The basic texts for our understanding of the construction of Roman harbours are found in Vitruvius's De Architectura 2.6.1, and 5.12.2-6, written c. 25 BC. A significant portion of these passages concerns the technology of building harbour structures in hydraulic concrete, a mortar of pozzolana, lime and water combined with aggregate, that could be placed in inundated wooden formwork in the sea (Vitruvian formwork, reconstruction). The extraordinary strength and durability of this material, along with its ability to set underwater, made it crucial to the infrastructure of the Roman Mediterranean. This passage in Vitruvius was the starting point for the Roman Maritime Concrete Study (ROMACONS), co-directed by the authors, which has begun to bring new procedures to the analysis of Roman hydraulic concrete in marine structures throughout the Mediterranean. An introductory report on the background to the study of Roman concrete, the ROMACONS research design and procedures for collecting and testing concrete cores, along with the first scientific results of the physical and mechanical testing of our samples, has recently appeared (Oleson et al. 2004).

The first two campaigns of the ROMACONS project in 2002 and 2003 focused on the collection and analysis of cores of Roman hydraulic concrete taken from structures in the sea (see the reports at this website). In 2004 we attempted a reconstruction of the procedures used by the Romans to build inundated formwork in the sea and to prepare and place mortar and aggregate in the forms. Although Vitruvius discusses the design and construction of formwork in the sea as well as the preparation of mortar, many details are missing from his presentation. Precisely how was the mortar placed in the forms? Did the formwork itself have to be constructed with tight seams or even caulked in some way to prevent seepage of the mortar before it had set? How were the forms actually constructed in the sea? Was aggregate added to the mortar before it was placed within the forms, or was it added later? Was the mortar and aggregate tamped, and if so, how? Why do there appear to be such variation in the formwork that has survived at various archaeological sites? Why does overlapping of boards appear in some forms and not others? How many man-hours were required to create a cubic meter of Roman concrete? How long would it take for Roman hydraulic concrete to set in sea and how long to reach its maximum strength? Our understanding of Roman imperial construction along with observation of Roman harbour works suggested that a fixed building protocol did emerge for the use of Roman hydraulic concrete, while at the same time analysis of the evidence for ancient formwork indicates that many variables influenced its physical configuration (Oleson et al. 2004: 216).

Given the paucity of textual evidence and the difficulty of interpreting the archaeological data, the authors decided that the best way to resolve some of these issues was to build a small pila somewhere along the Italian coast using materials and tools available to Roman builders. For financial, logistical and practical considerations it had to be small by Roman standards (approximately 2x2x2 m), and constructed in a location that permitted the bulk of this block to be underwater while it cured, but with its surface always above the high tide mark.

Finding a suitable site was easy; gaining permission from the Italian authorities to construct a concrete block that might well last for centuries was a far more difficult task. Fortunately, the primary sponsor of ROMACONS, the Italcementi Group, maintains a marine testing station at the Brindisi branch of the Lega Navale (an Italian yachting association) in the Sino de Ponente of the harbour of Brindisi, Italy. At this location Italcementi scientists analyze the mechanical and physical properties of various mixes of modern hydraulic concrete (Italcementi test barge). Through the efforts of Dario Belottii and Massimo Borsa of Italcementi, permission from the club’s president, Admiral P. Fadda, was obtained for our experimental archaeological project to proceed at this facility. The work was carried out from 13 to 21 September 2004.

The location was ideal for our test pila (view of pila in marina). The approved site was near a concrete quay leading to a local rowing club and within 1 m of a floating dock of the Lega Navale in a depth of water of c. 1.7 m at high tide (location prior to construction). The concrete quay provided a workspace for our mortar trough and the storage of our building tools during the day. About 3 m above the site, the bulk quantities of lime, pozzolana, tuff and sand were stored in a section of the yacht club’s parking lot (raw materials). Using a rope and a basket, quantities of the building materials were easily lowered to the quay as required (lowering materials). Transport of the mortar from the trough to the formwork (once it had been constructed) required only a short walk. Moreover, the pila can be routinely checked over time by personnel from Italcementi in the course of their work on their own experiments. Our site was also deep in the well-protected harbour itself and thus not vulnerable to sea conditions, and the tidal variation also was ideal, c. 50 cm between low and high tides.

Itacementi generously supplied the materials necessary for our experiment: a special order of slaked lime putty (grasello di calce) selected to best match the composition of our lime samples taken from Roman structures (lime bag), pozzolana from Bacoli in the Bay of Naples (pozzolana), tuff from the same area (to be used as aggregate) (tuff), beach sand (to seal the modern contaminated mud and to provide the appropriate Vitruvian foundation for the first mortar layer), 0.30 m wide planks (c. 1.0 Roman foot) and beams of varying lengths for the exterior frame of the formwork (lumber). Unfortunately, we were unable to obtain the green, uncured wood most likely used for this purpose in antiquity. The planks and beams were made of a laminated kiln-dried reconstituted wooden material with a very low specific gravity, which complicated our efforts at placement in the sea.

The design of the formwork was based on an interpretation of the method described by Vitruvius 5.12.2 (Vitruvian form, reconstruction). Vertical piles at the corners were intended to support edge beams onto which vertically driven sheet pile planks could be fixed (Brandon formwork design). But due to the excessive buoyancy of the planks and beams available to us, it was impossible to drive the piles into the harbour mud. Changing the design strategy surmounted this problem. The planks were pounded in the harbour floor first, and the horizontal frames were then secured to them. Thus the formwork was constructed hull or shell first with a frame added later, in the sequence followed by Roman shipwrights (formwork walls). Although this modification worked in this instance, it is unlikely that formwork erected in this fashion could have withstood the force of the sea in an exposed site during the time it took to build it.

The seams between the planks were generally tight except in the east wall, since our first efforts at sheet piling were the least successful. Gaps also existed in the NE and SW corners. We came to realize how difficult it was to keep the planks tightly together as we drove them into the bottom. Sub-bottom obstacles such as rocks and miscellaneous harbour debris sometimes made the task very difficult, as it surely was for the Roman builders. We realized quickly that such unseen irregularities in the construction surface constitutes one reason why ancient formwork is often very irregular in design, In addition it became obvious that features such as overlapping planks on ancient formwork are a response both to the difficulty of determining the exact dimensions of a form prior to its completion and to the need to ensure a tight joint between all the planks. These ancient design anomalies represent efforts to repair unexpected problems that arose during construction. If a short gap or a poor joint required patching, a plank could simply be hammered into the seabed along the inside face of the existing wall (overlapping planks). Even with repairs, our seams were hardly watertight and would not become so even if swelling of the wooden planks occurred. We were concerned about the possible seepage of mortar through the seams in the walls of the form. Construction of the form required 37.5 man-hours (formwork complete).

Using the Vitruvian formula of two parts pozzolana and one part lime (lime and pozzolana), we mixed the mortar by hand in a quayside trough using mattocks, shovels and rakes—variations of which were in the kit of Roman harbour builders. We made a decision to use seawater for our mix rather than fresh, since we believe that was standard operating procedure in Roman times. Vitruvius was fastidious in his specifications of the character and quality of the pozzolana and lime that went to make up hydraulic concrete for use in the sea. If fresh water had been essential to the mix rather than easily obtained salt water, we believe he would have specified its use. We prepared a very stiff mortar that would not dissipate when dropped into the inundated form, and one that potentially would be stronger (prepared mortar). Such mortar is far too viscous to be poured down a tremie tube like modern concrete. A solution to placing such mortar into an inundated form was to use wicker baskets with two drop lines on the handles and a trip line attached to the base (basket with mortar). The design of the baskets was based on those illustrated in Roman construction scenes and on the baskets recently found at Pisa in the excavations of the Portus Pisanus. Once a basket had been lowered into the form, it was easily tipped and emptied by pulling on the trip line (basket being lowered) (Oleson, 1985 and illustrated in Hohlfelder, 1987)

We discovered it was possible to lower the basket into the water then easily manoeuvre it to the desired location by taking advantage of the buoyancy of the basket, and then to deposit the mortar precisely by use of the trip rope (basket after dumping). The mortar, which appeared to be only slightly heavier than water, came out of the wicker basket with a gentle tug, while efforts to empty a plastic bucket failed. The impermeable walls of the plastic container created too much suction. One man could lift, transport and dump a basket full of mortar (c. 15 kg), but only with difficulty because of the awkward shape and the need to untangle the three ropes. Two men accomplished the task easily (carrying basket). Adding a third man did not increase efficiency and made manoeuvring the basket more difficult. The deposits of mortar kept the form of the basket container after their release (lumps of mortar).

Once a layer of mortar reached a thickness of approximately 0.20 m, the surface was tamped by individuals using rakes and standing on planks placed on the external crossbeams of the form (tamping). The caementa, irregular pieces of aggregate broken from blocks of Bacoli tuff with a sledge, were measured by volume (tuff in trough), then thrown into the form piece by piece in an effort to cover the entire surface of the mortar unseen beneath the murky water in the form (throwing aggregate). Once distributed in this random manner, the aggregate was then tamped in the same way the mortar had been. This protocol was repeated until the concrete mix finally breached the surface.  The concrete became more visible in the murky water as the form began to fill allowing to verify that the practice of randomly throwing aggregate piece by piece did result in the even distribution pattern we had hoped to achieve (aggregate near surface).

By the end of 20 September (day 8), we had placed the last layer of mortar and aggregate and tamped it in place by walking on the soft mixture (tamping). The following day the concrete had set sufficiently to be walked on without any of the aggregate shifting or sinking deeper into the mortar (detail of concrete, walking on mortar). A final layer of mortar was place over the aggregate, levelled using trowels, and then paved with local tuff blocks (finishing surface, paving blocks). On Tuesday, 21 September our pila of c. eight cubic meters was finished after 273 man-hours of work moving 10 metric tonnes of materials. As far as we can tell this is probably the first such structure built with these materials and  techniques in the last 1,600 years or longer (finished pier).

Many of the questions we had hoped to answer with this experiment have been addressed. Trip baskets for the delivery of the mortar work efficiently and easily. The Vitruvian mortar retains its integrity in the sea with very little seepage between the planks of the form. The work of building a form and filling the form went smoothly and quickly even for a small team of completely unskilled labourers encountering a range of problems and finding their solutions for the first time. There is much more to add to the story of the Brindisi pila, but we shall wait until we have taken our first core and tested it in the laboratories of Italcementi before publishing a fuller, more scientific account. The pila will remain in its present form until March 2005, when the formwork will be partly removed (if that proves to be possible), and it will be cored for the first time. Another core will be taken in September of 2005 and probably one the following September as well to determine the rate of curing. The experiment appears to have been successful, but we shall have to wait sometime to learn its complete results.From Brindisi, we went on to Santa Liberata, where we cored one of the largest known Roman concrete pilae (see the next report).


We are deeply indebted to Dr. Luigi Cassar of Italcementi Group who has supported our research in so many ways since its outset. We also thank Mr. Dario Belotti, Ms. Isabella Mazza and Mr. Massimo Borsa of Italcementi for their invaluable assistance particularly in surmounting numerous logistical and practical problems, such as finding and purchasing the proper raw materials so this project could occur. We are also grateful to the Lega Navale and its president, Admiral P. Fadda, for providing the venue for this experiment and for every possible courtesy and consideration while we daily threatened to disrupt the normal activities of this club. The Italcementi Group, the Loeb Classical Library Foundation (Harvard University), and the University of Victoria and the University of Colorado, Boulder provided generous financial support. Lastly we thank Mr. Francesco Retta, Mr. Fabrizio Orlandino, Mr. Peppino Brescia, and Mr. Mario Colucci, all of Italcementi, Brindisi, who volunteered to work with us as their regular schedules permitted and share the travails of the ancient Roman harbour builders.


Hohlfelder, R.L., 1987, “Caesarea Maritima, Herod the Great’s City on the Sea.” National Geographic 171.2: 260-79.

Oleson, J.P., 1985, “Herod and Vitruvius: Preliminary Thoughts on Harbour Engineering at Sebastos, the Harbour of Caesarea Maritima.” Pp. 165-72 in A. Raban (ed.), Harbour Archaeology. Proceedings of the First International Workshop on Ancient Mediterranean Harbours, Caesarea Maritima, 24-28.6.83, BAR Int. Series 257. Oxford.

Oleson, J.P., Brandon, C., Cramer, S.M., Cucitore, R., Gotti, E. and Hohlfelder, R.L., 2004. “The ROMACONS Project: a Contribution to the Historical and Engineering Analysis of Hydraulic Concrete in Maritime Structures.” IJNA 33.2: 199-239.