from Poetovio, feature the form transitu instead of transitui, and occur within the fourth declension.⁹⁶ This phenomenon may be influenced by the existence of neuter u-stem nouns, in which both -u and -ui forms of the dative were considered normative.⁹⁷ In Pannonia, confusions between the dative and ablative in third-declension nouns may also be attributed to the phonetic convergence of the endings -e and -i, while in the fourth declension, the analogy between -u in both dative and ablative may have facilitated the interchangeability of these forms.⁹⁸

Overall, 94 % of the case confusions in divine names and epithets in Pannonia involve the incorrect use of another case in place of the dative; only 6 % concern confusions between other cases. One such instance involves the substitution of the ablative for the accusative (coded as abl. pro acc. in obiecto directo). This particular case also displays the loss of the final -m,⁹⁹ suggesting a phonetic rather than purely morphological cause. The sole example of this type in the current corpus comes from a building inscription (LLDB-44534; AE 1936, 132) discovered at Carnuntum, which constitutes the earliest known epigraphic evidence for the cult of Jupiter Dolichenus at this site. The inscription, commissioned by a local iuventus (youth association), (iuventus colens Iove(m) Dolichen(um)… fec(it)) recording the construction of a wall and gate for the sanctuary on Pfaffenberg during Hadrian’s reign. In this case, the expected accusative form Iovem Dolichenum appears as Iove Dolichen, suggesting that the final -m was omitted, thereby rendering the form morphologically identical to the ablative. While this may reflect phonetic erosion in colloquial Latin, it also highlights the intersection between orthographic simplification and case confusion in provincial epigraphic practice.

A comparative overview of the case confusions involving divine names in Pannonia, juxtaposed with data from the wider Roman Empire as recorded in the LLDB database, reveals further insights. In the provinces outside Pannonia during the 1st-4th c. AD, errors indicative of the developing nominative-accusative-ablative merger account for 18 % of all case confusions. By contrast, in Pannonia such confusions constitute only 7 %. When considering those mistakes that point to the development of the three fundamental Vulgar Latin cases – namely, nominative-accusative-ablative, accusative-ablative, and genitive-dative confusions – the average frequency stands at 29 % across the Empire, but only 17 % in Pannonia.¹⁰⁰

These findings may also be compared with the patterns observed across the Empire¹⁰¹ and those observed by B. Adamik in a non-religious context.¹⁰² In his study, the two Pannoniae are not treated separately, but are included within the broader Illyrian provinces. According to his figures, the merger of the genitive and dative (25.4 %) was nearly as advanced as that of the accusative and ablative (14.9 %), yielding a ratio of 1.7.¹⁰³ If the Danubian provinces are considered as a unified linguistic region, nearly all types of case confusion are attested,¹⁰⁴ although the substitution of the nominative for the dative remains the most frequent. Some patterns, however, appear to be unique to Pannonia, such as confusions between the ablative and dative,¹⁰⁵ and between the accusative and dative.¹⁰⁶ In both Pannonia and Dacia, the most frequent confusions involve the dative being exchanged for the nominative¹⁰⁷ or genitive.¹⁰⁸

Writing Technique Errors (errores non grammaticae)

This set of cases includes approximately 15 non-linguistic errors, which are primarily related to text editing issues, such as repetition of formulae, omissions, and substitutions. Inscriptions often contain errors for a variety of reasons, with one major factor being inadequate knowledge of the Latin language by the engravers. Examples of such mistakes are evident where inscriptions were misapplied to altars, often due to the copying or misreading of literary texts.¹¹³ One such error is found on the altar of Marcus Ulpius Petuernus in Intercisa, where the letter ‘g’ was carved instead of ‘c’, spelling Invig|to instead of Invicto (LLDB-2201; CIL III, 10310). This can be classified as a hypercorrection error, which is typically attributed to simplified or archaic writing practices, where all /g/ sounds were represented by the letter ‘c’, and the lack of proper understanding of a sign’s phonetic value by the carver. In other instances, the engraver noticed the error and attempted to correct it, as seen in the case of Menitrae instead of Mithrae (LLDB-28111; CIL III, 3474).¹¹⁴ Finally, letter repetitions or omissions frequently occur when the engraver loses track of where a new line begins, or they forget to carve one or two letters. These mistakes underscore the mechanical nature of the epigraphic production process in certain contexts and highlight the potential disconnect between the composition and execution of inscribed texts, particularly in provincial workshops.¹¹⁵

Substratum vel adstratum

Local or non-Latin linguistic interferences in Latin contexts, as J. Herrera Rando demonstrates in this volume through several Lusitanian examples, are also important to consider. Such linguistic influences may manifest in vowel changes, inflectional forms, or individual epithets. In Pannonia, only a small number of the 135 inscriptions display code-switching, and these are primarily confined to the names of deities, while the remainder of the text remains in Latin. As noted in the introduction, Pannonia was home to a diverse population, including numerous groups of Eastern origin settled during the 2nd and 3rd c. AD. The cohors I Hemesenorum, composed of Syrians, introduced the cult of their patron deity from Hemesa to Intercisa. Five inscriptions in and around Intercisa mention the name of Deus Elagabalus, two of which were explicitly erected by members of the same cohors I Hemesenorum. Most notably, during the reign of Septimius Severus, this unit constructed a temple in Elagabalus’ honour (AE 1910, 141).

In relation to this cult, three examples from Pannonia illustrate the hypercorrect use of the diphthong <ae> in place of /e/. Two inscriptions from Intercisa feature the form Aelagabalo (LLDB-72920; RIU-5, 1106; and LLDB-516; AE 1910, 141), one of which (AE 1910, 141) is the dedicatory inscription of the temple. Despite the limited attestations, the high variability in the spelling of the god’s name is further demonstrated by the variant Alagabalo from Brigetio (LLDB-6883; CIL III, 4300). This local e > a shift represents a specific instance of phonetic code-switching, in which the Greek-transcribed Hemesian name was rendered into Latin. Although multiple variants reflect how the Syrian theonym was pronounced and perceived within local communities, Roman inscriptions consistently employ the form Elagabalus.¹¹⁶

Evidence of Greek adstratum influence is observable in several Pannonian inscriptions, particularly in onomastic variation. In theonyms the influence is evident in the form Ariagne instead of Ariadne on two Pannonian inscriptions, both from the vicinity of Aquincum (LLDB-6228; RIU-3, 938; LLDB-151315; TitAq I, 202). This spelling reflects the existing Greek variants Αριάδνη and Αριάγνη within Latin local inscriptions.¹¹⁷ Another case of Greek influence may be observed in the spelling Priepo (LLDB-117178; AE 1991, 1319) for Priapus, where the shift involves the accented long /a:/ vowel. Though less common, this variation likely stems from Greek phonetic influence. Additionally, the form Asculepio found in an inscription from Aquincum (AE 1937, 182) may also be the result of Greek Latin interaction. This variation may indicate a phonetic development—possibly the assimilation of the Greek dative Ἀσκληπιῷ or arise from orthographic ambiguity between Greek and Latin representations of the theonym Asclepius. While such forms are relatively rare, they nonetheless attest to the presence of Greek linguistic influence in the Latin epigraphy of Pannonia, particularly in urban centres such as Aquincum, where Greek-speaking populations were more likely to interact with Latin inscriptional practices.

Celtic’s substrate or adstrate effects are also attested, as seen in the spelling Serana for Sirona. This form preserves the archaic *ðer- root, whereas the restored pronunciation *Tsirona or *Stirona (meaning “stellar” or “astral”) derives from ster- with the suffix -ona.¹¹⁸ A Celtic phonological feature may also explain the epenthetic /i/¹¹⁹ in Marimogio (LLDB-147053; AE 1988, 933; and LLDB-147052; AE 1977, 628). The deity’s name appears in several variants – Harmogius, Marmogius, Armogius, Marimogius, and Maromogius – all associated with the cult of Mars Marmogius, worshipped in the region between the Noricans and Tauriscans in south-western Pannonia, near Poetovio.

B. Fehér and A. Gonda attribute the merger of short /e/~/a/ sounds primarily to the influence of the Celtic substratum, particularly in the context of personal names and ethnonyms of Celtic or Germanic origin. They distinguish this phenomenon from similar vocalic developments observed elsewhere in the Roman Empire, arguing for a regionally specific substrate effect in Pannonia.¹²⁰ In the LLDB database, /e/~/a/ misspellings are most frequent in and around Pannonia during the 1st-3rd c. AD.¹²¹ As a substrate effect, this phenomenon likely had a lasting impact on regional phonological development. In the present corpus of divine names, three inscriptions reflect this vocalic variation.¹²² Notably, these instances affect names of non-Pannonian origin, suggesting that Pannonian Latin influenced the phonological realisation of borrowed or foreign elements. The alternation between /a/ and /e/ in proper names and ethnonyms – particularly those of Celtic or Germanic derivation – is, to date, documented exclusively within the Pannonian epigraphic record. This further supports the interpretation of a localized substrate effect shaping the regional Latin pronunciation and orthographic conventions.

The worship of Roman deities such as Diana, Liber Pater, Libera, Terra Mater, and Bona Dea – gods often associated with indigenous cults – was widespread among the civilian population. However, deities named solely in Latin and belonging to the Roman tradition are rarely associated exclusively with Pannonia. The cults of certain deities of Celtic origin, such as Sedatus,¹²³ Jupiter Teutanus,¹²⁴ and Jupiter Depulsor,¹²⁵ who are also attested elsewhere in the Empire, are found in the province as well. Among the well-known Celtic-origin deities worshipped across the western provinces, Epona is likewise attested in Pannonia. Additional Celtic traditions are reflected in the collective designation Dii Augurales (RIU-1, 1), referring to a group of Celtic-tradition deities venerated in the Savaria region.

In contrast, among the Celtic mother goddess groups, such as the Nutrices and Silvanae attest to cults of distinctly local origin in Pannonia.¹²⁶ The personalised epithets of Silvanus further reflect local adaptations and significance.¹²⁷ Similarly, in the southern part of the province, a divine pair worshipped under the names Vidasus and Thana appears, although their original names remain unknown.¹²⁸ This type of localised, indigenous cult – frequently syncretised into Roman religious expression – is notably absent from military contexts. Deities such as the Nutrices, Baltis,¹²⁹ and various goddesses of local healing springs, worshipped under multiple names, are found almost exclusively in civilian settings. Their names reflect the integration of non-Latin phonetic and religious elements into Latin epigraphy, illustrating the coexistence of multilingual and multicultural practices within the broader Roman religious landscape.

Local traditions also indicate that certain indigenous deities, identified with Jupiter Optimus Maximus, were worshipped on hilltops: in Carnuntum, Carnuntinus was worshipped on the Pfaffenberg; in the vicinity of Aquincum, Teutanus on the Gellérthegy (Gellért Hill); and in Poetovio, Depulsor on the Panoramaberg. Similarly, in south-western Pannonia, the deity Culminalis was venerated. Local importance was also attributed to Fortuna, who was worshipped with the epithet Karnuntina, often identified with Nemesis.¹³⁰ From Savaria, Itunus and Ituna (AE 2000, 1192) – that is, the Dii Itinerarii – are the only provincial features known to derive from a Latin name. Equally significant at the local level was the cult at Aquae Iasae, where an altar was dedicated to the Fortunae Iasonianae (AquaeIasae 76). The Semitatrices, who are not attested elsewhere, likely belonged to the cultic life of the indigenous population and are depicted on an altar in Savaria (CIL III, 4174).

A review of the cults attested in Pannonia reveals a somewhat unexpected pattern: contrary to expectations that the linguistic interpretation of local cults would present the greatest difficulties, or that these names would exhibit the most linguistic variation – the divine names and epithets associated with local cults show fewer linguistic irregularities. However, these names are known exclusively from Roman epigraphic sources; their equivalents in the original local languages are unknown, as there are no data indicating the continuity of the cults. No continuity of cult places from the pre-Roman period is attested in the Pannonian provinces, nor are epichoric epithets of local gods attested in the northern part of Pannonia. This phenomenon is not unique to Pannonia. Comparable tendencies have been observed in other Roman provinces with strong indigenous religious traditions. In such cases, epigraphic sources rarely preserve pre-Roman linguistic features, and even when native names are retained, they appear in Latin orthography, often stripped of their original morphological and phonological characteristics.

Ethnic Composition and Social Status of the Inscriptions of our Corpus

Religious inscriptions provide direct insight into the religious life of the upper provincial leadership, the army, and the urban elite, as these groups were almost the only individuals able to commission the erection of altars.¹³¹ The extent to which this inscriptive material reflects the language use of the broader population and changes in the Latin language is therefore limited. Nevertheless, the linguistic specificities of these inscriptions may reveal not only aspects about the transformation of religious practice, but also a departure from classical norms and strict grammatical conventions. A significant number of divine names and epithets show deviations from the classical, literary language and may reflect features of everyday spoken language.

Most of the worshippers recorded in these inscriptions can be recognised as Romans of different social strata, although the onomastic evidence provides some additional details. The largest group represented in the present corpus (135 inscriptions) consists of dedications commissioned by soldiers (41 inscriptions). However, the ethnic and social composition of the military stationed in Pannonia varied significantly over the three centuries of our era. An additional 14 inscriptions involve high-ranking inhabitants: six refer to municipal magistrates (decurio coloniae, duumvir and sexvir coloniae); three mention the governor or his staff (beneficiarius consularis); five include religious positions (sacerdos, collegius Fortunae, etc.); and only one inscription mentions a person of freed status (libertus).

The origin of the soldiers can be identified from their names in only a few cases. Although origo functioned as a cohesive factor within the army, in the corpus under study only one person, Atilius Primus, specifies his place of origin: domo Sergia Marsis, indicating that this centurion came from the city of Marruvium or its surrounding territory in Italy.¹³² This centurion also held the position of interprex of legio XV Apollinaris, likely serving in situations where the legion encountered non-Latin-speaking populations.¹³³

In certain specialised units of the tactics type (Syrians, Mauri), the predominance of the ethnic group from which the unit was originally recruited was largely preserved.¹³⁴ The question of whether there was a connection between such ethnically organised troops and the gods they worshipped has been raised previously.¹³⁵ In some cases, this phenomenon can be observed, particularly among the Syrian soldiers stationed in Intercisa and in relation to the local cult of Elagabalus.¹³⁶ The cohors I Hemesenorum was originally recruited from Hemesa, and subsequent reinforcements continued to consist predominantly of Syrian soldiers. As a result, the cohorts stationed in Intercisa retained a strong Syrian connection and identity for generations due to the constant recruitment of Syrian replacements. However, soldiers with Thracian names can also be identified in the epigraphic record, such as Aurelius Lipor (CIL III, 3393).¹³⁷

From the reign of Hadrian onwards, legionary recruitment in Pannonia was likely composed primarily of military families settled in the canabae, or of native peregrine inhabitants of the province, contrary to earlier practice. Soldiers from other provinces were recruited into the legions of Pannonia either as temporary replacements or as vexillationes.¹³⁸ While many of these soldiers were of extra-provincial origin, some were locally recruited from Pannonian settlements. One such individual is Lucius Minicius Honoratus (CIL III, 13497), whose gentilicium suggests provincial origin. The nomen Minicius is attested both along the ripa (the frontier zone of the Danube) and in Sicily, indicating the potential mobility or broad geographic dispersion of certain Roman family names. Another example is Publius Aelius Respectus (CIL III, 3910), who dedicated an altar to Mitrhe and probably came from the area around Mogio (CIL VI, 32536).

During the Severan period, the legio II Adiutrix included many soldiers of Italic origin who had joined the legion during, or as a result of, the expeditio Urbica.¹³⁹ From the second half of the 2nd c. AD onwards, the number of Roman citizens serving in these units increased, although their origins were not exclusively Italic. In our epigraphic corpus, the decurion of ala Brittonum, Marcus Domitius Secundinus (CIL III, 3305), who was of Briton origin, is recorded.¹⁴⁰ Marcus Simplicius Quietus, the tribune of cohors III Batavorum (AE 1935, 163), made a dedication to the goddess Vagdavercustis, written as Deae Vag/daevercus/ti. A person with the same gentilic, Simplicius Super, is also attested in the case of a decurio from Germania Inferior (CIL XIII, 8805), possibly a relative, suggesting connections and networks extending into other regions. Indeed, the cohort came from Germania Inferior, where Vagdavercustis was venerated as its dea patria.

The personal names of the civilian population in some cases only provide limited indications of the origin of the individuals who commissioned the inscriptions. Aemilius Hermes (TitAq I, 49), who fulfilled his vows to Dis Pater and Proserpina, bears a Greek cognomen,¹⁴¹ identical to that of an individual mentioned in a Dacian funerary inscription (CIL III, 1228), who was a libertus. Lucius Catius Messorinus and Lucius Catius Messianus (AE 2001, 1636), who dedicated an inscription to Silvanus Silvestris, bear the cognomen Catius, which is attested in northern Italy and Gallia Narbonensis. Some local people of Pannonian families also appear. The nomen Mamilius, found on the votive plate dedicated to Liber Pater and Libera from Bodrog, is known from Savaria and Scarbantia, suggesting that the dedicator, Mamilia Firmina (AE 1974, 515), likely belonged to a family from western Pannonia. It cannot be excluded that the tablet itself was produced locally. However, since this nomen is also common in Italy and Hispania, the family’s ultimate origins most likely trace back to the western part of the Empire.¹⁴² Among the civilian population, there were also individuals of Syrian origin, such as the assumed eastern, Syrian origin of Ulpius Petuernus (CIL III, 10310), based on his surname.¹⁴³

Conclusions

The examples cited above illustrate the diversity of divine names and epithets in Pannonian inscriptions, most of which date from the late 1st to the 3rd c. AD. This corpus contains a variety of votive monuments, primarily altars (101), construction inscriptions (4), relief (1), tabulae (6), basis of statue (1), as well as instrumentum domesticum (1) and a votive triangular piece.

Adopting the categorisation of bilingual phenomena presented in the introduction to this volume, we might extend the observations on the code-switching developed by J. N. Adams, A. Mullen, M. J. Estarán Tolosa and G. Tozzi to address linguistic interactions. Their framework helps to highlight how, within a Latin text, phonetic or morphological characteristics of the secondary language may alter words of the primary language without introducing full sentences, phrases, or parallel content in a secondary language. A general overview of the religious inscriptions from Pannonia compared with those from other regions of the Roman Empire reveals similar patterns of vowel changes in divine names. However, 25 % of all vowel changes in Pannonian inscriptions occur within divine names.¹⁴⁴ Although confusions do occur in Pannonia, the votive texts tend to be highly formulaic and conservative; consistent changes are primarily observed in the names of the gods, while only a few inscriptions in this corpus display significant linguistic features. The vowel fusions and blends typical of Pannonia are also attested in divine names at the imperial level. Likewise, consonant substitutions in divine names mirror those found in other lexemes throughout the region. Consonant changes, which account for 17.8 % of the total changes in divine names, are slightly less frequent than vowel changes, but their distribution parallels that found in other parts of the province.¹⁴⁵ This suggests that there is no exceptional treatment of divine names in terms of phonological shifts.

In the examination of all case and declension-related anomalies – including errors in morphological formation and syntactic agreement – within the broader category of nominalia, we observe a frequency of 7.6 % in Pannonia Inferior and 5.9 % in Pannonia Superior. At first glance, the epigraphic material from Pannonia does not exhibit overtly localised linguistic features. However, a more detailed analysis reveals a noteworthy pattern: nearly half (47%) of all attested case and declension confusions pertain specifically to divine names and epithets.¹⁴⁶ The distinction of divine names is primarily characterised by confusions in declension and case. It is noteworthy that the prevalence of such confusions in Pannonia is notably higher compared to other words within the province. Remarkably, the name of gods, and their deviations from standard forms, is rarely attested in the nominative,¹⁴⁷ due to the structure of votive dedications. However, although the confusions of divine onomastic sequences are higher in Pannonia than in other provinces, the majority of the data cannot be explained through substrate causes, as both the names involved and the changes in which these confusions occur are also attested in other areas of the Empire.

Although the worship of some pre-Roman deities can be found in Pannonia, the most widespread linguistic phenomena observed in the local corpus also appear in inscriptions from other provinces. While some cults of Celtic origin persisted in Pannonia, the linguistic interpretation of their spelling in relation to the substratum is often difficult, as they do not diverge from the confusions observed in other divine names, e.g. Mithras and Jupiter Dolichenus, which are conversely much higher. In fact, some divine names, such as Vidasus, Thana, Belenus, and Teutanus,¹⁴⁸ appear in Latin inscriptions without any apparent linguistic modification. This further suggests that substrate influence was minimal or inconsistent.

The majority of the inscriptions were donated by Roman citizens and slaves. However, Roman naming conventions may have concealed the actual origins of the dedicators, making it difficult to identify foreign or Pannonian ancestry, as local inhabitants adopted Roman names and epigraphic formulae early in the process of integration. As a result, the language of the inscriptions alone does not reliably indicate the dedicator’s ethnic or linguistic background. Nevertheless, the influence of bilingualism – including Celtic and Pannonian substrata – can occasionally be discerned in onomastic elements, toponyms, and certain theonyms within the epigraphic record. In comparison to other regions of the Roman Empire, Pannonia presents a particularly fruitful context for examining the complex multilingual and multicultural interactions among diverse groups and their effects on onomastics,¹⁴⁹ toponyms, and divine names.

Comparable phenomena are attested in other frontier provinces of the empire, such as Britannia, Dacia, and the Germaniae, where the Latin of inscriptions similarly interacts with local linguistic substrates. In these contexts, as in Pannonia, divine names and cult epithets seem particularly sensitive to regional variation and linguistic interference, likely because they operate at the intersection of local religious tradition and Roman epigraphic convention. Crucially, the linguistic differences observed here do not simply reflect contact between distinct languages such as Latin, Greek, or local tongues, but also arise from the sociolectal interplay of various evolving varieties within the same language. This dynamic exists in constant dialogue with the more conservative linguistic forms preserved in epigraphic registers, offering valuable insights into the fluidity of language use in late antique society.

Abbreviations

Semmelweis University, vagasitunde6@gmail.com, ORCID: https://orcid.org/0000-0003-2320-3225. The present contribution has been prepared within the framework of the  HORIZON-ERC-2022-ADG project n° 101098102, entitled “Digital Latin  Dialectology (DiLaDi): Tracing Linguistic Variation in the Light of  Ancient and Early Medieval Sources”, and the NKFIH (National  Research, Development and Innovation Office, Hungary) project n° K  135359 entitled “Computerized Historical Linguistic Database of Latin  Inscriptions of the Imperial Age” [http://lldb.elte.hu/]