3.1 Analysis by the ratio of vocalic versus consonantal changes

Let us take a look at the provinces ranked by the ratio of vocalic versus consonantal changes in the 1^st century (Table 5.1, Chart 1a). Of course, at that time, in addition to the fact that the province of Dacia did not yet exist, most of the Danubian provinces, due to their relatively recent integration into the Roman Empire, were at a rudimentary level of Romanisation and Latinisation. Accordingly, on the one hand, there are provinces that have such a low number of data that the assessment of the situation is completely uncertain: such as Moesia Inferior and Moesia Superior, where the number of data forms is therefore enclosed in brackets. The western half of the region is the one from which we already have sufficient data from the 1^st century: Noricum and Pannonia Superior, as well as Pannonia Inferior, which has less data but with some caution can still be evaluated. The two Pannonias (which at that time were administratively still a single province) seem to be uniformly conservative in vocalism, but Noricum clearly has an innovative vocalism, since in Pannonia the rate of vowel changes is much lower, almost half (23.1% and 24.2%) of the expected rate (43%), while in Noricum the rate of vowel changes is slightly higher (47%) than the expected rate of vowel changes (43%). This expected ratio corresponds to the average distribution of vowels and consonants in Latin, according to which about 43% of Latin sounds are vowels and about 57% are consonants.²⁴

In the 2^nd century, we already have enough data from all Danubian provinces to assess the situation (Table 5.1, Chart 1b). It can be seen that the Danubian provinces do not behave uniformly in terms of the development of vocalism, so they do not show a typical grouping compared to the control provinces. At the same time, it is also noticeable that each province has moved towards a more intensive vocalism development. Dacian Romanism, which appeared in the 2^nd century, immediately jumps to the top of the provinces with its vocalism that is more intense than expected (51.4%), while Noricum also remains slightly innovative in vocalism (44.4%), and Moesia Inferior also enters the scene in the same league (44.6%). Pannonia Inferior shows a completely balanced development at this time, as the rate of vowel changes (42.7%) is essentially the same as the expected rate of vowel changes (43%). Moesia Superior (40.3%) and Pannonia Superior (36.6%) appear at the other end of the axis with their slightly conservative vocalism.

After that, in the 3^rd century (Table 5.1, Chart 1c), we can see that the Danubian provinces are surprisingly uniformly innovative in the field of vocalism, since all of them have a higher than expected (43%) vowel change ratio (with values between 44.9% and 52.4%). This way they align in a relatively sharp contrast to all the control provinces, which show a slightly or strongly conservative development at this time, meaning they were more innovative in terms of their consonant system. We will see what the real dialectological value of this apparent uniformity is in the subsequent analysis of vowel mergers.

Turning to the analysis of the 4^th century (Table 5.1, Chart 1d), we can immediately face the biggest problem of the region: due to the unfavourable historical conditions, Romanism in the Danubian region began to erode, or even disappeared completely, as in Dacia, which was evacuated at the end of the 3^rd century at the latest.²⁵ Where we still have enough data, such as in Moesia Inferior (61%), and in Pannonia Inferior (61%), it appears that these areas have moved further towards a system with an even more radically innovative vocalism. It is possible that this happened in the other Danubian areas as well, where a sufficient Latin-speaking population may have remained, but the low data density no longer allow reliable conclusions to be drawn regarding the investigated phenomenon, except perhaps in Pannonia Superior, which seems to have the most innovative vocalism (64.3%), and where the number of data forms, 28, is just below the exclusion limit, which is 31.

3.2 Analysis by the percentage of E/I and O/U confusions

Now let us move on to the actual subject of my presentation, the examination of the vowel mergers appearing in E/I and O/U misspellings in the Danubian provinces. First, I examine the percentage of E/I and O/U confusions counted against all vocalic errors and rank the selected provinces in each century accordingly (Table 5.2, Chart 2a–d). If you look at diagram 2a), you can see that the rate of vowel mergers is low everywhere, well below the expected rate (71%), which is in accordance with the rudimentary nature of the change process. This expected 71% ratio corresponds to the average proportion of e-i and o-u vowels in the Latin vowel system that could have been affected by the vowel merger.²⁶ At that time, according to the rudimentary level of Romanisation of some of the Danubian provinces, we have no data on E/I and O/U confusions in the two Moesias, and neither in the province of Dacia, which did not exist at the time. We have scattered data on the phenomenon from the other Danubian provinces, of which probable statements can only be made regarding Noricum (6.3%) and Pannonia Superior (7.1%) in light of their sufficient amount of data; these were neighbouring provinces, meaning that the matching developments might have been related. According to this, this area was the most conservative among the examined provinces, but it was not far behind the conservatism of Central Italy (Samnium 9.6%, Roma 9.5%), while it was significantly behind its southern neighbours, Venetia et Histria (23.1%) and Dalmatia (15.7%), which seem to be main locations that boosted the spread of vowel mergers in the region.

Table 5.2.

e/i + o/u faults: percentage of all vocalic errors

After that, if we look at the 2^nd century in diagram 2b), it can be seen that vowel mergers became more common in all areas from which we had data from the 1^st century. The Danubian region does not behave uniformly in terms of change. It seems that an innovative eastern region was emerging, where Moesia Inferior (19.8%) and especially Moesia Superior (26%) started to catch up with the most innovative provinces in terms of vowel mergers, i.e. Dalmatia (27.8%) and Venetia et Histria (32.2%). At the same time, a more conservative western region began to take shape, where vowel mergers spread very slowly and were still relatively rare: the most underdeveloped were Pannonia Inferior (7.4%) and Dacia (7.8%), while Noricum (11.1%) and Pannonia Superior (11.8%) seem to be less underdeveloped compared to them.

After that, in the 3^rd century (see diagram 2c), we can see that compared to the 2^nd century, the frequency of vowel mergers continued to rise slightly, in fact everywhere, even in the very conservative Dacia (13.1%) and Noricum (13.7%). A decrease can be seen only in the case of Moesia Superior (20.4%) and in Bruttium et Lucania (12.5%), but in the latter the extremely low number of data forms from the 3^rd century does not allow us to judge the situation well. Apparently, there was also a more conservative and a more innovative group within the Danubian provinces at this time, but they were mostly not next to each other, such as Moesia Inferior (35.5%), which caught up with the innovative Dalmatia (31.2%) and Venetia et Histria (37%), and Pannonia Superior (26.6%), which was catching up with them. The two most conservative provinces, Noricum (13.7%) and Dacia (13.1%) were not next to each other either, as the area between them was occupied by the innovative Pannonia Superior (26.6%). Pannonia Inferior (15.7%) and Moesia Superior (20.4%), on the other hand, were right next to each other and to Dacia, so it might be possible to talk about a relatively uniform development in relation to them, while Moesia Superior, with its higher ratio of 20.4%, still seems to move away from the group in a more innovative direction.

After that, if we look at the developmental characteristics of the areas of the Danubian territories with an acceptable amount of data in the 4^th century in the light of diagram 2d), we can see that the remaining Danubian provinces still did not behave uniformly in terms of the frequency of vowel mergers. In this regard, Moesia Inferior (61.1%), ahead of Dalmatia (50.7%) and Venetia et Histria (58%), is considered a clearly radical innovative area. At the same time, Pannonia Superior (44.4%) can be regarded an innovator catching up, and Pannonia Inferior (20%) can be considered conservative and lagging behind, but the low numbers warrants for caution. At this point it is worth mentioning that these three provinces were located in the same order and in roughly the same place on the frequency scale in the 2^nd and 3^rd centuries (2b and 2c), which may indicate permanently preserved differences in the territorial development of the affected areas. This circumstance can already prompt us to be cautious about the seeming dialectological uniformity of the Danubian region, which we could have deduced from diagram 1c, where we could see that the Danubian provinces were surprisingly consistently innovative in the field of vocalism and in this they seemed to be distinct from the other control provinces.

3.3 Analysis of the E/I to O/U ratios

The picture painted so far should be refined by an analysis of the E/I to O/U ratios in the selected areas in all four centuries, see charts 3a, 3b, 3c and 3d of Table 5.3. While evaluating the rates of o-u and e-i mergers as displayed there, I use the expected ratio of 1.9 as a reference number. This reference number indicates that the number of E/I confusions in the data pool is approximately the double of O/U confusions. This number also indicates that the two changes took place more or less at the same time and with the same intensity, since the number of the e-i sounds is approximately the double of o-u sounds in Latin (for comparison, Herman used 2 as his reference number, which is almost the same).²⁷ Consequently, a numerical value considerably higher than 1.9 indicates that the merger of front vowels e-i was more advanced than that of back vowels o-u in a given area, or in other words that the merger of back vowels had just started or was underdeveloped compared to the merger of front vowels.

Table 5.3.

e/i to o/u ratio

Table 5.4.

e/i/o/u to é/í/ó/ú ratio

In line with this, let us take a look at the data from the 1^st century (Chart 3a). At that time, we have no data from the province of Dacia, which did not exist yet, but there are no examples of vowel mergers either from Moesia Inferior or Moesia Superior. There are only a couple of examples from Pannonia Inferior, Pannonia Superior and Noricum, only for E/I confusion, this is indicated by the purely illustrative technical number 12 in parentheses in diagram 3a). There is one example of O/V confusion from contemporary Dalmatia and Venetia et Histria each, but there are also 10 to 11 E/I confusions, so the indicator number of these two areas is high (10 and 11, respectively). What is also interesting is that in Rome, where the rate of vowel mergers was relatively low (9.5%), the ratio of E/I and O/V confusions compared to each other is 2, so essentially we can observe the expected value (1.9) here, which means that the city of Rome and perhaps Samnium (1) and Apulia et Calabria (1.7) were at the forefront of back vowel merger. This suggests that the merger of back vowels, which at that time almost did not affect Northern Italy at all (the proportion of Venetia et Histria is 11),²⁸ presumably started to spread from central and southern Italian areas.

If we now consider the 2^nd century with the help of diagram 3b), it can be seen that the merger of back vowels began to spread almost everywhere, but with varying intensity. The Danubian provinces do not show a unified picture in this regard, as they are at different stages of development, and those Danubian provinces showing similar trends are not next to each other. There are only one or two scattered examples of O/V confusion in Pannonia Inferior and Moesia Superior, so the ratio of these two areas is relatively high: 6 and 5.5. There are 2 clear examples of O/V confusion from Dacia, in addition to 8 E/I confusions, which resulted in a ratio of 4, but the low number warrants for caution. Pannonia Superior is considered innovative to a lesser extent (2.7), and Moesia Inferior to a greater extent (2.4), which in the case of the latter is confirmed by the relatively significant number of pieces (17:7). Noricum is an outlier, as almost only O/V confusions are reported here, but the vast majority of the examples are recorded in the Dolichenus epithet, and albeit these happen to be on locally made silver cult objects, it cannot be ruled out with complete certainty that in fact the Dul- name variant was intended, which originated from an external location.²⁹ It is also noteworthy that Northern Italy represented by Venetia et Histria (2.8) not only caught up with Rome (3), but even left it behind in terms of the merger of back vowels, so it can be assumed that Venetia et Histria may have played an important role in spreading this innovation to the Danubian provinces.³⁰

Turning to the developments of the 3^rd century, at first glance, in the light of diagram 3c), it appears that some kind of unification trend is emerging among the Danubian provinces, somewhat similar to what is depicted in diagram 1c), since almost all Danubian provinces are located next to each other on the diagram (only Venetia et Histria appears to be wedged between Moesia Superior and Pannonia Superior). However, if we take a closer look at the differences in the ratios of these provinces, the structural fragmentation of the region immediately becomes clear. In terms of o-u merger, Dacia is the most conservative (11:2=5.5), Pannonia Inferior (21:5=4.2) and Moesia Inferior (31:8=3.9) clearly show more innovative tendencies, Noricum and Moesia Superior are even more innovative (however, the low numbers in these two provinces warrant for caution: 6:2=3 and 7:3=2.3, respectively), and Pannonia Superior (20:9=2.2) was slightly closer to the expected ratio (1.9).

If we only look at the Danubian provinces and take into account aspects of Romance linguistic geography, meaning that we look for traces of asymmetric eastern and symmetric western type development trends in the Latin material, where in diagram 3a–d the former is indicated by the high ratio and the latter by the low ratio, then we could come to the conclusion that while Pannonia Superior (2.2) clearly followed the symmetrical western type of development, Dacia (5.5) clearly took the path of the asymmetrical eastern type of development. Moreover, it seems that Dacia clearly moved forward in the direction of the asymmetrical eastern type system compared to the 2^nd century, since its ratio did not decrease but increase: 3b) 4 > 3c) 5.5. If we also consider the well-known fact that the Romanian language developed from one of the branches of Balkan Latin and is clearly of an asymmetric eastern type (where there is only e-i merger), we could come to the conclusion that in the development of Latin in Dacia in the 2^nd–3^rd centuries, we found the antecedent of the relevant characteristic of the later Romanian language. However, such a conclusion may prove to be hasty if the wider context is not taken into account. Not only Dacia (5.5), but also Dalmatia (7.2) shows asymmetric eastern development tendencies at this time, and Dalmatia even more strongly than Dacia. Moreover, not only Dacia, but also Dalmatia seems to move towards the eastern asymmetric system in the 3^rd century compared to the 2^nd century: 3b) 4 > 3c) 7.2. We now know about Dalmatia that later it turned toward a symmetrical Western type of development, which is clearly visible in the Latin inscriptions of the area already in the 4^th century, 3d) 52:21=2.5. We can infer the same from the extinct Dalmatian language.³¹ While in the case of Dalmatia later Latin material from the 4^th, 5^th and 6^th centuries is also available, in the case of Dacia it is not, so we cannot say in what direction the Latinity of Dacia and its population that emigrated from there developed further in terms of vowel mergers in the 4^th and even in later centuries.³²

Turning briefly to the 4^th century, in the light of diagram 3d) it can be concluded that Pannonia Inferior (4) and Pannonia Superior (3) among the Danubian provinces with an amount of data that can still be assessed were probably among the areas catching up in terms of o-u merger at the time (although the low figures warrant for caution for both provinces), similarly to Venetia et Histria (3.3) and less similarly to Dalmatia (2.5). The only area with reliable data, Moesia Inferior, clearly shows a western symmetrical development trend (9:13=0.7). About the other Danubian areas it is not possible to make a reliable statement.

3.4 Analysis of the E/I/O/U to É/Í/Ó/Ú ratios

Since Herman also described the relationship between Danubian provinces and the other areas he studied in terms of whether or not there is a difference between stressed syllable and unstressed syllable development of vowel mergers, I will also examine the selected provinces in this regard. Of course, I also use an expected ratio here, 1.7, which expresses that in Latin, 62.8% of vowels in general occur in unstressed syllables and 37.2% in stressed syllables (62.8 : 37.2 = 1.69 = 1.7).³³ To this expected ratio of 1.7 I will compare the observed ratios of the studied areas in all four centuries. If the observed ratio in a given area is higher than the expected ratio, vowel mergers were much more intense in unstressed syllables than in stressed ones, and vice versa, if the observed ratio is lower than the expected ratio, the vowel mergers were much more intense in stressed syllables than in unstressed ones. If the observed ratio is the same as the expected ratio (1.7) or very close to it, the vowel mergers took place with the same intensity in stressed and unstressed syllables.

If we look at the chart displaying results from the 1^st century (Chart 4a), we can see, just as before, that there is either no data from the provinces of the region at all, as shown by the case of Moesia Superior and Moesia Inferior (not to mention the province of Dacia, which did not exist at that time), or if there is, the small amount of data does not allow for drawing conclusions, as shown by the example of Pannonia Inferior (1 piece of data) and Pannonia Superior (2 pieces of data). As before, in the 1^st century only the data of Noricum let us draw some sort of conclusion, namely that the stressed syllable may have been affected by vowel mergers to a slightly greater extent than the unstressed syllable (2:2=1), in view of the overall unequal occurrence of vowels in stressed and unstressed syllables, as expressed by the ratio 1.7; however, due to the relatively low number of data, it is reasonable to formulate this conclusion with caution. What seems remarkable is that in most of the control provinces the stressed syllable seems to have been affected by vowel mergers to the expected degree or close to it: this seems clear in Rome (2.3), but it seems probable in Dalmatia (1.2) as well. Venetia et Histria (3) seems somewhat more conservative, as here the unstressed syllable was more affected by vowel mergers than the stressed syllable. The low number of data forms in the other areas does not allow us to draw conclusions about them.

After this, if we turn to the analysis of the 2^nd century with the help of diagram 4b), we can see that the Danubian provinces did not behave uniformly at that time. A lagging group can be discerned where the stressed syllable is essentially untouched by vowel merger, including Noricum (where the technical number 10 means that we only have examples of unstressed syllables) and Dacia (9), where there is only one example of stressed syllable confusion at this time. This is followed by a group catching up, where the stressed syllable was already clearly affected by vowel merger, to a lesser extent Moesia Superior (3.3) and to a greater extent Moesia Inferior (2.4). Finally, there was a group where the stressed syllable was even more affected by vowel merger than expected: this includes definitely Pannonia Superior (1.2) and perhaps also Pannonia Inferior (0.4), where the low number of data forms warrants for caution.

In the 3^rd century (chart 4c), it seems that the levelling of the Danubian territories had begun, since vowel merger appears in the stressed syllable in Dacia (3.3) and Noricum (3), but it is still clearly far from the expected ratio (1.7). At the same time, it can be seen that the Danubian provinces do not behave uniformly in this respect, since, for example, Pannonia Inferior (4.2) and Moesia Inferior (3.9) seem to be more conservative than Dacia (3.3) and Noricum (3), where the stressed syllable seems to be more resistant to vowel merger than in other Danubian areas. In addition, a group can be detected that can be regarded as almost caught up, where the ratio of vowel mergers approaches the expected value: primarily Pannonia Superior (2.2) with a high number of data, and Moesia Superior (2.3) with a lower number of data can be classified here. We cannot therefore speak of neither complete levelling, nor a territorial unification trend.

In the 4^th century (chart 4d), in those areas from which an acceptable amount of data is available, it can be seen that both Pannonia Superior and Pannonia Inferior were lagging behind in terms of vowel mergers in stressed syllables, Pannonia Superior to a greater extent (7), and Pannonia Inferior to a lesser extent (4) – as far as the a relatively small amount of data allows any kind of conclusion. Only Moesia Inferior (1.8) seems completely caught-up, with balanced development, since its ratio essentially corresponds to the expected ratio (1.7).³⁴

It can already be pointed out that the 4^th-century Danubian provinces with sufficient data still available for them display a very similar pattern in terms of all three aspects of vowel mergers, with one exception, they essentially follow each other in the same order. Moesia Inferior has the highest overall vowel merger ratio (2d: 61.1%), the lowest front-back vowel merger ratio (3d: 0.7) and the lowest unstressed-stressed vowel merger ratio (4d: 1.8). As a result, Moesia Inferior can be considered the most western type Danubian region in terms of the development of vowel mergers. The Danubian province with the second highest overall vowel merger ratio is Pannonia Superior (2d: 44.4%), with also the second lowest front-back vowel merger ratio (3d: 3) and the highest unstressed-stressed vowel merger ratio (4d: 7). As such, Pannonia Superior can be considered an area catching up to the Western type in terms of the development of vowel merger, except for the development in stressed syllables, in which this area may have been the most behind. Finally, Pannonia Inferior can be considered the most underdeveloped representative of the Danubian provinces, with the lowest overall vowel merger ratio (2d: 20%), the highest front-back vowel merger ratio (3d: 4) and the second highest unstressed-stressed vowel merger ratio (4d: 4). Pannonia Inferior can therefore be considered an area lagging behind Western development in terms of vowel merger, except for the development in stressed syllables, in which respect this area may have been slightly more innovative, at least compared to Pannonia Superior. In any case, we cannot give an answer to how these areas and the entire Danubian region could have continued to develop in the centuries to come, supposing there was any room for their development in the chaos of the migration period, due to the complete lack of sources.