193 Arheološki vestnik 71, 2020, 193–215; DOI: https://doi.org/10.3986/AV.71.05
The top-ranking towns in the Balkan 
and Pannonian provinces of the Roman Empire
Najpomembnejša antična mesta
balkanskih provinc in obeh Panonij 
Damjan DONEV
Izvleček
Rimska mesta balkanskih in podonavskih provinc so bila doslej le redko del raziskav širših mestnih mrež. Namen 
prispevka je prepoznati glavne značilnosti mestnih sistemov in na podlagi najpomembnejših mest provincialne mestne 
hierarhije poiskati njihovo vpetost v ekonomijo provinc v času severske dinastije. Avtor se osredotoča na primerjavo veli-
kosti prvorazrednih mest z ostalimi naselbinami, upošteva pa tudi njihovo lego in kmetijsko bogastvo zaledja. Ugotavlja, 
da moramo obravnavano območje glede na ekonomske vire razumeti kot obrobje rimskega imperija. Glavna bogastva 
obravnavanih provinc so bili namreč les, volna, ruda in delovna sila, kar se jasno izraža tudi v osnovnih geografskih 
parametrih prvorazrednih mest: v njihovi relativno skromni velikosti, obrobni legi in vojaški naravi.
Ključne besede: Balkanski polotok; Donava; principat; urbanizacija; urbani sistemi
Abstract
The Roman towns of the Balkan and Danube provinces have rarely been studied as parts of wider urban networks. 
This paper attempts to identify the principle features of these urban systems and their implications for the economy of 
the provinces at the time of the Severan dynasty, through the prism of the top-ranking towns in the provincial urban 
hierarchies. The focus will be on the size of the first-ranking settlements in relation to the size of the lower-ranking 
towns, their location and the agricultural riches of their hinterlands. One of the main conclusions of this study is that, 
from an economic perspective, the region under study was a peripheral part of the Roman Empire. Its main assets were 
its timber, wool, metallic ores and labour force. This is reflected in the basic geographic parameters of the first-ranking 
settlements: their relatively humble size, their peripheral locations and militaristic nature.
Keywords: the Balkan peninsula; the Danube; High Empire; urbanization; urban systems
INTRODUCTION:
CONCEPTS AND GOALS
One of archaeology’ s many borrowings from the 
broader field of Human Geography is the concept 
of urban or, more generally, settlement systems. 
The attraction that this notion holds for the ar-
chaeologist is in part created by an allusion to an 
underlying scheme in the distributions of settle-
ments discovered in the field. Settlement maps – an 
iconic artefact of archaeological knowledge – are 
more than meaningless site-distributions or mere 
reflections of the distribution of natural resources. 
The concept of urban system supersedes that of 
urban geography by postulating a set of stable and 
structured relationships between the individual 
towns of a given area. This implies that a group 
of settlements that belongs to a certain geographic 
or economic area constitute an integral organism. 
Changes in one segment of the system are bound 
to impact the integral structure. Therefore, the 
history of individual towns is not determined 
solely by “endogenous” factors, but it is part of a 
broader socio-economic dynamic.

194 Damjan DONEV
On the surface, there is nothing controversial in 
this position. After all, all good histories of individual 
towns place a strong accent on the participation of 
the urban communities under study in the wider 
socio-economic processes and conditions.
1
 It is 
unnecessary to adopt the concept of urban systems 
to learn that towns are not isolated islands. The 
conceptual muddle has in part arisen because ar-
chaeologists often need to be reminded that towns 
– and settlements in general – are essentially places 
at which social and economic processes intersect, 
not just distinct site-categories.
2
 The distribution 
of towns and settlements does not merely reflect 
the distribution of population, but is also highly 
sensitive to the distribution of services, produc-
tion and consumption and wealth. Although some 
degree of ambiguity is unavoidable, it can be said 
that there are patterns in the urban geographies 
that cannot emerge outside a certain set of social 
and economic relations. From this point of view, 
towns are the silent source material for the study 
of the societies that have created them.
The concept of urban systems is not entirely 
unproblematic when applied to pre-industrial socie-
ties. It has been devised by economic geographers 
who study modern societies and, in its original 
form, does not account for the different realities 
in the pre-industrial era. Although no one would 
be so obdurate to deny any connections between 
settlements that belong to the same geographic 
or administrative unit, the degree of interaction 
between individual towns and settlements, even 
in the case of advanced societies like the Roman 
Empire, is a controversial topic.
3
 Many archaeolo-
gists and historians of antiquity would argue that 
the town-countryside relationship is much more 
important than the town-town relationship for the 
study of the towns of Classical Antiquity. No less 
problematic is the notion of an enclosed entity 
entailed in the concept of urban systems. This is a 
minor issue when dealing with national economies 
in an Early Modern context, but it becomes ever 
more difficult to define the unit of analysis as we 
move back in time.
4
The analytical tools that accompanied this concept 
have been of a far greater practical significance 
1
  See for example, Mossé 1973.
2
  Cf. Abrams, Wrigley (eds.) 1978.
3
  Finley 1977; Rich, Wallace-Hadrill (eds.) 1991; 
Cornell, Lomas (eds.) 1995; Bowman, Wilson (eds.) 2011; 
Hanson 2016.
4
  Johnson 1980.
than its theoretical implications. Although these 
techniques have been adopted by archaeologists 
for several decades, not everyone has embraced 
them wholeheartedly and their usage is still rela-
tively limited.
5
 This is not the place to review 
the differing opinions on the topic of geographic 
analysis in archaeology.
6
 Obviously, the main point 
of this disagreement has been the extent to which 
these ideas and techniques can be applied to the 
distant past without distorting historical reality. 
In any case, there is no apparent reason to discard 
the whole approach on theoretical grounds. More 
importantly, there are not many alternatives to the 
traditional descriptive approach that does not allow 
us to look into the possible structure of the urban 
geography and its ramifications for the society or 
time-period under study.
In this paper, we shall demonstrate the still sig-
nificant potential of some of these techniques for 
the study of ancient societies and their economies, 
on the example of the urban hierarchy in the Bal-
kan and Danube provinces of the Roman Empire 
in the second and third centuries AD.
7
 Settlement 
hierarchies are a topic familiar to most archaeolo-
gists. This is a facet of settlement geography that 
has been studied by archaeologists long before the 
adoption of the concept of the urban system.
8
 Its 
popularity in regional studies suggests that many 
archaeologists believed that settlement geographies 
were more or less structured – otherwise hierarchies 
would have been very shallow or imperceptible – 
even if they lacked the conceptual and analytical 
apparatus to elaborate this point. Among other 
things, this aspect of settlement geography has 
attracted much attention, precisely because of its 
5
  Orton, Hodder 1976; Johnson 1977; Lock (ed.) 2000; 
V erhagen 2018; for criticism, see, Hodder 1986; Tilley 1994.
6
  A useful summary of the early views is given in Earle, 
Preucel 1987; this debate has continued within the context 
of GIS and digital technology, Van Leusen, Gaffney 1995; 
Wheatley 2004.
7
  Most of the data has been collected by the author 
of this paper within the frames of the project “An Empire 
of 2000 cities: economic integration and urban networks in 
the Roman Empire” , coordinated by Luuk de Ligt and John 
Bintliff. The size-figures for the towns of the southern part 
of Epirus, Lower Macedonia, coastal and eastern Thrace 
have been kindly provided by Michalis Karambinis and 
Rinse Willet. The period of the Severan dynasty has been 
chosen, because it is believed that the urban system of the 
Empire reached a point of maturity at that point of time. 
Admittedly, this is not applicable to all provinces of the 
Empire, but it fits the developments in the study-area.
8
  For example, Adams 1965.

195 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
relevance to the organization of administration, 
economic and religious life in the past.
Understandably, a detailed analysis of the urban 
hierarchies in all Balkan and Danube provinces 
would be out of place for this paper.
9
 Therefore, 
the focus shall be on the top-ranking settlements 
in the urban systems of the individual provinces. 
As we shall shortly see, every province in this 
study area had one or more urban settlements that 
were clearly set apart from the rest of the towns. 
One of the goals of this survey is to identify and 
examine the parameters that distinguish the top-
ranking towns from the other settlements in the 
urban hierarchy. Apart from practical reasons, the 
choice to focus on the top-ranking settlements is 
also motivated by the special place of this settle-
ment category in the urban hierarchies. Scholars 
that study urban systems have observed that the 
size of the top-ranking settlements, and especially 
their relation to the settlements of lower rank, is 
an important index of the nature of the economic 
9
  Donev 2020.
relations in the regions under study.
10
 This is a 
measure of the degree of economic and politi-
cal centralization in society. To be sure, looking 
solely at the top-ranking settlements cannot reveal 
everything about a given urban system, but it can 
be highly indicative, not only of the relationships 
between the towns that constitute the system, but 
also of the place and role of the study-region in 
the wider urban constellation.
11
There are other, equally interesting questions 
that surround the emergence of first-tier settle-
ments. Did they perform any special functions, 
not shared by the rest of the urban settlements? Is 
there anything in the locations of this settlement 
category that gave them the necessary advantage 
over the average provincial town? Were the first-
ranking settlements self-sufficient or were they 
10
  The concept of primate city was first introduced by 
Jefferson (1939) and was quickly adopted across the social 
sciences. Berry 1961; Vapnarsky 1969; Timberlake (ed.) 
1985; Chen 1991; Chase-Dunn, Willard, Willard 1994; 
Chase-Dunn, Manning 2002.
11
  A good example is the study by Smith 1985, 121–167.
Map 1: Map of the study-region and the size of the top-ranking settlements.
Karta 1: Obravnavano območje in velikost prvorazrednih naselbin.

196 Damjan DONEV
dependent on some special arrangements to secure 
their sustenance? It is important to stress that the 
goal is not to offer an explanation for the rise of 
these particular towns on the top of the urban hi-
erarchy. The factors that shape urban geographies 
are often unextractable from the particular histori-
cal developments and, obviously, these cannot be 
considered for every regional centre. By focusing 
on a few aspects of the urban geography, we hope 
to draw the profile of the regional capitals and 
infer possible implications for the nature of the 
provincial societies and their place in the Roman 
Empire.
The study-area includes all Balkan provinces 
with the exception of Achaea (Map 1). The small 
segment of the tenth Italian region that extends 
into Istria and the Julian Alps is left out of the 
analysis, because the great majority of the towns 
in this region, including the top-ranking town of 
Aquileia, are located on the Apennine peninsula. 
Noricum is likewise excluded because it belongs 
to the Alpine region, although, unlike Italy, it had 
much in common with the Balkan and Pannonian 
provinces. Taken as a whole, this fairly large area 
does not represent a fully meaningful analytical 
unit. There are profound differences between the 
frontier provinces along the Danube and Dacia, 
and the demilitarized provinces in the southern 
and western parts of the peninsula. This study shall 
also point to a number of similarities between the 
urban hierarchies in the individual provinces but, 
in principle, each province represents a discrete 
urban system.
The size-figures refer to the period of the High 
Empire or, more precisely, the time of the Severan 
dynasty – arguably, for many provinces in this 
study region, a period of prosperity. Admittedly, 
a diachronic approach would have been much 
more promising than focusing on one point in 
time, but the scarcity of data does not allow us 
to venture into this direction. Even within a 
synchronic approach, the margin of uncertainty 
is great, especially for towns like Dyrrhachium, 
Byzantium or Thessaloniki, occupied continu-
ally until the present day. Because of this, it is 
necessary to consider alternative size-estimates 
for some of these towns. In order to simplify the 
discussion, the possibilities will be limited to the 
minimum and maximum size-estimate for every 
town included in the analysis. In non-controversial 
cases, the difference between the two alternative 
estimates is negligible.
WHICH ARE THE TOP-RANKING
SETTLEMENTS?
The urban and urban-like settlements of the 
Balkan and Danube provinces – here defined as all 
agglomerations greater than 5 ha and with evidence 
of public buildings or formal institutions – group 
into three or four size-categories. Whether these 
settlements are spread across pre-defined size-
ranges or plotted against their rank on a logarithmic 
scale – Rank Size analysis – the outcome is the 
same. The specific size-ranges may vary between 
provinces, but in general, there is a strong posi-
tive correlation between juridical status and size. 
Settlements could rarely aspire to grow beyond 
the threshold of 15 ha, without an autonomous 
status. Excluding a small number of obscure poleis 
or municipia, most of the settlements that belong 
to the lowest size-range lack autonomous status. 
Technically speaking, they were not towns, but 
this fact does not necessarily exclude them from 
the urban system. The fact that this size-category 
represents at least 70% of the settlement systems 
hints, at its importance for the urbanization of the 
provinces. In every province, the second-tier set-
tlements are represented by the largest among the 
autonomous towns. Normally, this group accounts 
for between 8 and 20% of all settlements included 
in the analysis. In some provinces, the remaining 
autonomous towns, together with a small group 
of prosperous vici, cluster into a distinct third tier. 
In others, there is less differential growth within 
juridical categories and they join the lowest tier 
in the hierarchy.
In all provinces, the urban hierarchies have a 
clearly perceptible top settlement-tier (Tab. 6). It 
has been somewhat surprising to discover that 
the number of settlements that belonged to this 
tier differs from province to province. Over half 
of the provinces have a single settlement on the 
top of the urban hierarchy. These are Dalmatia 
with Salona, Macedonia with Thessaloniki, Thrace 
with Byzantium and Dacia with Apulum. It is 
possible that this model was more widespread. If 
more conservative size-estimates are adopted for 
Dyrrhachium, Apollonia and Durostorum, both 
Epirus with Nikopolis and Moesia Inferior with 
Novae will join this group.
A different pattern emerges in the Pannonian 
provinces, Moesia Superior and, if the high size-
estimate for Durostorum and the low size estimate 
for Nikopolis are accepted, Moesia Inferior and 
Epirus. In these provinces, the second largest set-

197 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
tlements occupy more than three-quarters of the 
area occupied by the first-ranking settlements and 
the two are lumped together in the same size-range 
or placed too close to each other on the rank-size 
graphs. In fact, in Pannonia Superior there are not 
two, but three settlements in the top size-range. 
These are the two legionary towns of Carnuntum 
and Vindobona, and Poetovio. In the other prov-
inces of this group, with the obvious exception of 
Epirus, the top-tier of the settlement hierarchy is 
reserved for the legionary towns: Aquincum and 
Brigetio in Pannonia Inferior, Viminacium and 
Singidunum in Moesia Superior and, conditionally, 
Novae and Durostorum in Moesia Inferior. In the 
case of Epirus, a lower size-estimate for Nikopolis 
will lower the size differential between the first- 
and second-ranking settlements to less than 20%, 
and the newly founded provincial capital will join 
the Corinthian colony of Apollonia in the top tier 
of the urban hierarchy. It should not be excluded 
that the second-ranking settlement was the Roman 
colony of Dyrrhachium – whose size estimate is 
extremely uncertain – and not Apollonia, or that 
all three towns belonged to the top rank.
12
This unusual distribution is almost an exclusive 
attribute of the urban hierarchies in the frontier 
provinces. The case of Moesia Inferior is problem-
atic because we have virtually no clue about the 
size of the canabae of Durostorum. The minimum 
estimate for this town is provisional. It is very 
likely that the true size-figure was much closer to 
the maximum than to the minimum estimate.
13
 
Epirus too is an unlikely exception. In this case, 
the maximum estimates for Dyrrhachium and, 
especially Apollonia, are meant to represent the 
improbable extremes. On the other hand, the 
maximum estimate for Nikopolis is more probable 
than the minimum.
14
Most urban geographers would interpret the 
presence of more than one settlement in the top 
tier of the settlement hierarchy as a sign of an 
immature urban system.
15
 In normal conditions 
12
  The Augustan walls of Dyrrhachium enclose an 
area of almost 50 ha, Gutteridge, Hoti, Hurst 2001; but 
accidental discoveries during modern construction suggest 
an extensive suburbium; Santoro, Monti 2008. On the other 
hand, geophysical research at Apollonia in the last couple 
of decades has shown that the town contracted during the 
Roman period; Lamboley et al. 2011.
13
  Ivanov 2003; Donevski 2009; Donevski 2015.
14 
 This has been clearly indicated by M. Karambinis 
in the project’s database.
15
  Johnson 1980, 234–236, 240.
and given the necessary time depth, the uneven 
circulation of people, wealth and ideas should result 
in a differential growth among the leading settle-
ments, leaving only one settlement in the top-tier. 
The close proximity of the top-ranking settlements 
in the frontier provinces – the legionary towns are 
often spaced at distances shorter than 100 km – 
eliminates the possibility that they were centres of 
distinct sub-regions. The time-factor could have 
also played an important role. In the time period 
that is the focus of this study, the urban systems 
of the frontier provinces were not older than a 
century to a century and a half. Compare this to 
the urban systems of the demilitarized provinces 
that had evolved over a period of several centuries 
prior to the reign of the Severan dynasty. The 
urban geography of the frontier provinces was 
chiefly outlined by the movements and actions of 
the Roman army at the time of conquest, and the 
military towns continued to play a vital role in 
the administration of the frontier provinces and 
the defence of the frontiers.
THE PRIMACY INDEX
The differences between the size of the built-up 
areas of the first-ranking settlements in this study-
region are not negligible, even if only the largest 
settlements in every province are considered (Tabs. 
1; 2). The maximum figures range from 70 to 200 
ha, the minimum, from 53 to 157. They cluster 
into three size ranges for both estimates.
16
 The 
majority of the first-ranking settlements belong to 
the middle size-range. They measure between 70 
and 100 hectares according to the minimum, or 
between 100 and 150 ha, according to the maxi-
mum estimates. Only the Dalmatian capital Salona 
belongs to the lowest size-range of 50 to 70 ha or, 
alternatively, to that of 70 to 100 ha. Depending 
on the size-estimates, the three largest towns in 
the study-area measure between 100 and 150 or 
between 150 and 200 ha.
Seen from an Empire-wide perspective, this 
part of the Roman Empire lacked very large towns. 
Admittedly, both the Pannonian capitals and 
Byzantium satisfy the criteria to enter Pounds’ A 
category,
17
 but the maximum estimates for these 
16
  The size-ranges have been determined by the Natural 
Break method; Osaragi 2002.
17
  Pounds 1969. This pioneering study of the urban-
ization of the Mediterranean in Classical Antiquity has 

198 Damjan DONEV
towns are, for different reasons, problematic.
18
 In 
this respect, the Balkan and Danube provinces stand 
much closer to the western European provinces 
than to the Eastern Mediterranean.
No apparent patterns emerge from the dif-
ferences in size between individual towns. This 
should occasion no surprise. Settlement size is 
the resultant of an unknown number of factors 
that are often historically and geographically 
specific. Nonetheless, overall, the military towns 
are larger than the civilian. The inclusion of the 
smaller top-ranking settlements would have only 
reinforced this pattern. Many of the civilian towns 
come close to the size of the smaller of the two 
legionary towns, but they never really outgrew 
them. The only exceptions are Byzantium and, 
if the high size-estimates are accepted, Nikopolis 
recognized five size-categories. Category A unites the 
largest towns in the Roman Empire, with built-up areas 
greater than 100 ha.
18
  The Pannonian capitals were double towns – see 
below – and, in theory, it is possible to treat them as 
two separate towns or deny the urban character of the 
military component; Piso 1991; Doneus, Gugl, Doneus 
2013. Byzantium is a difficult case for obvious reasons. It 
is comforting that the size-estimate accepted here – cor-
responding to the area of the town enclosed by the Sev-
eran wall – can be transformed into a population figure 
that matches the population estimates for pre-4
th
 century 
Byzantium, derived from historical sources, Mango 1985; 
Mango, Dagron (eds.) 1995.
and Poetovio. Severan Byzantium might have been 
incomparably smaller and less important than the 
towns of Constantine I or Justinian I, but the fact 
that it was the only civilian town that joined the 
group of largest towns in the wider region, without 
the presence of the military, is indicative of its 
potential for growth. At the same time, the fact 
that only a few civilian towns managed to cross 
the 100-ha threshold underscores the economic 
imbalance between the military and civilian sectors. 
The scope for growth in the autonomous sector 
was relatively limited, and this might be related 
to the size and importance of the army.
So far, we have looked at and compared the 
sizes of individual towns. The size-distributions of 
individual towns can hint at important aspects of 
the societies studied, but they force the observer 
to view the urban system under a very narrow 
angle. Absolute size-figures are insensitive to the 
degree of centralization in the system or the de-
gree of urbanization. For example, the scarcity of 
civilian towns larger than 100 ha could have been 
compensated for by the presence of a large number 
of medium- and small-sized towns. It is therefore 
crucial to ask how dominant the top-ranking set-
tlements were in the provincial urban hierarchies.
There are a few different ways of expressing 
this relation, known as the primacy index.
19
 In 
19
  See the bibliography cited in fn. 10.
Province
Tow n
Min. size (in ha)
Max. size (in ha) 
Data source 
Thrace Byzantium 150 200
Mango 1985; Mango, Dagron (eds.) 1995;
R. Willet (pers. comm.)
Moesia Inferior Novae 100 115 Tomas 2007; Donevski 2009; Boyanov 2010
Macedonia Thessaloniki 80 110
Karambinis (pers. comm.) 
(see fn. 34)
Epirus Nikopolis 80 150 Michalis Karambinis (pers. comm.)
Dalmatia Salona 53 72 Cambi 1991
Moesia Superior Viminacium 77 100 Popović 1967, 29–49; Zotović, Jordović 1990
Pannonia Inferior Aquincum 157 187 Zsidi 2003
Pannonia Superior Carnuntum 140 185 Doneus, Gugl, Doneus 2013
Dacia Apulum 100 158 Diaconescu 2004
Tab. 1: Size-estimates for the first- ranking settlements in the provinces of the study-region.
Tab. 1: Ocena velikosti prvorazrednih naselbin v obravnavanih provincah.

199 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
this study, we have calculated the ratio of the 
first to the second-ranking settlement – in this 
paper, primacy index A – and the fraction of the 
total urban area represented by the first rank-
ing settlement – primacy index B (Tabs. 3; 4). In 
essence, both approaches are derived from the 
Rank-Size rule or Zipf’s Law.
20
 This is a case of 
size-distribution in which the size of the town 
ranked nth always equals the quotient of the size 
of the largest town in the system and rank n. In 
effect, the largest town is two times larger than 
the second largest town, three times larger than 
the third largest town and so forth. There is no 
single theoretical explanation for this regularity. 
It is most commonly observed in the urban hier-
archies of highly developed countries or, at least, 
in countries with a long urban tradition.
21
 The 
20
  Johnson 1977; Drennan, Peterson 2004; Marzano 2011.
21
  Berry 1961; Krugman 1996.
Rank-Size rule is of interest to this paper, because 
it provides us with a measure of centralization. 
To geographers, the Rank-Size rule epitomizes 
systems in which people, resources and services 
are optimally distributed in space. It is a state of 
equilibrium, in which the forces or centralization 
are counterbalanced by those of disintegration. This 
implies that the primacy index A – the ratio of 
the size-figure for the largest to that of the second 
largest settlement – in these systems must equal 
2. If this ratio is smaller than 2, the disintegrative 
forces prevail, if they are greater than 2, the oppo-
site, centripetal forces prevail. We will shortly see 
that this index cannot encapsulate the properties 
of the integral system, but it has the potential to 
shed light on aspects that are easily lost from sight 
if the focus is on the overall shape and position 
of the rank-size graphs.
Although the percentage of the total urban area 
is a variable dependent on both the size of the 
Province 
Tow n 
Min. size  (in ha)
Max. size (in ha)
Data source 
Thrace (2) Ainos 100 150 R. Willet (pers. comm.)
Thrace (3) Philippopolis 75 105 Djambov, Mateev 1980
Moesia Inferior (2) Durostorum 50 110 Donevski 2015
Moesia Inferior (3)
Troesmis/ Mar-
cianopolis
32 70 Alexandrescu, Gugl 2016; Tačeva (ed.) 2004
Macedonia (2) Amphipolis 30 60
Karambinis (pers. comm.) 
(see fn. 34)
Macedonia (3) Beroia 40 60
Karambinis (pers. comm.) 
(see fn. 34)
Epirus (2) Dyrrhachium 48 94
Gutteridge, Hoti, Hurst 2001;
Santoro, Monti 2008
Epirus (3) Apollonia 67 80 Lamboley et al. 2011
Dalmatia (2) Doclea 24 32 Wilkes 1969
Dalmatia (3) Iader 22 32 Suić 1981
Moesia Superior (2) Singidunum 60 75 Vujadin Ivanišević (pers. comm.)
Moesia Superior (3) Scupi 40 44 Mikulčić 1971
Pannonia Inferior (2) Brigetio 130 160 Borhy et al. 2004
Pannonia Superior (2) Poetovio 110 180 Horvat et al. 2003
Pannonia Superior (3) Vindobona 105 135 Kronberger, Mosser 2002
Dacia (2) Sarmizegetusa 32 75 Diaconescu 2004
Dacia (3) Romula 60 64 Diaconescu 2004
Tab. 2: Size-estimates for the second- and third-ranking settlements in the provinces of the study-region.
Tab. 2: Ocena velikosti drugo- in tretjerazrednih naselbin v obravnavanih provincah.

200 Damjan DONEV
Province Tow n Primacy index a Primacy index b Projected
Thrace Byzantium 1.3 0.18 0.24
Moesia Inferior Novae 1.04 0.13 0.22
Macedonia Thessaloniki 1.83 0.05 0.19
Epirus Nikopolis 1.59 0.22 0.26
Dalmatia Salona 2.21 0.08 0.21
Moesia Superior Viminacium 1.32 0.19 0.25
Pannonia Inferior Aquincum 1.17 0.18 0.22
Pannonia Superior Carnuntum 1.01 0.21 0.25
Dacia Apulum 2.10 0.16 0.21
Province Tow n Primacy index a Primacy index b Projected
Thrace Byzantium 1.5 0.21 0.24
Moesia Inferior Novae 2 0.24 0.22
Macedonia Thessaloniki 2 0.09 0.19
Epirus Nikopolis 1.19 0.18 0.26
Dalmatia Salona 2.20 0.10 0.21
Moesia Superior Viminacium 1.28 0.23 0.25
Pannonia Inferior Aquincum 1.20 0.28 0.22
Pannonia Superior Carnuntum 1.25 0.24 0.25
Dacia Apulum 1.66 0.17 0.21
Province Tow n
Ratio of second- to
third-ranking town
(min. estimates)
Ratio of second- to third-
ranking town
(max. estimates)
Thrace Ainon 1.33 1.42
Moesia Inferior Durostorum 1.56 1.57
Macedonia Amphipolis/Beroe 1 1
Epirus Dyrrhachium/Apollonia 1.39 1.17
Dalmatia Iader/Doclea 1.09 1
Moesia Superior Singidunum 1.5 1.70
Pannonia Inferior Brigetio 2.16 2.13
Pannonia Superior Vindobona 2.33 1.92
Dacia Sarmizegetusa 1.03 1.17
Tab. 3: Primacy indices for minimum estimates.
Tab. 3: Indeks prednosti za najmanjše ocene.
Tab. 4: Primacy indices for maximum estimates.
Tab. 4: Indeks prednosti za najvišje ocene.
Tab. 5: The ratio of the size of the second- to the third- or, in the case of Pannonia Superior, the third- to fourth-ranking 
settlement.
Tab. 5: Razmerje velikosti med drugo- in tretjerazrednimi naselbinami oziroma tretje- in četrtorazrednimi naselbinami 
Zgornje Panonije.

201 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
largest town and the total number of towns in the 
system, it too can be projected on the basis of the 
Rank-Size rule. Given that these two parameters – 
the size of the largest town and the total number 
of towns – are known, it is possible to predict the 
total urban area and the percentage of the largest 
town. For the provinces studied in this paper, the 
largest town needs to occupy between one quarter 
and one fifth of the total urban area to satisfy the 
Rank-Size rule. If in reality, this percentage is lower 
than predicted, the largest town has not reached its 
optimal size; if it is higher, it has grown too large.
On the surface, the first approach appears too 
imprecise, because it focuses only on the largest 
two settlements in the system.
22
 This is particu-
larly inconvenient for this study-area, because of 
the presence of two or more towns in the top-tier. 
Predictably, primacy index A suggests that these 
systems have a very shallow hierarchy. In all prov-
inces with multiple towns in the top-tier, the ratio 
of the largest to the second largest settlement is 
lower than 1.3. According to this index, some of 
the largest towns in the study region were the least 
dominant in their provincial urban hierarchies. 
As discussed in an earlier paragraph, this is con-
sidered a characteristic of systems that have been 
established recently or were so poorly integrated 
that they essentially functioned as two or more 
separate sub-systems. However, it suffices to look 
at the differential between the second- or third- 
and the next-ranking settlement to discover that 
this is a simple effect of the multiple top-ranking 
settlements in these provinces (Tab. 5). The ratio 
of the second to the third largest settlement or, 
in the case of Pannonia Superior, between the 
third and the fourth ranking settlement, increases 
sharply in all provinces with multiple top-ranking 
settlements. In fact, in the Pannonian provinces, it 
is greater than two – a trait of systems dominated 
by a primate city. At the same time, this ratio 
drops below 1.5 in the case of the provinces with 
a single settlement in the top-tier.
Going back to the primacy index A, it was equally 
surprising to discover that the smallest among the 
first-ranking settlements were the most dominant 
in the urban hierarchies of their provinces. The 
primacy index was the highest in the case of 
Dalmatia, a province whose first-ranking settle-
ment, Salona, was the smallest among the group 
of first-ranking settlements in this region. Still, in 
no other province was the largest settlement more 
22
  Cf. Chen 1991.
than twice the size of the second-largest settle-
ment. Technically, Salona approaches the model 
of the primate city, but the rank-size graph for this 
province is pronouncedly concave.
23
 This might 
not be entirely out of place, because other sources 
indicate that Salona was the head of a dendritic 
regional system and her main role was to channel 
the natural resources that flowed from Dalmatia 
to Italy.
24
 In this respect, Salona does match the 
profile of the typical primate city. The strong but 
asymmetric links with Italy could be one of the 
factors that prevented her from growing into a 
true primate city.
The only other province in which the primacy 
index is close to two, with both the minimum and 
maximum size estimates, is Macedonia. Similar 
primacy indices have been recorded for Moesia 
Inferior and Dacia, but not for both size-estimates. 
In some of these provinces, Macedonia and Moesia 
Inferior, the primacy index follows the Rank-Size 
rule, i.e. the largest settlement is exactly twice the 
size of the second largest town. This is an unlikely 
scenario for Moesia Inferior but, in the case of 
Macedonia, it might be related to the maturity of 
the system, predating the Roman conquest by at 
least a couple of centuries. However, as in Dal-
matia, the Rank-Size rule is broken immediately 
after the second-ranking settlement, and nearly 
all lower ranking settlements are larger than pre-
dicted by this law.
With the exception of Apulum in Dacia, a town 
that belongs to the group of the largest top-ranking 
settlements, the largest towns in Macedonia and 
Moesia Inferior do not exceed the threshold of 100 
ha. This negative correlation between settlement-
size and primacy index A is predetermined chiefly 
by the phenomenon of multiple towns in the top-
tier of the urban hierarchies. However, even if the 
hierarchies with multiple top-tier settlements are 
left out of the analysis, the smaller first-ranking 
settlements will continue to be more dominant 
in their provincial hierarchies than their larger 
counterparts. This seemingly confusing relation 
points to the fact that there was a difference of scale 
in the size-ranges of the first- and second-order 
settlements between the frontier and demilitarized 
provinces. Both the first- and the second-tier set-
tlements were proportionally smaller than their 
counterparts in the frontier provinces.
23
  Donev 2020.
24
  Alföldy 1965; Bojanovski 1974; Cambi 1991.

202 Damjan DONEV
The paradoxical situation in which the smallest 
towns are the most dominant in their provincial 
hierarchies is rectified, once we calculate the frac-
tion of the largest town in the total urban area 
of the provinces. There is a positive correlation 
between the size of towns and their percentage in 
the total urban area, albeit not particularly strong. 
For example, both Byzantium and Viminacium 
represent about the same percentage of the total 
urban area of their provinces, although Byzantium 
was twice the size of Viminacium. This is obviously 
related to the fact that primacy index B takes into 
account the number and size of the rest of the 
towns in the system.
25
 Therefore, calculating index 
B makes sense only in relation to the projections 
of the Rank-Size rule.
The last columns on the right in Tabs. 3 and 4 
estimate the percentage of the largest settlement in 
the total urban area on the basis of the Rank-Size 
rule. We were surprised by the relatively narrow 
range of variation between the individual provinces. 
The provincial systems look far less uniform when 
the individual size-figures are compared. The 
projected index fluctuates between 19 and 26% 
for the minimum size-estimates, and even less – 
between 21 and 25% – for the maximum estimates. 
It is correlated positively to the size of the largest 
settlement, and negatively to the number of towns 
in the system. Comparing these projections with 
the actual percentages of the largest towns in the 
total urban areas, it becomes evident that, in nearly 
every province and in nearly all scenarios for the 
size-estimates, the largest towns occupy a fraction 
of the total urban area smaller than predicted by 
Zipf’s Law. There are only a few exceptions. One 
is Novae and the towns of Moesia Inferior at their 
minimum size – a scenario that is not very likely 
and does not deserve a thorough comment. The 
other is Aquincum and the towns of Pannonia 
Inferior, again with the minimum estimates only. 
This is the only town which has outgrown its 
optimal size by 6%. This does not contradict the 
rank-size graph for Pannonia Inferior, although 
even Aquincum cannot be readily described as 
a primate city. The other two towns that come 
25
  Obviously, its major disadvantage is that there is 
no guarantee that the list of towns and especially town-
like settlements are complete for every province, still less 
that each province has been studied equally well. Luckily, 
the settlements that are most likely to be missing are the 
smallest, and their share in the total urban area cannot 
have been greater than few percent, cf. De Ligt 2012.
close to the projections of the Rank-Size rule are 
Carnuntum, Aquincum’s counterpart in Pannonia 
Superior and, somewhat surprisingly, Viminacium 
in Moesia Superior. Both towns fall short of the 
projected values by one or two percent.
The share of the rest of the first-ranking settle-
ments in the total urban areas of their respective 
provinces is smaller than predicted by the Rank-
Size rule. This is particularly pronounced in the 
maximum size-estimates, a scenario in which every 
first-ranking settlement in the study-region is at 
least 4 to 5% smaller than the optimal projection. 
From this perspective, none of the first-ranking 
settlements has grown to its optimal size. This 
could either imply that a large proportion of the 
wealth produced within the system was trapped 
in low-ranking, micro-regional centres or, more 
likely, that most of it escaped the confines of the 
province via the capital and the other major centres.
It should be noted that the differences between 
the projected and estimated indices were variable. 
Ironically, the greatest differential – greater than 
10% – has been recorded in Macedonia and Dalma-
tia, the provincial systems with the most dominant 
first-ranking towns according to primacy index A. 
The first-ranking settlement in these provinces 
might have been twice the size the second-ranking 
settlement, but the large number of towns of me-
dium and small size have preconditioned a primacy 
index B lower than 10%. A similar differential has 
been calculated for the towns in Moesia Inferior, 
with the maximum size-estimates. In the rest of the 
provinces, the differences between the projected 
and estimated figures for primacy index B range 
between 4 and 6%.
THE FACTOR OF FORMAL FUNCTION
Earlier it was mentioned that all towns that 
belong to the second tier of the provincial urban 
hierarchies were autonomous towns. Because they 
did not possess urban territory and a land-owning 
elite, the subordinate central places were much 
smaller than the official towns. What traits dis-
tinguished the first-ranking settlements from the 
rest of the autonomous towns? W ere they assigned 
some special role or privilege that gave them the 
advantage over the second-tier, autonomous town?
Indeed, only three out of 16 top-ranking settle-
ments lack a special role in their respective urban 
systems (Tab. 6 ). However, the reader should recall 
that the top-ranking status of Dyrrhachium and, 

203 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
especially that of Apollonia, is uncertain. Thus, 
Byzantium is the only town in the study-region 
whose status of a top-ranking settlement is un-
disputable and who did not perform any special 
political function prior to the fourth century. The 
Thracian capital was in Perinthos and the seat of 
the provincial assembly in Philippopolis. This is 
another indication that, from a regional perspective, 
Byzantium had been an important centre prior to 
the founding of Constantinople.
26
 All other first-
ranking settlements in this study-region either 
grew next to legionary bases or were the provincial 
capitals and, quite often, both.
The legionary towns were by nature composite 
agglomerations. This is related to the social compo-
sition of these communities. They consisted of the 
garrison unit – usually, a legion plus an auxiliary 
unit –, the civilian community, composed of the 
families of soldiers, craftsmen, petty merchants 
and various service-providers and, in the case 
of double towns, the autonomous community of 
Roman citizens, located at a distance of about 2 
26
  Clearly, the new capital was on a scale completely 
different from that of its predecessor, but to argue that 
the town had been fairly insignificant prior to the reign of 
Constantine I is an overstatement; cf. Mango 1985, 118–119, 
who is nonetheless right to observe that Byzantium had a 
poor water supply and unprotected hinterland.
km from the legionary camp.
27
 Each of these sub-
communities occupied a physically discrete segment 
of the conglomeration: the army was stationed in 
the fortified camps, the civilian community that 
followed the army lived in irregular agglomera-
tions that surrounded the army camps on all sides, 
whereas the landowning elite, if present, was based 
in the municipium, located just outside the military 
zone.
28
 The legionary camp, taken in isolation, would 
already rival the second-tier settlements in some 
of the demilitarized provinces. If we add to these 
the municipium – which would alone qualify as a 
second-tier settlement – and the canabae – extend-
ing over areas of at least 70–80 ha – the dominant 
position of the legionary towns in the provincial 
urban systems becomes easily comprehensible. 
Obviously, the complex topography of these towns 
is merely a symptom of their size and segmented 
social structure. Notwithstanding the negative 
connotations of living on the edge of the civilized 
world and in the shadow of an army camp, the 
permanent presence of the army did attract large 
civilian communities, which included not only the 
regular army followers but also veteran soldiers 
and their families. The permanent garrisons would 
have been potential consumers and, although their 
27
  Piso 1991, 137–141.
28
  Doneus, Gugl, Doneus 2013, 179–182.
Province Tow n Administrative role
Thrace Byzantium /
Moesia Inferior Novae Legionary base/provincial capital?
Moesia Inferior Durostorum Legionary base
Macedonia Thessaloniki Provincial capital
Epirus Nikopolis Provincial capital
Epirus Dyrrhachium? /
Epirus Apollonia? /
Dalmatia Salona Provincial capital
Moesia Superior Viminacium Legionary base/Provincial capital
Moesia Superior Singidunum? Legionary base
Pannonia Inferior Aquincum Legionary base/Provincial capital
Pannonia Inferior Brigetio Legionary base
Pannonia Superior Carnuntum Legionary base/Provincial capital
Pannonia Superior Poetovio Procuratorial seat
Pannonia Superior Vindobona Legionary base
Dacia Apulum Legionary base/Provincial capital
Tab. 6: A list of the top-ranking settlements in the study-area and their role in the administration of the provinces.
Tab. 6: Seznam vodilnih mest na obravnavanem območju in njihova vloga pri upravljanju provinc.

204 Damjan DONEV
impact on the provincial economies is a matter 
of debate, this must have been one of the factors 
that propelled their growth.
29
 In addition, because 
of their strategic importance, the legionary towns 
received generous acts of munificence from the 
emperors on a regular basis.
30
The differences in size between the individual 
legionary towns is determined chiefly by the pres-
ence or absence of autonomous towns near the 
legionary bases. These were absent in the Moesian 
provinces and, as a result, the legionary towns in 
Moesia were smaller than their Pannonian coun-
terparts or the Dacian capital.
31
 A comparison 
between the size-figures for the double towns 
in the Pannonian provinces will reveal that the 
provincial capitals, Carnuntum and Aquincum are 
approximately 20% larger than the second legion-
ary towns of Vindobona and Brigetio. There is a 
similar differential between the size of the capital 
of Moesia Superior, Viminacium and the second 
legionary town of Singidunum, although neither 
was a double town. Thus, despite the uniformity 
of the legionary towns – the combined areas of 
the legionary camp and the canabae are similar in 
most frontier provinces – the additional function of 
provincial capitals always resulted in a greater size.
The effect of the role of provincial capital on 
the size of the settlements is most plainly evident 
in the demilitarized provinces. With the notable 
exception of Thrace, the first-ranking settlements 
were always the provincial capitals. The absence 
of other potential factors of growth – like the 
permanent garrisons in the frontier provinces 
– contributed to the much greater differential 
between the first- and second-ranking settlements 
in these provinces and a higher primacy index 
A. It is a sensible objection that the presence of 
the provincial governor and his relatively small 
staff could not have had the same effect on the 
urban economy as the large garrisons, but the 
indirect effects of their presence should not be 
underestimated. T o live in a close proximity to the 
provincial governor must have been advantageous 
in the eyes of the provincial elite, and had they 
decided to move into the capital, it would have 
inevitably drawn other urban-minded segments 
29
  Erdkamp (ed.) 2002.
30
  Lengyel, Radan (eds.) 1980.
31
  Claims have been made that both Novae and Du-
rostorum were double towns, but the evidence adduced in 
support of this thesis is not very convincing, Tomas 2007; 
Donevski 2009; Boyanov 2010.
of the provincial population.
32
 This would have 
automatically extended the economic base of the 
capitals. The case of Poetovio demonstrates that 
even the presence of the financial administration 
can be a powerful attractor. Because of their status, 
the provincial capitals were probably the exclusive 
locations for a certain number of services.
33
 Finally, 
like the legionary towns, the provincial capitals 
were more likely to draw attention and support 
from the central government than the ordinary 
provincial town.
Admittedly, we are still far from understanding 
the specific reasons that led to the clustering of 
people and wealth near the bases of the provincial 
government. It can be argued that most of the 
provincial capitals were already large prior to their 
promotion. Indeed, the case of the Thracian capital 
Perinthos – which is not among the three largest 
towns in this province – warns against assigning 
too much weight to the factor of formal status. 
However, the history of most of the capitals in 
the demilitarized provinces speaks to the contrary. 
Thessaloniki experienced a period of growth as a 
provincial capital, surpassing the former capital 
of the Macedonian kingdom, Pella.
34
 Prior to the 
Roman conquest, Salona had been an insignificant 
emporium of the Greek colonies in the Adriatic, 
whereas Nikopolis was a newly founded settlement. 
Apart from a few exceptional cases, the role of 
provincial capital guaranteed a dominant position 
in the urban system.
In conclusion, the top-tier of the urban hier-
archy in this entire region was to a great extent 
determined by political and strategic factors. Once 
chosen to become a legionary base or a seat of 
the provincial governor, the town’s place in the 
top-tier of the hierarchy was not only secured, 
but it was also unlikely to be challenged. The 
gap that separated the first- from the second-tier 
settlements was even more insuperable than that 
separating the autonomous towns and the subor-
dinate central places.
32
  Wilkes 2002; Ferjanić 2002.
33
  This was certainly the case for legal services, Eck 
2000, although this issue remains unexplored.
34
  Karambinis (forthcoming), Urban Networks in Early 
Roman Macedonia and Aegean Thrace.

205 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
GEOGRAPHICAL POSITION,
GATEWAY COMMUNITIES?
If the first-ranking settlements in this study-
region were largely chosen by the will and whim 
of the conqueror, was there anything in their lo-
cation that preordained their place in the urban 
hierarchy? What are the patterns of distribution of 
the top-ranking settlements in this study-region? 
Their location in relation to the provincial ter-
ritories and main axes of communication are of 
particular interest.
It is immediately apparent that the great majority 
of the top-ranking settlements are located on the 
edge of the study-region (Map 1). Only two out 
of 16 towns have a central position in relation to 
the provincial territories. These are Apulum and 
Poetovio, a non-capital. The rest of the top-ranking 
settlements are either located on the coast or on 
the Danube frontier. From the point of view of 
governing the provincial territories and, especially, 
providing services to all corners of the provinces, 
these are not the optimal locations. The peripheral 
locations of the top-ranking settlements reflect a 
very different logic. Although geometrically pe-
ripheral, the sites of all top-ranking towns were 
well-connected to the Empire-wide network of 
roads and cities. The chief concern of the central 
and provincial government would have been to 
ensure an uninterrupted flow of taxes and resources 
from the provincial capitals to Rome and, in the 
frontier provinces, the interconnectivity of the 
military bases on the frontier and their continuous 
supply with food and other necessities. It has to 
be remembered that the Danube was not only a 
frontier, but an important interregional corridor.
35
 
Seen from this angle, the locations of the legionary 
towns on the frontier were analogous to those of 
the large port-towns in the demilitarized provinces. 
They all occupied locations that were potential 
mediators between the region they governed and 
the outside world.
On the surface, our top-ranking settlements are 
reminiscent of the profile of the gateway towns, 
observed by regional geographers in an Early Mod-
ern colonial context.
36
 Briefly, the gateway town 
is a settlement that appears in the contact-zone 
between two areas at different levels of economic 
development. In our study-region, this is paralleled 
both by the port-towns, which at the time of the 
35
  Whittaker 1994.
36
  Burghardt 1971.
conquest stood on the edge of an area that was still 
largely uncivilized, and the legionary towns, the 
bases for future conquest and mediators between 
the Empire and the barbaricum. In contrast to the 
traditional central place, the gateway is, like the 
great majority of our top-ranking towns, located 
on the edge of its territory. Empirical observations 
have demonstrated that the gateway towns were 
always larger than the typical central place. Another 
striking similarity between the gateway model and 
our first-order settlements is the structure of the 
urban population. Gateway towns have been char-
acterized as places of trade, commerce, building 
industry and transportation. They are often set 
in opposition to the older, more cultured cities. 
Admittedly, census-data for our study-period is 
lacking and we have to think in terms different 
than building industry or wholesale. Nevertheless, 
it is surely likelier to discover merchants, traders 
or concessioners in the epigraphic record of the 
port-towns and canabae legionis than in those of 
the average provincial town. Most striking of all, 
is the similarity in topography. Like some of the 
legionary towns, the gateway communities live in 
segmented conurbations. In fact, some have even 
been described as twin towns or double towns, a 
duality that has been linked to a functional segre-
gation between administration and trade.
Although this parallel is intriguing and worth 
exploring, the differences between these two dis-
tant historical contexts should not be lost from 
sight. According to this model, gateway towns 
are transitory phenomena, typical in periods of 
territorial expansion, whether conquest or colo-
nization. As the border is pushed farther out, the 
gateway town loses its function and is transformed 
into a traditional central place. We could think of 
Poetovio and, especially Sarmizegetusa, – even if 
it was not a first-ranking settlement – as former 
gateway communities.
37
 However, the majority of our 
port-towns continued to function as links between 
Rome and her provinces, as long as Rome was the 
centre of the Empire. They had a permanent role 
in the Empire-wide urban network that might have 
been instrumental to their size and prestige. The 
37
  Poetovio was not only a former legionary base, but 
also had a composite topography, with several districts 
strung along the Amber Road; Horvat et al. 2003. On the 
other hand, the fragmentation of the territory of the Dacian 
colony – Piso 1995 – is an exact replica of the territorial 
redefinition of the gateway community after its loss of 
status and transformation into a traditional central place, 
Burghardt 1971.

206 Damjan DONEV
legionary towns, on the other hand, are in many 
ways specific and difficult to fit into any model. 
Because of the Limes, they could not service the 
outward hinterland like the model gateway town. 
They could certainly channel the external trade 
of the Empire, but too little is known about the 
intensity of this relationship to see it as a potential 
factor of growth.
38
 As centres of administration, 
these towns looked inwards and were more akin 
to a former than to an actual gateway. The Roman 
military outposts at the time of the Late Republic 
would have resembled the gateway communities 
more closely than the legionary bases founded 
after the introduction of the static Limes.
The patterns of distribution of the first-ranking 
settlements also reflect the variable importance of 
the different segments of the Balkan coast and the 
Danube Limes. It is striking that none of the top-
ranking settlements was located on the western 
38
  Wielowiejski 1990; Gabler, Vaday 1992; Krekovič 
2001; Kuzmová 2005; Galestin 2005.
Black Sea coast. In contrast, two, possibly four, 
top-ranking settlements were located on the East-
ern Adriatic and Ionian coasts. The key difference 
between the two littorals is obvious. Whereas the 
Adriatic and Ionian coasts faced the centre of the 
Empire, the Black Sea was a peripheral region, 
only partly integrated into the economy of the 
Roman Empire. At the time of the High Empire, 
Byzantium would have been far more important as 
a station on the continental roads that linked Asia 
to Europe than as a maritime base in the traffic 
between the Black Sea and the Mediterranean. The 
variable strategic importance of the different seg-
ments of the Danube Limes could also explain the 
differential growth among the legionary towns. Is 
it by chance that double towns have only emerged 
in the Pannonian and not in the Moesian section of 
the Limes? Pannonia was not only much closer to 
Italy and Rome than Moesia, but it was crossed by 
two roads of a major interregional importance: the 
Amber Road and the road that led from Aquincum 
across the Carpathian Range. On the other hand, 
Province 
Tow n
Urban population
(min. area × 150)
Arable land  
(in ha) (15 km)
Land deficit (in ha)
Land deficit 
incl. rural population
(in ha)(75%)
Land deficit
incl. rural population
(in ha)(50%)
Thrace Byzantium 22 500 23 150 + 650 - 66 850 (3.9) - 21 850 (1.94)
Moesia Inferior Novae 15 000 29 950 + 14 950 - 30 050 (2) - 50 (1)
Moesia Inferior Durostorum 7 500 27 950 + 20 450 - 2 050 (1.07) + 12 450
Macedonia Thessaloniki 12 000 34 800 + 22 800 - 25 200 (1.72) + 10 800
Epirus Nikopolis 12 000 8 000 - 4 000 - 52 000 (7.5) - 16 000 (3)
Epirus Dyrrhachium? 7 200 21 500 + 14 300 - 7 300 (1.34) + 7 100
Epirus Apollonia? 10 050 31 500 + 21 450 - 8 700 (1.2) + 11 400
Dalmatia Salona 7 950 8 300 + 350 - 23 500 (3.83) - 7 600 (1.91)
Moesia Superior Viminacium 11 550 20 250 + 8 700 - 25 950 (2.28) - 2 850 (1.14)
Moesia Superior Singidunum? 9 000 17 650 + 8 650 - 18 350 (2.04) - 350 (1.01)
Pannonia Inferior Aquincum 23 550 15 600 - 7 950 - 78 600 (6.04) - 31 500 (3.01)
Pannonia Inferior Brigetio 21 000 31 450 + 10 450 - 52 550 (2.67) - 10 550 (1.33)
Pannonia Superior Carnuntum 21 000 32 900 + 11 900 - 51 100 (2.55) -9 100 (1.27)
Pannonia Superior Poetovio 16 800 41 350 + 24 550 - 25 850 (1.62) + 7 750
Pannonia Superior Vindobona 12 000 29 700 + 17 700 -18 300 (1.61) + 5 700
Dacia Apulum 19 500 34 950 + 15 450 - 43 050 (2.23) - 4 050 (1.11)
Tab. 7: Population to land ratio for the top-ranking settlements in the study-region.
Tab. 7: Razmerje med prebivalstvom in zemljiščem prvorazrednih mest na obravnavanem območju.

207 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
the Moesian section of the Limes faced a limited 
segment of the barbaricum, essentially a pocket 
of enemy territory trapped between the Empire 
and Dacia. The tribes that lived on the other side 
of the Pannonian frontier, the Marcomanni and 
Quadi, were sedentary societies that had developed 
a certain level of economic relationship with the 
Roman Empire and, according to some historians, 
were even considered as a possible addition to the 
Empire.
39
 In contrast, the Moesian sector of the 
Limes bordered on areas dominated by nomadic 
people, like the Iazyges or the Sarmatians and, 
until the Gothic invasions of the mid-3
rd
 century, 
they were of a lesser strategic importance than the 
Pannonian Limes.
The importance of the global urban constella-
tion for the regional urban developments is best 
illustrated by the changes in the urban geography 
after the establishment of the new capital on the 
Bosphorus. Whereas the Dalmatian coast and the 
western Balkans in general entered a long period of 
urban decline and restructuring, the eastern part 
of the peninsula, especially Thrace and the Black 
Sea region experienced a period of prosperity and 
urban growth that, despite the chronic instability 
caused by barbarian invasions, lasted until the end 
of the 6
th
 century.
WERE THE FIRST-RANKING 
SETTLEMENTS TOO LARGE FOR THEIR 
TERRITORIES?
Discussing the primacy index and the degree 
of centralization in the urban systems, it was ob-
served that individual size-figures reveal little if 
seen in isolation. Towns that were relatively small 
proved to be the most dominant in their respec-
tive urban systems. Equally important is the ratio 
of the size of the town, expressed as a projected 
population figure, to the agricultural productivity 
of its hinterland. This will give us an idea of the 
self-sustainability of the top-ranking settlements 
and it will filter through the factor of agricultural 
productivity as a determinant of the size of towns.
The population estimates are derived by mul-
tiplying the size-estimate of the built-up area by 
150. In view of the assumption that larger settle-
ments have higher population density, these are 
fairly conservative estimates.
40
 In order to test the 
39
  Birley 2000.
40
  Marzano 2011, 196–228; Hanson 2016.
sustainability of these towns in optimal conditions, 
we have used only the minimum estimates. For 
similar reasons, the outreach radius of the towns 
has been extended to 15 km. This is equivalent 
to the so-called market catchment, enclosing the 
area from which it would have been viable to bring 
agricultural goods on the town-market and return 
in one day.
41
 The amount of arable land within this 
catchment was calculated with the help of Google 
Earth on the basis of the modern plough-zone. 
Another simplification was to set the annual grain 
consumption rate of a grown-up individual at 1 
ha.
42
 In reality, the total consumption rate must 
have been somewhat lower, because the children and 
elderly probably consumed less, and calories would 
not have been obtained from grain only. However, 
our goal is not to offer a precise reconstruction 
of the agricultural production and consumption 
in these towns, but to make a rough assessment 
of their sustainability. Could these towns secure 
their subsistence from the resources available 
in their hinterlands or were they dependent on 
a much larger hinterland or the support of the 
central government? In order to widen the error-
margin and account for the possible mistakes in 
the calculation process, it is advisable to look 
only for cases in which the deficit is far beyond 
the agricultural capacity of the urban catchment.
A simple comparison between the estimated urban 
population and the amount of arable land in the 
market catchments shows that most towns could 
comfortably live off the agrarian potential of their 
hinterlands (Tab. 7: Column 5). This result does 
not contradict the analysis of the primacy indices. 
The great majority of the first-order settlements 
were neither too dominant in the urban hierarchy 
nor were they too large for their hinterlands. The 
only exceptions are Aquincum and Nicopolis. These 
two towns have little in common, except for the 
fact that their agricultural territories are extremely 
limited because of their locations. Nikopolis is 
located on an isthmus, whereas Aquincum is too 
close to the northeast corner of Pannonia Inferior. 
The theoretical surplus was also dangerously low 
in the case of the port-towns of Salona and Byz-
antium. Despite the big size-differential between 
these two towns, they were equally oversized in 
41
  For a detailed discussion of the concept, see Bintliff 
2002.
42
  Foxhall, Forbes 1982. Of course, this is dependent on 
the estimated productivity of ancient agriculture, Garnsey 
1998; Erdkamp 2005.

208 Damjan DONEV
relation to their hinterland. The market catch-
ments of the rest of the top-ranking settlements 
could easily sustain a settlement twice their size.
The problem with these calculations is that 
they do not take into account the factor of the 
rural population, who would have also been de-
pendent on the resources available in the urban 
hinterlands. The figures (Tab. 7: column 5) would 
have been roughly accurate were the majority of 
the urban dwellers farmers themselves. However, 
this could not have been the case for the provin-
cial capitals and legionary towns.
43
 Therefore, 
realistic projections for the grain demand in the 
market catchments of the towns cannot be made, 
unless the rural population is accounted for in the 
equation. Unfortunately, there is very little data 
about the possible size of the rural population in 
the hinterland of the Balkan towns. Nonetheless, 
making a few educated guesses might open an 
important insight into these relations.
Most scholars who have studied the economy of 
the Roman Empire through the lenses of urbanization 
argue for urbanization rates of slightly over 10% 
in most European provinces, with the exceptions 
of Italy and Baetica.
44
 This could very well apply 
to the frontier provinces in this study-region, but 
in the more urbanized demilitarized provinces in 
the south of the peninsula, it is sensible to expect 
higher urbanization rates, at least in the region of 
20 and probably closer to 30%. To avoid making 
separate calculations, the population to land ratio 
has been estimated for hypothetical urbanization 
rates of 25 and 50%. The first figure is a notional 
average for the entire study-region, the second, a 
notional maximum urbanization rate. Theoretically, 
even higher urbanization rates are possible, but only 
at the expense of agricultural productivity. In this 
scenario, the advantages of sustaining a smaller 
rural population would have been cancelled by a 
lower agricultural output.
Once the rural population enters the equation, 
none of our towns is self-sustainable, even at a 
relatively high urbanization rate of 25%. The deficit 
ranges from seven to eight square km of arable land 
in the case of the smaller top-ranking settlements, 
to over 50 sq. km, in the case of the double towns 
43
  Cf. the occupation profile of the population of 
Herculaneum, De Ligt, Garnsey 2012.
44
  In this, and in the majority of other studies of ancient 
urbanization, the rate of urbanization is expressed as the 
fraction of the projected total population. An exhaustive 
bibliography can be found in Hanson 2016, fn. 657.
or some of the port towns. Obviously, if we lower 
the urbanization rate, the deficit will rise further. 
Still, in absolute terms, these are not particularly 
high deficits. Afterall, these were autonomous towns 
whose administrative territory was often much 
greater than the market catchment. However, it 
should not be forgotten that the addition of extra 
agricultural land would have amounted to nothing 
without a rural population to cultivate it.
If the urbanization rate is increased to 50%, 
the sustainability problem disappears for over 
half of the settlements, and for the other half, 
the land-deficit is reduced to less than 10 sq. km. 
The only exceptions are Aquincum, Byzantium 
and Nikopolis, towns whose micro-locations are 
extremely marginal in relation to the productive 
land in their respective provinces. Obviously, an 
urbanization rate of 50% is unrealistic for any 
pre-modern society, but only in so far as this rate 
is pertinent to the entire territory under study. It 
is equally unrealistic to expect that the density of 
rural population was equal in any given segment 
of the countryside. Admittedly, it is a sensible 
assumption that the urban market was a strong 
attractor, both for the urban and rural segments 
of the population. However, the fact that most 
of this land was owned by the urban elite would 
have posed a limit to the concentration of rural 
dwellers in the urban hinterlands. A look at the 
settlement patterns in the agricultural and mar-
ket catchments of the Roman towns in this area 
will reveal a countryside dominated by villae or 
farmsteads.
45
 With the exception of Thrace and, 
possibly Macedonia, nucleated settlements are 
rare.
46
 But even if the survey data is not deemed 
very reliable, a rise in the rural population in the 
urban hinterland is unlikely to offset the very high 
population densities documented in the urban 
settlements. Hence, the proportion of the rural 
population in the total population of the town 
and its market catchment must have been lower 
than implied in the general urbanization rates.
47
45
  Carnuntum: Ployer 2009; Apulum: Oltean 2007; 
Aquincum: Zsidi 2003; Apollonia: Davis et al. 1998–2002; 
in this case, the pattern dates to the Hellenistic period.
46
  The evidence is almost exclusively epigraphic: Gerov 
1980; Papazoglou et al. 1999; Martemianov 2012.
47
  Even if half of the rural population in an urban 
territory with an urbanization rate of 10% lived in the 
market catchment, the urbanization rate in this catchment 
would have still been almost 10% higher than the general 
urbanization rate. For example: 5000 (urban pop.) / [5000 
(urban pop.) + 22500 (half of the rural population)]= 18.1%. 

209 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
High local urbanization rates were not the only 
possible solution to the land deficit of the top-
ranking settlements. In the case of the legionary 
towns, the sustenance of a large segment of the 
population – the army units – would have been 
guaranteed by the provincial governor. In fact, 
grain for the legionary towns could have been 
requisitioned from those parts of the province that 
lacked a large urban population. On the other hand, 
the big maritime settlements, like Byzantium or 
Nikopolis would have had access to the agricultural 
produce of an area much wider than their market 
catchments. Finally, the epigraphic record from 
the countryside reveals that the economic bases 
of the first-ranking settlements were much larger 
than their urban territories, because members of 
their elite could own land anywhere in their native 
province or beyond.
48
As expected, the inclusion of the rural popula-
tion in the calculations of the land to population 
ratio has pushed nearly all of the first-order set-
tlements above the threshold of self-sustainability. 
However, the land-deficit was not particularly 
high, even if the urbanization rate is set at 25%. 
This is in accord with our earlier observation that 
none of these towns was so large as to present a 
serious logistical challenge for the local or pro-
vincial authorities. There were no towns on the 
scale of Alexandria, Antioch or Ephesus in this 
study-region. All of our top-ranking settlements 
could secure their grain-supply either from within 
their urban hinterlands or from the provinces to 
which they belonged.
CONCLUSION
Seen from an Empire-wide perspective, the first-
ranking settlements in this study-region cannot 
be qualified as metropoleis. In fact, almost none 
would qualify as a first-ranking settlement on an 
Empire-wide level. This is understandable in view 
of the short urban tradition in most corners of the 
study-area. With the exception of the coastal zone 
colonized by the Greeks and the areas controlled 
by the Epirote and Macedonian monarchies, 
Classical urbanism was introduced in this region 
Mathematically, this phenomenon is encapsulated by the 
function of local density: as the area of interest decreases, 
local density increases, see Conolly, Lake 2006, for a brief 
introduction and references. 
48
  Donev 2020.
shortly after the Roman conquest. Proper roads 
were absent prior to the conquest, and population 
density would have been too low to support large 
urban populations in many parts of the area.
49
 
Although important roads passed through the 
study-region and its legions played a key role in 
the defence of Italy, from an economical point of 
view, the Balkan and Danube provinces occupied 
a peripheral corner of the Roman Empire. The 
most important assets of this region would have 
been the army recruits and, possibly, its ore and 
timber resources. This would not have been the 
right climate for the emergence of very large cities.
An interesting particularity of the urban systems 
in the frontier provinces of the study-region was 
the presence of two or more towns in the top-
tier of the urban hierarchy. The size-differential 
between these towns was too small to place them 
into separate size-ranges. Although slightly unu-
sual at a first sight, this arrangement is highly 
revealing of the nature of the urban systems in the 
study-region. The urbanization of the area and its 
later development were tightly controlled by the 
provincial government. It is true that the urban 
systems in the frontier provinces had existed for 
only about a century to century and a half at the 
time of the Severan dynasty. The lack of greater 
time-depth could have been an important factor 
for the survival of these arrangements, but it is 
questionable if the legionary towns would have 
grown at a different pace had the Limes survived the 
third century crisis. The upkeep of both legionary 
towns, despite of their size and proximity, was a 
purely strategic decision, unaffected by spontane-
ous, bottom-up tendencies.
The calculation of the primacy indices showed 
that none of the first-order settlements exhibited the 
traits of a primate city. Intriguingly, the towns that 
resembled this model the most were the smallest 
among the first-ranking settlements in the study-
region. However, it suffices to consider the integral 
urban system to conclude that this resemblance 
is superficial. From the vantage point of the larg-
est towns, the urban systems in the study-region 
were fairly decentralized. This can be seen as an 
outcome of two complementary tendencies. On the 
one hand, the poor level of connectivity between 
the second-tier settlements limited the scope for 
differential growth in the system. The differences 
in size between these towns are minor, reflect-
49
  Moesia Inferior: Gerov 1997, 121–209; Duch 2017; 
Dalmatia: Dzino 2010.

210 Damjan DONEV
ing primarily the variable distribution of natural 
resources or distinct historical developments. On 
the other, the excessive growth of at least some 
of the first ranking-settlements would have been 
prevented by their role in the political economy 
of the High Empire. Only a smaller proportion of 
the wealth produced in their respective provinces 
would have been kept in these towns. The bulk 
would have been sent to Italy and Rome.
50
 In the 
frontier provinces, the emergence of a primate 
city was impossible due to the presence of two or 
more towns in the top-tier of the urban hierar-
chies. These urban systems were tailored primarily 
to the needs of the military and their evolution 
would have been to a large extent controlled by 
the provincial authorities.
The absence of a perceptible spontaneous dy-
namic in these systems is also underlined by the 
fact that nearly all of the top-ranking settlements 
had been assigned the role of provincial capitals, 
legionary bases or both. The second-order towns 
could never aspire to compete with the provincial 
capitals or legionary towns, still less to take over 
their place in the urban hierarchy. This feature of 
the provincial urban systems is not unique to the 
Roman Empire. The modern national capitals, at 
least those in the Old World, are also the largest 
towns in their respective countries. In fact, it 
is possible that this scheme was introduced for 
the first time in this region by the Romans. The 
stability of the provincial administrative system 
resulted in a clear articulation of top-tier settle-
ments in the urban hierarchies of the provinces. 
The political centres of earlier and later tribal 
alliances and monarchies were more transient 
than the Roman provincial capital. Cogently, the 
urban hierarchies in these societies would have 
been relatively shallow.
One of the major differences between the first-
ranking settlements in the Roman provinces and 
those in the modern nation-states is that the former 
had been conceived of primarily as nodes that linked 
the provinces to the centre of the Empire rather 
than as independent administrative centres. This 
is most apparent from the peripheral locations of 
nearly all top-ranking settlements included in this 
analysis. They were located either on the western 
and southern Balkan coast or on the Danube. This 
special role in the Empire-wide urban network 
brought them close to the colonial capital in the 
Early Modern era – hence, possibly the hints of 
50
  Cf. Marzano 2011, 196–228.
primacy among some of these towns – or the 
gateway communities observed in the New World 
during the colonial period. The specifics of the 
region and time-period studied do not allow us 
to identify the top-ranking settlements with either 
of these models. Nevertheless, these parallels help 
us understand certain aspects of the top-ranking 
settlements that are otherwise indistinguishable. 
On a global level, their main task would have 
been to convey the wealth and resources extracted 
from the provinces to Rome. This role must have 
given them the decisive advantage over the other 
provincial towns but, at the same time, it must 
have acted as a growth-limiting factor.
In view of their modest size, the sustenance of the 
top-ranking settlements in the study-region could 
not have presented a serious logistical challenge. 
Their population was only a small fraction of the 
population of towns like Alexandria, Antioch or 
Ephesus. The potential grain supply problems that 
were identified in this study were caused primarily 
by their marginal locations. The top-ranking settle-
ments had a number of possible solutions at their 
disposal, even if it is granted that the urbanization 
rates in the urban hinterlands were lower than 
25%. These included direct intervention by the 
provincial government, access to agricultural goods 
produced outside the province and landownership 
outside the urban territory, a mechanism that 
might have had a negative impact on the growth 
of lower-ranking towns or the emergence of towns 
in densely populated rural areas. In theory, the 
bare existence of oversized towns is a potential 
factor of differential growth.
This panoramic overview of the top-ranking set-
tlements in the study-region could not do justice to 
the specific developments of the individual towns 
and it probably blurred a number of important 
differences. We may only hope that the insights 
opened from this perspective outweigh the likely 
simplification of certain developments that are 
pertinent to individual or groups of towns. The 
argument that each of these towns had its unique 
development is certainly valid, but it is not very 
helpful and should not prevent us from drawing 
analogies or making generalizations. In the end, 
there are numerous monographs that offer detailed 
studies of individual towns. What is missing is an 
analysis of the relationships between the towns 
in the region and their role in the Empire-wide 
urban network.

211 The top-ranking towns in the Balkan and Pannonian provinces of the Roman Empire
ABRAMS, P ., E. A. WRIGLEY (eds.) 1978, Towns in soci-
eties. Essays in economic history and historical sociology. 
– Cambridge.
ADAMS, R. McC. 1965, Land behind Baghdad: a history 
of settlement on the Diyala Plains. – Chicago.
ALEXANDRESCU, C-G., C. GUGL 2016, The Troesmis-
-project 2011–2015 – research questions and metho-
dologies. – In: C-G. Alexandrescu (ed.), Troesmis – a 
changing landscape. Romans and the others in the Lower 
Danube region in the first century BC - third century 
AD. Proceedings of the International Colloqium Tulcea, 
7th–10th October 2015, 9–23, Cluj-Napoca.
ALFÖLDY , G. 1965, Bevölkerung und Gesellschaft der 
römischen Provinz Dalmatien. – Budapest.
BERRY , B. J. L. 1961, City-size distributions and economic 
development. – Economic Development and Cultural 
Change 9.4/1, 573–588.
BINTLIFF, J. 2002, Going to market in Antiquity. – In: 
E. Olshausen, H. Sonnabend (eds.), Zu Wasser und 
zu Land. Verkehrswege in der antiken Welt, Stuttgarter 
Kolloquium zur historischen Geographie des Altertums 
7 (1999), 209–250, Stuttgart.
BIRLEY , A. 2000, Marcus Aurelius. A biography. – London.
BOJANOVSKI, I. 1974, Dolabelin sistem cesta u rimskoj 
provinciji Dalmaciji (Dolabellae systema viarum in pro-
vincia romana Dalmatia). – Djela ANUBiH 47, Centar 
za Balkanološka ispitivanja 2, Sarajevo.
B OYANOV, I. 2010, Municipium Aurelium Durostorum 
or vicus Gavidina. – Archaeologia Bulgarica 4/2, 53–59.
BORHY , L. et al. 2004, Brigetio, Ergebnisse der 1992–1998 
durchgeführten Ausgrabungen (Munizipium, Legions-
lager, Canabae, Gräberfelder). – In: M. Šašel-Kos, P. 
Scherrer (eds.), The Autonomous Towns of Noricum 
and Pannonia / Die autonomen Städte in Noricum und 
Pannonien. Pannonia II, Situla 42, 231–251.
BOWMAN, A., A. WILSON (eds.) 2011, Settlement, 
urbanization and population. – Oxford Studies on the 
Roman Economy 3.
BURGHARDT, A. F. 1971, A hypothesis about gateway 
cities. – Annals of the Association of American Geo-
graphers 61/2, 269–285.
CAMBI, N. (ed.) 1991, Antička Salona. – Split.
CHASE-DUNN, C. S. MANNING 2002, City-systems 
and world-systems: four millennia of city growth and 
decline. – Cross-cultural Research 36/4, 379–398.
CHASE-DUNN, C., C. WILLARD, A. WILLARD 1994, 
Cities in the central political/military network since CE 
1200: size-hierarchy and domination. – Comparative 
Civilizations Review 30, 104–132.
CHEN, X. 1991. China’s city hierarchy, urban policy and 
spatial development in the 1980s. – Urban Studies 
28/3, 341–367.
CORNELL, T., K. LOMAS (eds.) 1995, Urban society in 
Roman Italy. – London.
DAVIS, J. L. et al. 1998–2002, The Mallakastra Regional 
Archaeological Project. [http://river.blg.uc.edu/mrap/].
DE LIGT, L. 2012, Peasants, citizens and soldiers: Studies 
in the demographic history of Roman Italy 225 BC– AD 
100. – Cambridge.
DE LIGT , L., P . GARNSEY 2012, The album of Herculane-
um and a model for the town’s demography. – Journal 
of Roman Archaeology 25, 69–94.
DIACONESCU, A. 2004. The towns of Roman Dacia: an 
overview of recent archaeological research. – In: W.S. 
Hanson, I. P . Haynes (eds.), Roman Dacia. The Making 
of a Provincial Society, 87–142, Portsmouth.
DJAMBOV , C., M. MATEEV , 1980. Donées nouvelles sur 
l’ ensemble du forum de Philippopolis. – In A. Fol (ed.), 
Pulpudeva: Semaines philippopolitaines de l’histoire et 
de la culture Thrace 3, 106–121, Plovdiv.
DONEUS, M., C. GUGL, N. DONEUS 2013. Die Cana-
bae von Carnuntum-Ein Modelstudie der Erforschung 
römische Lagervorstädte. – Wien.
DONEV, D. 2020, The busy periphery: the urban systems 
of the Balkan and Danube provinces (2
nd
–3
rd
 c. AD). 
– Oxford.
DONEVSKI, P . 2009, Archaeological excavations in Silistra 
(Durostorum). – Bulitinul Muzeului Judeţean T eleorman 
1, 105–130.
DONEVSKI, P. 2015, A comparison between Novae and 
Durostorum in Lower Moesia: topography, defensive 
system and legal status. – In: L. V agalinski, N. Sharankov 
(eds.), LIMES XXII: Proceedings of the 22nd International 
Conference of Roman Frontier Studies, Ruse, Bulgaria, 
September 2012, 163–168, Sofia.
DRENNAN, R. D., C. E. PETERSON 2004, Comparing 
archaeological settlement systems with rank-size graph: 
a measure of shape and statistical confidence. – Journal 
of Archaeological Science 31, 533–549.
DUCH, M. 2017, Economic role of the Roman Army in the 
province of Lower Moesia (Moesia Inferior). – Gniezno.
DZINO, D. 2010, Illyricum in Roman Politics. – Cambridge.
EARLE, R., R. W. PREUCEL 1987, Processual Archaeo-
logy and the radical critique. – Current Anthropology 
28/4, 501–538.
ECK, W. 2000, Provincial administration and finance. – 
In: A. Bowman, P. Garnsey, D. Rathbone (eds.), The 
Cambridge Ancient History XI. The High Empire A.D. 
70–192, 266–292, Cambridge.
ERDKAMP , P. (ed.) 2002, The Roman Army and the Eco-
nomy. – Amsterdam.
ERDKAMP , P . 2005, The Grain Market in the Roman Em-
pire: a social, political and economic study. – Cambridge.
FERJANIĆ, S. 2002, Naseljavanje legijskih veterana u bal-
kanskim provincijama. I–III n.e. – Posebna izdanja 79. 
Balkanološki institut SANU, Belgrade.
FINLEY , M. 1977, The ancient city: from Fustel de Coulan-
ges to Max Weber and beyond. – Comparative Studies 
in Society and History 19/3, 305–327.
FOXHALL, L., H. A. FORBES 1982, Sitometria: the role of 
grain as a staple food in Classical Antiquity. – Chiron 
12, 41–90.
GABLER, D., A. H. VADAY 1992, Terra Sigillata im 
Barbaricum zwischen Pannonien und Dazien. – Acta 
Archaeologica Academiae Scientiarum Hungaricae 44, 
83–160.
GALESTIN, M. 2005, Barriers for Barbarians. – In: Z. Visy 
(ed.), LIMES XIX, Proceedings of the XIXth Internati-
onal Congress on Roman Frontier Studies held in Pécs, 
Hungary, September 2003, 221–226, Pécs.
GARNSEY , P. 1998, Grain for Athens. – In: W. Scheidel 
(ed.), Cities, Peasants and Food in Classical Antiquity: 
essays in social and economic history, 183–200, Cambridge.
GEROV , B. 1980, Beiträge zur Geschichte der römischen 
Provinzen Moesien und Thrakien I. – Amsterdam.

212 Damjan DONEV
GEROV , B. 1997, Beiträge zur Geschichte der römischen 
Provinzen Moesien und Thrakien II. – Amsterdam.
GUTTERIDGE, A., A. HOTI, R. H. HURST 2001, The 
walled town of Dyrrachium (Durres): settlement and 
dynamics. – Journal of Roman Archaeology 14, 391–410.
HANSON, J. 2016, An urban geography of the Roman world, 
100 BC to AD 300. – Roman Archaeology 18, Oxford.
HODDER, I. 1986, Reading the Past. Current approaches 
to interpretation in Archaeology. – Cambridge.
HORV AT, J. et al. 2003 = J. Horvat, M. Lovenjak, A. Dolenc 
Vičič, M. Lubšina T ušek, M. T omanič Jevremov, Z. Šubic 
2003, Poetovio, development and topography. – In: M. 
Šašel-Kos, P. Scherrer (eds.), The Autonomous Towns 
of Noricum and Pannonia / Die autonomen Städte in 
Noricum und Pannonien. Pannonia I, 153–189, Situla 41.
IV ANOV , R. 2003, Durostorum. – In: R. Ivanov (ed.), Rim-
ski i ranovizantijski selišta v Bâlgarija II, 75–86, Sofia.
JEFFERSON, M. 1939, The law of the Primate City. – Ge-
ographical Review 29, 226–232.
JOHNSON, G. A. 1977, Aspects of regional analysis in Ar-
chaeology. – Annual Review of Anthropology 6, 479–508.
JOHNSON, G. A. 1980, Rank-size convexity and system 
integration. A view from archaeology. – Economic Ge-
ography 56/3, 234–247.
KREKOVIČ, E. 2001, Some comparisons between Germanic 
and Sarmatian limes in Pannonia. – Acta Archaeologica 
Academiae Scientiarum Hungaricae 52/1–3, 103–107.
KRONBERGER, M., M. MOSSER 2002, Vindobona-legio-
nary fortress, canabae legionis and necropolis. – In: P. 
Freeman, J. Bennett, Z. T. Fiema, B. Hoffmann (eds.), 
LIMES XVIII: Proceedings of the XVIIIth International 
Congress of Roman Frontier Studies held in Amman, 
Jordan, 573–584, Oxford.
KRUGMAN, P. 1996, Confronting the mystery of urban 
hierarchy. – Journal of the Japanese and International 
Economies 10, 399–418.
KUZMOV Á, K. 2005, Samian ware on the north Panno-
nian frontier and behind it. – In: Z. Visy (ed.), LIMES 
XIX. Proceedings of the XIXth International Congres of 
Roman Frontier Studies held in Pécs, Hungary, September 
2003, 979–986, Pécs.
LAMBOLEY, J-L. et al. 2011 = J-L. Lamboley, F. Drini, 
F. Quantin, S. Shpuza, S. Verger, A. Skenderaj, E. Fol-
lain, M.-H. Barrière, P. Lenhardt, V. Bereti, A. Islami 
2011, Apollonia d’Illyrie (Albanie). – Mélanges de l’École 
française de Rome 132/1, 242–262.
LENGYEL, A., G. T . RADAN (eds.) 1980, The Archaeology 
of Roman Pannonia. – Budapest.
LOCK, G. R. (ed.) 2000, Beyond the map: archaeology and 
spatial technologies. – Amsterdam.
MANGO, C. 1985, Le développement urbain de Constantinople 
(IVe–VIIe siècles). – Travaux et memoires du centre de 
recherche d’histoire et civilization de Byzance 2, Paris.
MANGO, C., G. DAGRON (eds.) 1995, Constantinople 
and its hinterland: papers from the Twenty-Seventh 
Spring Symposium of Byzantine Studies, Oxford, April 
1993. – Aldershot.
MARTEMIANOV, A. P. 2012, Vici i komai pervih vekov 
nashei eri na teritorii trakiiskih zemel. – Drevnosti 
2012, 40–52.
MARZANO A. 2011, Rank-size analysis and the Roman 
cities of the Iberian Peninsula and Britain. – In: A. 
Bowman, A. Wilson (eds.), Settlement, urbanization 
and population, Oxford Studies in the Roman Economy 
2, 196–228, Oxford.
MIKULČIĆ, I. 1971. Teritorija Skupa. – Živa Antika 21, 
463–484.
MOSSÉ, C. 1973, Athens in decline, 404–86 B.C. – London.
OLTEAN, I. 2007, Dacia. Landscape, colonization and 
Romanization. – London.
ORTON, C., I. HODDER 1976, Spatial analysis in Archa-
eology. – Cambridge.
OSARAGI, T . 2002, Classification methods for spatial data 
representation. – Centre for Advanced Spatial Methods 
Analysis. Working Papers, 3–19.
PAPAZOGLOU, F . (ed.) 1999, Epiri, Macedoniae, Thraciae, 
Scythiae. – Inscriptiones Graecae X II.2, Berlin.
PISO, I. 1991, Die Inschriften vom Pfaffenberg und der 
Beriech der Canabae legionis. – Tyche 6, 131–169.
PISO, I. 1995, Le territoire de la colonia Sarmizegetusa. – 
Ephemeris Napocensis 5, 63–82.
PLOYER, R. 2009, Veteranen im Hinterland von Carnun-
tum: das Gräberfeld von Mannersdorf am Leithagebirge 
und die Besiedlung des Leithagebietes in römischer 
Zeit. – In: A. Morillo, N. Hanel, E. Martín (eds.), LIMES 
XX: Estudios sobre la frontera romana (Roman Frontier 
Studies), 1437–1446, Madrid.
POPOVIĆ, V. 1967, Uvod u topografiju Viminaciuma. – 
Starinar 18, 29–49.
POUNDS, N. J. G. 1969, The Urbanization of the Clas-
sical World. – Annals of the Association of American 
Geographers 59/1, 135–157.
RICH, J., A. WALLACE-HADRILL (eds.) 1991, City and 
country in the ancient world. – Leicester-Nottingham 
Studies in Ancient Society 2, London.
SANTORO, S., A. MONTI 2008, The city of Durres and its 
history, between environmental evolution and cultural 
transformations. – In: J. Erzsébet, F . Redő, V . Szeverényi 
(eds.), On the Road to Reconstructing the Past. Computer 
Applications and Quantitative Methods in Archaeology 
(CAA). Proceedings of the 36th International Conference. 
Budapest, April 2–6, 2008, 491–495, Budapest.
SMITH, C. A. 1985, Theories and Measures of Urban 
Primacy: A Critique. – In: M. Timberlake (ed.), Ur-
banization in the World Economy, 121–167, New York.
SUIĆ, M. 1981, Zadar u starom vijeku. – Zadar.
TAČEVA, M. (ed.) 2004. Vlast i socium v rimska Trakija 
i Mizija. – Sofia.
TILLEY, C. 1994, A phenomenology of landscape: places, 
paths and monuments. – Berg.
TIMBERLAKE, M. (ed.) 1985, Urbanization in the World 
Economy. – New York.
TOMAS, A. 2007, Inter Moesos et Thraces. A contribution 
to the studies of the rural hinterland of Novae in Lower 
Moesia. – Archaeologia 58, 31–47.
VAN LEUSEN, M., V . GAFFNEY 1995, Postscript – GIS, 
environmental determinism and archaeology: a parallel 
text. – In: G. Lock, Z. Stančić (eds.) Archaeology and 
Geographic Information Systems: a European perspective, 
367–382, London.
V APNARSKY , C. A. 1969, On Rank-Size distributions of 
cities: an ecological approach. – Economic Development 
and Cultural Change 17/4, 584–595.

213 Najpomembnejša antična mesta balkanskih provinc in obeh Panonij
Naselbine, predvsem urbani sistemi, predstavl-
jajo temeljno izhodišče za preučevanje antičnega 
urbanizma. Takšen pristop ponuja teoretičen okvir 
za analize razmerij med posameznimi mesti, hkrati 
pa omogoča povezovanje mestne geografije z eko-
nomskimi razmerami na preučevanem območju. 
Kljub nekaterim očitnim težavam, na katere so 
naleteli sodobni ekonomski geografi pri uporabi 
tovrstnega koncepta za pretekle družbe, številne 
študije vendarle dokazujejo ustreznost in vrednost 
takšnega koncepta. Čeprav niti arheologija niti 
antična zgodovina ne ponudita dovolj podrobnih 
informacij, kot jih pisni viri modernih družb, smo 
nekatere analitične pristope sodobnih geografov 
vendarle vključili v analizo antičnega urbanizma. 
Cilj naše razprave je pokazati, da lahko sistema-
tično preučevanje urbanih sistemov privede do 
pomembnih zaključkov o antičnih družbah in 
njihovem gospodarstvu.
Predmet našega prispevka so urbani sistemi 
balkanskih in donavskih provinc v času severske 
dinastije; iz analize sta izključeni provinca Ahaja 
in Deseta italska regija na severnem Jadranu. Ker 
pa je popolnoma nemogoče, da bi v tako kratkem 
prispevku obdelali celoten mestni sistem obravna-
vanega območja, je študija osredotočena na prvo-
razredna mesta. Primerjava velikosti posameznih 
provincialnih mest, kmetijski potencial njihovega 
zaledja ali njihovo mesto v mestni mreži celotnega 
rimskega imperija jasno odsevajo naravo celovitih 
mestnih sistemov.
Vendar pa obravnavano območje naše študije 
za analizo ne predstavlja povsem smiselne celote. 
Mejne province se v nekaterih pogledih povsem 
razlikujejo od provinc, v katerih ni bilo legij, niti 
ni jasnih dokazov o gospodarskih stikih med njimi. 
Nedvomno je torej, da obstajajo pomembne razlike 
med posameznimi provincami, zato je edini možen 
pristop za preučevanje prvorazrednih naselbin 
njihova obravnava v okviru urbanih geografij 
posameznih provinc.
Podatke smo črpali predvsem iz topografskih 
študij, deloma pa iz informacij, ki jih ponujajo 
zračni posnetki. Podatki še zdaleč niso popolni 
in zahtevajo veliko mero domnevanja, kar pa ne 
oslabi glavnih zaključkov študije.
Primerjava velikosti naselbin s statusom ali 
videzom mesta je razkrila tri do štiri naselbinske 
razrede za vsako posamezno provinco na obrav-
navanem območju. Ni presenetljivo, da število 
mest od najnižjega do prvorazrednega upada. V 
večini provinc je prvo stopnjo mestne hierarhije 
zasedalo le po eno mesto. Pa vendar sta zaradi 
posebnih razmer v nekaterih mejnih provincah 
najvišji razred dosegli dve legijski mesti. Glede na 
to, katere ocene velikosti smo vključili v analizo, 
so bila v Zgornji Panoniji kar tri najpomembnejša 
mesta, kar pa je odsev statičnosti urbanih siste-
Najpomembnejša antična mesta 
balkanskih provinc in obeh Panonij 
Povzetek
VERHAGEN, P. 2018, Spatial Analysis in Archaeology: 
 Moving into New Territories. – In: C. Siart, M. Forbri-
ger, O. Bubenzer (eds.), Digital Geoarchaeology. New 
Techniques for Interdisciplinary Human-Environmental 
Research, Natural Science in Archaeology, 11–25.
WHEATLEY , D. 2004, Making space for an archaeology 
of place. – Internet Archaeology 15. [https://eprints.
soton.ac.uk/28800/].
WHITTAKER, R. C. 1994, Frontiers of the Roman Empire: 
a social and economic study. – Baltimore.
WIELOWIEJSKI, J. 1990, Carnuntum als Umschlagplatz 
römischer Importe nach dem Norden. – In: H. Vetters, 
M. Kandler (eds.), Akten des 14. Internationalen Limes-
konresses 1986 in Carnuntum, 753–763, Wien.
WILKES, J. J. 1969, Dalmatia. – London.
WILKES, J. J. 2002, A roman colony and its people (c. 50 
B.C. – c. A.D. 150). – In: E. Marin (ed.), Longae Salonae 
I, 87–105, Split.
ZOTOVIĆ, L., Č. JORDOVIĆ 1990, Viminacium1. Nekro-
pola ‘Više Grobalja’. – Belgrade.
ZSIDI, P . 2003, Aquincum. Ergebnisse der topographischen 
und siedlungshistorischen Forschungen in den Jahren 
1969–1999. – In: M. Šašel-Kos, P. Scherrer (eds.), The 
Autonomous Towns of Noricum and Pannonia / Die 
autonomen Städte in Noricum und Pannonien. Pannonia 
II, 209–230, Situla 42.

214 Damjan DONEV
mov v obmejnih provincah, kjer so bili v ospredju 
predvsem vojaški in strateški interesi.
Po ocenah velikosti se mesta med seboj precej 
razlikujejo, velikostne kategorije pa se gibljejo 
od 50 do skoraj 200 hektarjev. Prvorazredne na-
selbine, obravnavane v tej študiji, niso posebno 
velike, če jih primerjamo s prvorazrednimi mesti 
celotnega cesarstva; v tem oziru so balkanske in 
podonavske province bliže severnoevropskim kot 
pa sredozemskim provincam. Z izjemo Bizanca so 
prvorazredna mesta v mejnih provincah običajno 
večja od prvorazrednih mest v provincah brez vojske.
Analiza absolutnih velikosti prvorazrednih 
mest nam torej omogoča le omejene zaključke. 
Da pa bi razumeli strukturo celovitega mestnega 
sistema, je ključno velikost prvorazrednih mest 
primerjati z velikostjo nižje umeščenih mest. Gre 
za ugotavljanje indeksa prednosti (primacy index) 
posameznih mest. V študiji smo uporabili dva 
načina ugotavljanja indeksa prednosti: najprej smo 
izračunali razmerje velikosti prvorazrednega mesta 
z drugim največjim, nato pa še razmerje velikosti 
največjega mesta s celotnim mestnim območjem. 
Dobljene rezultate smo primerjali z ocenami t. 
i. zakona porazdelitve rangov naselij (rank-size 
rule) in tako dobili izmero centralizacije v mestni 
geografiji. Po prvi analizi so bili rezultati nekoliko 
nepričakovani, saj so pokazali, da sta bili največji 
mesti obravnavanega območja, Karnunt in Akvink, 
najmanj vodilni v njunem mestnem sistemu, za 
bistveno pomembnejšo pa se je nepričakovano 
pokazala najmanjša med prvorazrednimi nasel-
binami, Salona, ki je imela dominantno vlogo. 
Nadaljnja analiza velikosti drugo- in tretjerazrednih 
mest in izračun razmerja med največjim mestom 
ter celotnim mestnim območjem pa sta pokazala, 
da je prvi rezultat zgolj posledica dejstva, da je 
več mest umeščenih v zgornje razrede hierarhije 
poselitve. Razlika med najmanjšim mestom v 
rangu najpomembnejših mest province in mestom 
naslednje stopnje je enako velika ali celo večja 
kot v provincah z enim prvorazrednim mestom, 
medtem ko največje prvorazredno mesto zaseda 
večji delež celotne mestne površine v primerjavi z 
manjšimi mesti. Kljub temu rezultati prve analize 
niso nezanemarljivi. Mestni sistem rimske province 
Dalmacije morda ni imel prvorazrednega mesta, 
a Salona je zagotovo bila najpomembnejše mesto 
v tej provinci in bi se, če ne bi bila podrejena 
interesom Rima in Italije, najverjetneje razvila v 
prvorazredno mesto.
Relativna enotnost indeksov prednosti posameznih 
provinc se lahko nanaša na različno velikostno 
hierarhijo mest v mejnih in nevojaških provincah. 
Prvo- in drugorazredna mesta v mejnih provincah 
so bila večja kot istorazredna mesta v provincah 
brez vojske. Mesta so bila v slednjih številnejša, 
največja mesta pa so zasedala manjši delež celotne 
urbanizirane površine.
Čeprav imajo nekatera mesta Spodnje Panonije 
ali Dalmacije določene vodilne značilnosti znotraj 
mestne hierarhije, nobeno od prvorazrednih mest 
v tej študiji ne dosega stopnje vodilnega mesta. 
Neenakomerni kontakti z Rimom, vojaški značaj 
mestnih sistemov mejnih provinc in čezmerna 
urbanizacija v nekaterih nevojaških provincah so 
bili glavni dejavniki, ki so zavirali razvoj vodilnih 
mest na obravnavanem območju.
Že ob hitrem pregledu tu obravnavanih mest 
opazimo tesno povezavo med upravnim statusom in 
vlogo mest ter njihovo velikostjo. Vsa prvorazred-
na mesta, z izjemo Bizanca, so bila legijski tabor, 
glavno mesto province ali pa oboje. Po eni strani 
to kaže, da je imel Bizanc pred preoblikovanjem 
v cesarsko prestolnico v 4. stoletju pomembno 
vlogo, najverjetneje zaradi svoje lege v globalni 
mestni mreži. Po drugi strani pa njegov primer 
kaže na relativno togo naravo mestnega sistema 
obravnavanih provinc. Rast mest je bila neločljivo 
povezana s formalnim statusom in privilegiji, ki so 
bili mestom dodeljeni glede na politične in strateške 
vidike. Manjša mesta so se zato le stežka razvijala 
in rasla. To je jasno razvidno tudi v velikostni 
razliki med avtonomnimi mesti in podrejenimi 
centralnimi kraji. Brez statusa avtonomnega mesta 
nobena naselbina ni mogla doseči drugo raven 
mestne hierarhije. In v istem smislu tudi povprečno 
provincialno mesto ni moglo konkurirati mestom, 
ki so ležala ob legijskem taboru ali provincialnem 
središču in sedežu finančne uprave.
Glede na lego in določene vidike mestne to-
pografije lahko veliko prvorazrednih naselij ob-
ravnavanega območja označimo kot prehodna 
mesta, ki so znana iz novoveških, kolonialnih 
kontekstov. Vzporednice lahko povlečemo ne samo 
zaradi formalnih podobnosti – lokacije, velikosti in 
mestne topografije – ampak tudi zaradi podobne 
vloge prehodnih skupnosti in tu obravnavanih 
prvorazrednih mest. Oboji so bili ustanovljeni, da 
bi posredovali med Rimom in provincami ali med 
Rimom in barbarikom. Primerjava prvorazrednih 
mest obravnavanega območja s prehodnimi mesti 
novega veka morda ni povsem ustrezna, vendar 
meče zanimivo luč na lego in vlogo obravnava-
nih provinc znotraj politične ekonomije visokega 
cesarstva.

215 Najpomembnejša antična mesta balkanskih provinc in obeh Panonij
S primerjavo velikosti prvorazrednih naselbin 
in njihovega kmetijskega zaledja znotraj mestne-
ga teritorija smo ugotovili, da na obravnavanem 
območju ni bilo pravih prednostnih mest. Če 
podeželskega prebivalstva v mestnih agrih ne bi 
upoštevali, bi skoraj vsa mesta bila samozadostna. 
Njihov položaj pa postane nekoliko bolj kočljiv, če 
upoštevamo podeželsko prebivalstvo. Toda tudi v 
tem primeru je treba stopnjo urbanizacije oceniti 
na manj kot 30 %, preden mestno in podeželsko 
prebivalstvo postane preveliko za lokalne kmetijske 
vire. Dokler stopnja urbanizacije ni manjša od 
15 %, so ocenjeni primanjkljaji razmeroma majhni. 
Menimo, da je treba ocene lokalne urbanizacije 
razlikovati od stopenj splošne urbanizacije in da 
morajo biti opazno višje od slednjih. Toda kljub 
nizki stopnji urbanizacije v mestnih zaledjih so 
imela ta mesta na razpolago številne mehanizme, 
s katerimi so nadomestila primanjkljaj zemljišč 
v agru. Velika upravna ozemlja, dostop do žita, 
pridelanega v drugih delih cesarstva, naklonjenost 
cesarja ali provincialnih oblasti ali posestva zunaj 
mestnih ozemelj je samo nekaj rešitev za morebitno 
pomanjkanje žita v zaledju prvorazrednih naselbin.
Velikost, status in lokacije prvorazrednih na-
selbin zrcalijo naravo celovitih mestnih sistemov 
v provincah obravnavanega območja in njihovo 
vlogo v gospodarstvu v času principata. Urbaniza-
cija balkanskih in donavskih provinc je neločljivo 
povezana z rimskim osvajanjem tega območja, ki je 
prišlo pod upravo rimske države, k osvajanju pa so 
cesarje vodili predvsem vojaški in strateški razlogi. 
Glavna naloga mest je bila zaščita in upravljanje 
novoosvojenih ozemelj ter pobiranje davkov in 
izkoriščanje naravnih virov. V času visokega ce-
sarstva so balkanske in donavske province ležale 
na periferiji imperija s središčem v Italiji in Rimu. 
Te razmere pa niso bile posebno ugodne za razvoj 
zelo velikih mest. Korenita sprememba regionalnega 
mestnega sistema v pozni antiki dodatno podpira 
tezo, da so bila prvorazredna mesta v obravnavani 
regiji na vrhu razvejenega mestnega sistema, ki je 
vse do 4. stoletja gravitiral k Rimu in Italiji.
Prevod: Anja Ragolič
Damjan Donev
Middle East Technical University
Settlement Archaeology Program
TR-Ankara
damjaned@gmail.com