The center cannot hold: A Bayesian chronology for the collapse of Tiwanaku

Abstract

The timing of Tiwanaku's collapse remains contested. Here we present a generational-scale chronology of Tiwanaku using Bayesian models of 102 radiocarbon dates, including 45 unpublished dates. This chronology tracks four community practices: residing short- vs. long-term, constructing monuments, discarding decorated ceramics, and leaving human burials. Tiwanaku was founded around AD 100 and around AD 600, it became the region's principal destination for migrants. It grew into one of the Andes' first cities and became famous for its decorated ceramics, carved monoliths, and large monuments. Our Bayesian models show that monument building ended ~AD 720 (the median of the ending boundary). Around ~AD 910, burials in tombs ceased as violent deaths began, which we document for the first time in this paper. Ritualized murders are limited to the century leading up to ~AD 1020. Our clearest proxy for social networks breaking down is a precise estimate for the end of permanent residence, ~AD 1010 (970-1050, 95%). This major inflection point was followed by visitors who used the same ceramics until ~AD 1040. Temporary camps lasted until roughly ~AD 1050. These four events suggest a rapid, city-wide collapse at ~AD 1010-1050, lasting just ~20 years (0-70 years, 95%). These results suggest a cascading breakdown of community practices and social networks that were physically anchored at Tiwanaku, though visitors continued to leave informal burials for centuries. This generation-scale chronology suggests that collapse 1) took place well before reduced precipitation, hence this was not a drought-induced societal change and 2) a few resilient communities sustained some traditions at other sites, hence the chronology for the site of Tiwanaku cannot be transposed to all sites with similar material culture.

Fig 1. Tiwanaku’s location.

Shown as a red dot in South America (a) and in the southern Lake Titicaca Basin (b, basemaps are from OpenStreetMap sources accessed with the tool MapTiler in QGIS 3). Labels identify the main excavation sectors mapped in blue (S2 File). The base map is an orthomosaic of aerial photos from the 1930s (S5 File).

Fig 2. Examples of residential occupations.

a: plan drawing of Mollo Kontu. b: Plan drawing of Muru Ut Pata (see S2 File). For photographs of Akapana East, structure 5, see [:Figs 10.10 and 10.19,43]. For photographs of Akapana East 1M, level 3, see [:Fig 10.6a,43], a late structure with redwares, interpreted here as a temporary occupation that reused existing blocks. The double-coursed stone foundations aligned a few degrees west of north follow the site-wide pattern for Tiwanaku-period residential architecture. In most cases, walls enclose smaller structures that face an open area, forming a residential compound or patio group, the standard spatial layout at the site [41]. These stone foundations helped slow rising damp from degrading adobe walls. Later, alignment changed with improvised structures, for example, in Mollo Kontu, as in the stones in blue in (a) and the reused single-file blocks in Akapana East 1M. It is unclear what superstructures these buildings had, if any.

Fig 3. Principal Tiwanaku-period monuments: the Akapana (a–d) and Pumapunku (e–g).

a: the Akapana, visible on the orthomosaic of aerial photos from the 1930s (see S5 File). Dimensions: 197×203×16.5 m [:22]. It just south of the earlier Kalasasaya and Sunken Temple (summarized in :106–109; see S2 File). b: the stepped entrance on the western side of the Akapana. c: a statue that once stood next to the entrance, called the Chachapuma, holding a trophy head [46]. d: area of the monument with the mouth of conduit for internal drainage system [:Photo D, 47]. e: outline of the Pumapunku [:Fig 3]. f: reconstruction of the northern portion of the Pumapunku’s andesite building [:Fig 20]. g: large scattered andesite blocks that were once part of the monument. For more photographs and drawings, see [, , , –52], though we should be wary of ideal reconstructions based on right angles, symmetry, and proportions [:Fig 4, 51]. Many reconstructions follow idealized visions that tend to overlook details that do not conform to modern architectural conventions [52]. Figs 5b and 5d are republished from [44] under a CC BY license, with permission from the Instituto de Investigaciones Antropológicas, UNAM, Mexico, original copyright 1992, courtesy of Linda Manzanilla. The photographs in 5c and 5g were taken by Alexei Vranich. Fig 5e and 5f are republished from [48] under a CC BY license, with permission from Alexei Vranich, original copyright 2018.

Fig 4. Representative examples of Tiwanaku redwares.

Ancient potters fired these ceramics at high temperatures, decorated them with red slips and painted designs, and innovated a series of new forms. a: hyperboloid drinking goblet (kero; event A111, midden). b: portrait vessel (wako retrato; A99, an offering from burial A96; see Fig 7). c: pedestalled burner (sahumador) from the same burial as (b). d: shallow bowl (height: 11 cm), with an everted rim (escudilla), plan view. e: hyperboloid bowl (tazón; D40, ash pit). f: bowl with everted rim (escudilla; D36, hearth). Scale is approximate, due to minor photographic distortions. All materials are from the Mollo Kontu sector [53] except (d), which is from a mass burial north of the Akapana [54]. Photographs by J. Augustine.

Fig 5. Tombs and violent deaths.

a, b: The capstone for a rectilinear tomb with adobe walls, plan view, from Mollo Kontu (events A96 and A97, Beta-275871, individual MK-05144 [:Fig 4.18–19], c: profile view of a tomb from Mollo Kontu (events A102/A103, Beta-275872, individual MK-05404) [:Fig 23]). These small tombs were made for infants (individuals MK-05144 and MK-05404). This tomb architecture is diagnostic of the Tiwanaku period [:183–185]. Scale bar only applies to a–c. d: Photograph of a mass burial north of the Akapana, taken while exposing the bones of individual 12, center, found face down with his arms and legs spread out. This young adult male had a stature of 171.5 cm and unhealed cranial wounds and jaw and tooth fractures [:177–178, Fig 13.4]. e, f: cranium of individual 7, from the same context as (d), with four puncture holes and the likely element used, a bronze point (45×9 mm), perhaps the tip of a staff found in the same context [,:173, Fig 13.15]. See [:21–22] for images of Individuals 3 and 4 from the base of the Headless Monolith, found face down in a trash pit. See [:Fig 18.5] for detail of cut marks on a right distal femur found on the Akapana. Photographs 5d–f by John Verano, used with permission.

Fig 6. Bayesian models of four residential sectors and a composite, site-wide model for permanent and temporary occupation.

The green and red curves are phase boundaries. The blue curves are KDE plots of dates associated with permanent occupation, which are robust of summarizing groups of dates [66]. The yellow curves are KDE plots of dates associated with temporary occupation. Below each KDE plot, hash marks indicate modeled medians of individual dates and contexts; the sample size for each model is listed on the right. For phases with few dated events, boundaries are not shown because they are not meaningful with such small samples. In these cases, the KDE plots can only be treated as visual guides to the distribution of sparse data.

Fig 7. Composite model for monuments and community construction projects.

(As in Fig 6).

Fig 8. Bayesian models of community architecture projects and site-wide residential occupations (Figs 6 and 7).

This Fig shows four single-phase models for contexts with redwares, burials in tombs, violent deaths (these are shown together but the models are independent), and all other contexts with human bone. The timing of collapse in the early eleventh century is best documented by the near-synchronous end to permanent residence, violent deaths, and the use and reuse of redwares (red probability curves). As in Fig 6.