This work is described in several English Heritage technical reports (see end note), but the time has come for wider presentation. This feature follows an invited seminar in Devizes, Wiltshire, in September 2004.

It has become clear that the archaeological history of Silbury is crucial to understanding the issues that face us today. There have been several antiquarian investigations and excavations, and perhaps others we know nothing about. Most significant for the present state of the hill are a shaft and two tunnels.

In 1776 Colonel Drax and the Duke of Northumberland dug down from the top. Accounts tell us only that they found a small sliver of oak and 'a man', presumably a skeleton, at the base of the shaft, though a newly discovered report says the hill consisted 'of chalk and gravel thrown together by the hands of men' and that 'there were many cavities in it'.

Dean Merewether supervised the excavation of a tunnel from the side of the hill in 1849. He noted the old ground surface, and many sarsen stones apparently placed on a primary central mound of turf, chalk and black soil.

More informative were Richard Atkinson's BBC funded excavations in 1968-70, broadcast in four Chronicle programmes. His tunnel (1968-9) mainly followed Merewether's. He never fully published his work, written up after his death by Cardiff archaeologist Alasdair Whittle.

After the collapse in 2000, English Heritage staff photographed the hole from the air and on the ground. Cementation Skanska sub-contracted Geotechnical Engineering to drill seven boreholes from the summit in 2001 and 2003. Analysis of the cores has given much new evidence. In 2000 we recorded the archaeological deposits inside the 1776 shaft, before it filled with collapsed material. From all this, we can summarise our archaeological knowledge as follows.

Small boulders of sarsen (local sandstone) may have marked this mound, and evidence for wooden stakes was found - though what purpose they served is not known.

This was all capped with alternating and sharply defined layers of clay/soil, chalk, clay/soil and chalk to bring the mound to around 5.25-5.5m high and 34-36m across.

Atkinson recovered a wealth of biological remains from the turf stack and from the pre-mound surface, including insects, land snails, pollen, seeds, moss and other plants, which survived because oxygen was excluded. The mound's huge weight compressed and protected the deposits from temperature change and moisture loss. There is also an unknown degree of iron pan formation, which must affect gaseous exchange within the mound.

The biological remains show Silbury I was built on mature chalk grassland containing plants such as salad burnet, small scabious, bird's foot trefoil and meadow buttercup, with very little woodland in the area. Dung beetles indicate numbers of livestock comparable to the present.

Mature chalk grassland, a habitat that requires carefully managed grazing, stands in sharp contrast to the mosaic of woodland and clearings known from other monuments across Neolithic England.

The new cores show biological remains are also preserved in the capping layers. There has been no deterioration since Atkinson's work, good news for the long term future of this unique Neolithic archaeology.

At the centre of the hill, in the 1776 shaft and a core close by we recorded a distinct layer of crushed chalk c 10.3m from the summit. If this is the top of the Silbury II mound, it would have been around 20m high.

However, Atkinson recorded crushed chalk above the inner dump. Its detailed description is so similar to what we interpret as the top of Silbury II, that this may also be the surface of an earlier mound, with a base diameter of c 46m. We suggest - very cautiously - that there were four stages of mound construction (see diagram). A more complex sequence inside the mound is supported by new analysis of the ditch system.

A rough chalk 'wall' revealed in the excavations on the summit in 2001 seems different from those recorded in 1969-70. The interpretation of all these features needs to be re-considered.

We still do not know when Silbury was built: the published radiocarbon dates from new excavations on the top of the hill have been withdrawn by the Oxford Radiocarbon Accelerator Unit because of a problem affecting bone. These samples, with a suite from Atkinson's tunnel, are being re-dated. The first reliable series of dates should be available by the end of 2005.

Silbury is covered with scars, many the result of archaeological and antiquarian excavations. Beside the tunnels, smaller excavations near the base of the mound were opened by a distinguished team including John Lubbock and William Cunnington in 1867, and in 1922 by Egyptologist Flinders Petrie. Alfred Pass dug ten 'shafts' in the ditch to the west and north of the mound in 1886.

Some scars, such as a ramp towards the summit in the south-west, were perhaps cut by the 1776 miners. Others, such as the remnant of a ring of trees around the summit, may be signs of 18th century landscape design. Some may indicate transient activities: 18th century newspapers reported festivals there with wrestling, bull-baiting and eight-a-side football, attracting 6,000 people.

Most distinctive are the steps in the slope close to the summit. Atkinson likened them to wedding cake tiers. On perambulation during survey, however, it quickly became clear that the ledge spirals downwards. Breaks of slope indicate similar, silted ledges further down. Atkinson's trenches across the upper ledges make it clear that they had been revetted by posts with iron nails; a coin and pottery suggested a date soon after 1010 AD, and Atkinson believed the mound had been fortified against the Danes.

He wrote to the landowner implying that he had also encountered an Early Medieval ditch on one of the lower breaks of slope on the northern face. It is unclear whether these ledges were entirely Medieval, or a reshaped older feature. A spiral step would help get materials to the summit, and spirals are part of Neolithic iconography.

Over 10,000 survey points on the mound alone allowed us to plot detailed contours. This emphasised that the structure is not in fact circular, but built in straight segments that may indicate radial walls or buttresses.

Only in the south-west does this not apply, where a concavity might represent an early collapse. There is no mound of debris at the base, but one of Pass's shafts here found great amounts of chalk rubble with sarsen boulders. Sarsen forms a component of the structure, and it maybe that these boulders are indicative of a collapse.

Silbury II was enormous, no less impressive, for example, than Newgrange and larger than Maes Howe. However, while Silbury later became the largest prehistoric mound in Europe, it is the 165m by 85m cistern or tank attached to the ditch that makes it unique. Yet we know hardly anything about it!

The earliest ditch appears to be that found by Atkinson, thought to have surrounded Silbury II and been backfilled and covered by Silbury III.

Two causeways still connect the mound to the hillside to the south: but these are not causeways in the usual sense, as both are steeply cut down by 3m and to cross them would need a bridge. The only likely explanation is that they are relict fragments of another ditch, the rest since cut away (phase B on the plan).

The 50m wide ditch signals another episode of activity (phase C). On the southern side it appears to be much narrower than elsewhere, but Atkinson's excavation suggested it extended beneath the road. Perhaps a structure prevented Roman road engineers building further south; much pottery and over 100 coins were found in the ditch here, and settlement and other activity in the area imply a Roman interest in, at least, the ditch and probably the site as a whole.

Pass's trenches and Atkinson's unpublished core and seismic data show the ditch is deeper immediately around the mound. This might represent a phase of recutting (D).

Of considerable size and depth, the ditch extension (phase E) reached west towards the Beckhampton stream. Today this stream is canalised, but in winter the ditch still holds much water; over a metre impeded our survey in April and May. Pass noted that even after a summer when the adjacent river Kennet had dried up, his trenches constantly held over two metres of water, while a well (or spring) in the south-west ditch terminal was remembered as being used in the 19th century.

The extension might have been no mere quarry, but deliberately formed, perhaps even constructed to retain water. Water infatuation is implicit in the location of many henges, while the massive palisaded enclosures at West Kennet, partly visible from Silbury, straddled the Kennet. In other parts of the world, cisterns are focal points of ritual and ceremony, and the mirror-like quality of standing water may have had symbolic implications-viewed from the mound summit, water in the ditch reflects the sky. Suggestions that the stream north of Silbury was dry in the Neolithic may need reconsidering. It is clear the ditch silts need further investigation.

For two days in May, as it dried out, we observed a massive vegetation mark crossing the ditch extension, some 10m wide and almost 50m long, oriented just off-centre of the mound. This likely represents a deeply excavated feature. It was cut on the surface by water meadow drains mapped in the 19th century, so on the face of it is unlikely to be modern.

It is important to see Silbury in its wider context. It rose alongside other constructions at Avebury, West Kennet and the Sanctuary, which presumably drew an extremely large number of people. Its precise location provides intriguing possibilities for a natural feature - tree, swallow hole, spring- to have been its original focus.

It is almost inconceivable that inhabitants of the Romano-British village to the east did not use or revere Silbury in some way; at least one burial was placed on the slopes.

The Early Medieval presence may have been considerable. This could have been defensive as Atkinson suggested but, given the earlier significance of the site, may have been of a more symbolic nature. With the attempts to stamp Christian authority on such places during the Conversion period it would perhaps be surprising if Silbury was completely ignored.

Like any monument (or parish church), Silbury has seen centuries of weathering, amendment and addition. Building phases can be detected from the excavations, but these represent only a snapshot of an intermittently changing structure.

Our research leads to two conclusions. Geophysical and geotechnical experts advise that, overall, the hill is stable and there is no identifiable risk of widespread collapse or slope instability. However, we now know that, unless attended to, there will be continuing, localised disruption of the mound as a result of the three major archaeological excavations.

Merewether's tunnel was not backfilled. Much of its original outer length seems to be filled with collapsed material, marked by depressions visible on the surface above, while deeper into the hill it was mostly incorporated in Atkinson's tunnel. However, there are thought to be significant voids: where Merewether's tunnel had already migrated above the line of Atkinson's, as the roof fell in raising both top and bottom of the passage, and where his side galleries were not assimilated by Atkinson's tunnel.

Atkinson backfilled his tunnel with excavation spoil bulked by imported material. However, the two boreholes that penetrated Atkinson's workings revealed substantial voids: filling was poorly executed. Without contrary evidence, we assume there are significant voids along the whole length of his tunnel.

New theoretical models show how the Merewether and Atkinson tunnel voids might behave. It is thought those close to the sloping surface of the mound will migrate upwards, widening into a cone shape as they reach the ground surface. Cavities deeper in the mound will also slowly migrate, but disruption of sediments will be confined to the area immediately above the tunnels. They will eventually cease to grow either when the cavity is completely filled with loose debris, or earlier if the tunnel roof achieves a state of equilibrium.

There is open space in Atkinson's tunnel where it undercuts the 1776 shaft. The shaft itself is now completely full. It is expected, assuming the Atkinson void below is filled, that this material will slowly settle down leaving a void of up to 1-2m below the polystyrene blocks, were these not to be replaced.

We have calculated potential future damage from collapses in these unsupported tunnels. As well as the obvious loss of historical and ecological information were subsidence to occur, the condition of some classes of biological evidence would deteriorate once moved.

Silbury Hill occupies some 15,300m2, of which c 270m2 (1.8%) has been disturbed by the horizontal tunnels. Collapse affecting the surface of the mound has already taken place over the outer 25m2 above the 1849 tunnel, leaving a ground area of around 250m2 (1.6%) above which some mound sediments are at risk.

The volume of the hill is about 239,000m3, of which 310m3 (0.13%) consists of likely tunnel voids.

Our calculations for future damage from collapse are based on the position in 1968-9. Further collapse has occurred since (as shown in a borehole), so the figures will slightly over-estimate what remains to be damaged, but the precise extent of this is unknown.

Silbury I is clearly at considerably more risk than the larger mounds of Silbury II and III. This is a particular problem, as some of the best preserved organic material is found within this inner mound.

The extraordinary preservation of biological remains is dependent on the exceptional, precisely maintained conditions in the centre of Silbury. If primary mound material collapses into voids, it would not undergo significant gaseous exchange with the atmosphere. However, the environmental deposits would be mechanically damaged and stratigraphically compromised. Serious collapses, exposing new surfaces, would involve further loss of material as the 'sealing' rind of decay assumed to have formed around the tunnels and voids, and protecting the material behind it, would be lost.

It is clear that 755-1,715m3 of the monument could be lost (0.3-0.7%), with the greatest destruction affecting some of the most important deposits. The timescale for this is not fully understood. Recognising that any 'cure' should not be more damaging to the archaeology of Silbury than the problems themselves, there are three possible strategies.

A difficulty with this technique is ensuring that all voids are filled - there is no real way of knowing if any remained. Additionally, drilling new boreholes could destabilise important biological remains through the introduction of air into otherwise well-sealed anaerobic environments.

This technique could cause more stratigraphic loss than remote filling. There is the risk that new workings may trigger further collapse. However, the significant benefits are a full archaeological record, something not achieved before, and far more control over the quality of backfilling. Importantly, there would be no new damage to biological deposits situated well beyond the edges of the existing voids.

The polystyrene in the summit crater would only be replaced with chalk once the voids at the base had been filled or supported, because of the additional weight loading. This would have the advantage of compacting the existing shaft fill. All these options are being considered by English Heritage.

All authors work for English Heritage: