BUTSER ANCIENT FARM ARCHIVE 1973-2007 Archivist Christine Shaw
Home History of the Project Research Publications Image Archive
Back to top

The Pimperne House

a review by R. W. Hedge, C.Eng., M.I.Chem.E., B.Sc.(Eng), ACGI, Past Chairman of the Friends of Butser Ancient Farm

Reynolds, in Experimental Reconstruction in Harding, Blake and Reynolds (1993), describes in detail the thinking behind his first Great Roundhouse project and used in the construction of the best available interpretation of the structure of the Pimperne House in Dorset, UK. This was undertaken in the Butser Ancient Farm Demonstration Area in The Queen Elizabeth Country Park, in the shadow of Butser Hill, where his first experimental work on Iron Age farming began.

Work on the Pimperne House began in 1976. The house was finally dismantled in 1990.

For images click here

There is incontrovertible evidence that the Pimperne House was thatched. In the UK, this dictates a minimum angle of 45 degrees for the thatch, in order to shed water properly and retain the thatch's integrity against water penetration and rotting. Practice seems to have settled on this as the traditional angle to use, for greater angles tend to increase the roof area and hence the amount of straw required. Reynolds notes that this angle is found in Africa, though he points out that, ironically, the roofs leak when it rains, which is rarely. The Pimperne House, which used 45 degrees for its pitch, was robustly watertight.

Once one accepts a pitched thatched roof with two circular rows of postholes as the basis for the structure, then only a right cone atop a right cylinder fits the requirements. Any student can establish the strength of both a cone and a cylinder by curling sheets of paper to these shapes and taping the edges. An upright cylinder made from a single sheet of paper will readily take quite large books disposed evenly across the top edge [Try this with different diameters]. Equally, it is very hard to distort a paper cone without imposing substantial crushing. Anyone caring to look at a text on airflows around such a shape, will see that it is a very effective way to split the airflows in a windstream, with very little in the way of asymmetric wind forces. This does not apply so clearly to the porch, which is a discontinuity. This discontinuity is discussed by Reynolds in relation to its impact on construction. The Pimperne House porch did not suffer any wind damage. However, the later Longbridge Deverill Roundhouse, constructed at Bascomb Down in 1992, had a different disposition with respect to the surrounding slopes and the valley. It "lost" its porch in the winter of 2000 due to it being lifted by aerodynamic effects and then dropped askew, to the extent the porch had to be taken down.

The next stage in developing the construct was to determine the positions of the posts for the load bearing ring and for the outer wall. Hedge, the author of this precis, separately superimposed a set of circles on the 1:50 drawing of the excavation, as provided with the Monograph, having estimated the "best" centre. Interestingly, on reading the Monograph closely, there was some divergence of approach between Blake and Harding and Reynolds. Blake and Harding in figures 17 and 18 offer an interpretation where they suggest two sets of postholes, slightly laterally offset, and corresponding to two stages of building. They acknowledge [page 31], however, that "This interpretation is certainly not demanded by the evidence (and indeed is considered unnecessary by PJR on the basis of empirical experience)". It has to be noted that all the authors accept that the pattern of holes relates to two series of posts, which at the very least implies some rebuilding. By colouring their selection of one of the sets of postholes, Blake and Harding emphasise the offset on which they draw. This is a known risk in having a potential for creating bias in data interpretation. It should be recognised that they have to leave "gaps" where the data is either absent or "blurred" by later(?) activities on the site and totally omit one half of the broken ring of slots and scoops (see below). Also their inner ring is quite erratic, despite their saying earlier on the same page that "The plan of the roundhouse reflects the meticulous attention to symmetry and detail which builders of Iron Age houses expended....". To continue the reasoning, readers without the plan need to know that the outer rings of holes (stake holes) clearly shows a pair of concentric rings. Outside the "complete" posthole circles, and not at first recognised as anything to do with the structure (but see later), there is a broken circular array of "slots and scoops".

By the strangest of coincidences or by the veracity of the data, Hedge clearly selected the same centre as Reynolds, as evidenced by the fact that the outer diameter for the "scoops and slots" that he found corresponds exactly to that deduced from Reynolds data, namely 15.8 m. Unfortunately, Reynolds does not quote this figure but gives a separation (radial) of 1.52 metres between this (broken) ring and the outer ring of postholes, given as having a diameter of 12.80 m (page 93). Twice 1.52 is 3.05, which added to 12.80 gives 15.85. Hedge, as an engineer, would question the use of 2 decimal places based on a 1:50 plan (Sadly, it can not any longer be checked if this is how Reynolds got his numbers, though he gives his metric values as conversions from Imperial "round numbers" of feet). Equally, this diameter of 12.8 metres is happily the arithmetic average of the values determined by Hedge for the diameters of the two separate concentric rings. The small difference between Reynolds value of 9.75 m for the diameter of the inner ring and the value got by Hedge can be attributed to differing choices in where to "set" the circle through the "scatter" of postholes and the increasing "error" contribution on smaller lengths on a 1:50 plan. Any student wondering at this expositon, should be reminded that this is the proper process to scrutinise and analyse interpretations offered against given data. Hedge is also somewhat proud of the fact that Reynolds, of all people, once said "You ... you bugger... you wouldn't believe anything !"

In a recent personal communication, [3rd January 2010], Professor Harding has given further insight into the preparation of ground plans before the days of electronic surveying and recording.

The first point he wished noted was the fact that in the Monograph it is not stated that the actual survey plan was 1:24 [as opposed to the 1:50 used for publication], at the time of excavation. He wryly noted that the RCHME Inspector advising them at the time considered this large scale somewhat "eccentric" and used a scale around half this. For the record Peter Reynolds used the 1:24 plan for the Butser Construction.

A further historic inconvenience was that in those days printed graph paper was to Imperial scales. Supply vagaries being what they were, some graph papers had divisions in tenths and others in eighths. Eighths was OK for 1:24 but tenths was hopeless, so sometimes [presumably for lesser parts of the survey] he and Peter had to resort to twelve tenths to the foot and on others to one tenth to an inch. This led to a lot of additional work in conversion and checking for publication.

We now come to the application of these figures to the final elected size of the house.

Inspirationally, or based on experience and intuition, considering the later interpretation of some slots made during construction, Reynolds states that he arbitrarily chose, "purely from the point of headroom" (page 95) the height of the outer wall to be 5 feet (1.52 m), which happens to be the radial separation between the ring of slots and scoops and the outer ring of stake holes, thus automatically accommodating the angle of 45 degrees required for the thatch. This, in turn, dictated that the height of the inner ring be 10 feet (3.04 m). A bit of geometry for the student to do !! However, he goes on to say (page 98) that having recognised the equality above, he used the pattern of slots and scoops to deduce that there were six principal rafters, coming to ground level and forming a double tripod. When it came to the stage of erecting this double tripod, once up, the apex was found to be six feet off-centre. "Actually centralising the cone required each of the six principal rafters to be moved with concomitant disturbance of the ground. In one particular instance the disturbance created was virtually the same as the original excavated disturbance." This was perhaps one of the compelling lessons at this stage.

Reynolds discusses many factors relating to the materials to be used, their collection, the time taken for various stages and the tools to be used. He concludes that nothing constructive could be learnt relevant to what the original builders experienced, although he records what was found in carrying out some operations. "The object of the study is the building itself ..." Also "Discoveries and decisions during the construction of the cylinder were numerous." Most of this relates to the quantity and type of material and the means of fixing and attaching members, large and small. The porch posed its own problems but no more than with any hip roof feature in modern thatched buildings.

"At this stage of construction, the base cylinder was complete. The application of the cone posed a totally different set of problems, the answers to which alone arguably justified the reconstruction at all." The rafters were in effect trees, weighing 740-985 kg, and of length 10.36 m (34 feet) to give an eave over the outer wall. The six principal rafters had to be longer and although their length is not given, it is easy to calculate from the dimensions above that they were about 11.2 metres or a fraction more. A significant factor was dealing with the tendency for such long rafters to sag. This required a horizontal ring beam to be inserted about one-third the way down the rafters from the apex. This was in addition to the main ring beam on the inner circle of support posts. It was also recognised that the rafters had to be disposed so as to counteract the thrust from each others mass. The discovery of the origin of the slots and scoops has already been mentioned. At the end of construction, the ends of the principal rafters were cut off just above ground level to prove they were not load bearing.

The thatching stage was less instructive from an archaeological stance. The only material available in quantity and at an acceptable price was modern wheat straw, so anything to do with preparation was not relevant to the past. Even using seven tonnes of straw, the thatch depth was about one-third of that recommended by professional thatchers. In fact, this all showed up in later lessons of maintenance, since the first thatch only lasted ten years, one-third of the expected life of a thatched roof. Reynolds gives some interesting background to straw and thatch preparation, past and present, for those interested in the topic.

The other main lessons during the life of the building related to which posts rotted, where and why. Except for the porch, which had to be rebuilt after eight years, only a few main posts suffered degradation due to specific factors and the building as a whole survived remarkably well, which showed up in the quantity of reusable material when the building was dismantled. This is discussed in the paper ‘Life and Death of a Post-hole’.

Dismantling took place over two weeks of work in September 1990 and careful observation and recording covering various features of the building. The whole was in good structural condition, except for the outer porch posts (the most exposed parts) and these had already been replaced because of rotting at the soil interface. Clearly they would have needed replacing again quite soon.

All visible support timbers were in good order and well secured. The worm infestation observed in the first years after construction had penetrated no more than 2 mm into the pith wood. This was found to be so for the rafters, also. Only one principal post had started to rot at the soil interface and no subsidence had occurred. Purlins, rafters and ties were all in an excellent state. The hazel of the purlins had set but, unlike a hurdle fence of comparable age, were not brittle and could be used to scale the roof during inspection. The rafters were hard and fit for reuse. Smoke blackening was confined to the upper third of all members.

The very detailed observations of the deterioration of the main structural posts showed that the early stages of decay did not give the full picture of how post holes appear in the archaeological record and it is the bridge between these two time scales that led to the paper "Life and Death of a Post Hole" (see publications). Very briefly, the early decay could be accelerated by any long-term ingress of moisture. However, even when this occurred, the first stage of decay left the bark intact and, where the pith wood rotted rapidly, an air gap developed which generally led to the remaining central hard wood being dry and long lasting. Thus the original stone infill lay up against the bark and no extra infill was inserted in the relatively short life span of the Pimperne construct. The essential hypothesis is that had rotting progressed, over the normal life span of such a house, then there would have been a growing gap that would be filled with the sort of objects and artefacts often found during post hole excavation. The presence of these objects as a post-abandonment feature has always been difficult to explain and the interpretation offered by Reynolds, based on these findings, is compelling.

The sort of degradation observed does not imperil the structure because all loads on the posts are essentially vertical and, as the hole is infilled, the residual post above ground level continues to sit on a firm base. This practice mimics the situation observed with what is seen with more recent cruck-framed farm buildings, where ultimately the support pillars end up on top of a raised stone plinth. In fact, this very process and its success is proof that the building design, both for the Pimperne construct and for the cruck-frame barns, results in vertical loading of the support pillars/posts, for otherwise these would become laterally displaced.

The porch structure is independent of the main structure and its deterioration pattern is not as regular. This may be why post holes around supposed porches are at best distorted and at worst obscure!

"Destroying" the outer wall proved to be the most arduous of all, requiring the use of sledge hammers. The wattle work, once the daub had been removed, was found to be dry and strong and had not become embrittled in the way that exposed field fences routinely and rapidly deteriorate. This must contribute to the stability and longevity of such constructs as this. In a similar manner to the main structural posts, the vertical stakes supporting the wattle were beginning to rot back to the surface but less markedly so in the time scale tested. One suspects the process would end at the ground surface, out of sight, and that the outer wall would rest happily on the ground, held in place by the rigid cylindrical shape.

An incidental observation over the life of the house was that the level of the inner floor varied according to the wear due to passage of people and to cleaning. The variety of patterns of the processes involved are detailed in the Monograph.

Another ancillary record is that after five years the soil fibre content of the inside floor had fallen to about 5%, compared with an initial value for the grassland, prior to construction, of some 24%.

In his final words, Reynolds unwittingly wrote his own epitaph, judging by statements in his many obituaries and in the eulogies on the day to celebrate his life's work:

"That it was an important construction, which fulfilled all the criteria of the archaeological evidence, is beyond question."

The best summary of the achievements from this construction is given on the final cover to the Monograph. "The Butser experiment yielded empirical evidence regarding the construction and maintenance of these impressive Iron Age houses, which in turn cast important new light on previously problematic aspects of the archaeological data".