Vous êtes sur la page 1sur 15

The Indented Trace in Later Greek Fortifications

Author(s): F. E. Winter
Source: American Journal of Archaeology, Vol. 75, No. 4 (Oct., 1971), pp. 413-426
Published by: Archaeological Institute of America
Stable URL: http://www.jstor.org/stable/502972
Accessed: 18-01-2016 23:20 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/502972?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Archaeological Institute of America is collaborating with JSTOR to digitize, preserve and extend access to American Journal
of Archaeology.

http://www.jstor.org

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

The

Indented

Trace

in

Later

Greek

Fortifications

F. E. WINTER
Over twenty years ago R. Martin described the
highly developed version of the indented trace used
in the walls of Gortys in Arkadia as a form "dont
aucune des realisations ne remonte au dela du milieu
du quatrikme siecle."l Martin's view of the history of
the indented trace was thus rather different from that
of R. Scranton, who had written in 1941: "Many [of
the walls employing the indented trace] are of dry
rubble, a fact which is a strong indication of, if it does
not prove, an early date for the construction. When further ... one reflects on the primitive character of this
method of enfilade, when used exclusively, he inclines readily to the tentative conclusion that it is an
early, let us say archaic feature. . . ." And later: "We
may be fairly confident that the complete or near complete dependence on the indented trace as opposed to
towers is an indication of a date earlier than the Persian Wars."2
Subsequent investigations have, I think, substantially
confirmed Martin's view that the more complex versions of the indented trace are late phenomena.3 Moreover, it is now fairly clear that neither polygonal nor
rubble masonry should predispose us toward an early
date for walls in which such masonry occurs; both
types were fairly common in specific regions and types
of construction during the Hellenistic period.4 At the
same time, while a date earlier than 350 for the indented trace at such sites as Epano Liosia in Attika,
Gortys, Samiko in Triphylia, Stylida, Samothrace, and
lasos in Karia, seems out of the question, at Halai in
Lokris enfilading was certainly achieved in places by
means of alternating faces and flanks of curtain, in a
system that can be dated to the archaic period on the
basis of excavational evidence.5
This state of affairs may seem to suggest that the
indented trace was used in so many different periods
of Greek military architecture as to make its occurrence
almost useless as an index of chronology. I hope to
show that such a view would be over-pessimistic. The

use of the indented trace can in fact sometimes assist


in dating the system in which it occurs, provided, a)
that we apply to extant examples of this device the
sort of precise definition and stylistic analysis that
Scranton brought to the study of Greek masonry, and
b) that we treat the device as only one among many
factors to be weighed in arriving at an absolute date
for a particular stretch of wall. Neglect of one or both
of these provisions has been responsible for much of
the uncertainty which in the past has often surrounded
the study of Greek city walls qua fortifications (rather
than as examples of masonry style, stoneworking techniques, and so forth).
For example, Scranton included in his list of walls
built on an indented trace (Greek Walls 186 List F)
not only such relatively simple, not to say primitive,
examples as Abai in Phokis and Halai in Lokris, but
also highly complex stretches such as those at Samiko
and Arkadian Gortys. It is true that at least the basic
idea of the indented trace is present in all four cases.
To suggest, however, that no further definition is
needed is rather like equating the earlier Hellenistic
towers of the South Gate of Perge in Pamphylia with
the towers of the Periklean circuit at Eleusis, on the
ground that at both sites we are dealing with curvilinear towers built of ashlar masonry. Again, Scranton
proposed a very early date for the Dema wall in Attika and an archaic date for the planning of the original mainland wall at Iasos in Karia. These dates he felt
to be in harmony with the use of the indented trace
and of rubble masonry. Yet in both cases there is far
greater provision for defensive sallies, through the
numerous posterns, than we have any reason to expect
in Geometric and archaic systems.6 Moreover, at Iasos
the scale of the towers, the numerous artillery-ports,
the complete integration of towers and indented trace,
and the archer-slots covering the approaches to the
posterns are all earmarks of Hellenistic work. In short,
there is nothing but the masonry style that is even sug-

1 R. Martin, BCH 71-72


(I947-48) 135-139.
2 R. Scranton, Greek Walls
(Cambridge, Mass. 1941) 155,
157.
3 See especially the detailed study of the Attic Dema (Epano
Liosia wall) by J. E. Jones, L. H. Sackett, C. W. J. Eliot, BSA
52 (1957) esp. I77ff, with discussion of other examples of the
indented trace for which an advanced date seems fairly certain.
4 Rubble masonry might occur in any period, in walls which
were built either hastily or of a material that precluded really
close fitting of the blocks; good examples may be seen at Gonnos in Thessaly and at Notion in Asia Minor, as well as at
lasos (infra). Polygonal masonry of Hellenistic date may be
found all the way from Etruria to Rough Cilicia. The great
virtue of Scranton's study lay in its (largely successful) attempt to bring some sort of order into the classification of
masonry styles; its weaknesses were in part those inherent in

any system of classification (which almost inevitably leads to


arbitrary distinctions), in part the consequence of trying to
build up a fairly elaborate chronological scheme on the basis
of evidence that was very sketchy (through no fault of Scranton's, certainly; lack of precise chronological data is the rule
rather than the exception in the study of Greek military architecture).
5H. Goldman, Hesperia 9 (1940) 381ff.
6 E.g. the wall around the Greek camp at Troy, as described
in II. 7 and I2; the extant walls of Old Smyrna and BuruncukLarisa, the archaic systems at Miletos, Samos and Eretria,
where there were no serious attempts to break up the Lydian,
Spartan and Persian siege-operations by means of repeated
sallies (cf. Her. I.I7ff, 3.54ff, 6.Ioi); the Samians did indeed
attempt a sally, where they had the advantage of high ground,
but remained within the line of walls and towers in the lowlying sectors beside the sea.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

414

F. E. WINTER

[AJA 75

gestive of early work; and against all the other features that indicate a late date, the evidence of masonry
style, such as it is, should never have carried much
weight.
Most of these objections to Scranton's conclusions
have been noted by other scholars. Yet they themselves
have not always avoided the risks of attaching too
much weight to one particular type of evidence, such
as the indented trace, without clearly distinguishing
between simple and highly complex combinations of
faces and flanks, or of trying to find an underlying
principle which would explain the use of the indented
trace at one point in a circuit and the introduction of
towers only a short distance away.
For example, we have seen that Martin, in his study
of the walls of Gortys, rejected an early date for the
long and very complex trace that occurs a short distance Nw of his Gate B. Such a date had been at least
implied in the grouping of this wall with others in
Scranton's List F. Martin, however, regarded the
stretch in question as a Hellenistic repair to a circuit
built ca. 370. It was in this connection that he observed that fully developed examples of such a trace
all seemed to be later than ca. 350. We may still wonder of course why the NE sector (especially Gate A
to postern on Martin's pl. xm) should be later than
350, when the not dissimilar stretch on the opposite,
or sw, flank of the enceinte (between Towers 4 and
5) can be dated ca. 370. 1 believe that Martin's relative chronology is correct, at least as far as the extant
stretches of wall are concerned. But I base this opinion
on general differences in concept and execution, rather
than on the use of the indented trace in itself. In brief,
on general grounds neither trace seems earlier than the
first half of the fourth century. Since it is most unlikely that the rubble and the carefully fitted stretches
of wall were built simultaneously, and since the rubble
is the intrusive element, I regard it as the later of the
two. Moreover, it is difficult to believe that the carefully
fitted masonry of the remainder of the circuit would
have needed repairing only a few decades after it was
built; thus the repair is probably no earlier than the
late fourth century. Of course there is still the possibility that the Hellenistic repair followed the same
trace as its predecessor of the earlier fourth century.
On this point neither the use of the indented trace nor
the styles of masonry involved provide us with much
significant information-still less with an absolute
chronology, which, in the absence of excavation, has
to be based on general historical background. Absolute
dates based on the overall history of a region some-

times seem more reliable than they really are; however, few will be inclined to cavil at Martin's admirable methodology.
Yet when Martin's observations on Gortys are applied by J. Pouilloux to the outer circuit at Rhamnous,
and used to support a post-35o date for the Attic fort,
the need for establishing definite categories of indented
traces becomes quite clear.' To me at least, the indented trace as used at Gortys has almost nothing in
common with the walls of Rhamnous. The scale of
the fort at Rhamnous, the masonry style, and the relative simplicity of the trace as a whole and of the gateplan, all appear to support the late fifth century date
that many earlier students had proposed. Since the
evidence at Thorikos now seems to confirm a similar
date for the fort there, Pouilloux' suggested chronology
for Rhamnous is perhaps weaker than ever.8 As far as
the extant remains at Rhamnous are concerned,9 the
limited use of faces alternating with relatively broad
flanks is surely stylistically even earlier than Martin's
first circuit at Gortys, dated by him ca. 370. Since virtually all the other archaeological and stylistic evidence
points in the same direction, it is highly misleading to
use Martin's criticisms of Scranton's indented trace
chronology as a guide in estimating the date of the
walls at Rhamnous.
Yet these arguments inevitably lead to the following question: what, specifically, are the characteristics
that permit us to classify the indented trace at Rhamnous as simple and undeveloped, the Gortynian circuit
of ca. 370 as (probably) an intermediate phase, and
the NE walls of Gortys and s walls of Samiko as a
highly-developed indented trace? Clearly, one significant feature is that already noted by Scranton, namely
that in the two last cases the jogs of the indented
trace have entirely replaced the towers that occur at
various points both in Martin's first circuit at Gortys
and in the outer circuit at Rhamnous. In itself, however, this feature is not very significant, even for the
purposes of a relative chronology; in fact, it led Scranton to favor a very early date for walls which others
have regarded as Hellenistic.
In examples of the indented trace where towers were
dispensed with altogether, or virtually so, and which
have for one reason or another been dated later than
ca. 350, we almost invariably find two further characteristics that are missing at Rhamnous. First, the
jogs are quite frequent, much more closely spaced
than is normally the case with towers; second, the
flanks that separate adjoining faces are relatively short,
so that no face projects very far beyond its neighbor.1'

7'J. Pouilloux, La forteresse de Rhamnonte (Paris 1954)


52-53.
8 H. Mussche, BCH 85 (I961) 176-205. It should be noted
also that Pouilloux never really explained how the small inner
circuit at Rhamnous, which he identified as the fort of the
late 5th century, could adequately have fulfilled the objectives
that inspired the fortification of Rhamnous, Thorikos and
Sounion in this period. Quite apart from the lack of space for
housing a permanent garrison, the size of the garrison suggested by the modest scale of the inner circuit would, I be-

lieve, have been insufficienteither to deny the use of this part


of the coast to Peloponnesian fleets or to provide a protected
beach whither Athenian ships could run for shelter from enemy squadrons. The large circuit, on the other hand, descending as it did all the way to the shore, could have served both
these purposes very effectively.
9 As opposed to those sections where Pouilloux could only
suggest the probable line and/or details of the trace.
10I cannot recall any system which has as many towers in
as short a distance as there are jogs in the stretches of in-

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

1971]

THE INDENTED TRACE IN LATER GREEK FORTIFICATIONS

415

Among the walls in which such a complex of faces


and flanks occurs, the Attic Dema now seems fairly
firmly dated to the later fourth century, and a Hellenistic date is highly probable for Gortys.11 Thus the
Samiko wall is also likely to be fairly late; and the
same may be said of the systems at Stylida and Samothrace (infra, n.
In the first four cases, especially,
the flanks of the 53)indented trace were quite literally a
substitute for towers. The short faces and frequent
jogs were probably designed to compensate for the
loss of range and visibility suffered by the defenders
when they were posted at the parodos-level of the curtains, rather than on the roof-platform of a tower. Obviously, a short face could be covered much more
effectively than one of greater length; and shallow jogs
or flanks not only made for greater economy (of space
as well as of time and material needed to build the
wall) but also reduced the danger of defenders posted
on the flanks of the trace accidentally hitting their
comrades on the adjoining faces or curtains.
Substantially the same sort of trace recurs in the
walls of Priene; in the circuit of the lower city towers
were generally replaced by a series of short jogs or
flanks, separated by relatively short faces or curtains.
It is only on the acropolis and near the point where
the w walls run dead against the cliffs that we find a
considerable number of closely spaced towers. Since
these sectors were much less open to assault than the
walls enclosing the city area on SE, s and sw, it seems
reasonable to conclude that the towers served a double
purpose: they not only protected the curtains but also
provided shelter for troops stationed some distance
away from the built-up areas of the city.12
If we look for other examples of circuits that rely
entirely or mainly on jogs rather than on towers
(Priene), or else contain fairly substantial stretches of
indented trace unbroken by towers (Samiko and Gortys), we shall find that such traces were exceptional,
and generally confined to areas where the wall either
descended, or traversed the upper reaches of, a long
and difficult slope. At Phthiotic Eretria, for example,
the steep descent from the citadel to the NE gate has
two jogs (of varying depth), the walls along the less
precipitous, but still steeply descending w sector a series of four or five; but the walls traversing the more

accessible N front run almost in a straight line, with


towers at regular intervals.13At Kremna in Pisidia the
walls along the edges of the "fortified terraces" seem
to have made extensive use of jogs, with relatively few
towers.14 At Erythrai, too, though towers are not uncommon, jogs also seem to have been frequently employed, where the walls ran along the crest of, or descended from, a ridge or hill.15 However, none of these
sites is really comparable with the Attic Dema, Gortys,
Samiko, Stylida and Samothrace, insofar as the length
of the curtains and the frequency of the jogs are concerned.
On the other hand, there is a considerable group of
sites in which towers and jogs form a fairly closely
integrated system. The curtains may vary in length,
and the towers and jogs may be closely spaced in one
area, more widely separated in another. Yet one usually
feels that in planning the trace the designer has taken
all factors into account: the nature of the terrain outside the walls, the size of the garrison available for
each sector of the circuit, the possibility of bringing up
reinforcements on short notice, and so forth. Western
Asia Minor offers some particularly instructive remains, and there is another well-preserved group in
Pamphylia and Isauria."1None of the circuits in their
present form can be much, if at all, earlier than the
middle of the fourth century; several of them belonged
to cities of the first rank; all are well enough preserved to allow fairly specific conclusions regarding
the principles of defensive planning involved.
As was the case at Priene, towers were sometimes
quite as numerous, perhaps more so, in inaccessible
parts of the circuit as elsewhere, e.g. on the crest of
Biilbiil Da' at Ephesos, on the acropolis hills at Samos
and Knidos, and in the more rugged sectors at Herakleia. Presumably in such cases provision of shelter for
a more or less permanent garrison and the difficulty of
rapid reinforcement of sectors so far removed from
the centers of habitation were among the decisive factors. Thus at Samos the walls on the Kastelli hill,
which is much closer to the center of the city and to
the harbor than the high ridge of the acropolis, have
relatively few towers. A similar contrast may be observed between the seaward and landward exposures at
Notion; but here the reason is evidently the almost in-

dented trace at Gortys, Samiko and the Attic Dema; and by


the later fourth century, at least, most towers projected much
more boldly beyond the adjoining curtains.
11 See the articles cited in nn. I and 3 supra.
12 Cf. the towers of the Peloponnesian wall of circumvallation at Plataia (Thuc. 3.2Iff); the towers of the Athenian
Long Walls, some of which were provided with chambers in
the ground storey (as I believe), where refugees sought shelter
during the Spartan raids of the Archidamian War (Thuc.
2.17); the chamber below parodos-level in the SE tower at
Phyle; the hollow ground-storeys in the towers at Gyphtokastro, and so forth.
13 See Stihlin, AthMitt 31 (190o6) i8ff, with plan, and
Das hellenische Thessalien (Stuttgart 1924)
I74ff.
14On Kremna, see C. Lanckoronski,Stidte Pamphyliens
und
Pisidiens H
with
As
far
I was able to
as
ff,
general plan.
II

judge in the springof 1968, the walls are still much as the
Austriansfound them some 8o yearsago.
15See the plan, RE 6 (I909) 581-582,and cf. Keil, JOAI
13

(1910)

5ff.

16The sites consideredin this articleare: in westernAsia

Minor, Assos, Pergamon,Aigai, Erythrai,Kolophon,Notion,


Ephesos,Samos, Miletos,LatmianHerakleia,Alinda, Priene,
Knidos;in southernAsia Minor,Perge,
Iasos, Halikarnassos,
Sillyon, Side, Kremnaand Isaura (the last in the northern
foothillsof the Taurus).The fortifiedsitesborderingupon the
inner end of the Gulf of Keramos,and those of Lykia and
most of Pisidia,have been omitted,in spite of their apparent
importance,becauseI have no first-handknowledgeof them,
and publishedinformationavailableto me does not indicate
any substantialconflictbetween the remainsin these areas
and the theoriesadvancedin the presentstudy.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

416

F. E. WINTER

[AJA 75

vulnerable nature of the slopes above the rocky coastline. At Halikarnassos, on the other hand, towers were
fairly common, not only in the lower areas around the
gates (e.g. for several hundred meters N and s from
the Myndos Gate), but also along the back stretch of
the acropolis and in the walls that enclosed the NE and
sw heights.
Clearly concepts of defensive planning might differ
from one site to another. At Priene, where the walls
may well have been built soon after the foundation of
the new town, the builders must have believed that
naturally strong positions needed little provision for
enfilading other than the jogs of the indented trace.
With the limited number of towers, hardly any of
them in the lowest-lying s sectors, little use of defensive artillery can have been contemplated. Probably a
determining factor was the heavy cost involved; for by
this time defensive artillery was in common use.17 At
Notion we are dealing with a site perhaps 25 per cent
larger than the built-up area at Priene, and provided with natural defenses of the same order. Here,
however, the indented trace, if it is understood to signify a stretch of wall with several, perhaps a large
number, of faces and flanks in succession, is not really
used at all. Instead, enfilading was achieved by means
of towers, which are fairly closely spaced along the
landward flanks, more widely separated above the sea.
Some of the towers on the landward side were probably designed to accommodate artillery, especially on
the N and NW, where the walls descended from the
upper slopes toward the plain, and where the main
entrances must have been located. In any event, the
designers of this circuit did not consider it possible to
dispense with towers altogether, even where the walls
looked down precipitous slopes to a rocky coastline
and treacherous shoals. Such a view of the importance
of towers is perhaps not surprising, for the terraced
plan and the surface remains on the hill at Notion
would seem to indicate a Hellenistic date for the town,
and presumably also for the enclosing walls.

Similar observations may be made at the much


larger sites of Halikarnassos and Samos. The present
layout of Halikarnassos presumably goes back to the
time of Mausolos,"8 who must, I believe, also have
been responsible for planning the circuit of the walls.
Some parts of the walls must have been rebuilt after
the siege of Alexander in 334; but this rebuilding could
have been limited to the short stretch of wall battered
down by Alexander's engines.19 Other repairs and additions, especially the building of more towers, may
have been carried out immediately before the siege.20
But there is no good reason to suppose that the Mausolan circuit was either superseded, or completely demolished and rebuilt, at any time after Mausolos'
death. In fact with the passing of the Hekatomnid
dynasty Halikarnassos must inevitably have been
eclipsed by cities in closer contact with the sea routes
or with the main roads leading inland from the
coast.21 The main lines of the Halikarnassian circuit
would thus belong substantially to the second quarter
of the fourth century-at least two decades, and perhaps 50-60 years earlier than the circuit of Priene.
Such a date would be quite in keeping with the loose
and rambling lines of the circuit; this feature is often
encountered in pre-Hellenistic fortifications where the
trace followed natural lines of defense wherever possible, even though much of the area enclosed would
not be used for building.22 Of course the question may
be raised as to whether it would really have been possible at Halikarnassos to leave outside the circuit any
of the areas that were actually included. Yet at Hellenistic Troizen the early acropolis was abandoned,
and a new line of wall built (facing uphill) along the
lower slopes; the walls of New Halos lay entirely in
the plain, enclosing a square area at the foot of the
old acropolis; and the Milesians of late Hellenistic
times, instead of trying to maintain a continuous circuit between the city and Kalabaktepe, relied instead
on the strong and intricate defenses of the s crosswall.23

17 See Winter, AJA 67 (1963) 377-378, and Phoenix 13


(1959) 187ff.
is See J. M. Cook and G. E. Bean, BSA 55 (1950) 89ff,

the evidence also suggests that Hellenistic circuits were much


more likely to be regular in outline, and fairly compact, than
their classical predecessorshad been; and these qualities often
resulted from the substitution of massive walls and other protective devices for natural defensive strength. In fact the engineers of the Hellenistic period rarely went out of their way
to include natural defenses within the circuit unless they had
some specific objective in view. For example, Demetrias was
founded not only to dominate the head of the PagasaicGulf but
also to command the roads leading inland, NW to Thessaliotis
and sw to the plain of Phthiotis. No other site, I think, could
have served these ends as well as that which was actually
chosen. Again, an important consideration in the extension of
the Pergamene circuit must surely have been the exercising of
more effective control of the roads leading up the flanking
river-valleys; and at Ephesos the security of the harbor depended upon control of the w end of the Biilbiil Dag massif,
which in turn necessitatedthe fortificationof the whole length
of the ridge.
23 Whatever the relationship of the Kalabaktepefortifications
to the city of Miletos in early times, there is an early Hellenistic
tower apparentlybelonging to a stretch of wall that descended
from Kalabaktepetoward the city; and in the earlier form of

on the circuit of Halikarnassos,and cf. Newton, History of the


Discoveries at Halicarnassus, Cnidus, and Branchidae (London
I862-63) general plan.
19 Arrian, Anab. 1.2I, mentions the demolishing of two
towers with the intervening curtain and the shaking of a
third tower; however, at 1.23.5, while the city itself is sacked,
the narrative surely implies that the site was left in a defensible state, i.e. the walls were not destroyed.
20 See Arrian, Anab. 1.2o.2, on the preparations made by
Memnon against Alexander's arrival.
21 E.g. Rhodes, Knidos, Miletos, Samos, Ephesos.
22 E.g. at Corinth, in some parts of the
early classical Athenian circuit, at Thebes, at Messene, at Akragas in Sicily, in the
Dionysian circuit of Syracuse and Epipolai, at Latmian Herakleia, at New Knidos, in the later fourth century at Phigaleia and Stratos. I do not deny that some examples certainly
belong, or seem to belong, to the Hellenistic period, e.g. in
NW Greece, at sites such as Kalydon, Palairos (Kechropoula)
and Vlochhs, at Demetrias, at Samos, in some measure at least
at Ephesos, in the enlarged circuit of Pergamon. However,

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

1971]

THE INDENTED

TRACE IN LATER GREEK FORTIFICATIONS

It is therefore not unreasonable to conclude that at


Halikarnassos it should have been possible to draw
the trace directly from the "Island" (where the Castle
of St. Peter now stands) to the SE foot of the acropolisridge, leaving the NE heights outside the walls.24 The
inclusion of these heights resulted less from sheer
military necessity than from the obsession of the earlier fourth century with the importance of taking advantage wherever possible of natural lines of defense.25
At Samos, on the other hand, where a superficial
resemblance to the Halikarnassian circuit might suggest the possibility of leaving the Kastelli heights outside the circuit, the inclusion of these heights was just
as vital to control of the harbor area as were the fortifications above the Salmakis promontory at Halikarnassos in relation to the harbor there. A torrent-bed
separates the Samian acropolis-ridge, on the one side,
from the Kastro and Kastelli hills on the other. As
soon as the Eupalineion was built, and the decision
taken to include the present acropolis within the enceinte, the extant line of wall, running from the Kastelli heights around the acropolis and down to the
shore w of the Kastro hill, formed the next logical
defensive perimeter. Such, surely, was the circuit laid
out by Polykrates' engineers in the sixth century; and
their Hellenistic successors could not afford to abandon
either the Kastelli heights or the acropolis, without endangering the security of their harbor or their watersupply.26
In short, the irregularities in the trace of Halikarnassos resulted from a more or less unthinking adherence to the natural contours of the site; at Samos,
on the other hand, difficult and irregular ground such

417

as the Kastelli heights was included because of general, rather than strictly local, strategic considerations.
We may therefore claim that the Samian circuit in
some ways represents a concept of defensive planning
more highly developed than that of Halikarnassos;
and this claim remains valid even if the Hellenistic
walls exactly follow the lines of the Polykratean circuit.
If we now try to place the circuit of Priene within
the same line of development, we shall probably conclude that it also is more "advanced" than the circuit
of Halikarnassos. Thus at Priene extensive use was
made of the indented trace wherever towers were not
strictly necessary; "irregularities" were in almost all
cases occasioned not so much by local features of the
terrain as by a desire to provide more effective coverage, either of the approaches to the gates or of a series of curtains in the indented trace.
The fortifications of Latmian Herakleia have long
been associated stylistically with those of Priene;27 and
the two circuits also seem, if we allow for differences
in surface area, terrain and so forth, to reflect somewhat similar concepts of defensive strategy. Actually,
if we take the area of the agora in each city as generally indicative of the size of the population, it appears that Herakleia, for all the length and complexity
of its circuit, was perhaps no more than 20-30 per cent
larger than Priene in terms of population. It is unlikely that the total population was larger than Io,ooo;
probably it was considerably less. Certainly much of
the higher-lying areas at Herakleia was quite unsuitable for building, being included within the circuit
on the basis of considerations similar to those that required the fortification of the acropolis at Priene.28

the SacredGatean armof wall seemsto have struckoff across


the interveningplain toward Kalabaktepe.Presumablythese
two piecesof wall were eventuallyjoinedto form the E linkand the city. If such was indeedthe
age betweenKalabaktepe
case,we are almostboundto supposethat therewas a similar
stretchof wall on the w. On the actualremains,see von Gerkan,
Milet I 8 (Berlin 1925) 26-44,
II 3 (Berlin 1935)
109-III

of "profit and loss" factors, did not involve the abandonment


of any facilities vital to the life of the city.
27 On the walls of Herakleia see F. Krischen, Milet III 2
(Berlin I922); Cook and Bean, BSA 52 (1957) 138-140, proposed to move the dates of both Herakleia and Priene to the
time of the Hekatomnid dynasty; this is plausible for Priene,
but seems to me out of the question for Herakleia-see next
120ff.
note.
24Note that our accountof the natureof the Halikarnassian 28 I have stated in my article on the Euryalos fort at Syracircuitis still valid, even if we regardthe presentNEwalls as cuse, AJA 67 (1963) 374-375 n. 38, that the walls of Herakleia
post-Mausollan,
recognizingthe Mausollanline in Newton's seem to me too advanced in design to have been built by the
"outerwall" (as Cook and Bean, n. I8 supra,believedthat Hekatomnid dynasty. At the same time it is difficult to resist
we should do). It is true that the views here adoptedabout some of the broader considerations that must have led Cook
the declineof Halikarnassos
in post-Hekatomnid
times,on the and Bean to associate these walls with Mausollos' regime. Anyone hand,and the generallymorecompact,less ramblingchar- one who has visited Herakleia, whether in the arduous manacterof Hellenisticcircuits,on the other,might seem to favor ner that was necessary before the mid-fifties, or by jeep from
a Mausollandate for the "outerwall" and a Hellenisticdate Milas, as I did in 1957, or directly from S6ke, as I did in 1962,
for the presentNEline. However,the outerwall seems to me must have felt very strongly that an independent Herakleia
would have had little importance as a port on the Latmian
hardly credible as the main NE bulwark of Halikarnassos in
either period; indeed Cook and Bean also recognize its weak Gulf, unless its presence was welcomed by the masters of Mileand vulnerable character.Therefore, whatever its date, I prefer tos or of Samos (or of both). Still less would there have been
to regard it as a fieldwork blocking the main road down from any question of fruitful relationships with such cities as Myus,
the pass NE of Halikarnassos, i.e. as comparable to the Ter- Priene and Magnesia, all of which lay on the far side of the
messian wall, AJA 70 (1966) x28ff.
Beg Parmak massif, the last two also on the far side of the
25However, the heights above the Salmakis promontory Maiander. Herakleia might well have been interested in imwould still have been included, in order to assure control of
ports from the rich agricultural lands of the Maiander valley;
the harbor.
but her own tortured landscape produced nothing with which
26 The situation at Samos is not really comparable with that to pay for these imports. On the other hand the
occupation of
at Athens and Troizen, where the construction of the later Herakleia by a powerful central government in Karia (whether
diateichismata, though certainly preceded by earnest weighing
located at Mylasa or at Halikarnassos) would have been en-

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

418

F. E. WINTER

At Herakleia, however, the fortifications were clearly


designed to make more use of defensive sallies and
defensive artillery than can have been contemplated at
Priene; in addition, there was a coastal and harbor
zone to be protected, as well as the landward exposures. Thus we find at Herakleia a complex pattern of zigzags, jogs, towers, bastions, and irregular
pointed salients, carefully planned to take every possible advantage of the natural features of the site.
Moreover, not only were the towers more numerous
than at Priene, but the walls were more massive, and
contained a larger number of gates and posterns. On
the whole, it seems to me that the walls of Herakleia
just as clearly represent a further development of the
underlying concepts of Priene as those of Priene do
over the design of Halikarnassos. In fact the Herakleian circuit, insofar as principles of military engineering are concerned, belongs to much the same
stage as do the Lysimachean walls of Ephesos. Presumably, then, they date from near the end of the
fourth century, whereas I should be inclined to date
the circuit of Priene in the forties or thirties of the
century.29

Comparison of the Herakleia walls with those on


Biilbill Da' at Ephesos is especially instructive, since
the terrain involved is quite similar. At Ephesos, as at
Herakleia, many towers were designed to carry artillery, whether "heavy" batteries such as must have been
installed in St. Paul's Tower, or the smaller number
of pieces that could have been accommodated in some
of the towers higher up on the long rocky ridge. Moreover, just as towers occur frequently even in the most
inaccessible sectors at Herakleia, so on Billbiil Da'9
more than 30 towers can be counted, in addition to a
series of short curtains separated by jogs, where the
wall starts to descend the E nose of the ridge. Finally,
closer examination reveals that the overall lines of the
trace were evidently conceived as a succession of recesses, of varying depth, flanked by projecting salitirely logical in terms of 4th century politics and geography.
The maritime routes around the sw corner of Asia Minor were
by this time increasingly dominated by the Rhodians; and in
any event the Knidian and Halikarnassianpeninsulas were too
far from the inland valleys around Mylasa to be practical outlets to the sea. Bargylia, at the head of the gulf of that name,
lay closer to Mylasa than did Herakleia. However, once assured of a foothold at Herakleia, the masters of inland Karia
would have been in easy contact both with the Maiander-valley
route leading inland and with the sea routes outside the entrance to the Latmian Gulf; and they would also have had the
power to support some sort of favorable modus vivendi with
the Milesians. Since Ps.-Scylax knows of a Herakleia going
back to the period before the campaigns of Alexander the Great,
Cook and Bean (ibid. 140) felt it possible to associatethe extant
circuit with the Hekatomnids.
Even if this date is not accepted, on the ground that the
walls cannot be so early, I think we must still regard the
fortification of Herakleia as designed primarily to ensure overland communication between the Latmian Gulf-Maiander valley area and central Karia. Even today there are many advantages for Milas in an "outlet" on what was once the Latmian
Gulf, but is now the great alluvial plain of the Maiander.These

[AJA 75

ents, i.e. once more a concept similar to that found


at Herakleia.
The walls of Kolophon, at least in theory, seem to
have been laid out in similar fashion.30 In relation to
the probable built-up area of the city, the trace is just
as irregular as those on Biilbiil Dai and at Herakleia;
and it shares with these sites both a considerable reliance upon towers and a tendency to substitute jogs
(or zigzags) for them, where the nature of the ground
permitted. Thus towers were most frequent in the
Nw sector, just outside Defirmendere village, and in
the NE and SE sectors, where the walls crossed the
Kabaklidere and its tributary, then recrossed the tributary close to the modern road between Deiirmendere
and Traqa villages. Between the NE and SE sectors, on
the other hand, i.e. on the slopes of the hill that rises
above the modern road opposite Traga village, towers
are few in number, alternating with or being entirely
replaced by a series of jogs; and the same type of trace
recurs on the slopes of the divides that separate the
Kabaklidere from its tributary on the east, and from
the De'irmendere valley on the west.
At lasos on the Gulf of Bargylia there are two separate fortifications, one enclosing the island (later peninsular) site, where the town of lasos seems to have
been located throughout its history, the other extending around most of the landward flanks of the long
barren ridge that rises above the shore behind the modern village.31 Most authorities are now inclined to date
this "mainland circuit" in the later fourth or early
third century;32 presumably then it is more or less
contemporary with the walls of Kolophon, Ephesos
and Herakleia. Like them, the lasos "fort" or "camp"
employs both towers and indented trace. Since towers
and jogs at lasos seem to me to form a system more
closely integrated than at the other sites, I believe the
third century date to be the more likely. However, the
relatively uncomplicated terrain traversed by the lasos
walls may have allowed a more regular alternation of
advantages are easily appreciatedby anyone who has traveled
from Ephesos and S6ke to Milas by the two alternate land
routes: in the one case inland to Aidin, then s to vine and
Yata'an and w to Milas, in the other directly across the plain
from S6ke, past Bafa, and down the "natural"route past Selimiye and the site of Euromos to Milas.
29 I.e. whether or not Cook correctly associates the occupation of the new site of Priene with the Karian satrapy, a preAlexander date for the walls seems to me very reasonable.Cf.
BSA (i957) 138-140 and The Greeks in lonia and the East
(London 1962) 181-182.
30 See Holland, Hesperia 13 (1944) 91ff.
31On Iasos see W. Judeich, AthMitt 15 (189o) I37ff, G.
Guidi, ASAtene 4-5 (1921-22)
346ff, Cook and Bean, BSA
(1957) 100-105. Recent Italian excavations on the peninsula
would seem to have proved conclusively that the city of lasos
was always located there, so that the long mainland fortification cannot be a city-circuit.
32Cf. my note in AJA 67 (1963) 374; in a lecture at the
British School in Athens in April 1967, A. W. Lawrence favored a date in 333, at the time of Alexander's campaigns in
Asia Minor.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

THE INDENTED TRACE IN LATER GREEK FORTIFICATIONS


419
1971]
towersand jogs. Again the systemwas designedto per- the walls were founded on a rocky shelf rising above
mit extensiveuse of defensiveartillery;for the scaleof the modern cart-track,so that few if any towers were
the towers, with their numerous large ports, is not needed. In general, the Assos circuit is reminiscent
otherwiseintelligible.
of those of Kolophon and Notion, especially in its
Insofaras the historyof militaryarchitectureis con- skilful use of natural defensive features and in the
cerned,the walls of Assos,especiallythe imposingtow- obvious importanceattachedto artillery-towersin the
ers of the main gate and the great U-shapedartillery- more exposed sectors.Yet the outlines are somewhat
tower that dominatesthe landwardapproachesto the less irregularthan at Kolophon,a featurewhich again
town, could hardly have been built before the last suggestsa date in the third centuryratherthan in the
quarter of the fourth century. Once more, I am in- later part of the fourth;and almost no use is made of
clined to favor a third-centurydate, probablyin the jogs as substitutesfor towers.
second half of the centuryafter the PergamenekingAigai and Alinda are two other sites in western
dom had been firmly established;for the kings of Asia Minor often mentioned in connectionwith the
Pergamon would surely have had a more direct in- development of fourth-centuryand Hellenistic city
terest in this ratherremote cornerof the Troad than architecture.Most authoritieswould, I think, agree
either Lysimachosor Seleukos Nikator. Lysimachos that the skilfully terraceddistributionof buildings at
could not have undertakena projectsuch as the fortifi- Aigai is as typicallyPergameneas the agoraof Assos.34
cationof Assos before301, when the defeatof Antigo- The layoutand buildingsof Alinda,on the otherhand,
nos left him master of western Asia Minor. On the although not dissimilarin arrangement,may well be
other hand, if the handsomeand carefully-builtnew appreciablyearlierthan those of Aigai, as R. Martin
walls are Seleukid work, they can scarcelybe earlier has argued.35Yet it does not seem to me possibleto
than the very end of Seleukos' reign, i.e. after the assign the walls of Alinda to pre-Hellenistictimes. I
defeat of Lysimachosin 281; for during the last dec- believe the great tower on the acropolisto be conades of the fourthcenturySeleukos'own energieswere temporarywith, if not later than, the towers of Heraalmostcompletelyabsorbedin Syriaand Mesopotamia. kleia; and the whole systemrelied far more upon towIn any event it seems not unreasonable,here as at No- ers than upon jogs, when enfiladingwas required.At
tion, to connectthe wall-circuitwith the remodelingof Aigai, the buildersevidentlyfelt thatnothingwas needthe interior of the city, which "shows a typical Per- ed but the verticalbarrierof the wall in most of the
gamene arrangement."33'
higher-lyingsectors.Elsewhere,enfiladingwas normalIf HellenisticAssos was in fact a sort of Pergamene ly providedby towers. Both at Alinda and at Aigai,
coastalfort, it is easyto understandthe extensiveprovi- such jogs as do occur usuallyseem to have been more
sion for artilleryat all points where the walls were or less incidentalresultsof the natural contours.
To conclude this survey of fortificationsin Westopen to assault. From the great U-tower on the N
around to the area of the main gate there is a more ern Asia Minorwe turn to the two sites that have been
or less regular succession of towers; the E gate is more intensivelyexploredthan any other exceptEpheflankedby towers;and the large tower of the NEangle sos. Both at Miletosand at Pergamonlong stretchesof
of the circuit was reinforcedwith internal crosswalls, wall have disappeared,or have been incorporatedin
presumablyfor the purposeof supportingbelostaseis. later work. Much, however,still remains;and a good
Between this tower and the village, however,the cliffs deal is known about the chronology of the various
on the crest of which the tracewas drawn made tow- sections.
ers almost superfluous.Similarlyfrom the w angle of
At Miletos the archaic walls on Kalabaktepeevithe circuit (beyond the main gate) s toward the sea, dently included both jogs and towers; for it does not
33 Cook, Greeks in lonia and East 189. Cook suggests (258259) that the walls should be associated with the regime of
Euboulos in the second quarter of the 4th century. As I have
noted above, this date seems a good 50 years too early. Moreover, the great gateway clearly belongs to a time when the
most important buildings lay s of the acropolis. Conceivably,
it could still be contemporarywith a predecessor of the Hellenistic agora; but it seems to me more likely that extant gate
and extant agora date from the same period, for it would have
been pointless to have spent time and money remodeling the
agora of a city which had to rely for security upon walls over
a century old (note that many cities seem to have made do
very well, during the Hellenistic age, with an agora that dated
from the early or middle part of the 4th century). I also note
that Cook thinks, as I do, that the residential quarters of Assos
lay mostly N of the citadel, facing inland. May Euboulos' stronghold not have been confined to the citadel and the area of the
modern village, i.e. to the regions N of the diateichisma that
descends from the citadel toward the main gate? Yet the gates
in this wall, since they face outward toward the N, would

seem to indicate that the N sectors were regarded as less important, and could therefore be abandoned in an emergency.
If this state of affairs actually existed, I would suggest that
it too reflects the conditions of Pergamene times rather than
of the 4th century. With the establishmentof a strong central
government at Pergamon, Assos and other settlements on the
coast would have acquired a new importance as coastal forts in
wartime and maritime staging-posts during the intervals of
peace. It was therefore vital that they be able to command the
shoreline and maintain access to the harbor or roadstead below,
even if land communications were temporarily severed. This
(in my view, new and Hellenistic) "seaward orientation" of
Assos as a staging-post continued into Roman times; cf. St.
Paul's last return journey to Jerusalem from Macedonia, Acts
20:13-14.

remains
at Aigaiweredescribed
in de4 Thevisiblesurface

tail by R. Bohn, Altertfimer von Aigai = Idal Ergdnzungsheft


2 (1889).

35 R. Martin, L'urbanismedans la Graceantique (Paris 1956)


357; cf. Cook and Bean, BSA 52 (1957) 143.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

F. E. WINTER

420

seem possible to associate all such features with the


repairs and additions that were certainly made later.36
On the other hand, even in later times the walls around
the central and northern portions of the peninsula may
never have made much use of towers, save at gateways
(e.g. the Lion Gate) or at the entrances to harbor and
roadstead (promontories at the outer ends of Lion
Harbor and "Theaterbucht"). In other sectors the trace
seems to have been highly irregular. In those parts
that can be followed with some degree of assurance,
it appears to have included a number of jogs and
zigzags; but these were the result of adhering to the
contours of hills and shoreline, rather than of systematic attempts to provide enfilading in spite of the
lack of natural defensive advantages.
The southern part of the site, however (i.e. s of a
line drawn from the Athena Hill to the Lion Gate),
was much more open to assault, and the walls were
more than once rebuilt during the Hellenistic and
Roman periods. It is interesting to observe that the sw
sector was laid out on the indented trace, with at least
two sally-ports and, probably, only one tower. No
doubt it was felt that the sea and the deep reentrant
curve of shoreline and wall provided adequate cover
in almost all stretches. On the SE flank, on the other
hand, the walls begin to veer sw away from the shore
in the vicinity of the Lion Gate, leaving a wide strip
of almost level ground between walls and sea. As this
ground became less swampy with the passage of time,
it would have been increasingly open to attack. Although the walls have not been fully excavated over
the whole stretch, trenches and soundings have indicated the general lines of the trace, at least in the
later phases. An indented trace in itself was clearly
not considered adequate. Instead, between the Lion
Gate and the next slight elevation to the s we find a
broad bay protected at either end by projecting salients; within this bay the wall exhibits a combination
of jogs and towers, as well as two sally-ports. There
was a third opening on the crest of the s elevation; and
between this gate and the SE angle of the circuit the
wall follows a zigzag, or sawtooth, course, with, apparently, two more sally-ports, and perhaps a tower at
the point where the trace advanced farthest toward the
SE. Finally, when in later Hellenistic times all thought
of trying to maintain a connection with Kalabaktepe
was abandoned, the s crosswall was built on a plan
that is best described as an indented trace with a tower
projecting from the outer end of each jog. This plan
obviously made for more effective enfilading of the
curtains, as well as providing for extensive use of defensive artillery. Moreover, von Gerkan reasonably
suggests that outside the main line of wall there was a
system of outworks such as are described by Philo
of Byzantion.7
In sum, the walls of Miletos would suggest that in
all periods the indented trace alone was considered
36 See n. 23 supra.

87 Milet II 3 53-54-

38 See the general plan in Altertfimer


Karten (Berlin

von Pergamon

I:

[AJA 75

satisfactory where there were steep slopes outside the


walls, or where the trace followed a deeply indented
shoreline protected both by the sea and by flanking
headlands. At least from middle and late Hellenistic
times, however, the addition of towers, sally-ports, and
other protective devices was regarded as desirable in
most situations, and essential where the walls were
easily approached.
The evidence of Miletos, Assos and Aigai would
seem to be confirmed by that of Pergamon.38 On the
crest and the upper slopes of the citadel hill, at least
as far as extant remains show, towers were rare save
at points of special importance, such as the gateway
halfway up the SE slopes of the hill.39 The outer circuit of the early second century, however, had a regular series of towers from the main gate around to the
tower commanding the Keltios valley; beyond this
point the wall turned w to ascend the steep slope to
the acropolis. The towers, to be sure, seem to have
been some 150-200 m. apart; but the intervening curtains were laid out either as deep indentations or on
a zigzag plan (tower
tower). This type
of trace would have effectively dominated the sloping
ground leading up to the walls, at least as far as infantry attacks and light machines were concerned;
really heavy machines would have been difficult to use,
even in the vicinity of the main gate. To judge from
the scanty remains, the walls that extended from the
main gate around the s and sw slopes of the hill, then
up the Selinus valley to a point below the Nw corner
of the citadel, were similar to those in the SE sector.
At least there were two towers at the angle where the
wall turned uphill; and enough remains s of this point
to suggest that the trace contained a series of deep
indentations, with towers some distance apart.
On the other hand, both at the Nw angle of the
circuit, and on the NE, below the modern parking lot,
the arms of wall that descended steeply from the citadel into the flanking valleys were laid out as a combination of indented trace and towers. In these sectors
each curtain was quite effectively covered by the next
one, higher up the hill. It may be remarked too that
the stretches of indented trace descended slopes which
were steepest at the top, gradually leveling out toward
the bottom. On such "concave" slopes, or on slopes of
"even" profile, there was adequate visual coverage of
all sections of curtain lower down the slope from the
higher ones. Had the slope been steeper at the bottom
than at the top, the extra height provided by the roofplatform (or third storey) of a tower would have
been essential, if the defenders were to be able to "see
over the top" of the resulting "convex" curve. But in
the actual situation at Pergamon towers were hardly
needed at all, except at points where abrupt changes of
direction produced boldly projecting salients, the converging flanks of which could be adequately covered
only by an angle tower, e.g. at the lower end of each
39 Along a line drawn through the Nw corner of the lower
agora and the NE corner of the gymnasium-complex, and projected up the slopes of the hill.

1913).

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

1971]

THE INDENTED TRACE IN LATER GREEK FORTIFICATIONS

421

sarily designed to carry artillery). " Much the same

arm of wall, and halfway down the Nw slopes (where


there seems also to have been a small gateway).
Before leaving Asia Minor to return to Old Greece,
we shall look briefly at the middle and late Hellenistic
walls of Perge and Sillyon (perhaps ca. 200, i.e. not
far in date from the outer circuit at Pergamon), Side
(middle to third quarter of the second century), and
Isaura (period of Augustus). The terrain traversed by
the extant walls of these sites ranges from long steep
(at times precipitous) slopes (at Isaura), through difficult to more gentle slopes (at Sillyon), to low coastline and almost level plain (at Side and Perge). Almost
everywhere, however, the designers relied exclusively on more or less regularly spaced towers, and
introduced few "defensive irregularities" save at the
SE and sw angles at Perge and at the s end of the land
wall at Side. Moreover, the towers are fairly close together, and it is clear from their scale and the size
of the openings in the walls that most of them were
designed primarily as artillery emplacements. Since
there could scarcely have been any possibility of mechanized assault against the walls of Isaura, and hardly
any in the case of Sillyon, we must conclude that by
about 200 B.C. military engineers, even when dealing
with sites of great natural strength, were increasingly
inclined to rely on a series of lofty (four to five storeys)
redoubts, equipped with numerous oxybeleis and perhaps some petroboloi as well. No doubt they came to
this decision partly on the basis of the difficulties that
an enemy would encounter in reducing one by one a
considerable number of such redoubts, partly owing
to the fact that, as long as their stores held out, each
redoubt could be held by a mere handful of troops
(more or less permanently quartered in the tower),
and partly on the added security that such redoubts
would have enjoyed, when the entrances were closed
and barred for the night, against the wiles of domestic
traitors.40 The indented trace offered none of these
advantages even on hilly ground. In open and accessible locations it was quite incapable of withstanding
Hellenistic siege-trains, unless, as in the southern
crosswall at Miletos, it was strengthened by the addition of towers (and perhaps also of outer defenses).
A review of later fortifications at sites in Old Greece
reveals much the same situation as we have found in
Asia Minor. The ancient walled sites of Phokis are
cases in point. Though there may be doubts regarding
the exact dates of many of these walls, no scholar
would now place the large group of walls built in trapezoidal isodomic masonry earlier than the battle of Chaironeia. It is thus interesting to note that almost without
exception the walls of this style, even when defending
strong and compact hill-sites, are laid out as a series of
more or less straight lines, with regularly and closely
spaced towers of considerable size (though not neces-

relatively large fortified area at Orchomenos, where the


jogs that do occur are usually associated with the defense of the posterns. A similar regularity of outline and
dependence upon towers, to the exclusion, or virtual
exclusion, of the indented trace, recurs at sites as widely
separated and as different in scale as Aigosthena,
Thisbe, Ay. Efthimia above the Amphissa plain, Pendeoria, Vidhavi, Soules and Mamaikou, and this despite the steep and inaccessible character of some of the
sites, e.g. Soules and Mamaikou, where there could
never have been any danger of assault by artillery and
machines. To be sure, at some of these West Lokrian
sites we also find a number of jogs, e.g. in the acropolis-circuit at Amphissa, the outer circuit at Glypha,
both the inner and outer circuits at Malandrino, and
Mamikou.42 However, we shall, I believe, look in vain
for a whole series of such jogs, without any intervening towers, of the type that we encountered at Gortys
and Samiko. Instead, the general effect is rather that
of a system in which towers were the normal method
of enfilading, although they might sometimes be replaced by jogs in recessed sections of the trace or at
points of greater than usual strength. This type of
planning is reminiscent of Pergamon, Miletos and Iasos, rather than of Gortys, Samiko, or the Attic Dema.
Sites in Thessaly also show a general preference for
towers, or for systems in which towers and jogs are
combined. On level ground especially, e.g. at New
Halos and in the plain at Palaiogardhiki, towers are
likely to be used exclusively. On slopes or along the
crest of a ridge, some towers may be replaced by jogs;
but it is exceptional to find a series of jogs uninterrupted by towers, as in the fieldwork (if such it is)
outside Stylida.43 More often, the walls were either
unflanked (where they were completely inaccessible)
or else provided with towers (which might be closely
or widely spaced, depending on the terrain and the
purpose of the circuit). For example, the towers are
fairly close together at the more accessible sites of Proerna, Phthiotic Thebes, and Gonnos, but much less
frequent at the top of the lofty hills of Klokot6 and
Vlochbs, both of which have many long unflanked
stretches of curtain. As at Miletos, lasos, and the Lokrian sites above, there is a tendency toward systems in
which jogs and towers are closely integrated, e.g. in the
SE sector of Vlochbs (where an arm of wall descended
toward the plain) and the sw sector at Pharsalos. At
Xyniai the tendency is most evident in the arms of wall
that descended toward the erstwhile lake; but it can
also be seen in other parts of the circuit. At times there
may even be a kind of fusion of towers and jogs to
form a series of shallow projections, presumably no
higher than the flanking curtains.

40 Cf. the arrangementsfor


closing and barring the entrance
of St. Paul's Tower at Ephesos, Winter, AJA 70 (1966) 136137 and pl. 38, figs. 28-30.
41 In general the walls of Drymaia, Lilaia, Daulis and Tithorea relied entirely on towers; however, there are exceptions
to the general rule, e.g. in the back stretch (from the point

of view, that is, of the modern village at the foot of the hill)
of the acropolisat Ay. Vlasis (Panopeus).
42 See site-plans in L. Lerat, Les Locriens de l'ouest (Paris
1952).
43 Y. B6quignon, La vallie du Spercheios (Paris 1937) 294ff
and fig. Ii.

may be said of the walls of Chaironeia,and even of the

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

422

F. E. WINTER

[AJA 75

The fortifications of Akarnania and Aitolia tell a


similar story. At least at sites which can be assigned
with certainty to a period not earlier (and almost certainly later) than the late fourth century, the jogs of
the indented trace are normally either integrated with
the towers, or, where a series of jogs occur in succession, confined to certain specific locations. Stratos is
a good example. Whatever the exact date of the existing walls, it seems unlikely that they can predate
the synoecism of ca. 314.44 Jogs and towers occur together: i) in the stretch of wall running from the
Zeus temple NE past the West Gate; 2) descending
from the Nw heights into the valley and again starting
up the w slopes of the acropolis; and 3) descending
to the high saddle SEof the acropolis.45Several jogs in
succession, without intervening towers, occur in the
stretch of wall just below the sw corner of the acropolis, or keep, and again s of the acropolis, in the first
175 m. of the great crosswall. It is significant, I think,
that both these arms of wall descend slopes that are
steeper at the top, but level out lower down, i.e. the
terrain is very similar to that below the acropolis at
Pergamon, where again the walls make several jogs
without intervening towers. In the NE and Nw sectors
of Stratos, on the other hand, jogs with projecting towers generally seem to occur at points where the slope
becomes steeper; in such cases the extra height of the
tower made it easier for the defenders to "see over
the curve" and down the steeper slopes beyond (see
above on Pergamon).
Similar planning is evident in the walls of Vloch6s
(probably a bit later than Stratos) and New Pleuron
(after 234). In both cases towers are sometimes found
even in the higher-lying sectors, e.g. around the acropolis at Pleuron and flanking the gate at the high western angle of the enceinte of Vlochbs;46 on the other
hand, we also encounter at Vloch6s long unflanked
stretches of curtain, in places where the wall ran along
the crest of a ridge or a rocky outcrop. The arms of
wall descending from acropolis to city-area at Pleuron,
however, relied for enfilading almost entirely on a
series of jogs; and the same is true of much of the w
and sw stretches of wall at Vlochos, where the line
descends from the high western angle, to cross the road
from Kenouryio and continue down into the valley
below Vloch6s village (in which the main ancient gate
seems to have been located). At both these sites, as at
Stratos, a succession of jogs seems to have been con-

sidered adequate where the wall descended even or


concave slopes; but where a rocky outcrop or the like
caused an even temporary convexity in the profile of
the slope, one or more towers were likely to be introduced.
With these conclusions in mind we shall not be
surprised to find only a limited use of jogs at Karavassaras (Limnaia) and Kalydon. In the former case the
circuit enclosed part of a hilltop, without many irregularities or significant changes in elevation. It is true
that there were "Long Walls" descending to the sea
on the north; but these walls are built on slopes of generally convex profile, i.e. steeper in the lower than in
the upper reaches, and so were unsuitable for extensive
use of the indented trace, at least according to the
theory that we are here proposing.47 At Kalydon, too,
much of the circuit clings to the upper slopes of hills
(those to the north and east of the valley in which the
Heroon lies). These high-lying sectors, broken into a
series of bays and promontories, were not the sort of
terrain in which jogs were likely to occur frequently.
However, where the walls descended to the West Gate
(from which the Sacred Way to the Laphrion issued),
we have once more a slope that is steeper at the top
than at the bottom; and again a series of jogs were
employed. Where the wall reascends toward the North
Gate, the Danish plan48 shows one tower or bastion
plus a series of jogs facing uphill. This same unusual
arrangement recurs, for good reasons, in the Attic
Dema; at Kalydon it proves to be easily explained by
the inward jogs in the rocky bank of the torrent-bed
upon which the walls were founded.49
Even Aitolia has a few relatively open and accessible
sites, such as Parav61la(Phistyon) and the sanctuary
area at Thermon. The walls at both these sites are
Hellenistic work (with the probable exception, in my
opinion, of the polygonal masonry on the acropolis at
Parav61a); and in both cases they run almost entirely
on regular lines, with closely spaced towers, and no
real use of the indented trace. The character of the
circuit at Thermon, in particular, is very reminiscent
of New Halos in Phthiotis, where again we find a
rectilinear enclosure lying right at the foot of a lofty
hill.50
Oiniadai is perhaps most instructive as an example
of a system in which the occurrence of a certain number of jogs is to be explained in terms of the natural
contours of the site rather than on the basis of a par-

44 Scranton, Greek Walls 94-96. Indeed the city wall may


well be appreciably later than the end of the 4th century;
Kirsten and Kraiker, Griechenlandkunde'(1962) 761, suggest a
3rd century date.
45 Probably also in the E sector, between the road and the
river bank.
46 On the Vlochbs gate see W. J. Woodhouse, Aetolia (Oxford I897) fig. opp. 188. Woodhouse shows a tower only on
the uphill side of the gatecourt; in my own field notes made
a number of years ago I drew the remains of towers on each
side, but I should not wish to insist on the correctnessof my
version without further checking.
470On such slopes the indented trace would probably have
been regarded as inadequate in the middle reaches, and un-

necessary (as towers would also have been) on the precipitous


lower stretches.
48 Cf. Kirsten-Kraiker,Griechenlandkunde' 770 fig. 191.
49 On the Attic Dema see the remarks of J. R. McCredie,
Fortified Military Camps in Attika (Princeton 1966) 63. At
Kalydon we are clearly dealing with simple adaptations of the
trace to a natural defensive line, i.e. that of the low cliffs overhanging the torrent bed; this line was felt to be strong enough
to make towers unnecessary.
50 This is the sort of situation that archaic and classical engineers would have done almost anything to avoid; but such
sites were not uncommon in Hellenistic times, as we may see,
for example, in the abandonment of the acropolis at Troizen
when the Hellenistic diateichisma was built.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

1971]

THE INDENTED TRACE IN LATER GREEK FORTIFICATIONS

423

ticular theory of fortification. Seen from a distance, the


site has a deceptively "regular" appearance; as one
draws closer, it becomes clear that it is very literally a
case of o6rav KodX7rovKat avaXwp?~1cw
c XT1 7
oWrdLpa"
&Z
it.51
&rLov
K7taLorvat,as Philo of Byzantion phrased
For this type of site Philo recommended the use of the
"double trace"; Oiniadai has nothing so elaborate, and
probably never needed it, being, as Thucydides remarked, "in the middle of a marsh formed by the
river Acheloos."52 Simply by adhering to the natural
contours of the site, the designers of the extant walls
were able to achieve a very strong defensive position.
Relatively few towers were needed, when we consider
the great length of the circuit; and those jogs that do
occur can all, I believe, be explained either as deviations imposed by the natural contours of the site, or
introduced in order to cross the upper rather than the
lower reaches of a bay, or the like. In fact, Oiniadai
is in so many ways sui generis that it tells us less than
might be expected about the basic principles of Greek
military architecture in any given period.53
On the basis of the above survey of indented-trace
systems in Greece and Asia Minor, I believe that we
may draw some definite conclusions.
I. It is important to distinguish between a series of
jogs introduced purely for purposes of enfilading, and
those which are partly or entirely the result of the
natural contours of the site. At Rhamnous, for example, the type of trace employed does not in itself seem
to afford any basis for dating the outer circuit to a later
period than the inner; in such cases problems of relative and absolute chronology must be resolved on
other grounds. The same may be said, as we have just
noted, of the jogs in the trace of Oiniadai, and even
(if we rashly tried to consider jogs in isolation from
towers, gates and posterns) of the Attic fortress at
Gyphtokastro, the relatively advanced date of which
nobody seriously questions.
2. It is also clear that there were different ways in
which a series of two or more jogs could be used,

quite apart from deviations enforced by the natural


contours of the site. In my opinion the following general categories may be established, in spite of the undoubted problems involved in determining the category to which a specific example belongs:
2a. Systems in which neither towers nor jogs are
particularly frequent, the projection of either or both
is minimal, and the towers may never have had an upper storey. Halai and Abai seem to me to fall within
this category; still other examples could be cited, e.g.
the inner circuit at Typaneai;54 and I believe that we
should also include sites where we find "fausses cremaillkres," i.e. slight outward jogs in the outer face
of the curtain achieved by a temporary increase in
the total thickness (
). In all the above
systems, as distinct from those in the next three categories, no great importance seems to have been attached to extensive provision for enfilading, whether
because the builders felt a simple rampart to be unscalable in itself, or because the ground outside the
walls was so difficult as to render the walls unapproachable.
2b. Systems which depend primarily upon regularly
spaced towers, but in which jogs or zigzags may be
substituted for towers (chiefly, no doubt, for reasons
of economy) in certain locations (e.g. descending an
even or a concave slope, or traversing the upper reaches
of a long steep slope, especially where the contour
lines curve slightly inward, or along an incurved shoreline; but rarely, I think, where level open terrain extended right up to the foot of the wall). The majority
of the combinations of towers and jogs discussed in the
preceding pages belong to this category.
2c. Whole stretches of wall laid out in a series of
relatively short faces separated by frequent flanks or
jogs, without any intervening towers. Examples are
rather rare, as far as I know (Arkadian Gortys, Samiko, Stylida, the wall across the pass at Thermopylai,
part of the Biilbiil Da' wall at Ephesos, Priene, where
the tower at the SE Gate is the only such structure

51 Philo, 86.5ff. It should be remarked that the


plan of the
walls at Oiniadai is not always easy to follow on the ground;
in this regard it is instructive to compare the more recent
plan of Kirsten, RE 17:2 (1937) 2217-18, fig. I, with the
older results of Heuzey and Powell (as reproduced,for example,
in the 1936 edition of the Guide Bleu).
52 Thuc. 2.102.
53 For instance, Kirsten, who regards the polygonal walls as
work of the late 3rd century, believes that earlier defenses, e.g.
during the 5th century when Oiniadai was the object of repeated
Athenian attacks before it was finally captured in 424, were of
a much simpler character, perhaps only earth ramparts, and
that the same was true elsewhere in Aitolia and Akarnania; cf.
his RE article cited in n. 51 supra. I am bound to say, however, that his arguments seem to me very forced at times. For
example, TroOrelXOv in Thuc. 2.81.2, used without any explanatory remarks, surely refers to a "conventional" city wall
of stone or brick, and anything else that may be implied by
Kirsten's "Umwallung" (p. 2227) seems an unwarranted inference on his part.
Perhaps this is the best place to mention two other fairly
well-known systems in which the indented trace appears, even
though they seem to add nothing to our knowledge of the

ways in which this trace was likely to be used. At Philippi the


walls were generally provided with regularly and closely
spaced towers, especially where they traversed virtually level
plain. On the E flank, however, several jogs were introduced,instead of towers, where the wall ascendedpast the theater and up
to the acropolis; and jogs appear again in the Nw sector, above
and below the modern road. Typologically, the trace seems to
show a greater reliance upon towers than does either Priene or
Martin's earlier 4th century sectors at Arkadian Gortys; but
towers and jogs are less completely integrated than, say, at
lasos or Pergamon. On the other hand, the extant walls of
Samothrace seem to me very close to those of Iasos, not only
in the integration of jogs and towers, but also in the provision
for the use of artillery and the occurrence of several posterns
within a relatively short distance. I therefore fully endorse H.
Seyrig's view that these walls cannot be earlier than the time
of Alexander; indeed they may be appreciablylater. Cf. Seyrig,
BCH 51 (1927) 353-368 on Samothrace;for the general plan,
etc., of the walls at Philippi, see P. Lemerle, BCH 62 (1938)
4ff.
54E. Meyer, Neue peloponnesische Wanderungen (Bern
1957) Plan I.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

424

F. E. WINTER

[AJA 75

along the E and SEflanks, perhaps Samothrace, though


here there are several towers); they seem never to occur in completely exposed locations, and are normally
found either descending a slope, or else flanked by
walls that stand on higher ground or form projecting
salients.
2d. A variant of the preceding, designed for level
ground or for projecting salients, where it was felt
necessary to provide more effective enfilading by means
of towers at each jog. This very complex type of trace
is extremely rare among extant fortifications (southern crosswall at Miletos, addition to Aspis wall at
Argos).55
3. Some conclusions may be reached regarding
chronology:
3a. Systems of the types noted under I and 2a above
may occur in almost any period.
3b. Type 2b, in my opinion, is probably never earlier than the time of Philip II; only a few of the systems that I would place in this category are likely to
be earlier than the later fourth century, and many are
demonstrably later (e.g. Pleuron, Dodona, Miletos;
probably also Aigai and Notion).
3c. Type 2c, in the simpler form which we find at
the w end of the N and s flanks of Gortys, does not
occur in any context demonstrably earlier than ca.
375, nor, I think, much later than ca. 250. The really
complex versions, such as Samiko, the Attic Dema, the
mid-northern sector at Gortys, Biilbill Da', probably
had an ever shorter vogue. None of them need be
earlier than ca. 335 or later than ca. 260. Martin may
well have been correct in associating 2b and 2c with
the rise of the Macedonian kings (or at least with the
engineers whom they employed)."
3d. From the later third century onward the "pure"
indented trace, i.e. a succession of three or more jogs
without intervening towers, seems to have become
steadily less popular, regardless of the nature of the
terrain, on account of the increasing reliance upon
multi-storeyed towers as artillery emplacements and
as semi-independent strong points or redoubts. Increased use of ditches and proteichismata was probably
also a factor; for such outer defenses, as we may see
in the walls of Manduria in ancient Calabria, permitted the use of a simpler trace in the main rampart.57
These conclusions may seem rather too general to
be of real value in the study of Greek military architecture. Moreover, even in their present form they
must be checked by other scholars against systems
known to them, which I have either overlooked, or
omitted from my survey because I have not personally

visited the sites. If my classification and general chronology should be found to have some validity, it will
at least be a small step forward in a very difficult
and confusing field of study.

55On the Argive wall see W. Vollgraf, BCH 31 (1907)


i44ff, and F. G. Maier, Griechische MauerbauinschriftenI
(Heidelberg 1959) 145-146.
56 See n. 2 supra.
57 Cf. N. Degrassi, FA II (1956) 179 no. 2578; B. Neutsch,
ArchAnz (1956)
267ff.
58 As for instance in Diels-Schramm,Philo = AbhBerlAkWiss,
PhilHistKl I12 (I919) 27 n. I.

which late Hellenistic pieces were actually mounted see my study


of the Euryalos fort at Syracuse, AJA 67 (1963) 363ff.
60 E.g. the prominence that Hermogenes evidently enjoyed
among Vitruvius' sources; the literal, verbal adherence of Roman and Byzantine writers to Hellenistic predecessors;Roman
borrowing from Hellenistic sources of the basilica-type and of
hypocausts (to say nothing of innumerable decorative forms
and details, ranging from moldings to the decorative use of
pilasters, half-columns, and arched niches).

59On the new types of artillery see Diels-Schramm,AbhBerl


(I9I8) no. 16 pls. 6-8; on the complexity of the systems in

UNIVERSITY

OF TORONTO

APPENDIX
Extant fortifications and theT rELXOroLiaL
tLav8pw8q,
Ecro7rVpyMa ovcrain Philo of
Byzantion (86.51f).
None of the editors of Philo's Mechanike Syntaxis
have tried systematically to explain his classifications
and recommendations in terms of extant fortifications;
and archaeologists familiar with the actual remains
have generally been content to comment on isolated
passages. Where no exact parallels were known, there
has been a tendency to assume that Philo's proposals
were examples of academic theorizing, and quite divorced from practical experience.58
This assumption seems to me to be unfair, not only
to Philo in particular, but also to Hellenistic scientists
and engineers in general. It is true that the Hellenistic
period produced a good deal of sterile academicism;
but not very much of it appeared in the fields of scientific and engineering research. Indeed it was in these
areas that most of the really original and creative work
of the Hellenistic age was accomplished. Nowhere,
perhaps, is this truer than in the sphere of poliorcetics
and military engineering. Consider, for example, the
attempts to overcome the low "volume of fire" of the
average oxybeles by developing a "semi-automatic repeater," and to eliminate the maintenance problems
of torsion-weapons by designing spring-operated and
compressed-air pieces to take their place. Again, compare the scale and intricacy of extant Hellenistic systems at Selinus, Syracuse, Miletos and Side, with the
relative simplicity of classical fortifications of the fifth
and earlier fourth centuries."9 Note too the extent to
which the later theory and practice of architecture and
military and civil engineering were dependent upon
Hellenistic sources of inspiration.60
In these circumstances, it seems to me likely that
Philo based his recommendations for the use of different varieties of teichopoiia on actual experience.
Presumably he contributed ideas of his own; almost
certainly he would have been interested in an "ideal"
system, rather than in debating the merits of such
modifications as would be introduced when the "ideal"
requirements of the military authorities were weighed
against economic and other non-military considera7rpLovor?, and Ao$ahr

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

THE INDENTED TRACE IN LATER GREEK FORTIFICATIONS


425
1971]
tions. Nevertheless,it should be possibleto recognize southernsectorwas providedwith a ditch, and to supin existing Hellenisticsystemsdesigns that bear some pose that this ditch extended into the area between
resemblanceto those proposedby Philo. On the other the SacredGate and our supposed"maiander-trace."63
hand, it would be surprisingif we encountereda cir- The additional protectionafforded by the outer decuit that adheredto a single Philonian teichopoiiato fenses, togetherwith the seriesof zigzags in the trace,
the exclusion of all the others. Philo's own advice, presumablycompensatedfor the loss of those adwould rather suggest that in vantages which a southwestwardcontinuationof the
7rpoopWVaT rovg r0rrovq,
each part of the circuit the system best suited to the "maiander-trace"
would have conferred.
local terrainshould be employed.
Another possible example of a trace related to
occursat Side, on either side
Among the sites mentioned above, Miletos is one Philo's "maiander"-type
with a variety of local conditions-and, as we have of the angle where the southernend of the land wall
seen, considerablevarietyin the planning of the trace. meets the southernsea-wall.It is true that there are
It may be that the form of Philo's q tp~vMtaLavspwo-qq
no towers in this stretch,whereasPhilo's "maiander7- 7rTE&vis reflectedin the traceof the first 200-250 m. trace"presumablywas providedwith towers. Yet the
of wall extendingsw from the Lion Gate.Certainlythe Side wall certainly does form a pattern that recalls
ground betweenthe walls and the sea could qualify as the stretch at Miletos, and might reasonablybe de7rE8tv';and between the salients at each end of the scribedas a "maiander."Perhapswe may also suggest
stretchthe trace is broken into a series of projections that this form of trace,used on level ground, was an
and recesses.The term maiandr6deswould surelyhave elaborationof the ..rt.
shapedbastionswhich fairly
been more logicallyappliedto a .JUL.r patternthan often occur on
in other Hellenistic syshilly
ground
to any other.
at Lepreonin Triphylia,Kokkoti in Phthiotems,
e.g.
However,thereare difficulties.Perhapsthe most obIt is also possiblethat the "maiander-trace"
was
vious is the fact that, while the succeeding350 m. of tis.64
in which the gate was
partly
inspired
by
gate-plans
wall (up to the areaof the SacredGate) and the whole set at the back
of a deep forecourt.65
of the s crosswalltraverseground that is just as much
It has been suggestedabove that the s crosswallat
in
as
that
which
our
"maiander-trace"
presumed
rr8tvr
occurs, the type of trace is altered. We have, first, Miletosis at least relatedto Philo's"tracewith oblique
somethingwhich, if it recallsPhilo at all (and if the curtains,"while the stretchof wall extendingNEfrom
limited excavation correctlyestablishedthe line fol- the SacredGaterecallshis "saw-toothtrace."Diels and
lowed by the wall),"6 is most reasonablyassociated Schramm, however, identified the "saw-toothtrace"
with his -EtXorod'a
rptovworg;and von Gerkanbelieved with the indentedtraceas we find it at Priene.66Philo
was at tells us that the "saw-toothtrace"was an inventionof
(correctly,in my opinion) that the s crosswall
'
least related to Philo's TELXorooda
AOX ra TE07ornvpyta Poyleidos, one of the engineers employed by PhilEXovaa.Philo, however, recommendsthe use of the ip II,67and this informationwould agree well with
"saw-toothtrace,"orav o Tro ,( KoXLdF,
and "the Martin'sconclusionthat the full developmentof the
trace with oblique curtains," TpLyvoLME78swELV.
Of indented trace should be associatedwith the Macecourseother engineers,including the designersof the donians. However, in Philo's scheme the "saw-tooth
Milesianwalls, no doubthad ideasdifferentfrom those trace"is to be reinforcedby, ratherthan a substitute
of Philo;62but one might at least expect some con- for, a seriesof towers.I am thereforemore inclined to
sistency of application.Part of the explanation lies recognizedthe "saw-toothtrace"in walls such as those
in the later date of the southerncrosswall,in which along the South Wadi at Dura-Europos,or at least in
great importance was attached to the provision of such stretchesas the wall that extends NE from the
numeroustowers. For the rest, it is perhapssimplest westerngate of Epipolaiat Syracuse(in a patternvery
to accept von Gerkan's suggestion that the whole similar to the Milesian wall under discussion).'6 At
61 Cf. von Gerkan, Milet II 3 37-39.

62 For example, Philo recommends that his "double trace"


V
be used STravK6X7rovS
K aiXvaXWp'ELS
Tf
XLepa
Smrov
CX?.
-6
Now nothing in Greek military
architecture
re5BeKTUQLYOVL.
calls the Philonian "double trace" as strongly as the northern
defenses of Selinus (F. Krischen, Die Stadtmauernvon Pompeji
[Berlin 1941] 29ff, pls. 11-19, 38-43) and the great battery,
together with the triangularbastion in front of it, in the Euryalos fort at Syracuse (Winter, n. 59 supra, esp. 383). In Italy,
Maiuri's restoration of the late Hellenistic walls at Pompeii
(MonAnt 33 [1930-31] I6I-I62 and fig. I2), if correct, may
represent a combination of Hellenistic "double trace" and Italic
agger. In Byzantine times, the design of the Land Walls of
Constantinople is surely based on something resembling the
Hellenistic "double trace." Among western mediaeval fortresses,
Carcassonneprovides an impressive example of the same sort
of design. Yet in none of these cases could one say that the
terrain was marked by "bays and inlets."

63 Indeed this region may have been provided with a ditch


even before the s crosswall was built.
64 See the plans: A. Blouet et al., Expidition scientifique de
Morde I (Paris 1831) pl. 50 fig. II: Stdihlin,AthMitt 31 (19o6)
3465 The w gate of Epipolai at Syracuse is a monumental example. In much the same way Philo may have adapted his
~jLtKVKXKWV
from gates with an open-fronted
retXoroda EK TWyV
court of semicircularplan, such as we find at Vloch6s in Aitolia and at Sillyon in Pamphylia.
66 Diels-Schramm, Philo note on Philo, 83.6-7.
67
Philo, 83.8-9.
8sPillet, without, apparently, any thought of Philo in mind,
described the Dura wall as "so formed as to resemble the teeth
of a saw" (Yale Expedition to Dura-Europos,First Preliminary
Report [New Haven I928] 9); though on much less broken
ground, the stretch of wall at Syracuse (Epipolai: area NE Of W
gate), shown in detailed plan by Mauceri, II Castello Eurialo

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

426

F. E. WINTER

Miletos we have a simplified form, in which most of


the towers were omitted; the same is true of the wall
at Syracuse.
The indented trace seems to me to be more suitable
as the ancestor of Philo's "trace with oblique curtains." In fact, I have suggested above that the Philonian version, with towers at each jog, was originally
developed for level ground such as that between the
city of Miletos and Kalabaktepe, where jogs alone
were inadequate, at least in Hellenistic times. Philo's
"maiander-trace"went one step further than the "trace
with oblique curtains," and would presumably have
been preferred by him for the s crosswall at Miletos.
He perceived, however, that the "oblique trace," with
its rather unusual relationship of towers to curtains,
would have been especially valuable in the defense of
a salient (rptyWvov Ea'Sc18tv);
and in fact something
nella storia e nell' arte2 (Rome 1939) pl. v fig. I, also seems
to qualify for the epithet "saw-toothed."

[AJA 75

very similar was employed in the triangular salient


added to the Aspis enclosure at Argos during the Hellenistic period.69
In sum, I believe that extant Hellenistic fortifications do in fact provide either reasonably good parallels, or plausible sources of inspiration, for most of the
types of trace listed by Philo. In the preceding pages
and Aoao T/aE'
r
only the LatavSpO8J',,
O7VPYtOa
rrptlovrv1
lxovurahave been discussed in any detail; but we have
also noted in passing possible parallels for the ideas
contained in the c
qjLtKVUKAovand S&rkA.Cerrwov not
were
these
always used as Philo
tainly
systems
recommended; but this is not surprising. The important point is that Philo's work was firmly based upon
the realities of military architecture of his day. In this
respect it was quite different from the "library-based
research" of Vitruvius, Vegetius and their Byzantine
successors.
69 See n. 55 supra.

This content downloaded from 150.216.68.200 on Mon, 18 Jan 2016 23:20:28 UTC
All use subject to JSTOR Terms and Conditions

Vous aimerez peut-être aussi