Dense pasts: settlement archaeology after Fox's The archaeology of the Cambridge region (1923)

Abstract Cyril Fox's publication The archaeology of the Cambridge region (1923) is celebrated as a milestone in the development of landscape archaeology. Its centenary invites reflection on Fox's approach to landscape and on the development of knowledge about the archaeology of the Cambridge region over the intervening years. Here, the authors compare the evidence available to Fox with the results of three decades of development-led archaeology. The latter have revealed very high numbers of sites, with dense ‘packing’ of settlements in all areas of the landscape; the transformation in knowledge of clayland areas is particularly striking. These high-density pasts have far-reaching implications for the understanding of later prehistoric and Roman-period land-use and social relations.


Introduction
Cyril Fox's influential volume The archaeology of the Cambridge region was first published in 1923. Its centenary provides an appropriate occasion to consider how approaches to landscape studies and settlement densities have changed over the years. Here, we offer an experiment in site-density data. Employing Fox's study as a datum, our aim is to draw attention to the high-density settlement patterns identified through 30 years of development-related fieldwork in the Cambridge region and the resulting flood of archaeological data. In particular, such landscape-scale investigations now have the potential for the near-total recovery of archaeological settlement patterns. The resulting high site densities raise pressing socioeconomic and environmental questions concerning the ways in which material culture and ideas were shared and natural resources exploited. The similar rise of development-led archaeology and the application of landscape-scale evaluation in many other parts of Europe and further afield is leading to a greater appreciation of the complexity of early settlement networks; in this article, brief comparisons are made with other landscapes in Southern Britain and Western Europe. There is also a broader historiographical dimension to our study. Fox's approach to the archaeology of the Cambridge region laid the foundations for his later book The Personality of Britain (1932). The broad characterisation of Britain into Highland and Lowland landscapes articulated there has cast a long shadow over studies of Britain's past (e.g. Gosden & Green 2021: 108-9), influencing regional and landscape archaeologies more generally.

Mapping: an engine of research
Ground breaking in its time and widely influential, first and foremost Fox's book has been a major resource for regional research in South East England ( Figure 1A). In acknowledgement of its contribution, the Cambridge Archaeological Unit (CAU) has initiated its own 'New Archaeologies of the Cambridge Region' series, with three volumes published to date (Evans et al. 2008Evans & Lucas 2020). Fox's study focused on an approximately 1900-square-mile (3058 km 2 ) 'frame', centred on Cambridge and extending for 20-25 miles (32-40km) beyond the city, encompassing portions of adjacent counties ( Figure  2A-C). A retrospect on the volume by its author (Fox 1947), and a biography of Fox (Scott-Fox 2002), provide further details of the survey's background and methods.
In recent years, several 'big data' studies have used digital mapping to mobilise the now vast body of information generated by development-led archaeology in England (e.g. Rippon et al. 2015;Smith et al. 2016;Gosden & Green et al. 2021). Among the earliest attempts at such distributional research was a nineteenth-century initiative by the Society of Antiquaries of London (e.g. Evans 1876Evans , 1893. Archaeological mapping, however, only came to the fore in the first decades of the twentieth century, with Fox (1947: 1-2) acknowledging the influence of O.G.S. Crawford, H.J. Fleure and J.P. Williams-Freeman. Yet, The archaeology of the Cambridge region was of an altogether different order. At 360 pages, it was much more comprehensive, with its approach considered to be unique in Europe (see Fox 1947: 16 on Oswald Menghin's review; the title of this section ('Engine of research') derives from Clark's 1933review of Fox 1932.
The volume was lavishly illustrated and Fox effectively crowd-sourced funding for the figures, raising £100. The visual material includes a set of accompanying period maps (plus nine maps reproduced in-text) and 37 plates showing material from the University Museum, many featuring five or more artefacts, amounting to a near-corpus of the region's findings at that time. Fox's main geographical and environmental sources were the Geological Survey and Tansley's (1911) The types of British vegetation. Essentially corresponding to permeable chalk, greensand or sands, and gravel geologies on one hand, and impermeable clay soils on the other, Fox sub-divided the region into 'open' and forested lands. These areas were respectively portrayed in white and green on the maps ( Figure 2B, C). Characterising them as areas  of 'primary' and 'secondary' settlement, Fox held that significant inroads were only made into the forested lands through "a certain standard of civilization" and really only occurred in Roman times (Fox 1923: 224-5, 315).
Beyond these designations, Fox exercised further landscape caricature, with the middle/ lower Ouse Valley deemed 'cold claylands'. In contrast, the upper reaches of the valley and that of the River Cam-plus the Ivel and the eastern fenland borders-were appreciated as fertile 'cornlands'; the lack of attested past activity from the fenland's marshes being attributed to malaria (Fox 1923: 9, 312-15). Based on subsequent findings and environmental research, Fox modified his open and forested categories. In the 1948 edition, he saw their interrelationship as one of gradation, respectively areas of 'easy' and 'difficult settlement'. Nevertheless, the potential of claylands for settlement was still severely underestimated (with the extent of forest cover exaggerated): Clay may have been the only, and therefore 'primary', soil on which man settled in a few areas but it must always and everywhere have been 'difficult'; iron-hard in droughts and gluey in the rains (Fox 1948: 20).

Low-density pasts
The discussion that follows is largely confined to the archaeology of the Iron Age and Roman period (c. 700 BC-AD 400). For our comparative purposes, we focus on these periods as they are characterised by high settlement densities and, often enclosed by ditches, their identification is relatively straightforward when compared with those of earlier periods. Details of the published sources for the site distributions discussed here are provided in the respective figure captions. However, owing to the nature of the data (e.g. duplicate entries in overlapping datasets), the quantification offered here should be considered indicative of trends rather than taken as absolute values. Indeed, given the pace of development in the Cambridge region, still further sites will have surely come to light since this article was written.
First, though, consider Fox's Bronze Age map ( Figure 2C), as it clearly expresses the dominance of the 'white' chalklands in the archaeological recording of the early twentieth century. Compared with the negligible finds from the heavy 'green' lands, more than 145 barrow symbols dot the length of the downlands up to the Brecks near Thetford. These barrows, in turn, were held to reflect the route of the Icknield Way, an early trackway running from Wessex to East Anglia along the chalk escarpment (see also Fox 1923: Sketch Map D, andEvans et al. 2018: 424-5). Recovered during quarrying and peat-digging, a distinct cluster of Bronze Age finds lay west of the Cam, extending from Cambridge north towards Wicken. Amid the map's indications, however, there are just four marked as a 'Living Floor/Settlement', located approximately 15-25km apart. Fox and other early practitioners were clearly aware that this settlement record was incomplete. Yet they do not seem to have conceptualised its extent and it is therefore difficult to appreciate just what kind of pasts they envisaged. After all, gauging the 'completeness' of an archaeological sample is contingent upon expectation. Ethnographic studies or colonial administrative records could have informed Fox and other early-day researchers of comparative settlement densities among 'tribes' or peasant communities (e.g. Steel 1955), but these were evidently not consulted.
With known settlements so widely spaced, the key question arose of how these sites could have shared rhythms of chronological change and material culture styles. How could distant Dense pasts: settlement archaeology 'dot communities' have been linked? Accordingly, transient metal smiths and potters served as mechanisms of transmission. Under the influence of palaeoeconomic approaches, in the 1970s and 1980s transhumant cycles also came to be widely invoked-at least for prehistoric periods-providing a deus ex machina explanation for the cultural integration of far-flung communities (Evans 1987). The acceptance of low-density pasts was furthered by a belief that the prime landscape locations were largely occupied by extant historic villages and towns (e.g. Hoskins 1973). In turn, the long-term continuity of these settlements reinforced ideas about the limited number of other past sites. This perception only began to be seriously challenged with the impact of extensive aerial photography (Taylor 1973: 28-33) and investigations in advance of early motorway construction. By the later 1970s and 1980s came a realisation that there was 'more', with documented contemporaneous settlements only 0.5-2 miles apart (0.8-3.2km ;Taylor 1985: 25).
A rare study of that era offering an explicit consideration of local settlement densities is a 1973 paper by Davidson and Curtis on an Iron Age site at Trumpington's Plant Breeding Institute. Based on its discovery settlement, amounting to only the ninth such site then known within the Cambridge area, the authors boldly announced: Supposing that all or most of the Iron Age sites in this area have now been located, we can consider the implications if they represent an approximation to the total economic exploitation of this area during that period.
Davidson and Curtis's map covered approximately 58km 2 ( Figure 1C). Today, the predicted number of Iron Age sites within that area would be in excess of 100: a more than tenfold increase (Evans et al. 2008: 183-6). This latter estimate derives from the results of extensive fieldwork arising from the extraordinary pace of development in the vicinity of Cambridge ( Figure 1B). For several decades, hundreds of archaeological interventions every year have been undertaken across Cambridgeshire, making the city's environs one of the most intensively investigated areas within Britain (see Smith et al. 2016: 386, figs 12.2 & 12.3). Development-funded fieldwork and, with it, the ability to sample-evaluate and excavate enormous tracts of land, is now common across much of Europe (e.g. Vander Linden & Webley 2012). The implications of the results of landscape-scale geophysical surveys and trial trenching are nothing less than revolutionary. While such techniques cannot recover all site types (e.g. biased to cut-feature archaeology), when compared with earlier fieldwalking surveys, for the first time there is the potential for near-total site recovery (Evans et al. 2008: 198-200 and see 181-6 for recent fieldwalking and cropmark recovery rates).
Within the Cambridge region, awareness of the potential archaeological significance of the claylands really only developed in the later 1990s. This appreciation was first prompted through investigations in advance of house building on the Isle of Ely and, then, more widely by aerial photography (Mills 2007) and during large-scale development north of Cambridge. Compared with gravel and chalk, the attraction of clay-based soils-for early communities having the plough technology to cope with them-was their fertility. These soils also have higher and more stable water tables compared with undeveloped or depleted gravel soils, which are more prone to summer drought. The delayed recognition of the extent of archaeology in the region's claylands is easily explained: heavy soils 'lock in' finds, making them less susceptible to detection during fieldwalking, and sites there are only rarely visible in aerial photography. In contrast, archaeological features in the 'open lands' have been much more readily apparent; regionally, the early detection of sites focused on the Cambridge chalk downlands, mirroring the situation in Wessex.

The land behind Cambridge
What constitutes a region? Potentially defined by geographical, environmental, political and historic land-use or other criteria, regions are not given. Fox's regional frame was centred on Cambridge, and its extent-determined by the 20-to 25-mile (32-40km) limit that could be reached by bicycle for site visits-maximised its geographical variability (Figure 2A-C). Recent national archaeological syntheses have used other definitions, and, accordingly, Fox's frame overlaps with theirs. Reading University's 'Rural Settlement of Roman Britain' project used Natural England's so-called 'Natural Areas', with the result that while most of Fox's region falls within their 'Central Belt', his chalkland-south portion lies in their 'East Region' (Smith et al. 2016: 15-16, fig. 1.5; see also Rippon et al. 2015). In turn, although the former broadly corresponds with the 'Central Province' defined by Oxford University's 'English Landscapes and Identities' project (EngLaId; Gosden & Green 2021: 56-7), intended to cross-cut set regions, their case-study transects bracket Fox's region (Green & Creswell 2021: 2, 26).
To appraise the degree of change in archaeological knowledge between Fox's time and the present, several case-studies are presented. Starting with the land around Cambridge, this involves the entirety of Fox's region. Therein, his 1923 map documents 222 Iron Age and 559 Roman-period entries (Figure 3, left). By 2010/2015 (see Table 1 for the basis of these data), based on aerial photography and fieldwork-including by the original Fenland Committee and later Fenland Survey programmes (Phillips 1970;Hall & Coles 1994)these numbers respectively had risen to 1309 and 1248: approximately six-and two-fold increases (Figure 3, right; Table 1). That said, Fox's map also includes hoards, spot-finds and other entries. If these are omitted, and comparisons restricted to the settlements, his numbers reduce to 11 Iron Age and 37 Roman-period sites versus 236 and 260 sites, respectively in the 2010/2015 dataset: the equivalent of approximately twenty-one-and seven-fold increases for the Iron Age and Roman period respectively.
Aside from along the south-western fen-edge, much of this increase in attested settlement activity has been documented along the river valleys and their edges, primarily in advance of quarrying and the expansion of urban centres. The updated distributions could be read as comparable to Fox's high densities along the chalklands. Yet, crucially, this has generally not been upon the downs proper, but its north-western skirtland. Comprising a band of more mixed geology, its northern extent corresponds to the eastern fen-edge and, towards the south-west, the Cam Valley. Having more level, terrace-like relief, the foot of the downlands would have also offered abundant water-sources (i.e. aquifer spring-lines).
Where, for both periods, a significant difference occurs is across the clay soils north of Cambridge: the 1085km 2 of the Bedfordshire and Cambridgeshire claylands. There, the increases in the numbers of settlements recorded by Fox and in the 2010/2015 dataset are fifteen-fold for Iron Age sites (3 versus 45) and sixteen-fold for Roman sites (6 versus 96).
Although these increases demonstrate substantial rises in settlement numbers throughout the region, they still under-represent the current state of knowledge (i.e. excluding evidence collected post-2010/2015), especially across the northern claylands. Consequently, our next case-study focuses on a 383km 2 swathe of the claylands north 'behind' Cambridge, encompassing all records from the Cambridgeshire Historic Environmental Record (HER; Figure 4, area outlined in blue). As of 'now' (2021), it has in the order of 415 and 590 Iron Age and Romano-British entries, respectively (not just settlements). However, through road-improvements and mass-scale house-building programmes, there six projects alone-together extending over more than c. 4074ha and amounting to a 10.6% area  sample-contribute 103 Iron Age and 56 Roman sites, and indicate average square-kilometre densities of 2.5 and 1.4 settlements, respectively (Table 2). For the purposes of 'framing', that figure's mapping (Figure 4, top, red-outline) also shows the wider site distributions known along the fen-edge to the north and, to the south-east, in the Cam Valley and downlands. These, however, are excluded from the density calculations used here on the grounds that these areas have yet to be subject to intensive, large-scale evaluations and, therefore, do not offer a comparable statistical basis. Within this study area, bounded by the River Great Ouse in the north and its Ivel tributary to the west, investigations of the northernmost stretch of the modern A14 route, and, to the south, at Love's Farm perched on the edge of the clays above St Neots' floodplain (Hinman & Zant 2018), emphasise the role of river transportation and its influence upon settlement densities. Indeed, Love's Farm has the highest Iron Age settlement density at four settlements per 1km 2 ; the Romanperiod level is also high, with near-continuous riverside settlement documented extending south from it (Hinman & Zant 2018: fig. 1.4; see also Evans et al. 2018: fig. 6.30).
Reflecting its Roman town-hinterland location, it is actually the Greater North West Cambridge area that had the highest Roman settlement density. By the same token, over much of their lengths, the A14 and A428's fieldwork closely followed the projected routes of major Roman roads: the A14, the Via Devana-linking Roman Cambridge and Godmanchester-and the A428 (running west from Cambridge), Margary's Road 231 (Evans & (Wright et al. 2009), and Longstanton/Northstowe situated on an inland gravel terrace, none of the six-projects' tracts amount to a 'pristine' claylands sample. While Cambourne's densities can be considered the most broadly representative, there will always be other factors, and geographical conditions are unlikely to have ever been the sole determinant of settlement. Acknowledging such caveats, and that averages invariably mask variability, we can calculate that some 957 Iron Age and 536 Romano-British settlements lay within the claylands' 'frame' outlined in red on Figure 4 (top). Indicating a doubling of the known HER Iron Age entries for this area, this seems entirely plausible. So does, too, that the calculated total number of Romano-British settlements is approximately half that of the Iron Age. Yet, what of the discrepancy that our calculated figure for Romano-British sites is slightly less than the HER's figures (536 versus 590)? This can be explained by its duplicate same-site entries and because it also includes stray finds and not just settlements. The estimate based on the results of the six development projects can, therefore, be considered the more sound. These claylands have alternatively been termed the Southern Plateau and Western Plain, reflecting the area's relief. Albeit a subtle topography, the area features small valley troughs and rises, and is bisected by streams and brooks. At points having mixed till-derived cover, the character of the clays varies from lighter Kimmeridge and Boulder Clay to the heavy Gault and Ampthill classifications; the latter were evidently avoided for settlement where possible. There are also 'inland' gravel terraces apparent in the distributions mapped along the route of the A14 works, where two-thirds of the Iron Age and Romano-British settlements lay densely packed on the gravels-amounting to less than half the route's length-while, on the clays, settlements were up to 4km apart (this, in part, reflecting 'distortion problems' of employing linear transects as a basis of area-representative samples).
Thus far, we have cited bald figures that take no account of variable period lengths or settlement duration. The Roman occupation of Britain lasted just over 350 years, compared with the approximately 750 years of the Iron Age. The latter's settlements are generally somewhat smaller than those of the Roman period (on average 1.3 versus 3.4ha, respectively, based on the results of the six projects). With plans ranging from sub-square to ovoid/ circular-some elaborated by internal sub-divisions and/or concentric circuits-many of the Iron Age enclosures are, nonetheless, small and simple. Some yield only meagre material assemblages and were probably not permanently occupied, their use possibly related to livestock.
The Roman-period settlements also show significant variety. Employing the 'Rural settlement of Roman Britain' nomenclature (Smith et al. 2016: 20-43), the majority are familial-scale farmsteads, with more than half having Iron Age origins. Other Roman-period sites fall into the more 'complex' farmstead category, whose larger plans often have further compound sub-divisions, plus a wider range of facilities and building types (e.g. aisled buildings, ritual settings, possible market functions). As well as a few known villas, there are also several large, village-sized roadside settlements (Smith et al. 2016: 192-206).
In any mapping of settlement densities, the scaling of symbols must reconcile cartographic legibility with on-the-ground realities (Evans et al. 2008: fig. 4.8). This was an issue of which Fox was fully aware (1947: 7; see Figure 5). Also, distributions are calculated based solely on the centre points of settlements, when-given the variability of their footprints-it is the interval between contemporaneous sites that is paramount.
While there are instances of Middle/Late Iron Age settlements 'bunching' in relation to local territory boundaries, they generally display relatively uniform distributions. Yet, where there was choice between gravels and clays, the Roman-period sites evince greater sensitivity to the possibilities of these soil types, preferring to locate settlement upon betterdrained gravels and farming the more fertile adjacent clays. As a result, during the Roman period, portions of the inland gravel terraces were packed with settlement. In compensation for the necessary scale and 'shape' of their off-terrace agricultural holdings, there was then a greater interval stand-off with neighbouring settlements on the surrounding clays. This is evident in the mapping from the Longstanton/Northstowe and North West Cambridge investigations, which both straddle the same inland gravel 'rise' (Figures 6 & 7). In the latter area, Roman-period settlement was near-spatially continuous, including a major roadside complex (Bunker's Hill, Site IV). In contrast, at Longstanton/Northstowe, the Roman-period settlements were more widely spaced (c. 1km apart), but the three main ones were all very large (Aldred 2021: 111-55): in the north, Fishpond Cottages' complex farmstead (4.3ha), then Inholmes, another of that designation (10.6ha), and, to the south, Beck Brook was an enormous crossroads settlement (16.5ha). Of a size rivalling that of contemporaneous small towns and featuring 'quality' buildings, the latter likely had an estate-administrative function.
Crucial is the much greater distance to the next-neighbour off-terrace settlements southwest of Longstanton/Northstowe. There, adjacent to the route of Via Devana/A14, the one saw both Iron Age and Roman-period occupation (TEA 38/Site 12); the other, a new Roman-period foundation, was probably a villa (Slate Hall Farm/Site 26; Evans et al. 2008: 174-81, figs 3.21 & 2.3). The intervening gap of 2-3km between the gravel terrace and these sites-probably used for agricultural production-was quite different from the situation in the Iron Age. Dotting the landscape more regularly and having an average spacing of approximately 400m, the earlier settlements display no equivalent clustering upon the gravels.

Near neighbours: high-density pasts
Fox and other early-day researchers could not have conceived of the scale of site densities identified by either the Longstanton/Northstowe or North West Cambridge projects. While the former's main Roman-period settlements would eventually be detected as cropmarks before development work, this was not true of the North West Cambridge area, and the full settlement densities in these 'heavy lands' have only been registered through recent multi-technique evaluation fieldwork.
The archaeology of the Cambridge region remains a remarkable achievement, drawing together a wealth of findings. In reference to Rackham's introductory landscape-as-library quotation, Fox's book greatly contributed to making the region's landscape more intelligible. Contingent upon the state of knowledge in the early twentieth century, today many of its interpretations are, of course, dated and explicitly reflect that era's emergent environmental/ geographical determinism. Nevertheless, landscape remains archaeology's abiding interpretative framework, and Fox's study continues to prompt and focus thinking. Its basic land-use units-river valleys, the chalk and claylands-remain valid, each demonstrating distinct landuse histories. It is just that, unsurprisingly, Fox's chronology for the claylands was much too recent.
With so much excavation now undertaken, the past has become a far more complicated place. By no means are such high settlement densities as presented here restricted to the environs of Cambridge, and comparably high-density landscapes are now known across much of south and east England (Figure 8, left; e.g. Smith et al. 2016: 21-2, 385-96;Aldred et al. in press). Indeed, based on site numbers alone, some of its Roman-period settlement densities were as high as anywhere in the Northwestern Roman provinces, even if villas did not develop to the same degree (e.g. Fulford 2020); only the Auvergne region in France has produced evidence of higher densities (Trément et al. 2018; see e.g. Mattingly & Witcher 2016;Figure 8, right).
As distribution maps fill with dots, gone are the days of simple either/or patterning, and, understandably, there might be a nostalgia for a time when matters were seemingly more straightforward. For example, Fox's neatly geometric Roman roads are a far cry from the lattice of routes attested by the North West Cambridge project (though, see Fox 1923: 319-20; Figure 9B). Similarly, Fox's annotation of his Sketch Map A, to provide an indication of the 'Beaker Folk's' arrival via the Wash-and the presumed long-distance connections facilitated by the Icknield Way-conveys a sense of narrative and even regionalities to his tightly clustered Beaker groupings ( Figure 9A). Now, however, as the dots accumulate and 'everything is everywhere', it is almost as if we lose our 'voice' and the ability to convey 'easy stories'.
Pending a comprehensive regional review, the challenges posed by such site numbers currently make it impossible to update Fox's study with complete confidence. Not only is there just 'too much', but also insufficient recent investigation of some areas. The downlands, for example, are well served by comprehensive aerial photographic coverage (Knight et al. 2018), but there has been insufficient excavation to modern standards. Trends and propensities are, however, apparent as we begin to grasp the dynamics of the claylands' colonisation. Substantial Neolithic settlements have yet to be identified on the claylands; nor are any early monuments such as barrows or henges known there. The Middle Bronze Age generally marks the first significant human activity and, thereafter, the claylands witnessed relatively 'light' settlement until the widespread Middle Iron Age influx. The means by which groups came to know and, eventually, to settle these 'heavy lands' require further research, as do the area's extra-regional connections and the differing character of some of the clayland settlements.
In reference to David Clarke's 1973 Antiquity article, these collective findings arguably amount to a further 'loss of innocence'. Based on vast-scale evaluation procedures, unknown in Clarke's time, an awareness of the sheer quantity of past settlement has now been achieved. This should be 'game-changing' (Evans 2012), but the implications have yet to be fully appreciated. When, over much of this region, people could have effectively waved to their neighbours, there is no longer a de facto need to evoke modes of mobility-wandering smiths or migrant potters-to account for widely shared material culture traits. Rather, the question of how such high settlement densities could have been maintained should command greater attention. Certainly, such densities have far-reaching implications, including the regulation of  territory/property, resource management and sustainability (e.g. wood supply) and, potentially, disease transmission (human and animal). They should even reconfigure the attention so often given to nuancing immediate settlement-location choice; in these 'landscapes of neighbours', a settlement might be where it was simply because so many other suitable adjacent locales were already occupied.