Almost 16 years have elapsed since the publication, in 1959, of Dr. Philip Corder's paper on the structure of Romano-British pottery kilns. During this time much new evidence has come to light, especially in the upper Nene Valley, and the series of experimental firings carried out at Wattisfield, Boston, Leeds, Barton-on-Humbers and Chichester has greatly enhanced our knowledge of the techniques employed. The present paper is primarily concerned with pottery kilns of the pre-Flavian period in the Nene Valley and seeks to bring together, under one heading, the new and largely unpublished material from this region.

In the mid-1950's attempts were made to solve Leicester's traffic problems by the construction of an inner ring road by-passing the present city centre around the clock tower. The projected line ran through the Roman and medieval towns of Leicester from the north-east corner to the medieval south gate, passing just west of the ancient site of the High Cross (FIG. I ). The engineering work for this road was expected to destroy substantial archaeological evidence of the history of the town, and the City of Leicester Museums, in association with the then Ministry of Works, planned a jointly-financed programme of excavations. The first was undertaken in Blue Boar Lane by Mr. J. S. Wacher. In 1961 Leicester Museums appointed a full-time Field Archaeologist to direct the work of rescue in advance of road building. The work recorded in this paper under 1961–4 was directed by M. G. Hebditch: that under 1965–6, and again under 1971–2, by Jean Mellor. In all its excavations, the Museum staff enjoyed the co-operation of colleagues in other departments of Leicester Corporation, especially those of the Engineer and the Estates Officer. A besetting problem has been the limited area for excavation available at any one time (because of irregular demolition and clearance of the site and the need to leave substantial areas for the support of adjoining buildings).

The products of the New Forest Roman pottery kilns have been known since the middle of the nineteenth century through the work of Akerman, Wise and Bartlett, and in this century through Sumner and Cunliffe; but, as yet, there has been no attempt to define the distribution of the individual types made in the New Forest. As a result certain late colour-coated wares in distant parts of Britain have been mistakenly attributed to the New Forest with little consideration of obviously different fabric characteristics. Products of the New Forest kilns have often been confused with those of the Oxfordshire kilns discovered at Sandford, Rose Hill and Dorchester, Cowley and Headington; the pottery is similar in form, and similar but by no means identical in fabric, to that of the New Forest and so a clearer distinction between the products of the two industries has become necessary. The aim of this article is to show the nature and distribution of the New Forest pottery types and the place of the industry in the development of colour-coated industries in Roman Britain.

Rivenhall lies 10 miles south-west of Colchester and a mile north of the main Roman road to London, in a sheltered valley adjacent to copious springs which feed a tributary stream of the river Blackwater (FIG. I). The subsoil geology is mainly boulder clay containing random gravel pockets; this is capped by brickearth in places. There are numerous prehistoric and Romano-British sites in the vicinity.

The late Roman defences which are found in many of the chief cities of France have long been remarkable for similarities of design. In the northern and western parts of France many of the cities were protected by thick, high walls, built in rubble concrete with a facing of ‘petit appareil’, interspersed with rows of tiles usually in double or triple courses. The foundations were of large unmortared stone blocks, which were, as often as not, re-used from earlier buildings destroyed in the barbarian raids of the later third century or else deliberately demolished to provide a solid foundation for the new walls. The whole circuit was strengthened by the addition of external towers, usually of one build with the walls. Fortifications of this type are also found in other provinces of the western empire.

The coins which form the basis for this discussion, summarized in Tables I to IV (facing p. 252), have been published in more detail in the Numismatic Chronicle (Southern France 1964, northern Italy 1971, northern France 1972) and in Britannia iii, 269-76, where full details of the nature, location and peculiarities of the groups may be found, together with a subjective estimate of each group's archaeological reliability. For ease of reference the full names of the groups, and the periods into which the coinage has been divided is repeated:

Mediobogdum is founded on comparatively level ground half-way down the west side of Hardknott pass, above the 800 ft. contour. It is an extremely wet site, peat having formed up to o·60 m thick in hollows between the rock outcrops or where collapsed walling has blocked the natural drainage. It was by chance in digging a drain through the retentura to improve the upper part of the site behind the granaries that the first piece of leather was found in 1965. This is thought to be part of a soldier's jerkin.

RIB 601 is a sandstone altar which was discovered in 1794 in the churchyard at Halton-on-Lune, three miles upstream of Lancaster. The stone, which is broken on its right-hand side, records a dedication made to Mars by a certain Sabinus and the men of a Numerus Barcariorum. It has been generally assumed that this is the same unit as the Numerus Barcariorum Tigrisiensium placed in the Notitia Dignitatum at ARBEIA (South Shields); this reference in the Notitia has been plausibly taken to apply to the late fourth century. This assumption of identity therefore entails postulating an earlier date for the unit's stay at Lancaster. As far as purpose is concerned, the usual assumption is that the function of the barcarii was one of transportation and lighterage. It is the purpose of this note to re-examine both assumptions.