2026
From Frontier to Flowerbed: Five Roman Stylobates in Filey, by Noah Fenwick
Visitors to the Yorkshire town of Filey are likely to remember its sandy beaches, framed by dramatically jagged cliffs and punctuated with ice-cream parlours, souvenir shops, and coffee shacks. They might also remember the clifftop Crescent Gardens, which separate the town’s high street from its esplanade. However, tourists pay little attention to the Roman archaeology these gardens conserve.
Figure 1: Frontal view of the Crescent Gardens flowerbed containing 5 Roman stylobate blocks (fifth hidden in top left). Author’s own image.
Figure 2: Three of the five stones within the Crescent Gardens flowerbed, Filey. Author’s own image.
Inside the gardens, nestled within an easily overlooked flowerbed, sit five large, roughly cuboidal blocks of dressed masonry, each with a stepped profile on all four sides, such that a subtly tapered upper tier rises from a broader base (Figs. 1-2). The heavily weathered blocks are still highly uniform, ranging from 38 to 56 cm in height, with an average base of approximately 83 cm². All that distinguishes one block from the others is a faint relief carving on its front face depicting a hound and a stag (Fig. 3).
Figure 3: Relief carving of a hound chasing a stag on one of the five stylobates in the Crescent Gardens, Filey. Author’s own image.
Beyond their general size and shape, the five stones share one highly distinctive feature: neatly carved into the top face of each is a centrally positioned square socket, measuring approximately 18 x 19 cm and cut roughly 6 cm deep. These sockets identify the blocks as stylobates, architectual elements designed to stabilise large, weight-bearing timbers that would have been set upright within them. Therefore, although careful curation has seamlessly integrated the stones into the garden’s decoration, the question arises as to how the five stylobates, originally functional rather than ornamental, came to rest in a flowerbed. As it happens, their modern arrangement is the result of a chance discovery in the mid nineteenth century made at Carr Naze, just north of the present-day town. The discovery was made in 1857, when heavy rainfall caused a landslip on the cliffs of Carr Naze that exposed Roman remains in the soil below. This accidental reveal prompted an excavation of the site on 12–13 October 1857, the results of which were published by Dr William Cortis one year later (Fig. 4).
Figure 4: Plan of the 1857 discoveries at Carr Naze, Filey, first published by Cortis in 1858. Image from: Birckstock et al, 2000, Illustration 65, p.84.
Cortis’ excavation report describes the five stones in situ, set into a floor of “puddled clay”. They were found arranged in a quincunx formation (four at the corners of a square and one in the centre, as on the fifth face of a standard modern dice), and enclosed on three sides by a surrounding wall. Whilst the 1857 excavation succeeded in bringing scholarly attention to Carr Naze, it was a product of nineteenth century archaeological practice, and much of the archaeology was displaced with significant gaps left in its interpretation. The five stones were relocated to Filey, and few of the small finds recorded at the time (coinage, pottery, bones, and shells) survive today.
However, Carr Naze has since been subject to two further archaeological investigations. The first, overseen by Simpson in 1923, was only partially recorded and never published. Nevertheless, it facilitated two important observations. Firstly, although doubt was cast on the reliability of multiple aspects of Cortis’ excavation report, the original quincunx arrangement of the stones within a central structure was confirmed. Additionally, Simpson highlighted similarities between features of the Carr Naze site and that of another Roman era clifftop complex (also excavated by Simpson) on the grounds of Scarborough Castle.
Then, in 1993–1994, a more comprehensive and thoroughly recorded excavation was undertaken by York Archaeological Trust in response to concerns over the risk posed to the site by coastal erosion. The resultant excavation report included a significantly amended site plan, providing much firmer grounds on which to interpret the nature and function of the site. (Fig. 5).
Figure 5: Site plan of York Archaeological Trust excavation of Carr Naze, 1993-1994. Image from: Birckstock, R. et al, 2000, Illustration 3, p.82.
This refined site plan reveals a small, rectilinear building at the centre of a walled enclosure. The quincunx arrangement of the stylobates within this relatively compact structure suggests that it took the form of a tower, thus requiring substantial structural support from its base. Additionally, excavation within the external enclosure revealed a cobbled pathway, likely linking the gated entrance to the tower itself, and a ditch running along the western side of the complex was found, separating the site’s single entrance from the mainland.
Figure 6: Plans of the Roman era clifftop sites of Huntcliff, Goldsborough, Scarborough, and Filey. Image from: White, 2022, Figure 5, p.60.
These details substantiate Simpson’s earlier assertion that the Carr Naze site shares key features with the Roman complex later incorporated into Scarborough Castle. As at Carr Naze, excavations at Scarborough revealed a compact clifftop complex defined by a surrounding walled enclosure and organised around a centrally positioned structure, most commonly interpreted as a tower. This comparison has since been extended to include two additional, similarly organised Roman era sites at Goldsborough and Huntcliff. Although these sites vary in their state of preservation, they share several of the same defining features (Fig. 6). Epigraphic evidence alludes to yet another similar site in Ravenscar (Fig. 7). The Ravenscar Inscription refers to a “tower and fort [having been built] from the ground up” and has conventionally been linked with Roman era earthworks discovered but displaced during the construction of Raven Hall in 1774. Although its modern habitation precludes excavation, the site’s topographical similarity to the other known sites supports the theorising of a fifth clifftop complex in Ravenscar. Thus, Carr Naze can be contextualised within a series of five clifftop installations along the Yorkshire coast (Fig. 8).
Figure 7: The earliest drawing of the Ravenscar Inscription (RIB 721) by Lionel Charlton. Image from White, 2022, Figure 2, p.47.
Figure 8: Map showing the Roman era sites of Huntcliff, Goldsborough, Ravenscar, Scarborough and Filey (Carr Naze) on the Yorkshire coast. Image from: White, 2022, Figure 1, p.46.
This contextualisation gave rise to the first theory of the site’s function. Shared architectural features have been used to suggest the sites’ derivation from a common plan, implemented with local variation. This inference, combined with their coastal topography, has led to the sites’ collective interpretation as a coordinated system of signal stations intended to monitor maritime movement, relay information along the shoreline, and provide early warning against seaborne threats. Within this model, centrally positioned, tower-like structures—such as that indicated by the stylobates arranged at Carr Naze—would have provided highly visible platforms from which to signal between neighbouring installations.
This interpretation is seemingly supported by the generally accepted dating of the clifftop sites to the late fourth and early fifth centuries AD, based primarily on stratified coinage from Carr Naze and corroborated by finds from the other excavated complexes. In this period, Britannia became increasingly vulnerable to seaborne attacks from Pictish raiders travelling south from beyond Hadrian’s Wall, and the development of coastal defences seems a plausible response to this vulnerability.
However, whilst the signal-station theory offers a superficially attractive explanation for the shared form and placement of these sites, it rests heavily on architectural analogy and topographical inference rather than on direct archaeological evidence for signalling activity. Thus, problems with this interpretation must not be overlooked.
One such problem becomes apparent upon closer consideration of the practical limitations of Roman signalling, derived from its inherent dependence upon visibility between stations, favourable weather conditions, and sustained human oversight. Even in optimal conditions, the vast stretches of irregularly indented coastline between the five supposed signal stations would have precluded direct visual contact between several neighbouring installations. Moreover, the frequent mist, fog, and heavy rain characteristic of the East Coast – today as in antiquity – constitute far from favourable conditions for signalling, further diminishing the feasibility of such a network.
It is, however, possible that the surviving sites do not represent the full extent of the original network. Coastal erosion has been particularly severe along this stretch of coastline, and Roman archaeology has not escaped the wrath of the sea. Since its excavation, the Huntcliff site has completely collapsed into the sea, and it has been speculated that a Roman installation which may have occupied the headland at Whitby has met the same fate, although no archaeological evidence survives to confirm this. The former existence of such a site would have bridged the gap between the two most widely separated installations in the sequence, Ravenscar and Goldsborough, improving the feasibility of visual communication along this stretch of coast. Nevertheless, this entirely speculative theory provides a dangerously tentative premise upon which to ignore such a substantial lack of evidence. Whilst it must be acknowledged that the loss of key evidence to coastal erosion may have warped the archaeological record, unsubstantiated assumptions should not be employed in compensation.
Instead, renewed attention to the stylobates themselves might offer an alternative perspective on the function of the Carr Naze site. Traditionally, the square sockets cut into the upper faces of the stones have been assumed to have received the full base of the wooden posts they supported; thus, each timber could only have measured approx. 19 x 18 cm2. More recently, however, it has been suggested that these sockets may instead have received only a tenon-like projection from the base of each post, rather than the full timber. Thus, each post could theoretically have been substantially broader, approaching the dimensions of the stylobate blocks themselves (Fig. 9). This would have allowed the central structure to bear considerably greater weight, increasing the height to which the tower could plausibly have been raised. A taller tower would, in turn, have significantly extended the visible horizon from the site. Although not facilitating communication between neighbouring installations, such height would have supported sustained observation of the surrounding seascape.
Figure 9: Visual demonstration of the potential difference between two methods of securing timbers within stylobate blocks. Image produced using ChatGPT 5.2 by author to demonstrate the suggestion of White, 2022, p.58.
It is thus conceivable that, rather than signal stations used to raise a local alarm against small-scale raiding parties, the tall towers may have been longer-range maritime watchtowers constructed to provide early warning of hostile vessels approaching Britannia from across the North Sea. Large seafaring fleets travelling from northern Europe would have relied upon prominent coastal landmarks such as Yorkshire’s high cliffs to navigate south towards more valuable targets. Elevated headlands such as Carr Naze would therefore have offered ideal vantage points from which to observe such movements well before landfall.
Crucially, the transmission of this information need not have depended upon a fixed system of visual signals. Although unlikely to outpace small raiding vessels, it is conceivable that a mounted messenger could travel fast enough to deliver a sufficiently early warning to suspected targets of larger, and thus slower, seafaring vessels – especially if they were spotted early enough by the taller lookout towers. Within this framework, the clifftop towers function not as signalling relays, but as points of detection, integrated into broader systems of communication that relied upon human mobility rather than continuous intervisibility.
Ultimately, the true function of the Carr Naze complex — and of the wider group of Yorkshire clifftop sites to which it belongs — remains elusive. In the absence of direct archaeological or literary evidence, any interpretation must remain provisional. Whether understood as elements of a signalling network, maritime watchtowers, or perhaps more likely, installations that fulfilled multiple roles over time, these sites resist definitive classification. Yet, despite this resistance to interpretation, the Carr Naze stylobates are themselves evidence of the value assigned by the Romans to the Yorkshire coast – a value which warranted investment in defensive infrastructure and a dedication of manpower.
Upon reflection then, whatever their original function, perhaps the five Roman stylobates are not devoid of modern purpose in the Crescent Gardens. From their flowerbed, the stones act as quiet reminders of the rich history and enduring value of the small stretch of coastline they once guarded.
Bibliography:
RIB = Roman Inscriptions of Britain.
Primary Sources:
Ammianus Marcellinus, History, Vol. III: Books 27–31, ed. and trans. J.C. Rolfe (Cambridge, MA: Harvard University Press, 1950).
Secondary Sources:
Birks, S.; Evans, J.; Lankester, P. (2000) Roman Filey: Excavations at Carr Naze (York: York Archaeological Trust).
Brickstock, R. et al. (2000) ‘Excavations on the Site of the Roman Signal Station at Carr Naze, Filey, 1993–94’, Archaeological Journal 157, 79–199.
Cortis, W. (1858) ‘Account of Roman Remains at Carr Naze, Filey’, Proceedings of the Yorkshire Philosophical Society 1, 29–36.
Donaldson, G. (1988) ‘Signalling Communications and the Roman Imperial Army’, Britannia 19, 1–29.
White, A. (2022) ‘“Turrem et Castrum”: Some Fresh Thoughts on the Roman Fortlets of the Yorkshire Coast’, Britannia 53, 41–68.
Southern, P. (1990) ‘Signals versus Illumination on Roman Frontiers’, Britannia 21, 233–42.
This post was written by Noah Fenwick, who recently graduated with a Taught Masters at Warwick on the Material Culture of Ancient Rome. Noah's interests include numismatics, iconography, and how the Romans engaged with collective memory through these mediums. He is also keenly interested in popularising classics in the modern world, and the curation of Roman material culture.