Abstract
The main goal of this paper is to provide a preliminary examination of the vulgar Latin grammatical gender system. The inscriptional evidence is studied closely to achieve this goal, this way the broad structure of the loss of the neuter can be observed and previously unknown dialectical differences can be discovered. Having done this groundwork further studies can investigate this material more thoroughly, and with a better understanding of the general features.
1 Introduction: problem statement, research history and methodology
The disappearance of the neuter in the Romance languages has always been a prominent issue for scholars of Latin language history, and especially of Vulgar Latin. No works have yet been published on the disappearance of the neuter with a focus on the inscriptional material. Yet with this material one could describe the main features and trends of this change in a systematic way, provide precise information on how it occurred, and discover what kind of spatial and chronological differences were observable. The aim of my paper is to investigate the presented issue in depth, using the data and analytical tools provided by the Computerized Historical Linguistic Database of the Latin Inscriptions of the Imperial Age (hereafter LLDB database), and to show the process of the disappearance of the neutral gender and the restructuring of the grammatical gender system in Vulgar Latin in general.
In Latin, the distinction between the three grammatical genders (masculine, feminine, neuter) was already present as an Indo-European heritage. In the Classical Latin period1 this subsystem was mostly stable, with only a few words showing gender variation, such as collus/collum, uterus/uterum. The plural of some masculine words could also be formed in neutrum, e.g., of locus (loci and loca). In addition, there were still some forms that showed gender variation in the Old Latin period, but these were normalized during the standardization of the Classical period, i.e., they were eliminated or at least avoided.2
By the time of Vulgar Latin,3 these fluctuating trends and other morphological changes in nouns led to the weakening of this system, which was based on the distinction of three grammatical genders, and this eventually led to the present state of the Romance languages, where instead of three genders only two remained, the masculine and the feminine. As a result, previously neuter words were most often subordinated to words of another gender (other scholars suggest that most often to the masculine gender).4 Based on the summarizing work of Väänänen and Herman, I can identify the main contours of the change and the words that were most exposed to the transformation of the linguistic subsystem.
These were all examples of cases where the confusion occurred between the masculine and the neuter. This is also connected to the fact that in the second declension5 there is no significant difference between these two genders, which is why in the middle of the 1st century AD there were already signs of the weakening of the neutral gender in e.g., Petronius,6 where we find balneus, fatus, vasus (instead of the classical balneum, fatum and vasum). There may also have been a change in favour of the masculine gender in the third and fourth declensions, for example salem, lactem (masculine [or feminine] gender accusatives instead of the neuter gender accusatives sal and lac in the third declension), and cornus, gelus and genus (masculine nominatives instead of the neuter gender nominative-accusatives cornu, gelu and genu in the fourth declension).
A decisive reference for both Herman and Väänänen concerning the frequency and distribution issues of the different gender confusions was the work of Bengt Löfstedt, in which he examined the language of legal Latin documents of the Lombard period and showed, based on statistical analyses of these documents, that despite all disappearing tendencies,7 the neutral gender was still detectable in the Latin of the period (7th–8th centuries). Moreover, the most significant feature of the neutral gender in the language of the Lombard documents that he discovered was that, after prepositions, the accusative case of neuter words was more resistant to the linguistic change that brought about the blurring of the distinction between the accusative and the ablative8 than the other two grammatical genders.9 This study, however, did not use extensive resources, thus geographically and chronologically it is very limited in the validity of its conclusions. Based on Loporcaro's relevant book however, it is clear that some form of neutral gender still existed in some Romance languages for a long time.10 This leaves us with the question to what extent this grammatical gender disappeared by the end of the 7th century.
The present paper, however, does not cover the investigation of possible grammatical gender differences in case-confusion after prepositions, because it requires a different corpus and method than that the LLDB database can provide as it only records substandard phenomena that deviate from the classical norm and allows for their distributional, statistical analysis. Meanwhile Löfstedt's analysis requires the inclusion of normative, standard forms, i.e., a comparative study of correct and incorrect variants.11 However, as it will become clear, the possibilities offered by the above-mentioned inscriptional LLDB database provides adequate and promising analytical possibilities for the study and description of the restructuring of the grammatical gender system and the gradual disappearance of the neuter.
Returning to the general trends of the change, besides the shift from neutral to masculine, there is also a phenomenon where the plural of neutral nouns (based on their suffixes)12 has been reinterpreted as a feminine noun in singular, such as the form folia instead of folium,13 and this development was also supported by the fact that there were previously neutral nouns with feminine equivalents, such as the armentum/armenta pair.14 Evidence of this change can be seen in the Romance continuations of certain words: see folium in it. foglia, f. feuille, s. hoja.15
Gaeng16 discusses the appearance of this change on inscriptions that (apart from one example) it appears only very late, in the 7th century, but other examples can be found in the inscriptions of provinces not examined by him, e.g., OSSA EXTERAE (= ossa extera, CIL III 9450,17 cf. LLDB-110906), which appears on a 3rd century inscription. It should also be noted that Gaeng's analysis was based on the ILCV's selection of Christian inscriptions, which did not allow him to consider the earlier appearances of this confusion on non-Christian inscriptions. Nevertheless, it is not only this type of neutral gender nouns that could have been transposed into feminine. Nouns of the singular form, originally neutral nouns, borrowed from Greek with the -μα suffix also became feminine in some cases in Late Latin, as it is shown, for example, by the expression diadema candida (Apul. Met. X 30).18
Although the literature is aware of the existence of these confusions, it still believes that the neuter remained relatively resistant to change until the emergence of the Romance languages.19 In addition, in some Romance languages, the trace of the -a suffix in the formation of the plural has been preserved, for example in the Italian braccio – braccia pair (from the Latin bracchium – bracchia); in Romanian -ora as a plural suffix has survived, as seen in the words timp – timpuri, piept – piepturi, which derive from the Latin tempus, -ora and pectus, -ora.20 The plurals of these previously neutral words all survived as feminine in Italian and Romanian, while the singular became masculine.21 In other words these nouns became an example of the alternating gender (genus alternans), where the gender of the noun changes depending on its number, i.e. masculine in singular, feminine in plural. The characteristics of such words and traces of neutral gender in Romance languages have been discussed in detail by Maiden, among others.22
Partly related to the grammatical gender confusion is the problem of the relative pronoun.23 The essence of the change was that the pronoun qui, quae, quod which previously had a distinction between three genders, increasingly lost this property and became a pronoun with a distinction between only two genders, like the interrogative pronoun quis, quid. Contrary to the tendency seen with nouns, it lost its feminine gender, which was then marked by the masculine form, rather than losing its neutral gender. There are many examples of this change in the inscriptional material, for example VALENTINA QI (CIL XIII 2430, cf. LLDB-111445), instead of the classical Latin Valentina quae. This change, however, since it is essentially independent of the weakening of the distinction of the neuter, is not particularly relevant to the present paper. Furthermore, the analogy underlying the change is based on the interrogative pronoun quis, quid, so this “error” is not really a case of gender confusion. The reason for its brief discussion is that it is tangentially similar to gender confusions and therefore it is important to distinguish it from them.24
For the purposes of data collection in the present study, words were considered “incorrect” and therefore data if their use in the individual inscriptions deviated from the classical Latin standard usage, and this deviation was in some way reflected in the use of gender. For this collection, I used the LLDB database,25 and in the study I have indicated the codes describing each deviation accordingly. The inclusion of the inscriptional material in the study is particularly justified by the fact that it is not only (more or less) well datable – thus allowing the study of temporal variation – but also well localizable, which allows for drawing linguistic-historical and dialectological conclusions.26 Nevertheless, no scholars to date have yet investigated the restructuring of the grammatical gender subsystem by using inscriptional material in sufficient depth,27 since it only has been discussed as a subsection of larger summary works,28 which has not necessarily allowed the problem to receive the attention it needs and deserves.
2 Analysis of the inscriptional material in the light of the LLDB database
A. Characteristics of gender confusions in the Early Imperial Period (1–300 AD)
I have displayed the gender confusions of the Early Imperial Period in the table above using the LLDB database (Table 1).29 Examining this, we can notice several features of the Latin of this period that are relevant to the subsystem under examination.30 On the one hand, at this time the gender confusion31 of the relative pronoun was clearly insignificant compared to other noun confusions, and on the other hand, even at this time, confusions involving the neuter made up the largest proportion of gender confusions,32 which is in line with the pattern that we can observe for the Romance languages. Although the confusion rate between the masculine and feminine gender is also high, it is mostly in line with the tendencies of change in certain word groups described by Väänänen and others.33 The significance of the characteristics of gender confusion examined becomes tangible when one looks at the Later Period and observe the changes that occurred there.
The average distribution of gender confusions in the Early Imperial Period (1–300 AD)
 | Chart 1 80% of the confusions do not involve the relative pronoun. |
 | Chart 2 (100% = 430) neutr. ∼ masc. 46% (197) neutr. ∼ fem. 20% (89) Total confusions involving the neuter 66% (286) Other confusions 34% (144) |
B Characteristics of gender confusions in the Late Imperial Period (301–700 AD)
By the Later Period, a substantial number of linguistic subsystems have changed, and the grammatical gender system was no exception, as can be seen in Table 2,34 when compared to the data from the Earlier Period. In terms of change over time, it is worth noting that while the proportion of relative pronoun confusions has increased to seven times compared to its previous level (1% > 7%), the proportion of true gender confusions has remained unchanged, at 4%, suggesting that the restructuring of the grammatical gender system has on average been unchanged in intensity throughout, and has not intensified over time. This is essentially in line with the observation of Löfstedt,35 although the replacement of neutrum forms by masculine forms is present in the language of Vulgar Latin documents (especially in singular, -um > -us), a sense of distinction between masculinum and neutrum is still clearly observable (for example in the use of prepositions mentioned above). Examining the distribution of gender confusions in Chart 4, we can notice that the proportion of confusions involving the neuter gender has increased significantly in the later, Christian Period compared to the early, pre-Christian Period, by a total of 8% (66% > 74%). This is perhaps the most important observation, since, as I indicated above, the neutrum disappears by the time of the Romance languages, which is clearly reflected in this trend of the increasing proportion of confusions involving this gender. To accurately evaluate the differences over time, it is also worth examining the dialectological differences that potentially emerge between the different areas.36
The average distribution of gender confusions in the Late Imperial Period (301–700 AD)
 | Chart 3 36% of Gender confusions do not concern the relative pronoun. Their overall ratio has increased by 0%. |
 | Chart 4 (100% = 260) neutr. ∼ masc. 49% (124) neutr. ∼ fem. 25% (64) Total confusions involving the neuter 74% (186) Other confusions 26% (74) |
3 Changes in the distribution of gender confusions by area37
To get an accurate picture of the territorial differences, it is necessary to examine the specific characteristics of each area, starting with Illyricum. By comparing the difference in the distribution of the gender confusions here with the average, we can note several characteristics of this area.38 In this area, the change in gender confusions began to accelerate in the Later Period, although not nearly to the same extent as in Hispania. In addition, as shown in Table 3, while in Illyricum the rate of confusion involving the neuter was 70% in the Earlier Period, close to the average of 66% (see Chart 2), yet this area has seen an increase of 20% in the neuter confusions (from 70% to 90%), which is more than double of the average increase of 8%. The rate observed in the Later Period is thus significantly different from the average of 74%. Observing changes in the internal distribution of confusions involving the neutrum, it appears that while on average there was only a 3% increase in the proportion of confusion between the male and neutral genders, in Illyricum this figure increased by 27%, nine times the average. Furthermore, while on average there was an increase in the proportion of confusion between the neutral and the female gender, their proportion here decreased by 7%, which could also be influenced by the fact that in the Earlier Period mixing between these two genders was more frequent than average in this area. This suggests that, by the Later Period, there was more uncertainty in the distinction between the neuter and the masculine gender only, and that words of the neuter gender were less likely to be reinterpreted as feminine. Due to the fact that in Illyricum, in addition to the proportion of gender confusions, the overall proportion of confusions involving the neuter also increased significantly in the Later Period, I can conclude that the disappearance of the neuter was already in full swing in the 4th–7th centuries, which predicts the complete breakdown of this gender later.
Distribution of gender confusions in Illyricum
 | Chart 5 (100% = 87) neutr. ∼ masc. 52% (45) neutr. ∼ fem. 18% (16) Total confusions involving neuter 70% (61) Other confusions 30% (26) |
 | Chart 6 (100% = 19) neutr. ∼ masc. 79% (15) neutr. ∼ fem. 11% (2) Total confusions involving the neuter 90% (17) Other confusions 10% (2) |
Distribution of gender confusions in Hispania
 | Chart 7 (100% = 31) neutr. ∼ masc. 22% (7) neutr. ∼ fem. 33% (10) Total confusions involving the neuter 55% (19) Other confusions 45% (14) |
 | Chart 8 (100% = 39) neutr. ∼ masc. 56% (22) neutr. ∼ fem. 31% (12) Total confusions involving the neuter 87% (34) Other confusions 13% (5) |
In the case of Hispania, the first striking feature is that in the Earlier Period (Table 4, Chart 7) the overall rate of confusions involving the neutrum was lower than average (55% instead of 66%). Moreover, the distribution of gender confusions within this period was also vastly different, since on average confusion between the masculine and neuter genders was more frequent, in the case of Hispania in the Earlier Period we see that mixing between the feminine and neuter genders was more frequent. This difference mostly disappears by the Later Period, when the proportion of confusion between the neuter and the masculine is almost the same as the average (56% compared to an average of 49%, see Chart 4). The relatively high number of confusion between the feminine and the neuter (31% compared to an average of 25%) remains, but in contrast to the average, this proportion has decreased, suggesting that the former high interchangeability of the past has not been reduced to a similar proportion to other areas. The most significant change over time in Hispania was the fact that the proportion of confusions involving the neutrum increased significantly by the later period, an increase ten times higher than the average increase of 3% (see Chart 4). This suggests that the process of neutral gender disappearance accelerated in Hispania, therefore the grammatical gender system missing the neutrum, which is a characteristic of the Romance languages, may have emerged earlier here than in the other areas examined.
In the case of Gallia and Germania, it can be seen from Table 5 that here, in both the Earlier and Later Periods, the overall confusion of neutral gender was relatively rare compared to the average (58 and 61 percent compared to an average of 66 and 74 percent). However, there is also a significant change in the proportions of the two mixing types (neutral gender and masculine gender, and neutral gender and feminine gender), with the proportion of confusions between masculine and neuter decreasing in the later period and the other increasing, which differs from the general increase observed in the average case. Therefore, the relatively low proportion of confusion of the neutral gender in Gallia and Germania suggests that this gender did not start to weaken as significantly here as in the other areas. On the other hand, however, we can see that the weakening was ongoing, which clearly indicates the vulnerability of this gender, but only suggests a less rapidly declining system.
Distribution of gender confusions in Gallia and Germania
 | Chart 9 (100% = 26) neutr. ∼ masc. 46% (12) neutr. ∼ fem. 12% (3) Total confusions involving the neuter 58% (15) Other confusions 42% (11) |
 | Chart 10 (100% = 18) neutr. ∼ masc. 39% (7) neutr. ∼ fem. 22% (4) Total confusions involving the neuter 61% (11) Other confusions 39% (7) |
In Africa (Table 6), the proportion of confusions involving the neuter was lower than average in the Earlier Period (54% compared to 66% on average, see Chart 2), but here we see a higher-than-average increase by the Later Period, bringing the African proportion close to the average. This indicates that the process of accelerating the disappearance of the neutral gender took place here too during the Late Imperial Period. Looking at the internal distribution of gender mixing, perhaps the most significant finding is that in Africa in the Later Period, mixing between the neutral and the female gender was more common than between the male and the neutral gender, whereas in the Earlier Period it was the other way round, but the difference between the two types was already smaller than the average rates (16% difference compared to an average of 26%). These results suggest that in the African areas the mixing of feminine and neuter gender was more important, and thus the transition of nouns from neuter to feminine gender may have been more frequent than it would be expected from the average rates.39
Distribution of gender confusions in Africa
 | Chart 11 (100% = 98) neutr. ∼ masc. 35% (35) neutr. ∼ fem. 19% (19) Total confusions involving the neuter 54% (54) Other confusions 46%39 (44) |
 | Chart 12 (100% = 57) neutr. ∼ masc. 33% (19) neutr. ∼ fem. 39% (22) Total confusions involving the neuter 72% (41) Other confusions 28% (16) |
Examining the distribution of gender confusions in Italy (Table 7), the most striking thing is that the proportion of gender confusions involving the neuter was already very high in the Earlier Period (78%), even compared to the average of 66% (see Chart 2). In the Later Period, however, the proportion in Italy was almost the same as the average (see Chart 4), which is due to the fact that the proportion of these gender confusions have decreased here. This decrease in the proportion is not particularly significant, so it is possible that it reflects only a small fluctuation. However, it is also possible that by the later period, the changes in the neuter in Italy have slowed down and the distinction of the neuter has partially strengthened.40 In addition, if we look at the internal distribution of these gender confusions, we find that in Italy the confusions between the neutral and feminine gender in the Later Period are not at all as significant as they are on average (where they were 20% in the Earlier Period and 25% in the Later Period, see Charts 2 and 4), and even here their proportion within the gender confusions has decreased. For Italy, it can thus be concluded that the disappearance of the neutral gender has always been present, but that this has largely been to the advantage of the masculine gender, at least based on the inscriptional evidence.
Distribution of gender confusions in Italy
 | Chart 13 (100% = 78) neutr. ∼ masc. 55% (43) neutr. ∼ fem. 23% (18) Total confusions involving the neuter 78% (61) Other confusions 22% (17) |
 | Chart 14 (100% = 45) neutr. ∼ masc. 58% (26) neutr. ∼ fem. 17% (8) Total confusions involving the neuter 75% (34) Other confusions 25% (11) |
Finally, looking at the distribution gender confusions in the city of Rome (Table 8), it is striking that the proportion of neuter confusions decreases slightly over time, but remains close to the early average of 66% (see Chart 2). Within these proportions, however, there is no change in the distribution of the two types of blending, which suggests that the rate of change of the neutral gender in Rome was constant (as indicated by the fact that the rate of confusion with the other genders did not change significantly by the Later Period, see Chart 16), i.e., the rate of disappearance of the neutral gender in this area during the Imperial period was constant. One can also add the fact that, as in Italy as a whole, the neutrum here was largely reduced in favour of the masculine gender. In any case, the confusion of the neutral gender within the gender confusions was always significant enough to be regarded as the realisation of the disappearance process.
Distribution of gender confusions in Rome
 | Chart 15 (100% = 92) neutr. ∼ masc. 55% (51) neutr. ∼ fem. 15% (13) Total confusions involving the neuter 70% (64) Other confusions 30% (28) |
 | Chart 16 (100% = 70) neutr. ∼ masc. 49% (34) neutr. ∼ fem. 15% (11) Total confusions involving the neuter 64% (45) Other confusions 36% (25) |
4 Dialectological characteristics of the gender confusions
This brings us to the comparison of the areas among themselves, which can show the relationship between the areas more accurately than comparing them with only the average rate. To do this, I have collected the most important types of gender confusions and their proportions in terms of the disappearance of the neuter by period in two tables (Tables 9 and 10). These tables make it easy to compare the spatial characteristics of the most important linguistically significant gender confusion (i.e., the confusion of the neuter) and to see the presence or absence of a linguistic link between the different areas.
Earlier period41
| Italy | Illyricum | Rome | Gallia and Germania | Hispania | Africa | |
| masc. ∼ neutr. | 55% | 52% | 55% | 46% | 22% | 35% |
| fem. ∼ neutr. | 23% | 18% | 15% | 12% | 33% | 19% |
| Ratio | 2.4 | 2.9 | 3.7 | 3.8 | 0.7 | 1.8 |
| Total | 78% | 70% | 70% | 58% | 55% | 54% |
Later period43
| Hispania | Illyricum | Africa | Gallia and Germania | Italy | Rome | |
| masc. ∼ neutr. Change | 56% +34% | 79% +27% | 33% −2% | 39% −7% | 58% +3% | 49% −6% |
| fem. ∼ neutr. Change | 31% −2% | 11% −7% | 39% +20% | 22% +10% | 17% −6% | 15% +0% |
| Ratio | 1.8 | 7.2 | 0.8 | 1.8 | 3.4 | 3.3 |
| Total Change | 87% +32% | 90% +20% | 72% +18% | 61% +3% | 75% −3% | 64% −6% |
The clearest dialectological link can be found between Rome, the rest of Italy, and Illyricum in the Earlier Period. It is evidenced by the fact, that the rate of confusion of the neutral gender is high in all three areas (>67%), and the proportion of confusions involving the two types of neutral gender is also higher than 2 in all these areas (2.3 for Italy, 3.7 for Rome and 2.9 for Illyricum, obtained by dividing the percentage of confusion of the neutral gender with the masculine gender by the percentage of confusion with the feminine gender). This similarity between these provinces is further strengthened by the fact that both major Romance languages originating from here (Romanian and Italian) have examples of a genus alternans, while it is not observable in other Romance languages.414243 Beyond this, we can clearly see a similarity between Hispania and Gallia and Germania, as both areas have a low overall confusion rate, which could easily be related to geography. However, a major difference between the two areas is that, while in Gallia and Germania, as in the other areas, the proportion of confusions between masculine and neuter is higher, Hispania is the only province where confusions between feminine and neuter are more frequent, which suggests that here, at that time, the distinction between feminine and neuter was rather uncertain. The fact that the proportion of confusions between the neuter and masculine gender is increasing suggests that, as in the other provinces, the distinction between these two grammatical genders is becoming uncertain.
By the Later Period (Table 10), the similarity between Italy and Rome is clearly preserved, although by this time the proportion of neutral gender confusions have become relatively high in all areas, but the close relationship between Italy and Rome is also evident from the fact that the proportions for these areas have declined by roughly similar proportions. Illyricum, which was previously somewhat similar to these areas, became even more closely associated with Italy and Rome for this period since it has a similarly high proportion of the two types of neuter confusions (7.2) compared to Italy (3.4) and Rome (3.3) (while in all other areas this figure is between a half and two). This is also due to the fact that in Illyricum the confusion between the neuter and the feminine became much less frequent. The similarity observed earlier between Hispania, Gallia and Germania is confirmed by the fact that in the Later Period the proportion of confusion with the neuter increased in both areas, so that the neutrum in both places must have been highly unstable. In these areas, the fact that the ratios of the two types of blending (masc. ∼ neutr. and fem. ∼ neutr.) are between one and two (1.8 in Hispania and 1.8 in Gallia and Germania) suggests that neuter words were reinterpreted as masculine and feminine in roughly similar proportions, even if they were somewhat more frequently masculine.44 It is advisable to examine the African confusions from a dialectological point of view. It may be worth comparing this area with Hispania because of the similarities in the case system change.45 The validity of this comparison can also be justified by examining the gender system, since we can discover that while the proportion of confusions involving the neutral gender was relatively low here (<60%) in the past, it has increased significantly in later times, even though not to the same extent as in Hispania. Furthermore, in the Later Period, only here is the proportion of feminine-neuter gender confusions higher than that of masculine-neuter gender confusions, which was true for Hispania alone in the Earlier Period. This may suggest a dialectological relationship and linguistic interaction between the Hispanic and African areas. All these observations can be used to draw important conclusions.
5 Conclusions
Overall, several significant conclusions can be drawn regarding the exact process of the disappearance of the neuter, which often contradict or rather correct some of the claims made by previous scholars. Firstly, it was clear that, while Löfstedt's (see above) studies would lead one to think that at the end of the Later Period, in the Longobard period, most Vulgar Latin speakers were still able to distinguish neutral gender words from feminine or masculine words, in the 8th century, the Latin speakers were not able to fully distinguish neutral gender words from feminine or masculine words. Table 10 and an examination of the individual areas suggests that this distinction have already become uncertain after the 3rd century and was not necessarily a well-functioning part of the linguistic system, even if it persisted in various forms for a long time. A further, equally unexpected result was the observation of a highly significant dialectological difference between the areas studied, in that in one part of the Empire, in Hispania, Africa and Gallia and Germania, the neuter words fluctuated to a roughly similar extent between masculine and feminine gender, while in another, clearly distinct part of the Empire, in Illyricum, Italy and Rome, the neutral gender words were largely reinterpreted as masculine. This second category of provinces might also relate to the fact that dialects with a genus alternans have only developed in these areas. All this means that the feminization of neuter words still needs further research, since the literature (especially Herman and Väänänen)46 indicates that this was not nearly as common as the masculinization of neutral gender words, i.e., a more precise and detailed exploration of the processes of feminization is needed. In addition, in some areas, the discovery of new data may help to refine our statements. This is particularly true for Gallia and Germania, where relatively little new data from other sources47 can already be of great importance for an accurate understanding of gender confusion. By completing this research, I can state that I have made the most important (temporal and spatial) features of the neuter gender decay process more accessible and added some important findings to their ongoing research using the methods of Herman with the help of the LLDB database. I hope that my results can aid any further research on this topic.
Acknowledgment
I would like to thank Béla Adamik for his invaluable help in the research and the writing process of this paper. I would also like to thank the reviewers of this paper for their useful remarks. The present study has been supported by the ÚNKP-22-2 New National Excellence Program of the Ministry for Culture and Innovation from the source of the National Research, Development and Innovation Fund, with the technical support of the National Research, Development and Innovation Fund, in the framework of the NKFIH project K 135359, entitled Computerized Historical Linguistic Database of the Latin Inscriptions of the Imperial Age.
For a discussion of the periodization, see Adamik, B.: The periodization of Latin: An old question revisited. In Haverling, G. V. M. (ed.): Latin linguistics in the early 21st century: Acts of the 16th International Colloquium on Latin Linguistics. Uppsala, June 6th–11th, 2011. Uppsala 2015, 640–652.
An example is the neuter word caelum, which previously had a masculine variant, see Adams, J. N.: Late Latin. In Clackson, J. (ed.): A Companion to the Latin Language. Malden–Oxford–Chichester 2011, 257–283, here 271–273.
Regarding the use of the term “Vulgar Latin” here cf. Adamik (n. 1) 646.
Herman, J.: Vulgar Latin. University Park, Pennsylvania 2000, 65.
The feminine words in this declension are relatively limited in number and are well defined, such as the names of trees. e.g., citrus, -ī (feminine).
Adams (n. 2) 271–273. See also Herman (n. 4) 56 and Väänänen, V.: Introduction au latin vulgaire. Paris 19813, 102.
Löfstedt, B.: Studien über die Sprache der langobardischen Gesetze, Beitrag zur frühmittelalterlichen Latinität. Uppsala 1961, 230.
Herman (n. 4) 52–54.
For example, according to Löfstedt (n. 7) 227, the preposition ad, which is followed by accusative in Classical Latin, is followed by the ablative of a masculine noun in 5 cases and in none of them by the ablative of a neuter noun. An inscriptional example of this replacement from the 7th–8th century is LLDB-111663: AD POPVLO = ad populum (from other areas we can also find examples of this confusion from neuter nouns LLDB-13154: AD () OFFICIO = ad () officium) – the corresponding LLDB series numbered datasheets can be retrieved and viewed in the LLDB search interface [http://lldb.elte.hu/]). Löfstedt found similar data for all other prefixes.
Loporcaro, M.: Gender from Latin to Romance. Oxford 2018.
For methodological problems of comparative analysis of correct and incorrect forms, however, see Adamik, B.: In Search of the Regional Diversification of Latin: Some Methodological Considerations in Employing the Inscriptional Evidence. In Biville, F. – Lhommé, M.-K. – Vallat, D. (ed.): Latin vulgaire – latin tardif IX: Actes du IXe colloque international sur le latin vulgaire et tardif, Lyon, 6 – 9 septembre 2009. Lyon 2012, 123–139.
The basis for this process was provided by the rare collective meaning of plural neutral gender forms ending in -a, which led to a shift towards singular feminine nouns ending in -a, see Väänänen (n. 6) 102 and 104.
See Herman (n. 4) 65.
For more examples of this change, see Väänänen (n. 6) 102–103.
See Herman (n. 4) 65; Väänänen (n. 6) 102; Maiden, M.: Morphological Persistence. In Maiden, M. – Smith, J. C. – Ledgeway, A. (eds): The Cambridge History of the Romance Languages. Cambridge 2011, 172–174.
Gaeng, P. A.: A study of nominal inflection in Latin inscriptions. A morpho-syntactic analysis. Chapel Hill 1977, 102–103, 124, 181 and 193.
Based on Galdi, G.: Grammatica delle iscrizioni latine dell'imperio (province orientali). Morfosintassi nominale. Rome 2004, 158.
For more details, see Väänänen (n. 6) 103.
Väänänen (n. 6) 103–104; Herman (n. 4) 65.
Väänänen (n. 6) 105.
Väänänen (n. 6) 104.
See Maiden (n. 15) 169–174.
Herman (n. 4) 66–67.
Regarding the relative pronoun see e.g., Gaspar, C. – Tantimonaco, S.: Relative Pronouns in Light of Epigraphic Evidence: The Province of Lusitania. Acta Antiqua Hung 59 (2019) 189–198.
I used the data from the LLDB database (http://lldb.elte.hu/) as of 23.10.2022 and used the dedicated module of the LLDB database to create the charts. See note 20 for the use of codes.
Adamik: In Search (n. 11) 125–126.
Neither Hofmann, J. B. – Szantyr, A.: Lateinische Syntax und Stilistik. München 1972, 8–12 nor Cupaiuolo, F.: Bibliografia della lingua latina (1949-1991). Napoli 1993, 269 is aware of such work.
In addition to the above-mentioned works, Pirson summarized gender confusions on the Gallic inscriptions, see Pirson, J.: La langue des inscriptions latines de la Gaule. Bruxelles 1901, 154–159. And in the areas covered by the third volume of CIL, Galdi has analysed changes in the noun system, on gender confusions see Galdi (n. 17) 298–310. Similarly on gender confusions in Hispanic inscriptions, see Carnoy, A.: Le latin d’Espagne. D'après les inscriptions. Bruxelles 1906, 226–227. Finally, on grammatical gender, but not primarily using inscriptional material, see Loporcaro (n. 10).
In Chart 1, I have marked with Genera relativi those gender confusions which, although affect the relative pronoun, reflect a fundamentally different type of trend (i.e. the disappearance of the feminine form in favour of the masculine form, by analogy with the interrogative pronoun quis, quid), according to their codes used in the LLDB database: masc. pro fem. relativo e.g.: LLDB-93106: VXOR () CVM QVO = uxor () cum qua, and fem. pro masc. relativo e.g.: LLDB-105092: TREBANO () CVM QVA = Trebano () cum quo. And the name Genera contains the same codes as shown in Chart 2, together with their distribution. The following are some illustrative examples of gender confusion: masc. pro neutr. (LLDB-106576: HIC MONIMENTVS = hoc monumentum), neutr. pro masc. (LLDB-111730: OMNE FLOS| = omnis flos), fem. pro neutr. (LLDB-101060: ITESTINAS = intestina), neutr. pro fem. (LLDB-79273: AR|A () QV|OT = aram quam [QV|OT = quod]), masc. pro fem. (LLDB-110323: PORTICVS (|) EXVSTOS = porticus () exustas), fem. pro masc. (LLDB-110314: DOLOREM () NEFANDAM = dolorem () nefandum), fem. pro communi (LLDB-110317: SACERDOTIAE = sacerdoti). Nominalia is used to denote all other codes that belong to the nominalia type (covering the variations of the noun subsystem) in the LLDB database but does not denote a gender confusion. For these, no examples are necessary as they are not particularly relevant to the subject of the study.
Regarding the fem. pro neutr. confusions, it seems worthwhile make a distinction among them based on whether they might be examples of the Romance genus alternans (seen in Romanian and in some Italian dialects). For future studies this might become a certainly productive research question, unfortunately the code structure of the LLDB database used in this paper doesn't allow for this distinction. Therefore, I will not separate the fem. pro neutr. category into further subcategories here.
For the main differences between this and other gender confusions, see the introduction.
From a methodological point of view, we can agree that the types of confusions where the masculine is used for a neuter word (masc. pro neutr.) and those where the neuter is used for a masculine noun (neutr. pro masc.), can be both seen as evidence for the disappearance of the neuter, since in both cases we see examples of the unstable distinction between masculine and neuter, which, as the Romance languages attest, led to the virtual disappearance of the neuter. The situation is similar with the confusion between feminine and neuter. That is, I regard these confusions as hypercorrection, suggesting the future disappearance of the neuter.
Typically, the originally feminine names of trees with -us ending in the II declension became masculine, cf. Väänänen (n. 6) 105. Some nouns showed an oscillation between masculine and feminine gender, such as arbor, cinis, fons or vallis (Väänänen [n. 6] 105). In addition, some abstract masculine nouns of the type -or, -oris also became feminine, which was analogous to the fact that a significant proportion of abstract nouns were feminine. This change also helped to distinguish these nouns from masculine occupational nouns with the -tor form (so-called nomen agentis). One example of this change is the originally masculine noun dolor, which in French, for example, continues to exist as a feminine noun, f. la douleur, but it. il dolore (Väänänen [n. 6] 105; see also Löfstedt [n. 7] 243–245). Finally, for some nouns, a separate form of the feminine gender has been developed (in the LLDB database this is indicated by fem. pro communi code), for example sacerdos m./f. (‘priest’, ‘priestess’) is accompanied by sacerdotia (‘priestess’) and sacerda (‘priestess’), which differ in their geographical distribution (Väänänen [n. 6] 106).
In Chart 3, I have also marked with the Genera relativi those gender confusions which are also related to other change tendencies of the relative pronoun, so they cannot be considered as “pure” gender confusions. Examples of these codes and those that fall under the name Genera are given in note 26. The exact distribution of codes collected under the name Genera in Chart 3 is shown in Chart 4.
Löfstedt (n. 7) 230.
For details on the evaluability of such differences, see Adamik: In Search (n. 11) 124–126.
The different areas contain the following provinces: Africa (Africa proconsularis, Numidia, Mauretania Caesariensis, Mauretania Tingitana), Gallia and Germania (Aquitania, Belgica, Lugdunensis, Gallia Narbonensis, Germania superior, Germania inferior), Hispania (Baetica, Hispania citerior, Lusitania), Illyricum (Dacia, Dalmatia, Moesia Inferior, Moesia Superior, Noricum, Pannonia Inferior, Pannonia Superior, Raetia), Italy (Regiones I-XI, without Rome, as I have examined Rome separately from Italy). Germania and Gallia were included as a single unit because of the small amount of data from the area and the occasional common administration of these regions.
In Illyricum in the Later Period, the known data are distributed between the provinces as follows: Dalmatia (9 data), Noricum (1 data), Moesia Inferior (5 data), Moesia Superior (2 data), Pannonia Superior (1 data), Raetia (1 data). The weakening and fragmentation of the area's Latin culture during the Migration Period means that a significant proportion of the data are all from Dalmatia.
The large number (8) of confusions with fem. pro communi code seen in this area alone also plays a role in the size of this percentage, which, with two exceptions, are all examples of the feminine noun sacerda (‘priestess’) used here instead of the standard noun sacerdos (‘priest’, ‘priestess’), a form known only from Africa, cf. note 23. Furthermore, this phenomenon is fundamentally different from gender confusion since here a distinction that exists in principle (i.e., the difference between the two genders of a noun, cf. sacerdos pius and sacerdos pia) becomes explicit and is reflected in the form and conjugation of the noun, rather than the word behaving as a gender that it could not previously have.
This would be consistent with the observations of Löfstedt (n. 7) 230.
I have sorted the areas in descending order according to the total proportions.
In the case of Italian, this is mostly observed in the case of some dialects, while in Romanian the alternating gender is productive in general. See Maiden (n. 15) 169–174, Loporcaro (n. 10) 33–60. This relation is only a conjecture at this point, as the verification of this needs a different kind of examination of the data.
I have sorted the areas in descending order by the total change.
On the Romance occurrences and characteristics of the feminization of the neutral gender, see Maiden (n. 15) 172–174.
Adamik, B.: The Transformation of the Case System in African Latin as Evidenced in Inscriptions. Acta Classica Debreceniensis 55 (2019) 13–36.
Herman (n. 4) 65; Väänänen (n. 6) 102.
This is because the inscriptions in the area have already been fully processed.
