11

Transmission and signaling

Targeting an enemy and collecting intelligence must go hand in hand with the ability to transmit the information to those who need it most. One of the Roman Army’s basic needs, therefore, was the ability to transmit intelligence. Although much of their communications traffic was carried by messenger, when speed of transmission was needed in emergencies, signaling was also an option. There is not a great deal of ancient commentary on the subject of signaling; there are slightly more than fifty references from all of antiquity, and less than half of those are Roman.¹ Roman signaling relied on fire signaling, except over very short ranges where voice, trumpet, or semaphore could be used. Archaeologists claim to have identified the existence of chains of signaling posts in Roman provinces along the borders, although there are as of yet no conclusively identified Roman beacons to prove how these systems worked.² Some have suggested that information concerning enemy forces was passed rearward to areas where major troop concentrations could be moved forward to threat-prone areas. Others conclude that long-distance communication by fires or torches was not feasible at all and that signaling was simply a local phenomenon between border forts. No one has suggested that a signaling system was used to keep the capital informed, as was true for the Assyrians or the Byzantines.³

Most scholars seem to agree that signaling systems were a normal feature of military control along the frontiers of the Roman Empire. One scholar, David Woolliscroft, has investigated how Roman frontier design incorporated signal towers, and in several publications he has attempted to illustrate how these frontier systems worked in areas along Rome’s borders, including Britain, Germany, and Raetia.⁴ There are very real differences between individual frontiers, because of local factors such as terrain, operational conditions such as weather, and the availability of materials and manpower, but the classic Roman Limes (pronounced lee-mace; literal meaning, boundary) was composed of a frontier line, with or without a barrier (like Hadrian’s Wall), an observation screen based on a system of watchtowers and a series of principle garrison forts backed by legionary bases.⁵ Within this arrangement, archaeologists have tried to identify the structures from which the signals were sent and received. Woolliscroft has added an additional dimension by doing intervisibility studies between the known and suspected sites.

In the absence of proof, all we can do is examine Rome’s technical capabilities and the techniques of signaling used by them according to the literary sources. Then we may examine the archaeological data to see where signaling systems may have existed and how they may have worked. This latter exercise is, of course, speculative. As Woolliscroft himself warns: “the fact that a signaling system was apparently both possible and desirable on a given Roman deployment does not necessarily mean that any such system actually existed.”⁶ But at least we can make an educated guess.

Messages by courier

The simplest form of communication between the field and headquarters was the dispatch. Emperors who conducted campaigns sent off dispatches in the traditional format to maintain the appearance that the Senate had a real share in the military policy of the emperor, and that he was not employing the army merely at his own whim.⁷ This was an informal process probably carried out by Augustus, but it soon became a courtesy expected by the Senate. Even in the third century ad, Dio Cassius thought it worth mentioning that Macrinus did not send a full dispatch to the Senate after his peace settlement with the Parthians.⁸ A dispatch, rather than containing delicate diplomatic intelligence or state secrets, was an official announcement that could be a vehicle for broadcasting the emperor’s achievements. The Emperor Gaius (Caligula) organized the delivery of his dispatches to and from Germany to produce a dramatic impression. The messenger was ordered to ride in a carriage at full speed into the forum and present the dispatches to the consuls in person at the temple of Mars Ultor, before a full meeting of the Senate.⁹ Trajan wrote frequently to the Senate, listing the names of conquered peoples and claiming that he had marched further than Alexander.¹⁰ Unlike intelligence that must be kept secret, the more widely the emperor’s view of what happened on military campaigns was disseminated, the better. Unlike genuine intelligence, which should be characterized by accuracy, information concerning political or military matters carried in dispatches, especially about successes, was spread even if it was not entirely accurate.¹¹

Fire and smoke signals

As an early warning method, or as a vehicle for passing on rudimentary instructions, fires can work well if the messages are simple and predetermined, for example, the one-if-by-land, two-if-by-sea variety made famous in the Longfellow poem about Paul Revere’s ride. Bonfires can provide both smoke signals for use in the daytime, and fire signals for use at night. Smoke signals are perhaps the less reliable of the two because in most places the wind will not allow the smoke to rise in straight enough columns or puffs, as they conveniently do in Westerns. Even fires lit at night are only useful if they can be seen. We know Caesar summoned troops by using smoke signals during the civil war,¹² and Frontinus tells us that “the Arabians” were warned of the approach of the enemy by fire and smoke signals, but we are given no historical context for the story.¹³

Tactical signaling had been employed in battles since the Republic, and the Romans had access to descriptions of Greek and Carthaginian signaling systems. In 134 bc at Numantia, we are told that Scipio Aemilianus operated a combined system of information transmitted by relays, posts, messengers, and fire signals.¹⁴ During the day, signals by red flags replaced smoke signals and seem to have been a great improvement. A later source reports Romans using daylight signaling by smoke and night signaling by fire, signal flags, semaphore, and military horns.¹⁵ There was a signaling station at Kossyra between Sicily and Africa, and it is thought that the system had been used by Dionysius of Syracuse (fifth/fourth century bc). Naval signaling was well known, but there was always the chance that the enemy might see the signals too. False signals and lights could also be used to mislead enemy ships.

As long-distance signaling systems, beacons have some very distinct advantages. They tend to be less affected by climate than other methods of signaling. A simple but powerful torchlight was less likely to be obscured or distorted by poor reception conditions. A bright light can be seen on a misty night even at a distance, which is why they were used in lighthouses in the ancient world. They also have distinct disadvantages. As mentioned before, they are limited to giving prearranged signals, such as “danger” or “the attack has begun,” and are not generally capable of dealing with the unexpected.¹⁶

The one documented case in Roman history in which a fire signal conveyed intelligence was the signal that announced the fall of Sejanus. It was a prearranged signal relayed from Rome to the Emperor Tiberius on the island of Capri.¹⁷ According to Suetonius, “he waited on a clifftop for the distant bonfire signals (announcing all possible eventualities), which he had ordered to be sent in case his couriers might be delayed.”¹⁸ This signal was for the emperor’s private use and was a back-up system. Fire signals in wartime are noted in Roman military records, and they were misunderstood by the recipients on a number of occasions. The other references to fire signals in Roman literature all point to military applications, not commercial, diplomatic, or peaceful purposes.¹⁹

In addition to the written evidence, we may have pictorial evidence for signaling on representations in the opening scenes of the panels on the columns of Trajan and Marcus Aurelius in Rome (see Illus. 2). The column of Marcus Aurelius is the less well preserved, but both show similar scenes, with a series of towers along the Danube River flanked by conical hay or straw ricks and a large woodpile. On Trajan’s Column the three watchtowers are portrayed with large torches projecting from the windows of each tower (Illus. 3). An early suggestion by Cichorius suggested that these were beacons prepared for action. The hayricks were designed to produce smoke plumes in the daytime, while the log piles would produce a bright light at night.²⁰ Webster went further and suggested that they might be parts of a semaphore system similar to that described by Vegetius.²¹ This suggestion has been repeated so often that it is usually accepted as established fact, but a more critical eye has been applied to this theory in recent years. It has been pointed out for example, that these flammable materials seem much too close to the tower itself to have been lit safely, and that they could be interpreted as piles of foodstores.²² David Woolliscroft, in his recent book on Roman military signaling, has pointed out that such equipment does not appear on all towers. He has gone further and actually performed experiments with log piles of various types of green and cured wood, and has discovered that such piles are “remarkably fireproof.”²³ Later scenes from Trajan’s Column show such structures being constructed from recently felled trees, whose wood would still be green and thus would not burn well.²⁴ Indeed, there are so many other uses that might explain these woodpiles that it is not necessary to turn them into signal beacons. They may be ballista bases, parts of temporary fortifications, or building platforms. Because they are often portrayed as having alternating layers, they may be woodpiles being seasoned for another use and may be close to the tower for security.

Illus. 2 Trajan's Column – the hay ricks. Reprinted by permission of the Library of the American Academy in Rome.Illus. 3 The signaling torches. Reprinted by permission of the Library of the American Academy in Rome.

The haystacks pictured in the scenes were initially interpreted as being used to ignite and create smoke by day or a very bright fire at night. Experiments have shown that such haystacks would bum with “remarkable ferocity.”²⁵ But just because they bum does not mean that this was their original function. First of all, they are pictured at ground level, which is not the optimum location for a signal beacon. Since only one of the towers on each of the columns is pictured with such a “beacon,” one wonders if this is really portraying a relay system. Woolliscroft has pointed out that the hayricks seem unnecessarily large when a fire basket mounted on each of the tower tops would have worked better, and their location is so close to the tower that it would have made them a fire hazard. In the end, they may simply be what they appear to be – hay bales used as animal fodder or bedding.²⁶ As for the torches being held out of the towers shown on Trajan’s Column, Southern has interpreted them as lamp-posts that lit up the tower’s surroundings at night.²⁷ Since all the towers face the Danube, perhaps they were used to guide shipping. It is not impossible that they were used for both purposes. As we shall see below, such beacon signals fit in well with some of the literary descriptions of Roman signaling techniques.²⁸

There are, as of yet, no archaeologically excavated beacon sites, but there have been a few suggested possibilities. Among the most promising are three structures found in association with Watchposts (WP) 1/47, 1/56, and 1/57 on the Roman frontier in Germany.²⁹ They consist of large post-holes surrounded by a ring of stone packing. It has been suggested that these features were built to carry fires laid round a central support. The original excavation report interprets them to be hayricks with central posts, like those on Trajan’s columns. But the excavations were done at the turn of the century and are not very thorough. We have no idea how many post-holes there were in total, and if they formed part of another structure. Indeed, Southern has pointed out that these may simply have been fires used to illuminate frontier crossings at night.³⁰

Another possibility is a burnt area found on a clifftop just outside of the Claudian coastal forlet of Martinhoe in North Devon, which has been interpreted as a beacon for communicating with ships at sea, but it could just as easily have been a bonfire.³¹ Since there is no stratigraphie proof that the area even relates to the Roman structure next to it, we cannot make any definitive statement about its use in Roman signaling.

In the Syrian desert, a number of circular rubble pads have been found, some of them located around towers. Although they may have been used as beacon bases, their function and date are uncertain. There are no bum traces, and they are only one stone thick, which certainly would not have provided height. It is also unclear why one would need a stone base for beacons when the surrounding countryside is neither damp nor flammable.³²

A potential multibeacon site was reported by Poidebard in 1934 on the Roman road to Palmyra in Syria. There he found four large standings – two on either side of the road. The stonework, however, encroaches on the line of the road, suggesting that the positions are post-Roman.³³

Our conclusion, then, must be that signaling by beacons was perfectly possible in Roman times, and it is the technique for which we would expect to find the most archaeological evidence. Yet, the fact remains that there are, as of yet, no conclusively identified Roman beacons. Ancient writers tell us the Romans used them, but their very nature makes them hard to identify on the ground, as it is almost impossible to distinguish them from an ordinary bonfire. The argument may be made, however, that watchtowers are good evidence for signaling, because, although they may have been used only as watchtowers, there is no point in having observers unless they can tell someone what they see. The question is how this communication system worked.

Codes

A number of ancient writers describe signaling techniques that attempt to get around the limitations of beacon signals by transmitting in full alphabetic codes capable of sending more complicated messages that were not prearranged. Such methods, were they ever used, would have had enormous advantages over the simpler beacon system. They would free the signaler from prearranged messages and allow him to send whatever message the particular situation required. More complex information could be sent faster than a messenger could carry it, especially over long distances.

However, more sophisticated signaling requires more and better trained soldiers to operate it. Alphabetic signaling requires literate operatives, who were not as common a commodity in the ancient world as they are today. Also, the more complicated the system the more probability that there would be a mechanical breakdown or human error. Since these more complicated systems are only described in theory by ancient writers, we have no way of knowing whether they were ever put into service. However, recent experiments in the field have shown them to be feasible.³⁴ For example, there is a technique described by the Greek writers Philon, Polybius (quoting Aeneas Tacticus), and Polyaenus. According to Polyaenus’s account, it was used by the Carthaginians in Sicily.³⁵ It involved using synchronized water clocks. Experiments demonstrated by David Woolliscroft on BBC television make a reasonable attempt to reconstruct such a system. Although at first doubts were raised as to whether the Romans had the precision engineering skills to build such a system and make it work reliably and quickly over long distances, Woolliscroft was able to build a smaller version of the apparatus and make it transmit messages accurately. This method of transmission is therefore entirely feasible, but we are still left in the dark as to if, when, or how the Roman Army may have used it in the field.³⁶

The oldest and most complicated code system mentioned by the Greek writers was transmitting cipher by torch combinations. The system is described by Polybius and is supposedly an improvement on a system invented by the Greeks, Cleoxenus and Democlitus.³⁷ It involved using two sets of five torches (designated left and right), which would be raised and lowered. In front of each group was a screen that hid the torches when not in use and allowed the receiver to see the light clearly when it was on and then obscure it when the torch was lowered. The code itself was a prearranged system of grouping the letters of the alphabet on tablets. Each tablet contained five letters except for the last one, which contained four. The first fire signal told which tablet was being referred to, the second torch told which letter on the tablet was being indicated. Using this system, a message might be sent in plaintext or further encoded. Since Polybius does not describe the mechanics of the torches themselves, the David Woolliscroft TV Programme experimented with two types of systems. One took a low-technology but labor- intensive approach that used simple hand-held torches; the second mounted a pivot bar supported by A-frames that could be used with or without counterweights to make them easier to lift. Both systems worked flawlessly. The torch method certainly works faster than any water clock system, but one must question the amount of manpower it would be feasible to devote to such a signaling system. And the A-frame set-up requires complex and cumbersome equipment, which, Woolliscroft admits, “must have lowered its usefulness to a point where one wonders if it could ever have had more than marginal military value.”³⁸ In the end, however, both systems are credible reconstructions of what the Romans might have done, even if we do not know whether either system was used in the field.

A simpler and perhaps more effective system is reported by the third-century ad writer, Julius Africanus, who reports that it was used in the Roman Army of his day.³⁹ This system, as constructed by Woolliscroft, uses “three torches designated left, right and centre” and each of them encodes a third of the alphabet each. Each torch represents eight letters. “The system uses varying numbers of flashes of the individual lights rather than combinations of different lights to transmit letters.” Thus, one flash on the left signals alpha, two flashes, beta, and so on.⁴⁰ This system, according to Woolliscroft, is much easier to set up, needs less manpower, and thus affords a greater possibility of range. His experiments yielded a transmission rate of 10–12 characters per minute, and this could be increased by rewriting the code so that the most commonly used letters were assigned the least number of flashes. One of the problems he first encountered was that the system had too poor a signal-to-noise ratio to be reliable. Woolliscroft experimented with reversing the system of Africanus and having the torches be visible all the time, and then lowering them behind a screen causing a momentary blinking out. He suggests that

Lateral inversion can then be cured by using separate transmitter and receiver codes, with left and right reversed, or even by senders simply turning their backs on the receivers to create a common left and right, although the latter method does make the checking of reception accuracy more complicated.⁴¹

Another problem to overcome is that a multitorch transmitter must be spread over a reasonably wide area in order for its separate lights to be resolved at any great range. This would make it difficult to mount on a tower. The width of a Roman tower top (3.5–4.5 m) would restrict the signaling range to 4km. If the Romans were required to signal from tower top to tower top at a range of up to 10 km, then this method becomes problematic. Still, Woolliscroft has taken the evidence and come up with a system that is actually functional at a 15-mile range, and with a bit of practice he could send readable codes without a glitch.⁴²

Julius Africanus also describes a flare system that operated by throwing fuel on a fire to make it flare up.⁴³ This would produce a very visible flare at night and a very visible plume of smoke during the day. Unfortunately, Africanus’s description is so unclear that we have no way of telling how the system worked. He does say that it would be used to transmit enemy numbers by sending one flare for every thousand enemy troops spotted. It has been suggested that chemicals might have been thrown onto fires to produce different colors, but there is no Roman evidence to show that they ever did this.⁴⁴

Semaphore

The fourth-century Roman author Vegetius and an anonymous Byzantine writer mention signals being sent by waving various objects, perhaps something resembling a modem semaphore. The Byzantine example suggests that battlefield standards were being used, but Vegetius is talking about something entirely different. He describes a tower-mounted beam that is moved up and down to transmit. Because he gives us no further detail, we are left to guess how the system worked. Woolliscroft has suggested the system used a binary code like a Morse system. The beam could have been made visible by projecting it from a tower at right angles to the observer, which would leave the torch silhouetted against the sky. As evidence, he sites the Roman tower at Barcombe B near the fort of Vindolanda, that are accurately aligned to face likely target sites.⁴⁵ He believes Vegetius’s method may have been put to use there. This might also explain the beacon lights protruding from the towers on Trajan’s Column.⁴⁶

Hadrian's wall: the stanegate phase

Before the construction of Hadrian’s famous wall in c. ad 122, a system of frontier control had already developed on the Tyne–Solway isthmus in Britain under the Emperor Trajan (see Map 20). The Stanegate is the name which, since medieval times, has been applied to the road that linked Corbridge and Carlisle (Map 21). It has been recognized that this road represents a stage in the development of the northern frontier that occurred sometime between the withdrawal of the Romans from southern Scotland in ad 100 and the inception of the fort phase of Hadrian’s Wall c. ad 125, at which point most of the forts on the Stanegate were eventually abandoned as redundant.⁴⁷ The Stanegate frontier probably consisted of alternating auxiliary forts and fortlets, and while some of the former probably dated back to Agricola’s governorship, the system as a whole came into operation in the very early years of the second century. Controversy still surrounds the exact date, form, or in some cases even the existence of this system as a formal frontier line.⁴⁸ If we accept its existence, however, we are better able to understand the signaling possibilities of this region.

Map 20 Defensive installations in BritainMap 21 The Stanegate Limes

What the Stanegate frontier represents is a defensive line used after the Roman Army pulled out of Scotland during the reigns of Hadrian’s predecessors, Domitian and Trajan.⁴⁹ The highway began to acquire an increasingly militarized appearance, and by Hadrian’s reign may have become a powerful frontier system in its own right. Only now, with more excavation and study, is the size and complexity of this system becoming apparent. Recent discoveries by aerial photography in the last few decades have detected a number of forts indicating that the system did not run just from Carlisle in the west to Corbridge, as was previously thought, but may have run from coast to coast, as did Hadrian’s Wall itself.⁵⁰ The frontier system in its final form seems to have been made up of four elements strung out together in a fairly regular pattern. Along the road itself are a series of large turf and timber auxiliary forts at roughly 7-mile intervals.⁵¹ Between each pair of forts was a much smaller fortlet measuring a little over three-quarters of an acre in size.This means the intervals between major installations are cut to about 3.5 miles. Woolliscroft reports only two such fortlets that are currently known with certainly: Throp, between Nether Denton and Carvoran, and at Haltwhistle Bum, between Carvoran and Vindolanda. Aerial and field evidence also suggests a third location at Castle Hill, Boothby, between Old Church Brampton and Nether Denton.⁵²

Could the elements of the Stanegate have been connected by a signaling system? According to Woolliscroft, yes. Some of the forts themselves are intervisible, and the rest are connected by a series of watchtowers. There is a difference between a signaling post and a watchtower with a capacity for signaling. A signaling post that functions simply as a link in a chain is positioned so as to be able to see its two neighbors. Its distance from them is decided by the maximum distance over which visual signaling is possible in normal conditions. To have more stations than is absolutely necessary slows down the transmission of a message.⁵³ As Woolliscroft has pointed out, these signaling towers do not occur at frequent enough intervals to form a regular observation screen, as they do on other frontiers, but rather occur only where a signaling link between two major sites is required. As seen in Map 22, Woolliscroft has mapped out the signaling links in the system, and they seem to work at least in the places where installations on the ground have been identified.

Map 22 The Stanegate – signaling.

It has been postulated that once the construction of Hadrian’s Wall began, the Stanegate system was incorporated into the design. If the wall was designed around a comprehensive, Stanegate-oriented signaling system, then all the wall installations, including the milecastles and turrets, should be directly intervisible with a Stanegate fort, fortlet, or signal tower. Any blind installation should be able to relay a signal to the Stanegate by a neighbor. And this is exactly the pattern Woolliscroft detected. The data suggests that each Stanegate installation had responsibility for a particular length of wall. It would seem that Hadrian’s Wall did have an efficient and comprehensive signaling system in its initial phase.⁵⁴

The Stanegate frontier was a defense against invasion that depended on auxiliary forces which could be combined to form a single substantial army. They could patrol the road, while the fortlets and towers allowed for reasonably tight surveillance of the area. The system could still have been easily infiltrated by small raiding parties, smugglers, or subversives, especially at night or in bad weather, but the signaling system would have allowed for surveillance of such movement and given enough advance notice to the defenders for them to respond.

The construction of Hadrian’s Wall in the 120s then is seen by scholars who have detected the earlier Stanegate system as a way of creating a more intensively policed line. It would have enabled the Romans to monitor these low- intensity threats without having any further dispersal of manpower. The wall was intended originally as only a thickening of the system, and not a replacement for it.⁵⁵ This does not mean that the system was built to be impermeable. It was vital that the exploratores, for example, be able to move through it for interception purposes. Even civilians could pass, with Roman permission, for reasons of trade or transhumance through the milecastle gates. The gates had to be frequent enough along the wall for convenience, and yet secure enough to allow close scrutiny of travelers or oversee trade, and to collect the import and export duties that contributed to the Empire’s income. What has puzzled scholars is why some of these gates stand on top of massive cliffs that would not have facilitated passage, or areas where accessibility was extremely difficult. The milecastles are not located at regular Roman-mile intervals, so we cannot argue that some surveyor was told to put a structure at every mile interval and then slavishly followed the instructions in spite of the terrain. It makes much more sense that the milecastles were sited for intervisibility and signaling.

Hadrian's wall: the wall fort phase

Most readers are familiar with this famous Roman structure, but for those who think of it simply as a wall, a closer description is necessary. In reality, it was a 76-mile stone wall, approximately 10 feet thick and although 11 feet high at the tallest surviving point, was probably 14–15 feet, possibly with a breastwork on top of that. It ran the entire length of the relatively narrow Tyne-Solway isthmus, from Wallsend on the northern bank of the Tyne, east of Newcastle, to Bowness on the southern shore of the Solway (see Map 23).⁵⁶

Map 23 Hadrian's Wall.

In front of the wall is a large V-sectioned defensive ditch, backed both by a road and a massive earthwork known as the Vallum. This consists of a flat- bottomed ditch flanked on either side by parallel earth ramparts.⁵⁷ At intervals of one Roman mile there are gateways guarded by small fortlets called milecastles. Between each milecastle is a pair of watchtowers or turrets. Because milecastle gateways probably also had towers, the wall probably had a screen of observation posts set at intervals of one-third of a Roman mile. Finally, at spacings of between 4 and 7 miles, the wall carried a series of 3–9 acre forts to house the noncitizen auxiliary units of 500–1,000 men each.⁵⁸

A great deal of ingenuity has been used by scholars to figure out how a signaling system would have worked laterally along Hadrian’s Wall (see Map 24).⁵⁹ The milecastles and turrets are not always intervisible with their neighbors. And why would they be signaling to others on the wall anyway, since in the beginning the lightly manned wall had no real reinforcements to send. A wall installation faced with trouble would have wanted to summon help from the nearest concentration of force from the nearest fort. At the time the wall was built, this would have meant the forts on the Stanegate to the south. As we have seen, Woolliscroft believes the deviations in milecastle spacings are consistent with an attempt to fit Hadrian’s Wall with a comprehensive signaling system oriented to the south. All of the oddities in the placing of milecastles on the wall are explicable in terms of this theory. The data suggests that in the siting of milecastles, the need to signal took priority over their usefulness as transit points.⁶⁰

Map 24 adrian's Wall – signaling, the wall fort phase.

As the wall was being completed, the system was subjected to a major redesign in which a new series of forts on the line of the wall itself was built to replace the older bases at Stanegate. The abandonment of the forts around which the signaling system had been designed would have forced the Romans to change the orientation of the signaling patterns. The Romans seems to have done rather well. The new forts would appear to have been deliberately sited in highly visible positions because, unlike the Stanegate forts, they were frequently built on hilltops. It is possible to map out a signaling system with the wall forts in place if one allows for relaying some of the signals through a second wall installation before they arrived at a fort or strategic tower.⁶¹

Hadrian's wall: the outpost forts

Roman control of the Tyne-Solway frontier did not stop at the wall. Part of the intelligence screen included outpost forts, of which five have been identified, at Birrens, Netherby and Bewcastle in the west, and Risingham and High Rochester in the east. These were large and obviously important forts, and each of the western forts had Cohors milliaria equitata, units which had 1,040 men, 240 of which were cavalry. These were the largest auxiliary units in the Roman Army, and as mounted troops they were ideal for conducting long-range reconnaissance. The more clandestine intelligence work may have been done by the exploratores and there is considerable evidence for these units at the outpost forts.⁶²

Given the function of outpost forts, it would seem to have been necessary for them to have signaling links back to Hadrian’s Wall. Recent archaeological work has discovered such links between the wall fort at Birdoswald and Bewcastle via two signaling towers, one on the hill called Gillalees Beacon and the other on Barron’s Pike (see Map 25).⁶³ The road between the installations on the western side, between Bewcastle and Netherby, has yet to be found, although Birrens and Netherby were linked by a branch of the Netherby-to-Carlisle road. A tower on Barron’s Pike opens up the possibility that there were lateral signaling links between these outposts.⁶⁴ Barron’s Pike tower also has direct intervisibility with the wall and the Stanegate fort at Carvoran.

Map 25 Hadrian's Wall-the outpost forts.

Cumberland coast defenses

The Cumberland coast defenses are part of the same system as Hadrian’s Wall (see map 20). They show design similarities and their signaling arrangements may be similar. The two share the same regular installation pattern, because the coastal mile fortlets and towers correspond exactly to the milecastles and turrets on the Wall line. Both have a series of forts, set at similar intervals, and in similarly prominent positions. The only thing missing was the wall itself, although there have been claims that the coastal works had a running barrier of its own (see Map 26).⁶⁵ There are significant gaps in our knowledge, however, which make this sector difficult to study. Unlike Hadrian’s Wall, the coastal defenses were built on “a truly rigid spacing system.”⁶⁶ The sites are almost always set apart by exactly one-third of a Roman mile. This system did not use the spacing shifts we see on the wall to facilitate signaling. The course of the line itself, however, seems to have been chosen with signaling in mind. Most of the coastal forts are sited in highly visible positions. Some of the frontier fortlets and towers are right on the foreshore, but others are significantly set back from the shoreline, which does not give them an optimal view of the sea. Instead, they are sited to be able to signal back to the forts.

Map 26 Cumberland coast defense

We are hindered in our understanding of this system because, unlike Hadrian’s Wall, we have yet to detect the full extent of the system. Nevertheless, Woolliscroft has constructed the most recent hypothesis of how a signaling system would have worked in this sector based on the most recent archaeological evidence and intervisibility studies (Map 27).⁶⁷

Map 27 Cumberland coast – signaling system

Stainmore Pass

In 1951, Sir Ian Richmond proposed the existence of a long-range Roman signaling system between the fortress of Legio VI at York and the presumed headquarters of Hadrian’s Wall at Stanwix. Seven structures along the Roman road over Stainmore included what he identified as two fortlets – Castrigg and Maiden Castle – and five signaling towers – Vale House, Bowes Moor, Roper Castle, Brackenber, and Barrock Fell.⁶⁸ This system has since been rejected. No two sites are of the same design, and in addition there are intervisibility problems.⁶⁹

At the time of Richmond’s publication only two of the sites, Maiden Castle and Barrock Fell, had been excavated sufficiently to provide a date, and only forty years later has excavation provided more evidence (see Map 28).⁷⁰ Since then, Bowes Moor was excavated, showing that it is not contemporary with the original two sites.⁷¹ More recently, aerial surveys by Higham and Jones have identified some structures on the central Stainmore that resemble first- or early second-century timber watchtowers.⁷²

Map 28 Stainmore pass

The excavations at Johnson’s Plain represent the first near total excavation of a Stainmore tower to be carried out. This excavation has revealed much about the site and this information may be applicable to other sites in the area. Some of the groups may be Roman signal and/or watchtowers, but the total absence of datable material makes sure identification difficult.⁷³ To date, not even Woolliscroft has been able to come up with a convincing model for their use.

Gask ridge system

The earliest land frontier in Britain, built in the ad 70s, fifty years before Hadrian’s Wall and seventy years before the Antonine Wall, was the Gask Ridge frontier system in Perthshire, Scotland.⁷⁴ This consists of a chain of fortifications that ran along the Roman road between the forts of Camelon (Falkirk) and Bertha on the Tay. The study of the signaling system along the Gask Ridge in Strathearn began at the turn of the last century, when D.J. Christison excavated Kaims Castle and eight watchtowers on the ridge.⁷⁵ None of these sites yielded any dating evidence. Another fortlet was excavated in the 1980s at Glenbank to the south of Ardoch,⁷⁶ and evidence for new towers keeps emerging, bringing the known or suspected number to twenty. Excavation continues at the present time (see Map 29).⁷⁷

Map 29 Gask Ridge installations.

The fortifications in this system are turf- and timber-built forts, plus smaller fortlets and watchtowers, all of which lie close to the Roman road. The three Roman forts on the line are Ardoch, Strageath and Bertha; they are auxiliary forts capable of holding a complete unit. The smaller fortlets are comparable in size to the milecastles on Hadrian’s Wall. The three at Glenbank, Kaims Castle, and Midgate are known for certain. The smallest and most common type of structure in this system, however, is the tower, which may have been used for signaling and observation. There are currently eighteen known and certainly more waiting to be discovered.⁷⁸ This system was, therefore, a fortified and carefully monitored road that protected the potential invasion route down Strathearn.

There has been much debate over the exact dating of this system. An early date would make it part of Agricola’s original installations, which were later replaced by glen-blocking forts.⁷⁹ Others support a date in the ad 80s, during the reign of the last Flavian emperor, Domitian, forty years before the construction of Hadrian’s Wall. This makes it the earliest frontier in Britain. For many years it was thought to be slightly later than the similar line in Germany, but the Germans have recently redated their frontier to the reign of Trajan (ad 98–117), and this would make it fifteen to twenty years later than the Gask system. The Gask system thus becomes the prototype Limes.⁸⁰

The system as we know it today seems to be missing many elements, but for the most part the majority of the installations along the Gask Ridge are intervisible with both of their immediate neighbors. Some of the elements are not intervisible, which suggests that there are more towers that need to be discovered.⁸¹ The system may also have continued further to the south, beyond Doune perhaps, to the Flavian fort at Camelon.⁸² Despite the gaps in our knowledge, it has been possible to reconstruct at least a skeletal signaling system for the Gask Ridge based, as with Woolliscroft’s other studies, on each installation’s field of view (see Map 30).

Map 30 Gask Ridge signaling.

Woolliscroft has hypothesized that there existed on the Gask Ridge a direct signaling system of the sort already found on Hadrian’s Wall and the northern part of the Wetterau Limes.⁸³ The positioning of the towers is always the clue. Many of the ridge-top towers lie behind the apparent optimum lookout line. If the Roman frontier garrison was supposed to be primarily interested in what was happening to the north, then why of the eleven known towers on the ridge itself do only two, Raith and Midgate, have a completely unimpeded view in this direction, even from the top of the tower? This layout would be potentially dangerous for a system so lightly manned, if it were specifically designed to control movements along the actual line from the north.

Woolliscroft’s answer is that the towers were “designed to look east/west along the system rather than to the north.” Anything coming from the north would have had to cross the Gask Ridge itself, an area where few modem roads cross, even today. (There are two minor north-south roads, one of which is only a single track.) It is telling, therefore, that all of the towers on the Gask Ridge, no matter how poor their views north, have superb views south over Strathearn, and this view would have been seriously compromised had the sites been positioned to gain a better view to the north.⁸⁴ The frontier’s intended function seems to have been to monitor the valley. The forts were designed for a defensive role, not as jumping-off points for an assault into the Highlands. The Gask Ridge is far enough back from the Highlands to give an hour’s warning of any attack. The Gask would therefore be “a back stop frontier and flank guard” to which the glen-blocking forts farther north could have acted as outposts, positioned as they were in the mouths of the passes likely to be used by invaders from the north.⁸⁵

It has been argued, by Donaldson for one, that because the Gask Ridge sites are so closely spaced, the signaling techniques employed on the system must have been limited to a range of one Roman mile.⁸⁶ But experiments conducted by Woolliscroft suggest that Roman visual methods were capable of much greater ranges. The important point here is that although the Gask Ridge system may have been used for signaling, it was not just a signaling system. The tower spacings make perfect sense as a line of observation posts. The overlapping fields of view would have allowed for tight surveillance and enabled the Romans to watch over every inch of the line. Signaling may have formed only a minor part of their role.⁸⁷ The role of these small frontier installations was to provide early warning rather than direct defense, and also the transmission of intelligence in an efficient manner. This system, then, resembles that on the Wetterau Limes.⁸⁸ There, too, many of the towers have a limited view to the north, but have superb views over their hinterland. Any invader entering the Fulda Gap (the most likely invasion route) could be monitored and conspired against from a relative safety distance in the surrounding hills, and the Limes could act, to some extent, as a defense in depth.⁸⁹ One would expect the Romans to have maintained intelligence cover ahead of their lines, in order to give advance warning of any major attack. In a fluid condition such as existed in Scotland in the Flavian period, any intelligence breakdown could put the Romans in jeopardy. An early warning system is only useful if it provides enough time for effective countermeasures to be taken. The Gask Ridge was far enough back from the Highlands to give an hour’s warning of an attack, but it was still close enough to provide a view of what was happening.

Antonine wall

Hadrian’s Wall was still undergoing extensive renovations when the Emperor Hadrian died in ad 138. The new emperor, Antoninus Pius, immediately decided on a new forward policy in Britain and sent a new governor, Quintus Lollius Urbicus, with orders to reoccupy southern Scotland and to construct a new wall across the Forth-Clyde isthmus (see Map 20). Moving the frontier of the province 100 miles further north implies to some that Hadrian’s Wall was not a success.⁹⁰ On the other hand, its replacement by another wall may suggest that Hadrian’s Wall had been a success but that it had become out of touch with the main centers of resistance to the Romans. The move north was carried out with enormous speed, and preparations began in ad 139.

The construction of the wall is the key to its function. The Antonine Wall at first seems so different from Hadrian’s Wall that it was thought unlikely that they operated in the same way. Although both walls share a basic framework of a rampart and a ditch, there are differences also in the garrison’s installations. The interfort spacings of the Antonine Wall were much shorter, varying from 1.56 to 3 Roman miles, as against 6 to 9 on Hadrian’s Wall.⁹¹ There are seventeen forts and several suspected forts.⁹² Many of these forts were too small to have held a full auxiliary unit, and there seems to be no sign of a fortlet and turret system (Map 31). But we do have enough fortlets (nine) to say there was probably a milecastle system.

Map 31 Antonine Wall – phases 1 & 2.

J.P. Gilliam presented a hypothesis that the Antonine Wall was initially designed as a copy of the secondary, wall-fort phase of Hadrian’s Wall. His theory also suggests that the system was originally held by just the six largest forts: Carriden, Mumrils, Castlecary, Balmuildy, Old Kirkpatrick, and one other.⁹³ The remaining forts would belong to a secondary phase. Some may have been planned by the time the wall was under construction. Gilliam suggested that a mile fortlet and even a turret system might come to light. In fact, the discovery of the fort at Inveravon and additional fortlets add credence to this theory.⁹⁴

Our knowledge of the basic anatomy of the wall is not very good. It has been predicted that there may be around 30 fortlets and possibly one or two forts still remaining to be discovered. We still do not know if the wall had a turret system, which means about 80 separate installations may be missing. There are fortlet gate towers, but they only occur at one per mile and that seems inadequate for a signaling system. Towers are such a universal feature of Roman frontiers that it should seem inevitable that they be found on the Antonine Wall. The likelihood of 84 stone towers remaining undetected seems slim. Wooden towers, on the other hand, seem possible, and they would be undetectable by aerial or surface observation unless they were surrounded by ditches. Such arguments from negative evidence remain weak, and thus the case for a turret system remains unproven.

In spite of these limitations, Woolliscroft has undertaken a survey, expecting to find a two-phase system as he had on Hadrian’s Wall.⁹⁵ This does largely seem to be the case. The forts can be easily linked together to form a unified chain and none requires more than two relays to communicate with either neighbor (Map 32). Of the nine known fortlets only one, Cleddans, is out of sight of a primary fort.⁹⁶ Of the 30 sites listed by Woolliscroft, only three are not in sight of a fort, although three more are not in touch with their closest fort. This leaves at most 8 out of the 39 fortlets (20 percent) requiring relays, each of which could have been achieved using just one other site. Because these relays could have been accomplished by other fortlets, it is not necessary to postulate a turret system. The other minor installations, whatever their function, can also be integrated into the system.⁹⁷

Map 32 Antonine Wall – signaling, phase one.

In the secondary system (Map 33), the introduction of new forts may have made some of the earlier sites redundant. It has often been assumed that the remaining fortlets went out of use, but a few have produced signs of modification rather than abandonment.⁹⁸ Sites such as Wilderness Plantation and Seabegs Wood may have remained in use because they were important signals relay stations. The secondary forts are mostly sited in highly visible positions on or near hilltops that are all intervisible with at least one neighbor. The signaling of the second-phase system was reduced to an interfort chain, but it remained secure and efficient. This arrangement is highly speculative because it is based on some locations that are yet to be proven. This guess, however, appears both plausible and consistent with other Roman frontiers.

Map 33 Antonine Wall – signaling, phase two.

Intelligence on the British frontier

The success of the Roman frontier system in Britain depended on a dispersal of forces along the border to permit intensive patrolling, observation, customs cover, and the scrutiny of persons wishing to cross. Such a diffuse deployment, however, would be extremely vulnerable in the event of a major attack. Large concentrations of forces had to be posted at strategic centers that would be able, at short notice, to intercept and destroy any external force or at least keep it at bay until help could be brought up from the legions stationed further to the south. The success of a flexible compromise between these two needs would have been dependent upon an intelligence system. If their intelligence was functioning properly, the Romans should have been able to launch preemptive strikes and fight back any large levy of barbarian troops before it had time to cause the Romans trouble. If Roman intelligence failed to detect signs of unrest or failed to warn the frontier forces, the dangers inherent in what was still a relatively diffuse troop deployment could be catastrophic. A provincial army that got caught with its component units dispersed might be swept aside.

Just such an occurrence seems to have happened in Britain in ad 367, during what is called “The Great Barbarian Conspiracy.” Roman control of the province was restored only when a relief force of field army units was sent over from the Continent under the command of the Count Theodosius.⁹⁹ We are told by the fourth-century historian Ammianus Marcellinus that Theodosius disbanded a force of locals called the Areani, who had been used by the Romans as a scouting force.¹⁰⁰ They had been paid off by the enemy however, and betrayed the Roman Army, which was caught off guard. This shows, once again, the danger of using locals for intelligence gathering. This is the same problem that plagued them in Germany under Varus. Had they used Roman exploratores, the betrayal would not have happened.

Wetterau Limes

Rome’s oldest continental land frontier passes through the provinces of Germania Superior and Raetia (see Map 34). The German Limes has been studied almost as thoroughly as Hadrian’s Wall, and enough information has been gathered for us to guess about the signaling arrangements. Almost all of the thinking on this system on the German Limes has been focused on the minor installations that include towers and small fortlets called Kleinkastelle. The small fortlets resemble the milecastles of Hadrian’s Wall and may have played a broadly similar role. But there are some significant differences. For example, unlike the milecastles, the Kleinkastelle lie slightly behind the frontier line and are not usually provided with a passage through it. They are also set at longer and much less regular intervals than their British counterparts. Unlike the milecastles, they do not appear to have been part of the original frontier design and some, at least in the older parts of the line, are built on top or alongside preexisting watchtowers. Not all of them are of uniform design and some of them are so big, they should not be classified as this type of structure at all.¹⁰¹

Map 34 The German limes

Not enough excavation using modem methods has been done along the Limes, and thus the best we can do is guess as to the dates of these monuments. The bulk of the Kleinkastelle seem to be of Trajanic or Hadrianic date (second century ad), although several may be earlier or later.¹⁰² It is probably not possible to trace the signaling system by discussing the development of the frontier itself. At best, one can study the Limes as it existed in its final form in the mid to late second century ad. Woolliscroft has studied a sector that stretches from the northwestern Wetterau north of Frankfurt to the fort known as Gross Krotzenburg on the north bank of the River Main (see Map 35).¹⁰³ This sector was chosen because it was an area large enough to be characteristic of the whole, and also because it was one of the few areas free of modem forests. The sector is 63 km long and contains ninety-six separate installations, including eight forts and fourteen Kleinkastelle.¹⁰⁴

Map 35 German Limes – study sector.

The frontier is arranged in the usual Roman fashion. All of the installations are arranged a little behind the line and as far as we know, there are no outpost forts on this part of the Limes. The line itself consists only of a road, later reinforced with a timber palisade, and later still with a ditch-and-earth rampart, now known as the Pfahlgraben.¹⁰⁵ The forts stand at reasonably regular intervals of about 6–7 km (4.5 Roman miles) apart. Unlike the turrets of Hadrian’s Wall, however, the watchtowers have no regular spacing system and their intervals may range from 500 m to more than 1 km. The most important fact is that they are always positioned (unlike Hadrian’s Wall) so as to be intervisible with both their immediate neighbors, and this seems to have been important enough a consideration to cause at least some of the spacing variations. Scholars studying this frontier have assumed that signaling was carried out laterally, with messages being relayed tower to tower along the frontier line.¹⁰⁶ Woolliscroft reports a model in the Limes museum at Aalen showing an attack on the Limes and a watchtower signaling directly back to a fort. Such a system would have been perfectly possible on the German Limes, although there are no written sources to tell us how the frontier worked, and as signaling procedures leave only the most tenuous of physical traces, there are obvious limitations to a purely archaeological approach.¹⁰⁷

What we must guess at is what would constitute good military practice. To be effective, any military communications system must fulfill certain basic criteria. First, it must work. Secondly, and most ideally, it should be as efficient as possible, which means it should transmit as much information as possible as quickly and as reliably as possible while using as little skilled manpower, energy, and capital equipment as possible. A military system should be as invulnerable to attacks as possible. This means that even more than a civilian system, it has to protect itself against human and equipment failure, enemy attack, and should contain fail-safe mechanisms that will allow the system to survive the destruction of even some of its component parts.¹⁰⁸ Individual sites can be reinforced, or they can be located in inaccessible places. The greater the redundancy or duplication a well-designed system contained, the greater the level of punishment the system would be able to withstand before it would fail, and so the more effective and reliable it would be under combat conditions.

Northern Zone

The section of the Limes studied by Woolliscroft breaks down into a northern zone and a southern zone. The possible signaling systems in these two areas differ greatly. The signaling in the northern zone appears to be a direct signaling system similar to the one on the Stanegate phase of Hadrian’s Wall (see Map 36). They both have forts set well back from the line itself and almost every minor installation in the zone enjoys direct intervisibility with a fort. The only obvious difference is that the German version lacks the Stanegate – Hadrian’s Wall system of rearward relay towers. Even so, only five of the zone’s fifty-one smaller installations require relays to signal a fort.¹⁰⁹ In every case the signal need only be relayed via one other site, and none of the five are dependent on just a signal relay site. All of them have a choice of at least two or more other sites that can be used (see Map 37). This safety feature means that none of the installations can be cut off by the destruction of any one site. As has been pointed out, this frontier was threaded in and out through undulating country and low hills in such a way as to facilitate communications while still providing most of the watchtowers with suitable viewpoints and keeping the line tactically strong as a whole.¹¹⁰

Map 36 The Wetterau study sectorMap 37 The Wetterau signaling system – norther zone

Woolliscroft argues that the northern zone is also well designed for facilitating communications between the forts themselves, although he points out that unlike the Stanegate and the wall fort phase of Hadrian’s Wall, none of the Limes forts are actually intervisible. However, all of the northern zone forts can communicate with their neighbors by means of single relay sites. There also seem to be very clearly defined spheres of responsibility for each of the forts. The site that serves as the relay link between the two forts stands on particularly prominent ground, which blocks any view past it from the sites in either direction. The relay thus becomes the only site common to both the fort’s field of view and it is sometimes the case that no other installation in one fort’s sector can see any of the sites in those neighboring forts. The northern zone thus consists of linked but otherwise almost independent building blocks that include a fort and a distinct set of frontier installations. These blocks are tightly integrated internally, but each is connected with its neighbors at only one point. Each Kleinkastelle stands on such a prominent point that no other installation within the sector, including the fort, can see past them into either of the blocks beyond.¹¹¹ If, as I have suggested, that redundancy is a plus in a signaling system, then the set-up in this area of the Limes would seem to increase the system’s vulnerability to attack. Here the interfort signaling chain contains no redundancy as only the relay Kleinkastelle can see into two blocks at once. Perhaps the advantage is that no installation is allowed to see anything for which it is not responsible.¹¹² The Kleinkastelle not only provide all the links in the interfort signaling chain, but Woolliscroft suggests that these sites may have protected the passages through the Limes, which gave the forts’ garrisons access to the outside world, and they may have facilitated civilian traffic through the frontier.¹¹³ In other words, the very fact that the relays were kleinkastelle is not a coincidence. They used these more powerful installations on key positions to keep the system going if the relay was destroyed.

Southern Zone

The southern part of the Wetterau Limes does not have as many parallels with the Tyne-Solway frontier in Britain. The signaling system in the south seems to resemble a lateral system, where three times as many installations cannot see a fort as on the northern sector (Map 38). This is probably due to considerations of terrain. The Limes in this sector crosses a series of four river valleys roughly at right angles and all four forts stand on low ground close to the rivers. Many of the minor installations stand on higher ground on the valley sides and can see down to the forts. One might point out that with little effort, this system might have been built with better tactical position had the line not been so straight. Had the Limes been allowed to follow a stronger, more twisting course to exploit the terrain as it does further north, then more sites could have been brought into positions that were intervisible with a fort. Another consideration is that unlike the northern sector, all the forts in the southern zone are built close to the frontier line.¹¹⁴ How, then, do we explain this set-up?

Map 38 The Wetterau signaling system – souther zone.

Woolliscraft gives a possible explanation by a comparison to the wall fort sites, where the forts are set well back from the line, usually on high or at least highly visible ground.¹¹⁵ This ensured that a maximum number of frontier installations could see them. When the wall forts were built on the line of Hadrian’s Wall, a large number of sites lost their direct view of a fort, and the configuration and problems of the southern Wetterau zone seem to be the same. Just a few changes could have shrunk the number of installations need for signaling relays. As a long-range signaling system, the area’s weaknesses are obvious. None of the forts is even intervisible. Unlike the northern zone, only one of the interfort links is accomplished by means of a single relay.¹¹⁶ The neutralization of any of these relay sites would have cut the interfort chain and made it vulnerable to disruption by enemy action.

The relay installations in the south are generally not Kleinkastelle, that is, fortified positions. They seem to be, for the most part, ordinary watchtowers. Indeed, there seems to be no logic at all behind the siting of the few Kleinkastelle to be found in the south, as they do not appear to fulfill any of the roles marked out for them further north. It is more difficult to discern clear borders between the spheres of responsibility of the forts in the southern zone.¹¹⁷ If the neutralization of any of these relay sites could cut the interfort chain and disrupt the system, as I have said, then how can one argue that this is a signaling system? And why, after having found such a logical set-up in the northern sector, should the southern sector be so different? These two Limes zones existed side by side and were contemporaneous. Do the two zones reflect two different designers or military units working within a loose overall plan? Because no evidence survives of the original builders, we have no way of testing this theory. And the evidence seems to suggest that whatever forces caused the emergence and concurrent operation of two different frontier designs on the Trajanic Wetteau Limes, there were still operating a generation later in the Antonine period.¹¹⁸

Perhaps, however, the wrong question is being asked. Maybe the Romans did not intend merely to make the lines short and efficient. The Wetterau loop does have the military effect of bringing a higher troop concentration into the area. The greater the length of the line, more would be the men and installations needed to guard the area. By having the Limes loop around the Wetterau rather than running straight across it, the Romans purposely made the line roughly twice as long as it need have been, and so there were twice as many forts and twice as many men in the area as would have been the case without the loop. The intervals between the forts in this area are shorter than almost anywhere else on the line, and a greater proportion is filled with full cohort or ala bases rather than the Numeruskastelle, which abound elsewhere. Behind this densely held line lay still more manpower in the form of the legionary fortress at Mainz and the pre-Limes fort of Friedberg, which both remained in use during the period. The shape of the loop means that any raiding party that penetrated the Limes from the north would still have to run a prolonged gauntlet between the Wetterau forts once inside.¹¹⁹

So how do these strategic considerations affect the signaling system? If we assume that the Romans would not deliberately build an inferior system, let alone continue using it if a better one were available at no great effort, then we must assume that there were two different signaling systems working side by side – one in the north and one in the south. The direct system in the north may be preferable, but it is also likely that the lateral system in the south was all that was possible. Woolliscroft has suggested the following reconstruction. The one potential weakness of a direct signaling system like the one in the north, is that its successful operation requires a condition that a lateral system does not. Because its signals are passed back from as well as along the frontier, a direct system operates in two dimensions and, as Roman signaling depended on clear lines of sight, this meansthat the system could only function in open country. Once there is the presence of tree cover, the direct system breaks down unless one is willing to deforest the area, which would take too much time and effort. The Romans may have been faced with the choice of clearing the trees or changing the signaling system. Unlike the direct system, the lateral system operates in one dimension only, because all signaling takes place along the line of the frontier. It has always been assumed that a corridor around the line would have been kept clear of trees so as to remove cover and enhance the view of the watchtowers. This means that so long as the frontier and its installations were confined to a narrow and absolutely straight line, the presence of forest was irrelevant. The installations would be able to see down this linear clearing and thus pass their signals freely along it. A frontier system adapted to forested conditions should then exhibit a number of characteristic features. First, there should be a straight line path of the frontier, and secondly, the close proximity of all installations, including the forts to the line itself. Thirdly, where the Limes does have to bend, it will do so at a fort, so that the fort has the best possible view along both legs of the line. Fourthly, to allow the relaying of signals, all installations should be intervisible with their immediate neighbors, no matter what effect it has on spacing. This is exactly the pattern we have in the southern zone of the Wetterau study sector.¹²⁰ The layout of the number of other parts of the line may be explained as an attempt to cope with dense forest. Although the immediate vicinity of the frontier was kept open in Roman times, the forts and frontier installations would have had to be able to see and exchange signals with each other. When we set the Limes in its environmental context, the signaling system in the southern sector makes much more sense.¹²¹ Woolliscroft points out that in areas outside of this study sector, where environmental data has been collected, heavily forested areas also seem to have long straight sectors of frontier with linear signaling systems.¹²²

The bulk of the signals traffic on a Roman frontier would probably have been between the individual line installations and their nearest fort. Communications of this type are inherently local in nature and are thus little affected by large-scale considerations. One might ask whether there was a trans-Wetterau signaling system. The ability to transmit across the Wetterau loop would have been a distinct advantage to the frontier’s long-range, interfort signaling system. Interfort signaling would have enabled large areas of the frontier to be brought rapidly to a state of alert in times of great crisis. Woolliscroft shows how such a system would be set up and that the whole of the northern Wetterau could be alerted from just two installations, and further that cross-loop signals could be initiated by all but one of the forts in the area. However, we have no evidence whatsoever for cross-loop signaling being used in practice, since signals protocols leave no physical trace.¹²³

Upper German-Raetian border

Woolliscroft and his team studied another stretch of the German Limes to the southeast in Baden-Württemberg. It is a 69 km stretch from Watchpost (WP) 9/120 north of Welzheim to WP 12/107 near the fort of Halheim (Map 39). It was chosen because the stretch is long enough to provide a representative sample and it is largely free of modern forestry. The bulk of the line lies in the Roman province of Raetia, and so it provides the opportunity to observe a different Roman province. Unlike the Upper German Limes, however, the frontier in Raetia was eventually given a stone wall, although one lighter in construction than Hadrian’s Wall. Woolliscroft assumes that the start of the so-called Teufelsmauer (Devil’s Wall) in the Rotenbachtal, near the fort of Schirenhof, marks the exact position of the provincial boundary. By comparing the two German sectors, we also get to see the Limes’ evolution over time, because the Raetian Limes was built at least a generation later, in the ad 150s, and by comparing the two any significant change in operational priorities might be detected in the design.

Map 39 The German/ Raetian Limes

As with the Wetterau region, the Raetian study sector contains a layout of two different patterns. The same elements (forts, towers, and Kleinkastelle) are still present, but the fort intervals are longer and less regular.¹²⁴ The bulk of the sector is laid out like the northern Wetterau zone, with forts set well back and the line making sound tactical use of the terrain. The first 8 km, however, runs perfectly straight, with its forts set right behind the line.

Like the northern Wetterau zone, the forts have been sited to enjoy a good view of the line, which often runs against their skyline. The line sites themselves are almost always sited to have a good view back from the Limes, even if this means having little or no view forward. In spite of the difficult terrain through which it passes, the frontier has, in Woolliscroft’s words, “been elegantly laid out to facilitate signaling,” for the entire sector comes close to a true direct system, although the terrain in some cases makes such perfection impossible.¹²⁵ At Aalen, for example, a complex series of hills and stream valleys makes a direct system impossible unless a fort was built so far up the southern side of the Rombach Valley that it would have caused tactical problems (Map 40). Instead, the Romans simply ran the line along a series of ridges to the north, on the shortest route to Buch, and set the fort itself on a high point just north of the river where much, but not all, of the line is in view.¹²⁶ This compromise solution seems to have been constructed with skill, and the fort still enjoys simple one-stage relay links with its neighbors.¹²⁷

Map 40 Upper German/Raetian border – river valleys.

The site at Schirenhof best demonstrates the Roman ability to adapt the defenses and signaling system to the terrain. The winding Rems Valley is almost a gorge at this point and the fort has been built on a plateau some 50 m above the valley floor on the south side of the river, in order to improve its view of the Limes. This puts the fort in visual contact with the line installations to its north and immediate west. A higher ridge obscures its view to the east. To the east, the fort would need a double relay to communicate with Böbingen. There may have been room for a fort on the top of the eastern ridge, and the far better view thus obtained would have allowed a one-relay signaling link with Böbingen. But putting a fort on a steep slope would have hindered such mundane tasks as provisioning the fort and would have affected the garrison’s mobility. There may have been a watchtower on the ridge top, but no evidence has so far come to light. The signaling restraints imposed by the fort’s actual position, coupled to the topographical complexity of the Limes line itself, which crosses a number of deep stream valleys in the area, may help to explain the unusual concentration of Kleinkastelle in the area.

The western end of this study sector is where we find the best demonstration of the skill with which Roman frontier designers could exploit difficult terrain to facilitate signaling. In the section where the Limes has to transfer from one topographical feature to another, that is, the Lein Valley to the Rems Valley, if the line had simply gone straight it would have crossed a series of ridges to the east of the Walkersbach before reaching the Rems slightly downstream of the fort at Lorch. Instead it loops around WP 12/1, then south at WP 12/5, to follow the ridge between two streams, and then turns east again at WP 12/11 and reaches the Rems Valley just upstream of Lorch at WP 12/14 (Map 41). The direct route would have been tactically less appropriate and the complex terrain would have made communications much more difficult. There is no easy route from the Lein to the Rems, but by utilizing the ridge on which WP 12/6–11 stand, the Romans placed the line on the most consistent run of high ground available, which puts the frontier towers on far better observation positions and greatly improves the signaling situation.

Map 41 Signaling – southern Raetian Limes, all installations used.

There are some amazing anomalies in this system, such as the 1.5 km stretch from WP 38 to 12/41 that is actually in a valley floor. It is a tactical absurdity that forces the line across far more difficult terrain. It surrenders much of the watchposts’ surveillance capability to the north and abandons a small part of the Rems Valley that must have been an important communications artery. There is no obvious military justification. The only rational explanation put forth so far has been that the frontier to the east in this area may have been built slightly before that to the west, and that the original line may have ended beside the Rems.¹²⁸

One of the scholarly considerations in studying this area is that much of the research was originally based on the findings of the Reichslimeskommission that mapped out the entire frontier and published it in a huge multivolume work.¹²⁹ Of the 125 mapped towers on the study sector, only forty-three (34 percent) are known with absolute certainty. There is some evidence for twenty-two more. If we build a model, it might follow one of two patterns: the first would include the fullest of mapped installations (Map 41), an alternative would be restricted to known sites (Map 42). If one looks at the entire corpus, ninety-two towers (74 percent) can communicate directly with a fort, while the remaining thirty-three (26 percent) can all have their signals passed to a fort via simple one-stage relays. Interfort communications are also relatively straightforward and the forts seem to have had clear-cut areas of responsibility. None of the forts is directly intervisible, but of the six interfort links on the sector, three can signal by one-stage relays. The remainder require relays by at least two other sites. It is notable that here, as on other frontiers, the relays for both interfort and line installation signals are always on the same side of the receiving fort as the site they are serving. Thus, although a fort could no longer see the source of a relayed signal, it would still know the direction from which the signal came.

Map 42 Signaling – southern Raetian Limes using only known or likely installations.

The interfort system on the Raetian Limes does show one small difference from that on the Wetterau zone, and that is that there are fewer indications that the interfort relays had been singled out and protected in any way. In the Wetterau, Kleinkastelle were used in this way, as what we would now called “hardened installations.” In this southern stretch, however, all the relays seem to be just ordinary towers, although some are larger than normal. To date, only four Klein- kastelle are known on the sector, compared to fourteen on the Wetterau Limes.

There is no significant change in the approach to signaling where it crosses the Upper German-Raetian border, and there is no sign of a change in signaling systems. The two provincial frontiers seem to have been integrated into a whole, and the very existence of a provincial boundary could certainly not be located from the signaling data alone. This suggests that military signaling could cut across administrative boundaries.¹³⁰ Woolliscroft has noted that building a model for a signaling system with only the forty-three excavated posts does not work very well. But by adding the twenty-two sites for which some evidence exists, he can make a reasonably good model that shows a direct signaling system. Most of the signaling towers have a clear view to the forts and interfort communication is possible.¹³¹

Ongoing research

The case for signaling systems on Roman frontiers will continue to be debated. Signal towers and relay stations continue to be uncovered in areas besides the three discussed above. On the Limes Palestinae, for example, numerous installations have been found from which signaling could have been accomplished. These have never been pulled together into a coherent signaling system and no intervisibility studies were reported by Gichon when he published them.¹³²

On the Roman Limes Arabicus a network of watchtowers for observation and communication was discovered at Rujm Beni Yasser, near the Roman legionary camp at Lejjun in Jordan. It has been suggested that here the Romans rebuilt and expanded an existing observation system.¹³³ More survey work has confirmed that an earlier chain of small Nabataean forts and watchtowers formed the initial framework of the Roman Limes. Excavation in search of Roman signal facilities failed to produce conclusive evidence, although S. Thomas Parker has mentioned that hundreds of these posts may have existed along the Arabian frontier and may have been successful in detecting infiltrators.¹³⁴

An experiment was carried out in 1982 by Parker trying to prove the theory of long-distance, linear communications along this Limes in central Jordan. Staff members manned fourteen forts and watchposts extending from north to south over a 35 km sector of the frontier. The observation posts were grouped together into three regions: northern, central, and southern. When the towers were manned at night, they were surprised to find that torches lit in the central towers were not generally observable from the southern locations – and intervisibility is the key element in such a system. The site, therefore, could not have been a major relay station whereby the northern forts communicated with the southern forts. Leaving aside the argument whether all these sites were Roman, ever occupied at the same time, or ever used for signaling, the archaeologists have obviously endowed this signaling system with a greater sophistication than it would have been capable of.¹³⁵ The central Jordan towers may have been part of an early warning system, but not as a static frontier line where every watchtower had a beacon.

Research on the defensive system of Dacia is still in its early stages. Investigation of the Dacian Limes began in the middle of the last century, but investigation of the forts is far from complete.¹³⁶ From the work done to date, we can see that the network of installations was based on arteries of communications with the lines of defense facing to the west, north, and east. For reasons of geography, the Dacian Limes presents some peculiarities. It differs from the Limes on other frontiers. The forts were located in the interior, in front of the passes, or at observation places between passes in order to forestall any attempt at penetration. The watch- or signal towers of this advanced line were placed mostly on heights, in front of the forts and sometimes between them.¹³⁷ One line of advanced watch and signal towers was identified at the end of the last century. Archaeological research has been carried on mostly in the northwestern sector, while field observations and the identification of an advanced line of towers has continued systematically in the northern sector and in the northern part of the eastern sector of the Limes. The advanced line here consists of towers made of earth or more commonly of stone, fortlets and even parts of a vallum. All of these fortifications were on heights in front of the forts at a distance varying between 1.5 and 15 km. Attempts have been made to link the towers of the advanced line with the forts to determine their function and even to date them. It has been argued by Gudea that the advanced line of watch- and signal towers existed along most of the Limes and was organized in an almost identical manner.¹³⁸ Work thus continues along the Dacian Limes and the preliminary evidence suggests that the signaling system will resemble that found elsewhere.¹³⁹

Roman SIGINT

The efficiency of the Roman command structure depended upon the communication of accurate intelligence, and nowhere was this of more importance than on Rome’s frontiers. Rome needed to send strategic, long-distance command communications, but this was done by messenger and not by signaling. For short-range combat situations, however, signaling is both attested to in the literary sources and highly probable between frontier installations. Although some skepticism has been rightly voiced about how some of these systems worked, archaeologists are able to make a better case for signaling systems, as more sites are properly excavated in frontier regions.¹⁴⁰ The requirements for such a system have remained the same since antiquity. A system must be capable of assimilating and transmitting information accurately and speedily, at all times and in all conditions. No system is foolproof, and under active service conditions any communications may be susceptible to failure. Thus back-up systems must also be built in where possible. It has been argued that signaling systems are always labor-intensive. Even if a chain of beacons were the only alarm system, it would require constant vigilance at every post along its length. The more complicated the message, the more it becomes necessary to train teams of competent operators. To operate a visual communications system on the scale visualized by some modem historians, the Romans would have had to be able to deploy large numbers of highly trained personnel. Because there is no term in Latin for a military signaler, one wonders just how many people were involved in this business. Frontier signals were probably very simple, probably just torches, and they acted to supplement messengers, not to replace them. They saved a bit of time by providing a speed-of-light alert mechanism and also acted as a back-up if the situation got bad enough that the messenger could not get through. If they were kept simple, the manpower problem might be overcome. Since watchtowers needed to be manned anyway, if the technique was kept simple, the crews would not have required any special expertise. Direct systems would remove the need for elaborate relay systems with their inherent dangers and those that were needed had to be kept to the minimum possible length.

Since all long-range signaling depended on visual techniques, direct communications could take place only between intervisible sites.¹⁴¹ Recent studies of these sites by Woolliscroft and others have created a paradigm for how signaling was used in Roman times. Because signaling leaves few physical traces, the possibility of finding actual bum marks seems unlikely. We must, therefore, rely on a combination of literary sources, sculptural evidence, and the intervisibility studies being performed on the Limes sites. With further excavation of new forts, fortlets, and towers along the Limes on Rome’s other frontiers, one day we may be able to see whether there was an overall design theory that went into these frontier installations. The evidence as it stands now suggests there was. Critics have said that “we are too ready to credit the Roman signaling system with greater technical proficiency than the ancient sources allow, and too willing to propose complex networks without adequate consideration of the technical problems involved.”¹⁴² Woolliscroft, however, has shown that even with a low level of technical ability, the Romans could have mounted an effective signaling system in northern Britain and Germany, and, indeed, the location of many of the installations can only be explained by their need for intervisibility with neighboring structures.

Finally, we must ask whether, for long-distance communications, visual networks were superior in terms of reliability, speed, and accuracy to a system of fast-riding couriers. G.H. Donaldson points out that “the effects of inclement weather on visibility, the inevitable dangers of garbling in the retransmission of a message by operators at every optical relay post, and the limitations on the quantity and quality of transmittable information imposed by the nature of the systems available make the written word carried by dispatch rider seem the better alternative by far.”¹⁴³ He completely doubts whether the Romans ever resorted to signaling for communications between Hadrian’s Wall and York, as Sir Ian Richmond has suggested.

Communications between watchtowers could have been done by sound-signaling rather than by sight. We know that the Romans used war trumpets such as the tuba and the cornu on the battlefield.¹⁴⁴ It is not impossible that such calls were also used to communicate between watchtowers. In fact, the mouthpiece of an instrument was recovered from the site of a watchtower on the German Limes.¹⁴⁵

There is no doubt that signaling was practiced on Roman frontiers. The only question which remains is how much signaling they did and how sophisticated the systems were. Woolliscroft has certainly given us an interesting thesis to test. Not only does he believe that Roman frontiers were equipped with comprehensive signaling systems, but that the requirement for such a system influenced the overall design of the defenses. If this hypothesis turns out to be true, then the Romans gave a much higher priority to intelligence gathering than we have ever given them credit for.

Notes

1. All 52 examples are listed in appendix one to D.J. Woolliscroft, Roman Military Signalling (Stroud: Tempus, 2001), pp. 159–71. See also W. Leiner, Die Signaltechnik der Antike (Stuttgart: W. Leiner, 1982).

2. 2 Woolliscroft, Roman Military Signalling, p. 28.

3. On the Assyrians, see C. Fries, “Zur Babylonischen Feuerpost,” Klio 4 (1904), pp. 117–21; Dvornik, Origins of Intelligence Services, pp. 19–20. On the fire system designed by Leo the Mathematician, see Dvornik, Origins of Intelligence Services, p. 142, although T. Pattenden, “The Byzantine Early Warning System,” Byzantion 53 (1983), pp. 258–99 does not believe this system was ever put into use. Cf. Volkjer Aschoff, Über den byzantinischen Feuertelegrafen und Leon den Mathematiker (München and Düsseldorf, 1980).

4. Woolliscroft’s study, Roman Military Signalling, is particularly good because he locates the possible signaling towers on foot, and he uses a method of low-level aerial photography to test the intervisibility of the sites. This is an improvement over Leiner’s study, Die Signaltechnik der Antiker, which discusses only the literary sources. See the bibliography for his earlier articles.

5. Woolliscroft, Roman Military Signalling, p. 14; G.D.B. Jones, “Concept and Development of Roman Frontiers,” Bulletin of the J. Rylands Library 61 (1978), pp. 115–44.

6. D.J. Woolliscroft, “Signalling and the Design of the Gask Ridge System,” PSAS 123 (1993), pp. 291–8. For a more detailed discussion of what a limes actually is, see B. Isaac, “The Meaning of the Terms Limes and the Limitanei,” Journal of Roman Studies 78 (1988), pp. 125–47.

7. Dio’s description of a dispatch sent by Hadrian shows that the traditional wording used by Cicero was maintained: “If you are in good health it is well, I and my army are in good health.” Dio 67.7.3; 68.29.1; 69.14.3; Suetonius, Caligula 44.2. See J.B. Campbell, The Emperor and the Roman Army 31 bc to ad 235 (Oxford: Oxford University Press, 1984), p. 148.

8. Dio Cassius 78.27.3.

9. Suetonius, Caligula 44.2. It was forbidden to ride in a vehicle in the Forum during the daytime.

10. Dio Cassius 68.29.1.

11. Campbell, Emperor and the Roman Army, pp. 148–50.

12. Caesar, BG 3.65. See Sulla’s attempt to use them in the Social War in 89 bc, Appian, Civil Wars 1.6.51.

13. Frontinus, Stratagems 2.5.16.

14. Appian, Spanish Wars 6.15.90–2.

15. Vegetius 3.5.

16. See the comments of Woolliscroft, Roman Military Signalling, p. 24.

17. W. Riepel, Das Nachrichtenwesen des Altertums (Leipzig: Teubner, 1913), p. 75. Cf. Suetonius, Tiberius 65.

18. Suetonius, Tiberius 65, trans. R. Graves, Suetonius: The Twelve Caesars (Harmondsworth: Penguin, 1957); Tacitus, Annals 6.39.

19. On the limitations of beacons signals see Polybius 10.43.1–10.

20. C. Cichorius, Die Reliefs der Trajanssäule (Berlin: G. Reimer, 1900), vol. 2, p. 20; Woolliscroft, Roman Military Signalling, p. 26.

21. Webster, Roman Imperial Army, p. 163.

22. P. Southern, “Signals versus Illumination on Roman Frontiers,” Britannia 21 (1990), p. 239; G.H. Donaldson, “Signalling, Communications and the Roman Imperial Army,” Britannia 19 (1988), points out the fire danger to the watchtowers; and E.W. Sockett made the case for the food stores in “A Note on Major Donaldson’s Article on Roman Military Signalling,” Archaeologia Aeliana 14 (1986), p. 187.

23. Woolliscroft, Roman Military Signalling, p. 26.

24. F. Lepper and S.S. Frere, Trajan’s Column; A New Edition of the Cichorius Plates (Gloucester, UK: Alan Sutton, 1988), pi. 88. Woolliscroft, Roman Military Signalling, p. 26, experimented with green hawthorne wood and found that when it was stacked in this manner it will do little more than smoulder, even when doused with gasoline.

25. See the comments of Woolliscroft, Roman Military Signalling, p. 28.

26. Website: Ibid., p. 28 rejects the suggestion of G.L. White, “Überlegungen zur Donaulimesdarstellung auf der Traianssäule in Rom,” Germania 67 (1989), p. 187, who believes the hayricks are really stacks of grain still on ear covered by leather caps designed to keep the top-most ears dry.

27. Southern, “Signalling versus Illumination,” pp. 233–42.

28. Woolliscroft, Roman Military Signalling, p. 47.

29. Website: Ibid., p. 29, citing E. Fabricius, F. Hettner and O. von Sarwey, Der Obergermanisch- Raetische Limes dees Römerreiches (Heidelberg: Verlag von O. Petters, 1894/1919), Abt. A, Band 1, Strecke Im 54 and Tafel 13, 16, 21 and 22.

30. Southern, “Signalling versus Illumination,” pp. 240–1. Woolliscroft rejects this interpretation because crossings at this point are entirely conjectural. He also points out that only the example at WP 1/47 had been fired, whereas if they had been used for routine illumination, one would expect all of them to have been used repeatedly.

31. A. Fox, “Martinhoe and Old Burrow,” in Studien zu den Militär grenzen Roms., Proceedings of the sixth International Limes Congress, Cologne, 1967, p. 20 and Fig. 2.

32. The original suggestions were made by A. Poidebard, La Trace de Rome dans le désert de Syrie: le limes de Trajan à la conquête arabe (Paris: P. Geuthner, 1934), pp. 59, 95–6. See the comments of Woolliscroft, Roman Military Signalling p. 29.

33. 33 Poidebard, La Trace de Rome dans le désert de Syrie, pis. 103–4.

34. See Woolliscroft, Roman Military Signalling, pp. 30–46.

35. Philon, Mechanica 7.8.55–57; Polybius 10.44.1–45.2; Polyaenus 6.16.2. Translations in Woolliscroft, Roman Military Signalling, Appendix 1.

36. Woolliscroft, Roman Military Signalling, pp. 31–5, Illus. 6 and 7, and whose says on p. 32: “It seems unlikely that ancient craftsmen could ever have made truly identical jars and even if they could, they would have found it more difficult to drill them in exactly the same place and with holes of exactly the same diameter.” Woolliscroft did try constructing jars of his own and constructed a system by which one could transmit messages with efficiency.

37. Polybius 10.45.6–47.4.

38. 38 Woolliscroft, Roman Military Signalling, p. 39.

39. Website: Julius Africanus, Kestoi 11, translation in Woolliscroft, Roman Military Signalling, Appendix 1, no. 52.

40. Woolliscroft, Roman Military Signalling, p. 44.

41. Website: Ibid., p. 45.

42. Website: Ibid.

43. Website: Julius Africanus, Kestoi 11, translation in Woolliscroft, Roman Military Signalling, Appendix 1, no. 52.

44. 44 Woolliscroft, Roman Military Signalling, p. 48.

45. See Woolliscroft, S.A.M. Swain and N.J. Lockett, “Barcombe B, A Second Roman ‘Signal Tower’ on Barcombe Hill,” Archaeologia Aeliana 5 (1992), pp. 57–62, where finds were made of a stone wall, Roman mortar, tile fragments and Roman glass.

46. Woolliscroft, Roman Military Signalling, pp. 46–7.

47. The idea of the Stanegate road as a frontier was first suggested by R.H. Forster and W.H. Knowles in “Corstopium: Report on the Excavations in 1914,” Archaeologia Aeliana 12 (1915), pp. 268–89. It was later developed by R.G. Collingwood and J.N.L. Myres in Roman Britain and the English Settlements (Oxford: Oxford University Press, 1936), pp. 126–7. A formalized list of the sites was presented by E. Birley, Research on Hadrian’s Wall (Kendal, 1961), pp. 71. C.M. Daniels warned against uncritical acceptance of the system in “Problems of the Northern Frontier,” Scottish Archaeological Forum 12 (1970), pp. 94–5. B. Dobson studied the problem of the “Stanegate system” in “The Function of Hadrian’s Wall,” Archaeologia Aeliana 14(1986), pp. 2–5. The idea was defended by N. Hodgson, “The Stanegate: a Frontier Rehabilitated,” Britannia 31 (2000), pp. 11–22. For a good discussion of its intelligence function, see D.J. Woolliscroft, “The Outpost System of Hadrian’s Wall,” TCWAAS 88 (1988), pp. 23–8.

48. Woolliscroft, Roman Military Signalling, p. 53. See also D.J. Breeze and B. Dobson, Hadrian’s Wall (Harmondsworth: Penguin, 2000), pp. 16ff; Dobson, “Function of Hadrian’s Wall,” pp. 1–30; Hodgson, “Stanegate: A Frontier Rehabilitated,” pp. 11–22.

49. It should be noted that controversy still surrounds the exact date, the form and even the very existence of the Stanegate as a formal frontier line, but for the system’s supporters, see Breeze and Dobson, Hadrian’s Wall, pp. 16ff.; Dobson, “Function of Hadrian’s Wall”; Hodgson, “Stanegate: A Frontier Rehabilitated” Woolliscroft, Roman Military Signalling, p. 53.

50. Woolliscroft, Roman Military Signalling, p. 53. See N. McCord and G. Jobey, “Notes on Air Reconnaissance in Northumberland and Durham II,” Archaeologia Aeliana 4 (1971), p. 120 and pi. 12 on the large mutliperiod fort detected at Washing Well, Whickham, on the south side of the Tyne valley. For the fort and store base found at South Shields, see S.S. Frere, “Roman Britain in 1985,” Britannia 17 (1986), p. 375. A fort long suspected to lie at Kirkbride on Moricambe has been uncovered. R.L. Bellhouse and G.G.S. Richardson, “The Trajanic Fort at Kirkbride: The Terminus of the Stanegate Frontier,” CW 2 (1982), pp. 35–50. A number of entirely new sites have been reported, including watchtowers at Easton and Far Hill and two forts to the south of the Wall fort of Brough-by-Sands. G.D.B. Jones, “The Solway Frontier: Interim Report 1976–81,” Britannia 13 (1982), pp. 283–98; N.J. Highham and G.D.B. Jones, The Carvetti (Gloucester: Alan Sutton, 1985), Ch. 2; A.S. Esmond Cleary, “Roman Britain in 1994, England,” Britannia 26 (1985), pp. 342–70.

51. Four of these are in the study sector surveyed by Woolliscroft: Old Church, Brampton; Nether Denton; Carvoran; and Vindolanda. Woolliscroft, Roman Military Signalling, p. 55. The Vindolanda roster tablet shows that these may have been unusually large garrisons. A.K. Bowman and J.D. Thomas, “A Military Strength Report from Vindolanda,” Journal of Roman Studies 81 (1991), pp. 62–73.

52. Woolliscroft, Roman Military Signalling, p. 55; see also F.G. Simpson, “Boothby, Castle Hill,” CW 2 (1934), p. 154; E. Birley, Research on Hadrian’s Wall (Kendall, 1961), p. 140.

53. See D.J. Breeze and B. Dobson, “Hadrian’s Wall: Some Problems,” Britannia 3 (1972), pp. 185–6.

54. Woolliscroft, Roman Military Signalling, pp. 66–7.

55. For a complete description of the system and this argument, see ibid., pp. 51–63.

56. Website: Ibid., p. 51.

57. On the construction and use of the Vallum, see D.J. Woolliscroft, “More Thoughts on the Vallum,” TCWAAS 99 (1999), pp. 53–65.

58. 58 Woolliscroft, Roman Military Signalling, p. 51.

59. R. Selkirk, “Roman Signal Stations,” Archaeology Today 8, 1 (1987), pp. 126–31.

60. Woolliscroft, Roman Military Signalling, pp. 63–4.

61. See especially Woolliscroft’s discussion of the Barcombe B site and how it fits into the Signalling system around Vindolanda, Roman Military Signalling, pp. 76–8.

62. 62 Both Risingham and High Rochester had a numerus exploratorum, RIB 1235, 1262, alongside their main garrison. At Netherby the very name of the fort was Castra Exploratorum, or fort of the scouts. See A.L.F. Rivet and C. Smith, The Place Names of Roman Britain (Princeton, NJ: Princeton University Press, 1979), p. 302. Woolliscroft, Roman Military Signalling, p. 81.

63. For the possible siting of a Roman signal tower on Barron’s Pike and its use in this system, see Woolliscroft, Roman Military Signalling, pp. 81–8.

64. For a full discussion of these sites, see Woolliscroft, Roman Military Signalling, pp. 79–88.

65. This took the form of ditches on the Cardurnock Peninsula in the north, and parallel timber fences of palisades further south, around Silloth. See G.D.B. Jones, “The Western Extension of Hadrian’s Wall: Bowness to Cardurnock,” Britannia 1 (1976), pp. 236–43; N.J. Higham and G.D.B. Jones, The Carvetii (Gloucester: Alan Sutton, 1985), pp. 30ff. Woolliscroft has discredited the Silloth palisade in D.J. Woolliscroft and G.D.B. Jones, “Excavations at Silloth and Fingland Rigg on the Cumbrian Coast,” forthcoming.

66. Woolliscroft, Roman Military Signalling, p. 89; D J. Woolliscroft, “Signalling and the Design of the Cumberland Coast System,” TCWAAS 94 (1994), pp. 55–64.

67. Woolliscroft, Roman Military Signalling, pp. 95–7.

68. Another problem exists in that the supposed tower at Brackenber does not resemble a Roman tower, and is situated in a blind hollow that most certainly does not suggest it was part of a Signalling tower requiring a field of view.

69. R.A.H. Farrar, “Roman Signal Stations over the Stainmore and Beyond,” BAR International Series 7li (1980), p. 211, who offers accurate site plans. According to Woolliscroft, Farrar mistakenly says that Roper Castle cannot see the fortlet of Maiden Castle and that the Punchbowl cannot see Brough Fort.

70. I.A. Richmond, “A Roman Arterial Signalling System in the Stainmore Pass,” in W.F. Grimes (ed.), Aspects of Archaeology in Britain and Beyond. Essays Presented to O.G.S. Crawford (London: H.W. Edwards, 1951), pp. 293–302.

71. P. Robinson, “The A66 Archaeology Project,” Current Archaeology 122 (1990), pp. 63ff. The problems have been discussed by Farrar, “Roman Signalling Stations”, pp. 211–32.

72. Higham and Jones, Carvetii, pp. 44ff. on the sites at Augil Bridge, Punchbowl, and Johnson’s Plain, between Brough and Maiden Castle. A fourth location was recognized earlier by J.K. St Joseph, “Air Reconnaissance in North Britain,” Journal of Roman Studies 41 (1951), p. 53. See Woolliscroft, Roman Military Signalling, pp. 98–102.

73. D.J. Woolliscroft, “The Roman Signal Tower at Johnson’s Plain, Cumbria,” TCWAAS 91 (1991), p. 19 and note 6. Woolliscroft, Roman Military Signalling, pp. 97–102.

74. On the dating of the Gask Ridge to the seventies, see Woolliscroft, with contributions, The Roman Frontier on the Gask Ridge. Perth and Kinross. An interim report on the Roman Gask Project, 1995, BAR British Series 335 (Oxford: Archeopress, 2002), pp. 9ff.

75. D.J. Christison, “Excavations Undertaken by the Society of Antiquaries of Scotland of Earthworks Adjoining the ‘Roman Road’ Between Ardoch and Dupplin Perthshire,” PSAS 35 (1900–01), pp. 16–43.

76. G.S. Maxwell, “Flavian Frontiers in Caledonia,” in H. Vetters and M. Kandier (eds), Akten des 14 Internationalen Limeskongresses 1986 in Carnuntum (Vienna: Verlag der Österreichischen Akademie der Wissenschaften, 1990), pp. 353–65.

77. See the Gask Ridge website: http://www.morgue.demon.co.uk/Pages/Gask/index.html, which lists all the forts, fortlets, towers, temporary camps and Roman roads on the Gask Ridge. Also D.J. Woolliscroft, etal., Roman Frontier on the Gask Ridge. Perth and Kinross, Ch. 1.

78. Website: Ibid., p. 1.

79. A.S. Hobley, “The Numismatic Evidence for the Post-Agricolan Abandonment of the Roman Frontier in Northern Scotland,” Britannia 20 (1989), pp. 73–74; Woolliscroft, “Signalling and the Design of the Gask Ridge System”, p. 291; Woolliscroft, Roman Frontier on the Gask Ridge: Perth and Kinross, Ch. 1.

80. Woolliscroft, Roman Frontier on the Gask Ridge: Perth and Kinross, p. 3. On the dating of the German Limes see K. Kortiim, “Zur Datierung der römischen Militäranlagen im obergermanisch-rätischen Limesgebiet. Chronoligische Untersuchungen anhand der Münzfunde,” Saalburg Jahrbuch 49 (1998), pp. 5–65.

81. Westmuir cannot be seen from Peel, for example. See Woolliscroft’s comments in “Signalling and the Design of the Gask Ridge System,” p. 294. Another important anomaly is Midgate, discussed on p. 302.

82. Website: Ibid., p. 293.

83. Website: Ibid., p. 295.

84. Website: Ibid., p. 300.

85. 85 Woolliscroft, Roman Frontier on the Gask Ridge. Perth and Kinross, p. 7.

86. Donaldson, pp. 352–3.

87. Woolliscroft, “Signalling and the Design of the Gask Ridge System,” p. 295.

88. See Körtüm, “Zur Datierung der römischen Militäranlagen,” pp. 5–65, which redates the foundation of the Limes to the early second century and seems to have been widely accepted.

89. Woolliscroft, “Signalling and the Design of the Gask Ridge System”, p. 300.

90. Breeze and Dobson, Hadrian’s Wall, p. 81. On the dating of the epigraphic and numismatic sources, see J.C. Mann, “The History of the Antonine Wall – A Reappraisal,” PSAS 118 (1988), pp. 131–7.

91. Woolliscroft, “Signalling and the Design of the Antonine Wall,” p. 153.

92. The eastern terminus and even the length of the wall are still disputed. See G.B. Bailey and D.F. Devereux, “The Eastern Terminus of the Antonine Wall: A Review,” PSAS 117 (1987), pp. 93ff.; for a list of the installations, see Woolliscroft, “Signalling and the Design of the Antonine Wall,” p. 160.

93. J.P. Gilliam, “Possible Changes in the Plan in the Course of the Construction of the Antonine Wall,” SAF1 (1976), pp. 5Iff.

94. There is general, if qualified, acceptance of Gilliam’s theory. See Woolliscroft, “Signalling and the Design of the Antonine Wall,” p. 153. Only nine fortlets are currently known: Dontochter, Cleddans, Sommerston, Wilderness Plantation, Glasgow Bridge, Croy Hill, Seabegs Wood, Watling Lodge, and Kinneil: ibid., p. 158.

95. Website: Ibid., p. 154.

96. Website: Ibid., p. 173.

97. Website: Ibid.

98. W.S. Hanson and G.S. Maxwell, “Minor Enclosures on the Antonine Wall at Wilderness Plantation,” Britannia 14 (1983), pp. 234ff. assume the disappearance of the earlier sites. Woolliscroft, “Signalling and the Design of the Antonine Wall,” p. 175. The site at Wilderness Plantation had its interior cleared and was cobbled over, but occupation continued: J.J. Wilkes, “The Antonine Wall Fortlet at Wilderness Plantation, Lanarkshire,” Glasgow Archaeological Journal 3 (1974), p. 57.

99. Ammianus Marcellinus 28.3–8.

100. Woolliscroft, Roman Military Signalling, p. 80 says: “The Areani were, then, at least a part of a regular intelligence service and they can probably be identified, if only by function, with the units of exploratores recorded in military inscriptions from earlier times.”

101. Website: Ibid., p. 103.

102. M. Mackensen, “Frühkaiserzeitliche Kleinkastelle an der oberen Donau,” Zivile und militärische Structuren im Nordwesten der römischen Provinz Raetien, 3rd Heidenheimer Archäologie-Colloquium am 9 und 10 Oktober 1987 (Heidenheim an der Brenz: Heimat-und Altertumsverein Heidenheim, 1988), pp. 13–32; W. Czysz, “Archäologische Nachuntersuchung am Kleinkastell Neuwirtshaus bei Hanau,” Neues Magazin für Hanauische Geschichte 6, 5 (1977), pp. 12Iff.

103. Woolliscroft, Roman Military Signalling, pp. 104–5.

104. The forts included are: Arnsburg, Inheiden, Echzell, Ober-Florstadt, Altenstadt, Marköbel, Rückingen and Gross Krotzenburg.

105. Woolliscroft, Roman Military Signalling, p. 105.

106. Website: Ibid., p. 109.

107. Woolliscroft discusses the merits of the lateral system versus the direct system, as seen on Hadrian’s Wall. But even he admits that since both of these models represent somewhat idealized positions, it is unlikely that either system would ever be found in a completely pure state on a real Roman frontier.

108. Woolliscroft, Roman Military Signalling, p. 109.

109. Kleinkastelle Griiningen and Langsdorf, and WP 4/63,4/65 and 4/67/8. See Woolliscroft, Roman Military Signalling, p. 111.

110. See Woolliscroft, Roman Military Signalling, p. Ill, who discusses the concessions that had to be made to the terrain, and who lists the sites that did not enjoy direct links to a fort.

111. As the best example of such a block, Woolliscroft suggests the Echzell sector. The limits of the fort’s sphere of responsibility are defined by the two Kleinkastelle at Widdersheim and Lochberg. Ibid., p. 116.

112. Woolliscroft makes the parallel with the situation at Barcombe B on Hadrian’s Wall, which serves as a one-stage relay between Housteads and Vindolanda. It too is sited so that it has visible contact with every installation in its own specific sphere of responsibility, but nothing whatsoever beyond it. Ibid.

113. At least three of the Kleinkastelle on the sector may be responsible to specific areas of local weakness, because they stand on weak sections of ground that lie out of sight of, and so outside, the immediate protection of the fort. Ibid., p. 117.

114. Website: Ibid., p. 118.

115. Website: Ibid., p. 121.

116. Website: Ibid.

117. This is especially true in the area between Rückingen and Krotzenburg. Ibid., p. 123.

118. Website: Ibid., p. 124.

119. Website: Ibid., p. 126.

120. The one real bend in this part of the frontier hinges on the fort at Morköbel. The presence of forest would explain why this fort was built so close to the limes and not further back, where it would be in open country and could have kept in touch with a much larger section of the line. Ibid., pp. 128–9.

121. Woolliscroft notes that there has been insufficient detailed environmental archaeology done in this region to confirm the exact relationship between the frontier and the forest at any one point. Therefore the relationship between forestry and the Roman Signalling system must remain hypothetical.

122. Butzbach just outside this study sector, and Murrhardt on the Antonine Outer Limes: H.U. Streckham, “Vegetationsgesichtliche Untersuchungen einer Römer-zeitlichen Torfbildung bei Butzbach in Hessen,” Saalburg Jahrbuch 17 (1958), pp. 6Iff., K.H. Knörzer, “Römerzeitliche Pflanzenreste aus einem Brunnen in Butzbach, Hessen,” Saalburg Jahrbuch 33 (1973), pp. 7Iff.; R. Knapp, “Die Vegetation der Umgebung von Butzbach in der Gegenwart und zur Römerzeit,” Saalburg Jahrbuch 30 (1973), pp. 133ff.; and R. Rösch, “Botanische Funde aus römischen Brunnen in Murrhardt Rems-Murr-Kreis,” AABW (1988), pp. 114ff. Woolliscroft also cites Zugmantel in the Taunus mountains; see F. Firbas, “Eine Flora aus dem Brunnenschlamm der Romerkastells Zugmantel,” Saalburg Jahrbuch 1 (1930), pp. 75ff.

123. Note his discussion of the three towers that he suggests may have been part of this system, allowing the entire northern Wetterau to be coordinated from one point. Woolliscroft, Roman Military Signalling, pp. 132–5.

124. They range from 6 km (Lorch to Schirenhof) to 13 km (Buch to Halheim). Ibid., p. 142.

125. The sectors around Welzheim, Böbingen and Buch are almost textbook examples, with almost every single line installation enjoying a direct view of their fort. Ibid., p. 143.

126. Website: Ibid., pp. 144–5, with photograph 75 and Map 76.

127. Website: Ibid., p. 144. One should note that some of the sites on this line of view are still conjectural, that is, they refer to sites that may have been there, but there is no archaeological evidence yet to confirm them.

128. The suggestion is made by N. Hodgson, “Relationships between Roman River Frontiers and Artificial Frontiers,” in Groenman-van Waateringe etal. (eds), Roman Frontier Studies 1995, proceedings of the 16th International Congress of Roman Frontier Studies (Oxford: Oxbow 1997), pp. 63, and Woolliscroft, Roman Military Signalling, p. 149.

129. 129 Fabricius, Hettner and Sarwey, Der Obergermanisch-Raetische Limes.

130. Woolliscroft, Roman Military Signalling, p. 150.

131. For a detailed discussion of the problems, see Ibid., pp. 150–41, especially on the problems on the limes stretch 8/9 and 12, pp. 152–3.

132. M. Gichon, “Towers of the Limes Palaestinae. Forms, Purpose, Terminology, and Comparisons,” Actes du Ixe Congrès International D’Etudes sur les Frontières Romaines, Mamaia, 6–13 September 1972 (Cologne: Böhlau Verlag, 1974), pp. 513–44.

133. V.A. Clark and S.T. Parker, “The Late Roman Observation and Signalling System,” in The Roman Frontier in Central Jordan (Oxford: BAR, 1987), pp. 165–81. S.T. Parker, “Archaeological Survey of the Limes Arabicus: A Preliminary Report,” Annual of the Department of Antiquities of Jordan vol. 21, pp. 19–31. S.T. Parker, “Towards a History of the Limes Arabicus,” in Roman Frontier Studies (BAR, 1979), pp. 865–78.

134. S.T. Parker, “Preliminary Report on the 1980 Season of the Central Limes Arabicus Project,” BASOR 247 (1982), pp. 1–26. The post at Yasser was established by Nabataeans (first century bc to first century ad), but then was not to be reoccupied during the fourth and fifth centuries in conjunction with the legionary occupation.

135. Clark and Parker, “Late Roman Observation and Signalling System,” pp. 165–81; S.T. Parker, “Exploring the Roman Frontier in Jordan,” Archaeology (Sept./Oct. 1984), p. 39. Although Parker claims the sites were all contemporaneous, see the criticism of Southern, “Signals versus Illumination,” p. 239.

136. N. Gudea, “The Defensive System of Roman Dacia,” Britannia 10 (1979), p. 71, who cites the earlier Romanian literature.

137. Website: Ibid., pp. 67–8.

138. Website: Ibid., pp. 73–7.

139. N. Gudea, “Die Verteigung der Provinz Dacia Porolissensis Zwischen Mauerperre und Verteidigung in der Tiefe,” in W. Groenman van Waaterings etal. (eds), Roman Frontier Studies 1995, proceedings of the 16th International Congress of Roman Frontier Studies (Oxford: Oxford University Press, 1997), pp. 13–24; Woolliscroft, Roman Military Signalling, p. 156.

140. See Malcolm Todd’s comments on the climatic difficulties for lateral Signalling along Hadrian’s Wall. Todd, Roman Britain 55 bc to ad 400; p. 142; R.A.H. Farrar raised the problems of intervisibility and the spacing of optical relay posts on the Stainmore Pass in “Roman Signal-Stations over Stainmore and Beyond,” Roman Frontier Studies1979 (Oxford, 1980), pp. 211–31; David Breeze has even questioned whether the remains strung along the Gask Ridge ever comprised a series of signal posts; The Northern Frontiers of Roman Britain (London, 1982), pp. 61–2. He is corrected by G.H. Donaldson, “Roman Military Signalling on the North British Frontiers,” Archaeologia Aeliana 13 (1985), p. 24, n. 5.

141. Woolliscroft, “Signalling and the Design of the Gask Ridge System,” p. 294.

142. Breeze and Dobson, “Hadrian’s Wall: Some Problems,” p. 186. On page 187 they suggest that an intervisibility study should be undertaken. That has now been done by Woolliscroft.

143. Donaldson, “Roman Military Signalling on the Northern British Frontiers,” p. 21.

144. See Webster, Roman Imperial Army, pp. 141–2.

145. D. Baatz, Die Wachtürme am Limes (Stuttgart, 1976), p. 48.