Glastonbury: today, tomorrow, 2,250 years ago

Guest bloggers Alison Wylie and Bob Chapman write...

The Glastonbury Festival has been run (almost) annually since 1970 on the Worthy Farm in the west of England. It began as a modest two-day greenfield festival; the first opened the day after Jimi Hendrix died. Despite some bad patches – infamous rain and mud some years, scorching sun in others – it has grown into the most famous music festival in the world.

 Glastonbury festival at its  muddiest !

Glastonbury festival at its muddiest!

It’s now a three-day Festival (with a capital ‘F’!), hosting over 500 acts on multiple stages, including the iconic Pyramid Stage modeled on the great pyramid of Giza. It’s an eclectic mix! Check out the Telegraph’s list of 100 best Glastonbury performances and the Festival’s history pages. You’ll find Paul McCartney (2004) and the Rolling Stones (2013); Bruce Sprinsteen (2009) and Stevie Wonder (2010); Joan Baez at that first festival in 1970 and then again in 2008, alongside Leonard Cohen and Jay-Z; some high profile country stars like Johnny Cash (1994) and Dolly Parton (2014); and even the English National Opera playing Ride of the Valkyries in 2004.

These days Glastonbury regularly draws 175,000 fans who throng the stages and set up a tent city that stretches as far as the eye can see. All of which requires large-scale infrastructure, from food outlets to toilets, to lighting and dance floors, to drainage systems and waste management.

  Everyone    camping   …

Everyone camping

You’d think that all this activity over 50 years would leave plenty of material evidence of the habits of British revelers for future archaeologists to find. But, in fact, the Festival’s organizers are proud that, every year, the site reverts back to farmland. The stages and sound systems are dismantled and removed, an army of volunteers clears the mountain of refuse, most of it ending up in recycling and organic compositing facilities.

  …and striking camp – leaving    mounds of garbage

…and striking camp – leaving mounds of garbage

After all is done, the ‘streams and hedges remained unpolluted’, and Worthy Farm goes back to being … a dairy farm (Glastonbury history, 2004). In fact, the Festival takes a regular ‘fallow’ year off to let the farm regain its equilibrium – next year for example.

 The Worthy Farm cows are pleased! ( Telegraph 7 July 2017 )   Check out this video of frisky cows heading back to their pasture.

The Worthy Farm cows are pleased! (Telegraph 7 July 2017)

Check out this video of frisky cows heading back to their pasture.

If we lost all the historical records of the Glastonbury Festival – the posters, photographs, news coverage, reviews, facebook posts, personal reminiscences – and assumed that the site had a continuous history of use as farmland for two thousand years, what could the archaeologist of the future tell us about what went on at Glastonbury? Perhaps not much if, indeed, the organizers of the Festival succeed each year in leaving such a tiny footprint on the site itself. This is an ‘information-destroying process’ with a vengeance, to use Elliott Sober’s term of art for processes that wipe away traces of the past. Glastonbury and cases like it are the stuff of philosophical nightmares about the limitations historical knowledge, a worry that even our best knowledge of the past may be riddled with errors and gaps we can’t detect. This is an anxiety that structures internal debate among archaeologists (see ‘Working Between the Horns of a Dilemma’, Chapman and Wylie 2016) and has attracted the attention of philosophers inclined to be skeptical about the historical sciences (e.g., Turner 2007; see Currie 2018). Even if we do find a few remaining scraps of the festivities at Worthy Farm, how much could we do with them? Sir Mortimer Wheeler, famous for his advocacy of scientific field methods, described this as the ‘Diogenes problem’ (1950; see also Smith 1955): we might find Diogenes’ tub in the town square of Athens but how would we ever infer that it had been occupied by a philosopher making a point about renouncing worldly goods?

But should we generalize from such cases to all archaeological sites and the knowledge claims based on them? Our answer, in Evidential Reasoning in Archaeology (2016) is a resounding NO. Epistemic modesty, a willingness to hold archaeologists’ claims open to scrutiny, is certainly warranted. However, to draw sweeping conclusions about what isn’t possible misses what archaeologists do accomplish, working creatively, systematically, doggedly with what Adrian Currie refers to as ‘scrappy remains’ in his 2016 ‘Swamp Zombies’ blog. Looking at how archaeologists build evidential claims about past lives and actions throws into relief the varied strategies they use to exploit the capacity of material traces to show us where we’re wrong, sometimes expanding our interpretive horizons dramatically beyond what we’d assumed to be plausible, or even possible.

For an example of the capacity of material evidence to unsettle our assumptions about the past, let’s return to Glastonbury, this time to consider a (much) earlier settlement: a small Iron Age village located some seven miles from the Festival site in the peatlands of the Somerset Levels.

   Somerset Levels    – flooded.

Somerset Levels – flooded.

This Glastonbury, which predates the Festival by some 2,250 years, is famous among archaeologists, not just for its extraordinary preservation (which we’ll get to in a moment), but for its remarkable history as the focus of intense scientific study for over a century. The site has, in a sense, two histories: one stretching back to a little village millennia ago, the other unfolding over a century of archaeological excavation, analysis and speculation. The open-ended potential of material traces to yield new and unexpected insights into the past is what makes possible this latter history, this afterlife of Glastonbury the Iron Age Village. Crucially, traces never travel alone; it is in mobilizing fellow-travellers – theoretical and empirical – and fine-tuning this scaffolding as we go, that they get traction as evidence. It’s especially striking that this is possible even in the case of ‘legacy data’ collected by previous generations of archaeologists. Often excavated and recorded using now dated techniques and informed by assumptions long since abandoned, these existing stores of archaeological data are nonetheless pressed into service, yielding answers to new questions no one envisioned (see ‘Working with Old Evidence’ in Evidential Reasoning). While the results are never definitive, we believe the history of wrestling with the Glastonbury legacy data counters any categorical pessimism about archaeology and the historical sciences.

What kick-started this ongoing process of reanalysis were the Victorian-era excavations undertaken by Arthur Bulleid (from 1892), later collaborating with St. George Gray (1904-1907). They excavated almost the whole of Iron Age Glastonbury, exposing the foundations of a number of extraordinarily well-preserved timber, wattle and daub structures, and recovering a wealth of organic and inorganic artefacts related to metalworking, textile manufacture and agricultural production. Although justly famous as a uniquely rich site, this Glastonbury was also baffling; Bulleid and Gray described it as an ‘amorphous agglomeration’ of habitation structures surrounded by a timber palisade, together forming what they hypothesized was a ‘lake village’ built on a platform of tree trunks, brushwood, clay, peat and rubble.

 Victorian era excavations - also flooded(!)

Victorian era excavations - also flooded(!)

  …and a    standard reconstruction    of Iron Age Glastonbury as a ‘lake village’.

…and a standard reconstruction of Iron Age Glastonbury as a ‘lake village’.

A remarkable find, Glastonbury was cited in archaeological textbooks for the next sixty years as a unique example of a well-preserved rural Iron Age settlement. New evidence became old evidence.

By the early 1970s the New Archaeology had burst on the scene. This was a self-consciously positivist research program advocated by a younger generation intent on using scientific methods and tools to ‘do more’ with their data. They were aggressively optimistic, confident that if sophisticated theory were brought to bear and explanatory hypotheses systematically tested, archaeology could extend our understanding of the human past well beyond catalogues of material traces of repetitive activity, artifacts fabricated and resources exploited; it should be possible to infer the social, economic and political organization of these prehistoric communities. In an excellent example of this optimism, British-style, David Clarke – a fellow of Peterhouse College at Cambridge (UK) – disinterred the Glastonbury monographs that had been languished largely unstudied in the Haddon Library of the Cambridge Faculty of Archaeology and Anthropology for over sixty years.

 David Clarke, the man behind the model(s)

David Clarke, the man behind the model(s)

Not content to rest with the conclusion that the data reported by the Victorian excavators are inscrutably chaotic, Clarke took up the challenge of exploring them in a variety of new ways. His point of departure was a quite general argument for the possibility that order might be discerned in this ‘amorphous’ mass – and that it could plausibly be treated as evidence of patterns of daily activity and social relations, even cultural symbolism. ‘Time’s arrow’, he cautioned, most certainly ‘left a continuous and destructive trace throughout its 2000 year trajectory’, generating a great deal of ‘background noise’ (1972: 805-6). At the same time, he was confident that the ‘repetitive redundancy’ of activities on the site over the long period of its occupation could be counted on to generate patterning that persists, a signal still detectable despite the noise. Although ‘discarded artefacts and rubbish’ would have been subject to ‘constant buffeting by animate and inanimate forces’ – AKA ‘post-depositional processes’ – given the long time span and (again) the repetitive nature of these forces, Clarke thought that they could be modeled as a type of ‘Brownian motion’; a ‘buffeted particle’ would be displaced in one direction and then back again such that it could be expected to ‘remain within a small radius of its point of deposition’ (1972: 806). The upshot: at least at a composite level, the ‘original site structure and activity patterns’ of lives lived at Iron Age Glastonbury should be evident in the internal structure of the archaeological deposit.

Having laid out these initial assumptions – Clarke was all about making his interpretive process explicit – he undertook to reanalyze Bulleid and Gray’s data, building what he described as a ‘multi-scalar’ lattice of models. At a ‘models of data’ level he disentangled four distinct occupational phases, drawing on the stratigraphic data reported (often inadvertently) by Bulleid and Gray, and he sorted the architectural features and artifacts they described into functionally defined categories. He used these to build what he described as ‘analytic models’ of spatial, formal, and temporal patterning in the associations between these distinct types of find. For example, he was interested in determining what types of building remains clustered together, and what tools or fixtures co-occur with them. Clarke then used these models of the internal relational structure of the archaeological record – described in other contexts as ‘conjunctions’ – as the basis for identifying distinct types of occupational structures, for example, major and minor houses, ancillary huts, storage features, work floors, stables and so on. These were, in turn, the basis for Clarke’s signature model of the Glastonbury settlement as comprised of seven ‘modular’ units: household clusters of structures that recur at each phase of the site’s occupation which, he hypothesized, represent a ‘social and architectural building block of which the settlement is a multiple’ (p. 815).

 Clarke’s modular unit settlement model for Glastonbury, Fig 3.4 from  Evidential Reasoning  (p. 114): (1) Major House, (2) Minor House, (3) Ancillary Hut, (4) Workshop Hut, (5) Courtyard, (6) Baking Hut, (7) Workfloor, (8) Granaries/Storehouses, (9) Stables, (10) Sties/Kennels, (11) Wagon Stance (adapted from Clarke 1972b: fig. 21.1).

Clarke’s modular unit settlement model for Glastonbury, Fig 3.4 from Evidential Reasoning (p. 114): (1) Major House, (2) Minor House, (3) Ancillary Hut, (4) Workshop Hut, (5) Courtyard, (6) Baking Hut, (7) Workfloor, (8) Granaries/Storehouses, (9) Stables, (10) Sties/Kennels, (11) Wagon Stance (adapted from Clarke 1972b: fig. 21.1).

And so, Clarke replaced Bulleid and Gray’s ‘amorphous agglomeration’ with an ambitious hypothesis about the settlement’s layout, tracing a stable organizational pattern through a four-part history of occupational phases. This he augmented with models of how the Glastonbury villagers might have supported themselves, given evidence of the resources they exploited, and of their relationships with other Iron Age settlements, given finds made of materials that had to have been traded into the village. In this way Glastonbury was situated as one node in what Clarke suggested might be a regional settlement hierarchy. We summarise the key features in Evidential Reasoning (2016: 113-18).

Clarke published these ideas in 1972, and his title tells us something about his intent: 'A Provisional Model of an Iron Age Society and Its Settlement System'. He meant his Glastonbury study to be a modeling exercise: explicit in its assumptions, experimental in its methodology, and provisional in its results. Clarke saw this as long-term and open-ended experiment, a matter of pushing ‘one interim model towards the limits of its potential’ and inviting re-examination of its ‘chain of assumptions’ (p. 802). He hoped that his models would be a catalyst for systematic testing and expansive rethinking.

Clarke died prematurely in 1976 so was unable to pursue the many lines of inquiry he saw emerging from his ‘provisional model’. Nonetheless, his work was hailed as a significant contribution, and was the inspiration for the analysis of other seemingly ‘amorphous agglomerations’ of data from European and Scandinavian sites. Within a decade, however, critics dug in their heels and challenged virtually every aspect of his data analyses and the interpretive models based them. One of the first, John Barrett (1987), questioned Clarke’s reliance on the excavation reports published by Bulleid and Gray. Despite Clarke’s own caveat that the results of analysis are only as solid as the supporting data, Barrett objected that he had ‘accepted at face value’ the summaries published by the Victorian excavators. He hadn’t returned to the primary data – their field records and finds – and he hadn’t scrutinized their field methods and recording techniques. Barrett didn’t do this leg-work himself; it fell to later critics to dig deeper into the archives and surviving collections. But he did note a number of discrepancies and ambiguities in how Bulleid and Gray reported the location of the finds they had recovered. For example, artefacts were often identified as associated with particular mounds, without any indication of where, in these mounds, they were found, calling into question Clarke’s models of the spatial and temporal relations between these finds.

In a related and ultimately more devastating critique, Barrett called into question Clarke’s ‘Brownian motion’ assumptions about post-depositional process. As the impetus for reanalysis it was perhaps plausible that, on balance, the ‘buffeting’ forces acting on the material left behind by Iron Age villagers would effectively cancel each other out over two millennia, leaving it roughly where it had been deposited. Applied to the Glastonbury site, however, this is an empirical assumption that Barrett found highly implausible. He drew on background knowledge – from basic geology and other sites in the region that had since been excavated – to make a case that the post-depositional processes affecting the Glastonbury deposits were not limited to small scale displacement. The repeated waterlogging and desiccation of a peat matrix over two thousand years was well known to be capable of significantly disrupting the stratigraphy of such a site. Fissures open up when the mud dries out that can transport small finds up or down through many depositional levels, unless they are stabilized by structural features like the mound platforms the stratigraphy of which, Barrett had already shown, had not been systematically recorded by Bulleid and Gray. In short, Clarke’s post-depositional model gave him reason to undertake a reanalysis of the Glastonbury data, but it was too crude to warrant his models of spatial/temporal patterning in these data. And if these empirical models were questionable, there was reason to be suspicious of the models of settlement layout and social organization he based on them. Clarke’s model wasn’t built on sand, but waterlogged peat was just about as worrisome.

Before long, John Coles and Stephen Minnitt (1995) took up Barrett’s line of critique and argued, on the basis of close scrutiny of the primary data, that Clarke’s analysis was bedeviled by pervasive ‘misunderstanding and misreading’ of the original excavation reports. They reanalyzed the site’s phasing based on the relationships between features and radiocarbon dates and a much richer body of regional comparisons than had been available to Clarke. The result was still a four-stage occupational sequence, but over a much longer time-period. They also reassessed Clarke’s functional categorization of the various structures and features that made up his posited modular units. They found the details     problematic. In some cases the structures Clarke identified as components of a settlement module were not contemporary with one another. And functionally diagnostic artefacts and features that Clarke had relied on to identify distinct types of building, for example, the finds he took to be characteristic of stables or granaries, turned out not to be reliably associated with ‘stables’ or ‘granaries’. They concluded on this basis that Clarke’s distinctive model of a basic architectural and social unit, a modular structure repeated across the site and over time, was unsustainable. Finally, they objected that the environmental reconstruction on which Clarke based his ‘economic’ models of land use and subsistence practices was untenable. They cited a range of geophysical and paleo-ecological data – some available when Clarke was in the thick of his reanalysis, some of more recent vintage – that call into question Clarke’s assumptions about the village having been built on a platform in a lake, and about the access its inhabitants would have had to arable land.

In short, Coles and Minnitt (1995) declared that their 3.0 reanalysis had ‘eliminated for good’ Clarke’s provisional models of the social, economic and architectural organization of Iron Age Glastonbury.

Is that it? Is this a negative object lesson that marks the limits of archaeological inquiry? Should we conclude that Clarke had spun a vast web of speculative models which, within twenty-five years, had collapsed under the weight of unsubstantiated assumptions – a warning to all who might venture beyond what the pessimists in archaeology describe as a ‘narrow empiricism’?

Examining this second history of Glastonbury – the ongoing debate about what can and cannot be teased out of an unruly mass of Victorian-era legacy data – we think a more optimistic appraisal is warranted. Judged on the terms in which Clarke proposed his exploratory reanalysis of these data and his ‘provisional’ models of Iron Age Glastonbury, his ‘experimental exercise’ was in many respects a tremendous success. Although the specifics of his relational models are problematic, Coles and Minnitt do not ultimately return a verdict of ‘amorphous agglomeration’. Clarke’s first-pass at reanalysis provoked them to dig deeper and, cross-examining the primary data, they identified a great deal of non-random patterning in the features, architecture and artifact assemblages reported for Glastonbury. Indeed, they ultimately proposed their own modular model of the settlement pattern at Glastonbury, recalibrated to a longer occupational timeframe and refined to reflect their recategorization of key features and structures.

 Minnitt & Coles’  Lake Villages of Somerset  (1996), Somerset County Council. Inset: Coles & Minnitt, Industrious and Fairly Civilized: The Glastonbury Lake Villages (1995), Somerset Levels Project and Somerset County Museum Service

Minnitt & Coles’ Lake Villages of Somerset (1996), Somerset County Council. Inset: Coles & Minnitt, Industrious and Fairly Civilized: The Glastonbury Lake Villages (1995), Somerset Levels Project and Somerset County Museum Service

As a pioneering demonstration that it might be possible to retrieve useable information from legacy data, Clarke challenged conventional assumptions about the limits of archaeological inquiry, and he catalyzed a whole new line of thinking about prehistoric social organization. Recent critiques of structuralist (modular) models of domestic space in Iron Age settlements renew Clarke’s call to make interpretive assumptions explicit so they can be systematically assessed, directing attention back to ethnographic sources as well as to the archaeological data (see Rachel Pope 2007).

This is all very much in the spirit of Clarke’s ‘experiment’. His ‘provisional models’ may have been wrong, but they provided a crucial scaffold for archaeologists to frame the questions they would need to ask if they were to build a richer, more credible understanding of Iron Age Glastonbury. The lesson we draw from this second Glastonbury history is that inquiry in a field like archaeology is necessarily iterative – in the sense Hasok Chang has described in Inventing Temperature (2004). In the absence of ‘infallible foundations’, there’s no choice but to bootstrap-construct empirical ‘fixed points’ that you well know will need to be continuously reconstructed, that will have to be kicked away once they’ve made it possible for you to learn enough to know where you’re wrong. What this history should caution against is ever resting content with the assumptions you rely on to get an exciting line of inquiry off the ground.

So: what could we expect archaeologists of the future to tell us about contemporary Glastonbury, the Festival? One answer is that even if the Festival organizers are brilliant at covering their tracks and leave no trace of all the infrastructure, the massive influx of fans and the good times had by all, we should we assume that this deliberate an ‘information-destroying’ process is the rule rather than an exception? Another is that we can’t predict what will be possible. That depends on whether anyone raises the relevant questions, and whether they are sufficiently creative and dogged and lucky enough to recruit the resources they need to address them. Delineating the possible is, itself, a path-dependent historical process, and one that is widely distributed. It depends on what scaffolding – technical, interpretive, practice – archaeologists can build or appropriate, often from fields that have little sense of and no particular interest in archaeological questions. Think nuclear physics, for the radiocarbon dating Coles and Minnit drew, and geology and paleo-ecology that informed Barrett’s critique. At its best, archaeology is a vibrant trading zone whose well-functioning depends on all the contingencies of institutional structure, funding, disciplinary cultures and social dynamics. The central point is, we shouldn’t treat worst cases – no traces, no relevant scaffolding – as setting limits on what we can learn about the past.


Thanks to Adrian Currie and the Extinct team! We’ve never done a blog post before and especially appreciate Adrian’s help wrestling this one into shape.



References cited


Barrett, J. C. (1987) "The Glastonbury Lake Village: Models and Source Criticism," Archaeological Journal 144: 409-423.

Bulleid, A. and Gray, H. S. G. (1911, 1917) The Glastonbury Lake Village, Volumes 1 & 2, Glastonbury: Glastonbury Antiquarian Society.

Chang, H. (2004) Inventing Temperature: Measurement and Scientific Progress, Oxford: Oxford University Press.

Chapman, R. and Wylie, A. (2016) Evidential Reasoning in Archaeology: Bloomsbury.

Chapman, R. and Wylie, A. (eds) (2015) Material Evidence: Learning from Archaeological Practice,  London: Routledge.

Clarke, D. L. (1972) "A Provisional Model of an Iron Age Society and Its Settlement System," in D. L. Clarke (ed), Models in Archaeology, pp. 801-870, London: Methuen & Co.

Coles, J. and Minnitt, S. (1995) ‘Industrious and Fairly Civilized’: The Glastonbury Lake Village: Somerset Levels Project and Somerset County Council Museums Service.

Currie, A. (2018) Rock, Bone and Ruin: An Optimist's Guide to the Historical Sciences, Cambridge, MA: MIT Press.

Pope, R. (2007) "Ritual and the Roundhouse: A Critique of Recent Ideas on the use of Domestic Space," in C. Haselgrove and R. Pope (eds) The Earlier Iron Age in Britain and the Near Continent, pp. 204-228, Oxford: Oxbow Books.

Smith, M. A. (1955) "The Limitations of Inference in Archaeology," Archaeological News Letter 6(1-7).

Sober, E. (1994) Reconstructing the Past: Parsimony, Evolution, and Inference, Cambridge, MA: MIT Press.

Turner, D. (2007) Making Prehistory: Historical Science and the Scientific Realism Debate, Cambridge: University of Cambridge Press.

Wheeler, M. (1950) "What Matters in Archaeology," Antiquity 25: 122-130.