Tiwanaku · Bolivia · -16.5544, -68.6733 · PYR record in the living atlas
The best reconstruction anyone has of Pumapunku fits on a table. In work published in 2018, the archaeologist Alexei Vranich modelled 140 pieces of andesite and 17 sandstone slabs and 3D printed them at four per cent scale, small enough to hold one in a hand, then rebuilt the structure the way you would rebuild a broken pot. He had to, because after roughly five centuries of stone robbing not one of the approximately 150 blocks of the standing architecture is still in its original place. The building had to be reassembled in miniature before anyone could say what it looked like. Even so, the model produces candidate configurations rather than an answer: the original height of the building and whether it was roofed are unknown, and no reconstruction has consensus behind it.
The disorder is a looting pattern
The photograph everyone takes at Pumapunku is of precisely cut blocks lying at odd angles across open ground, and the caption it usually gets is wrong. The scatter is not the debris of an ancient catastrophe. It is the documented result of five centuries of treasure-hunting, and of colonial quarrying that carried Tiwanaku stone away to build in La Paz and other towns. Pedro Cieza de León described the ruins around 1549 and wrote that the blocks and gateways seemed humanly impossible to move, which is worth registering for two reasons: the reaction is nearly five hundred years old, and it was aimed at a site the Spanish were at that moment in the process of dismantling.
The platform at Pumapunku measures roughly 116 by 167 metres at the base, with a core of sifted, layered soils and gravels. How much of it has been examined is the single most useful number about this site. A drone survey in 2016 established that the Pumapunku complex covers about 17 hectares, of which only about two had been explored, and it identified two further buried platforms there. A geophysical survey around 2007, using ground-penetrating radar, magnetometry, induced electrical conductivity and magnetic susceptibility, found wall foundations, water conduits, terraces and gravel pavements buried in the ground between Pumapunku and the Kalasasaya, none of which has been excavated. Almost everything written about this site rests on a fraction of it.
A building put back together at four per cent
The measurements came mostly from Jean-Pierre Protzen's fieldwork across several visits in the early and mid 1990s, supplemented by the records of Léonce Angrand in 1848 and Max Uhle in 1893, which preserve blocks that have since disappeared entirely. The one thing that did stay put was the sandstone foundation, whose slabs still carry recessed geometric outlines showing where the andesite once sat. Those outlines are the plan.
The result matters more than the technique. Reassembled, the andesite structure is immediately recognisable as the same design tradition as the earlier and nearby sites of Chiripa and Pucara. Pumapunku is not an orphan. It is the late, expensive end of a local architectural lineage with visible ancestors, and the blocks turn out to be standardised: many are exact replicas of one another, including matched left-hand and right-hand versions. That is a workshop producing to a specification, which is a considerably more impressive claim about the builders than the one usually made on their behalf.
What the famous H-blocks were actually for is still argued through architectural modelling: doorways, screens, or some other structural system. Nor is it settled whether Pumapunku was ever finished. Excavation shows a ceremonial function involving feasting and sacrifice, but whether it was a completed working temple or a building still going up when Tiwanaku's political order failed is unresolved, and what happened to it at the collapse, ritual decommissioning, ordinary abandonment or deliberate wrecking, is open too. Deliberate destruction of monuments is evidenced elsewhere at Tiwanaku; the sequence at Pumapunku specifically is poorly documented.
Metal poured into stone
The joinery is where Tiwanaku stops resembling anything else in the Americas. Blocks were locked together with I-shaped cramps of a copper-arsenic-nickel bronze, an alloy characteristic of this region. Heather Lechtman's analysis found that the technique was not uniform across the city: the cramps in the south canal at Pumapunku were cast in place, with molten metal poured directly into sockets carved to receive it, while those at the Akapana were cold-hammered from ingots of the same alloy and then fitted. Two different metalworking traditions, or two different dates, inside one site. The floor of the upper court carried a green plaster made of lime, sand and ground malachite, so the finishes were as considered as the joints.
The stone itself travelled. The red sandstone came from the quarry at Kausani, roughly 10 kilometres away. The andesite came from about 90 kilometres away, near the Copacabana peninsula, which means across water. Unfinished blocks survive showing the stages of the work. The quarries are the answer to the question the precision keeps raising.
The largest sandstone slab at Pumapunku measures about 7.8 by 5.2 metres and 1.1 metres thick. Its mass is another matter. The figure of 131 tonnes is quoted everywhere, with a second slab at about 85 tonnes, and both are volume-and-density calculations rather than measurements. Published estimates for the largest blocks run from tens of tonnes to over 100, and no block at the site has ever been independently weighed. Give the dimensions, which anybody can check, and treat the tonnage as an estimate.
On whether the cutting could be done at all: experimental work has shown that surfaces of this quality can be produced by skilled workers using period-appropriate methods, stone-on-stone abrasion and copper and bronze tools, both abundantly evidenced at the site. What has not been reconstructed is the full production sequence for the largest blocks, which is a narrower and more interesting gap than the one usually claimed. Protzen and Nair's The Stones of Tiwanaku (Getty Conservation Institute, 2013) is the standing technical treatment.
The date keeps being corrected, in both directions
Arthur Posnansky argued from about 1910 to 1945 that Tiwanaku was 11,000 to 17,000 years old, on an archaeoastronomical argument about the orientation of the Kalasasaya. That figure still circulates. It did not survive contact with radiocarbon. The correction, though, went too far in the other direction as well: Carlos Ponce Sanginés proposed a founding date of 1580 BC in the 1970s, and researchers have treated it as unreliable since the 1980s. A statistical reassessment of the accumulated radiocarbon dates now places the founding of the site around AD 110, with a 68 per cent range of 50 to 170, and the ceramics agree. At Pumapunku, the deepest layer of mound fill dates to AD 536 to 600, so the stonework there is later still. A 2023 Bayesian reassessment by E. J. Marsh, in the Journal of Archaeological Science: Reports, used updated calibration curves to model monumental construction at both Pumapunku and the Akapana as ending by about AD 720: a short, intense building programme rather than a slow accretion.
The most direct evidence against a deep-antiquity Pumapunku is not a date at all. Vranich's excavations found the platform fill laid straight onto sterile Pleistocene sediment. There is no earlier cultural layer underneath the structure for a lost civilisation to have left. Above it, at the periphery, Inca-period occupation layers sit over the Tiwanaku horizon, which is the other end of the same stratigraphic argument.
It is worth saying plainly that the alignment Posnansky measured is real. The Kalasasaya's eastern and western doorways do track solstice sunrise and sunset, the sun does rise through its main gateway at the equinoxes, and Pumapunku's principal gateway faces roughly east, consistent with equinox sunrise. The orientations are genuine. What fails is Posnansky's inference from them to a date fourteen thousand years too early.
There is a second correction that visitors are rarely told. Much of what is photographed at Tiwanaku today, the Kalasasaya walls in particular, was rebuilt in the 1960s under Ponce Sanginés, and that reconstruction has been criticised both for not being sufficiently grounded in evidence and for not matching the quality of the original masonry. This is worth knowing before drawing conclusions from the alignments. And the Gateway of the Sun, the single andesite slab carved with a staff-bearing figure and rows of attendants, has been held since the late twentieth century not to be standing where it originally stood. An alignment claim about a gate that has been moved is not a claim about anything.
The concrete claim, and what the petrology says
One alternative account of Pumapunku sounds more technical than the rest and is therefore the most persuasive to a general reader. Joseph Davidovits and the Geopolymer Institute argue that the blocks are not quarried stone at all but geopolymer concrete cast in moulds, which is the same claim they make about the casing stones at Giza. It has been tested. Petrological analysis has found nothing consistent with a synthetic or cast origin: the blocks match their natural quarry sources both chemically and in crystalline structure. The unfinished blocks left at the quarries are the other half of the answer.
Fields that hold the heat
Tiwanaku sits at about 3,850 metres, and the constraint at that altitude is not soil but night frost. The answer was the suka kollus, raised planting platforms separated by water-filled channels. The water absorbs heat through the day and gives it back after dark, and cold air drains off the mound into the canals instead of settling on the crop. The measured effect is modest and real: Diego Sanchez de Lozada and colleagues found soil on 50 centimetre raised mounds at 3,810 metres running about one degree Celsius warmer than the ground in nearby flat fields.
The system has been tested by rebuilding it. Alan Kolata began rehabilitating fields at Tiwanaku and near Puno in 1984, and Clark Erickson ran an experimental programme in the Llanos de Moxos from 1990 in which reconstructed plots grew manioc and maize without heavy upkeep after the first season. Yield figures from these trials are quoted very widely and very inconsistently, and we are not going to pick one. What is defensible is that the technique worked well enough that Aymara and Quechua-speaking communities took it up again, and that at its height the city supported a population usually estimated in the low tens of thousands in country that otherwise supports very few. Published figures range from over 10,000 to several times that, and the peak is put anywhere between the fourth and ninth centuries, so treat any single number you see as one scholar's reconstruction.
Tiwanaku has a longer excavation history than its fringe reputation suggests. Carlos Ponce Sanginés and the Instituto Nacional de Arqueología de Bolivia worked the site through the 1960s and 1970s and produced the first absolute dates for it. A Bolivian season worked the southern sector in 1989. Alan Kolata ran a programme across the 1980s and 1990s, and Alexei Vranich has dug here from the late 1990s into the 2000s.
The state kept an address offshore
In 2019 Christophe Delaere, José M. Capriles and Charles Stanish published underwater excavations at the Khoa reef, off the Island of the Sun in Lake Titicaca. The deposits included ceramic feline incense burners, killed juvenile llamas, and ornaments of metal, shell and worked stone, laid down repeatedly rather than once. The authors read them as cyclically performed state ceremonies, and argue that this ritual programme was part of how the Tiwanaku polity consolidated itself politically. The religion had a location on the water, and it was maintained on a schedule.
The land is not finished either. The same drone-based topographic survey reported in 2016 mapped more than 411 hectares to a margin of error under four centimetres and picked up structures nobody had recorded, among them a buried stone temple and around a hundred substantial circular and rectangular buildings that may be domestic compounds, along with ditches, canals and roads. What ended it was probably water: the evidence points to increased aridity in the basin around AD 1000, which would have broken the raised-field system the whole population depended on. Political dominance had gone by the eleventh century and the collapse is generally put between about AD 1000 and the middle of the twelfth century, with the exact end of it a range rather than a date. The city was empty for roughly five centuries before the Inca arrived and explained the ruins as the work of giants.
What our engines found
We ran the same test here as everywhere else, with the settings fixed in advance and identical at every one of the fifty sites we cover: a 200 km search radius, at least 12 km between sampled sites, 15 sites, a 2 per cent tolerance, and 2,000 Monte Carlo runs against a null model resampled from real catalogued sites in the same region. The settings are frozen because the answer moves if you tune them, and tuning them after seeing the answer is exactly the trick we object to elsewhere. Tiwanaku returned 7 golden-ratio pairs where chance predicts 10.9, giving p = 0.8666.
That is fewer than chance would predict, and it is not significant. The control points the same way: the arbitrary ratio 2.0 scored 1.73 times its expected count against phi's 0.64, so nothing here singles out the golden ratio.
The collinear test is the one that does return something: 13 three-site alignments against 3.9 expected, p = 0.0080. Before anyone builds on that, the honest caveat is terrain. Sites string out along river valleys, coastlines and narrow necks of buildable land, and geography produces straight lines without anyone planning them. It is a number worth investigating, not a conclusion.
We publish the misses because a tool that only reports its hits is not measuring anything. Every figure above is reproducible from the same coordinates, the same settings and the same seed.
One limit worth stating plainly, because it is the one most often glossed over. This test compares positions between monuments. It says nothing about any building's own proportions, its height, base, slope or perimeter. That is a separate question, and at several sites a genuinely open one.
Explore it yourself
Open the Tiwanaku record in the living atlas to explore the complex in 3D, run the alignment and ley line engines across the Titicaca basin, and set the sky engine to an equinox sunrise over the Kalasasaya. The Tiwanaku site-emblem tee carries the Gateway of the Sun in the Pyramason Chronicle style, with the site name and the exact coordinates.
Sources worth reading
Protzen and Nair, The Stones of Tiwanaku: Architecture, Construction, and Tiwanaku Stone Masonry (Getty Conservation Institute, 2013). Kolata, The Tiwanaku: Portrait of an Andean Civilization (Blackwell, 1993). Janusek, Ancient Tiwanaku (Cambridge University Press, 2008). Vranich, "Reconstructing ancient architecture at Tiwanaku, Bolivia: the potential and promise of 3D printing", npj Heritage Science (2018). Marsh, "Dating the ebb and flow of Tiwanaku and post-collapse material culture across the Andes", Journal of Archaeological Science: Reports (2023). UNESCO, Secrets of Tiwanaku revealed by drone.