How Long Does a Memory Last? Bayesian Chronological Modelling and the Temporal Scope of Commemorative Practices at Aeneolithic Monjukli Depe, Turkmenistan

Abstract

In this paper the history of one house and a human burial in the prehistoric settlement of Monjukli Depe, Turkmenistan, serves as a case study for the use of Bayesian chronological modelling to approach the reach of past memories. The method combines relative and absolute chronological data and aims not only at a more precise and robust chronology of past events, but also allows estimations of duration of particular processes. However, chronological models must be constructed with care, since the prior archaeological information significantly affects the output. The comparison of three alternative models for the Aeneolithic settlement of Monjukli Depe shows that prior information in modelling has a considerable impact on duration estimates for periods of the settlement history. The modelling chronology for Monjukli Depe allows the tracing of commemorative practices at a generational scale—the memory of Monjukli Depe House 14 was transmitted over several generations of inhabitants long after the house destruction. It is clear that houses possessed a great value in the social life of the settlement since local building histories were remembered over a long time.

Introduction

In recent decades, archaeological research has paid increasing attention to memory work as seen in approaches to material culture and its engagement in social practices of creating, maintaining, abandoning and transmitting memories (see contributions in Hofmann et al. 2017; Mills & Walker 2008b; Van Dyke & Alcock 2003). However, studies that trace singular acts of remembrance and especially their temporal scope are rare. This paper takes a closer look at one such case at the site of Monjukli Depe in south Turkmenistan. There, a burial of a more than 50-year-old male individual was found in an open area within the settlement. The position of the burial shows a clear spatial reference to a house that had existed on this spot in an earlier phase of the settlement but had long been destroyed by the time of the burial event, so that its ruins were no longer visible on the surface at the time of interment. The commemorative act associated with the burial—the remembrance of the particular structures of the past settlement by those performing the interment—seems to have had a considerable temporal depth. In terms of stratigraphy, the period between the destruction of the old house, complete burying of its ruins under settlement debris and the burial event included the construction, use and destruction of a later domestic structure in its vicinity. But how long did this last in calendrical terms? Is it possible to measure the length of the collective memory of inhabitants of Monjukli Depe?

This question can be approached by Bayesian modelling of radiocarbon dates. The method, which incorporates archaeological information into the statistical analysis of ¹⁴C measurements (Buck et al. 1992), has been used to estimate more precisely dates of individual events and the duration of specific processes. However, there are several possibilities to interpret the history of events as they are captured in the stratigraphic sequence. I will examine different scenarios for the settlement history and compare the estimations of time periods between the house destruction, disappearance of the ruins from the settlement surface and the later burial. I will then use these results to reconsider the commemorative potential of the houses in Monjukli Depe in general.

Bayesian chronological modelling and memory studies

The use of Bayesian chronological modelling in analyses of radiocarbon dates has experienced a tremendous upsurge in the last decade. The application of the method has been stimulated by the need for more precise chronological estimates of the events and processes for which radiocarbon data are available. By their nature, calibrated ¹⁴C data are not exact year readings of past events, but rather probability distributions that can span several hundreds of years. Bayesian chronological modelling combines archaeological information with radiocarbon dating evidence: the relative chronological information relevant to the temporal order of dated samples serves as prior information in estimations of more precise date ranges—posterior density estimates—for dated samples or events encompassed in a sequence of radiocarbon dates (for the method, see Bayliss, Bronk Ramsey et al. 2007; Buck 2004).

The new possibilities of Bayesian chronological modelling in evaluating chronological data and the rapid adoption of the method in archaeology have raised calls for a ‘third radiocarbon revolution’ (Bayliss 2009). The application of the method is geared not only towards the precision of chronological issues. Bayliss, Bronk Ramsey et al. (2007) see significant potential in the possibilities to approach time as perceived and experienced by past people, as this method not only allows us to pinpoint past events on a calendrical scale but also to approach the duration of particular time periods that may help understand people's experiences of temporal processes. More recent studies point to the potential of Bayesian chronology in household archaeology, which can be used to pinpoint everyday material practices more accurately in time and determine the tempo of their change (Marciniak et al. 2015; Overholtzer 2015).

The temporality of daily life and the perception of time by past people is closely tied to another topic with an inherent temporality—memory practices. These involve transmission and transformation of social practices and often a close engagement with the surrounding material world. Together, these make memory practices predestined to address diachronic questions in archaeology (Mills & Walker 2008a). As Ruth van Dyke (2019, 210) notes, memories are always spatially situated, evoked by particular places and associated with them. Especially the house often plays a fundamental role as an organizing unit for social practices, having a durability that transcends generations of its inhabitants, generates memories and spatially anchors memory practices (Joyce 2000). While there is an increasing number of studies on the spatiality of memory and the intersection of memory with material practices of past communities, few of them explicitly address the temporal scope of memory work involving finely tuned chronological issues. Recently Bayesian radiocarbon modelling has been applied in studies dealing with long-term memory transmission in Indigenous oral traditions in the Pacific Northwest region of Canada and in Papua New Guinea (Edinborough et al. 2017; Urwin et al. 2021). Mortuary practices imbued with commemorative potential have been addressed from the perspective of memory studies and Bayesian chronology. Seminal works for the application of Bayesian radiocarbon modelling methods in archaeology, devoted to prehistoric multi-burial barrows in Great Britain (Bayliss, Benson et al. 2007; Garrow et al. 2014; Whittle et al. 2007), offer valuable discussions of memory practice despite being primarily methodological in nature. In addition to multiple burials in barrows, redepositions of disarticulated human remains from Chalcolithic and Bronze Age burial and settlement contexts have been analysed with respect to memory issues using Bayesian chronological modelling (Booth & Brück 2020; Brück & Booth 2022). The focus in these studies was on the question of inter- and intra-generational memory of the deceased.

In this study, the object of investigation is also a burial. In addition to the question of the relationship between the deceased and the persons performing the ritual, the focus is on the shared memory of a domestic space in a settlement that changed over time. By modelling the radiocarbon chronology of the Monjukli Depe settlement, I will address both methodological aspects and the potential of Bayesian analyses for the interpretation of memory practices and their durations.

Monjukli Depe and burial MDB 11 in the ruins of House 14

The small mound of Monjukli Depe in the foothills of Kopet Dag in south Turkmenistan (Fig. 1) was occupied in the sixth millennium bce during the Neolithic Jeitun period and, after a long hiatus, reoccupied in the Early Aeneolithic in the fifth millennium bce. The site was discovered and excavated in Soviet times (Berdiev 1972). Fieldwork was resumed from 2010 to 2014 by a team from the Freie Universität Berlin. The Aeneolithic layers were extensively investigated, yielding substantial portions of superimposed and well-preserved mudbrick architecture (Bernbeck & Pollock 2016; Pollock et al. 2019). They give a comprehensive insight into the life of several generations of the past communities inhabiting this settlement. The small-scale focus of the excavation and documentation methods of the Monjukli project provides a good basis for establishing precise stratigraphic relations of the dated contexts.

Figure 1. Map of the Kopet Dag region with the locations of Monjukli Depe and two eponymous Neolithic (Jeitun) and Early Aeneolithic (Anau) sites.

The excavators divided the site-wide stratigraphy of the Aeneolithic period into four chronostratigraphic macro-units or strata (Pollock and Bernbeck 2019). This division is based mainly on major changes in the stratigraphic sequence in the largest and most extensively excavated trench in the central part of the settlement, Unit D. The finely differentiated contexts allow for a higher resolution of temporal-spatial entities than this division and offer several options for structuring the chronology of the settlement history of Monjukli Depe. The Aeneolithic sequence in the central part of the settlement begins with the erection of House 14 along with some other houses in the centre of the mound in the earliest Aeneolithic phase, Stratum IV (Fig. 2). These events marked the resettlement of the mound after an 800-year occupational hiatus with only sporadic human presence that followed the initial Neolithic occupation (Heit 2019; Pollock & Bernbeck 2019). Similar to most houses of the Monjukli Depe settlement, House 14 had a square plan with two massive opposing buttresses. These architectural elements not only had a static function as roof supports but also possessed a certain symbolic meaning for the inhabitants. This is evident in their elaborated plastering and, in the case of House 14, the painting of the front and side of one buttress (Bernbeck & Pollock 2016, fig. 7). In several cases objects including animal bones, horns, or stones (‘corner deposits’: Pollock & Bernbeck 2019, 77–9) were deliberately placed up against these buttresses during the use but also after the abandonment of the houses.

Figure 2. Settlement plan of the early phases of the Aeneolithic occupation at Monjukli Depe: Strata IV and III.

Unlike the other houses of the settlement, House 14 was destroyed after its abandonment by a fire that seems to have been set deliberately (Kubelková 2019). After the destruction of House 14, other neighbouring houses seem to have been abandoned as well. However, other parts of the settlement probably remained inhabited, and the ruins of the abandoned houses were frequently visited and used for non-dwelling activities such as waste disposal or animal penning, including the ruins of House 14, whose upper wall parts remained partially visible until the end of the after-use phase (Cereda forthcoming; Heit 2021; Öğüt forthcoming; Pollock & Heit 2023).

After some time, the houses of the earliest settlement phase were rebuilt or new buildings were erected in this area. These events have been attributed to Stratum III. One of the new houses was House 10 (Egbers 2019), located to the northeast of the former House 14 and partly superimposed over the ruined walls of this structure. The remaining space of the old House 14 to the southwest of House 10 was levelled and transformed into an open area. In this exterior space in late Stratum III—the time preceding further spatial transformations, such as the establishment of a large garbage dump (‘Eastern midden’) in this area in Strata II and I—a burial of a 50-year-old man in an L-shaped shaft grave took place (MDB 11: Rol 2019). The burial was placed directly on top of one of the buttresses of the former House 14, the walls of which were completely buried under settlement deposits at the time of the interment (Fig. 3). But the position of the body suggests that the burying party was still well acquainted with the house configuration.

Figure 3. Burial MDB11 (left) and its position on the ruins of House 14 (right). The walls of the house were no longer visible at the time of interment (burial drawing: Nolwen Rol).

That the placement of the buried person was probably not accidental is also indicated by the obvious symbolic significance of the buttresses in the microcosm of the Monjukli houses. They served as spatial anchors for special activities, not only during the use of the houses but also after their abandonment, as evidenced by structured depositions in the area of the buttresses of House 14 and other houses. This raises several questions: could the old man and the people who buried him have experienced the past house in use and remembered its setting? Or did these people at least observe the longstanding ruins of the building? Or did the burial rather take place after all contemporary witnesses of the house had passed away and the memories of it were transmitted to the next generation(s)? To answer these questions, it is necessary to determine both the duration of time between the destruction of House 14 and the burial as well as the length of time between the complete burying of the house ruins and the interment. Did it encompass a few years, or rather one or even more generations?

Eventful time: modelling process and tools

Time and its modes have been the subject of theoretical discussions in archaeology for some time. Shanks and Tilley (1987) were among the first to offer a radical critique of the chronological understanding of time in archaeological practice. They juxtaposed substantial time, shaped by social practice and experienced by past people, and abstract time, constructed by capitalist chronometry and used in chronologies. Later works continued discussion on multiple time rhythms, temporality of material culture, and scalar problems in addressing temporal issues in archaeology (e.g. Harding 2005; Ingold 1993; Lucas 2005; Robb & Pauketat 2013; Thomas 1996). The experience and perception of time by past people and the intersection with the material world is another topic on the agenda of temporality studies and an essential aspect of memory practices.

How can memory—experienced time—be grasped with the chronological means available to the archaeologist? Following Bayliss, Bronk Ramsey, et al. (2007, 2–3), our chronological understanding needs to be transformed when approaching time as experienced by past people. The first step that a Bayesian chronological modelling makes towards experienced time is an ‘eventful’ understanding of the chronology, which is broken down into a series of single events placed on a progressive timescale (Bronk Ramsey 2009b, 338). Consequently, as part of the modelling the temporal processes need to be described in terms of sequences of events. The currently most frequently used radiocarbon modelling program, OxCal (Bronk Ramsey 2009b), utilized also in this study (v 4.4.4), has with phase and sequence two key tools which structure the temporal processes in a chronological model. A phase is understood as a time segment between two hypothetical major events (boundary events) in which the ¹⁴C dated events are supposed to have taken place (Bronk Ramsey 2009b, 343–5). The sequence command defines a relative order for events and phases created by boundary events. The second feature which allows approaching experienced time is the estimation of durations; the lengths of individual phases and sequences can be accomplished by the span command, the duration between single boundary events can be estimated by the interval and difference commands.

With the agreement index (A), the OxCal program offers a diagnostic tool for testing the consistency of a model with the ¹⁴C evidence. It permits statements to be made as to whether the entire model (A_model) is reliable and how well individual determinations (A_i) fit into it (Bronk Ramsey 2009b, 354–7). This tool is important in comparing and selecting the most suitable from several juxtaposed models. As a general rule, the agreement index must be above 60 per cent to meet the requirements of a consistent model.

The attribution of several events to one phase for which a temporal proximity is presumed is an interpretive operation and the chronological phasing itself an essentially arbitrary decision made by the archaeologist based on prominent features of the stratigraphic record. In the case of Monjukli Depe, the sequence of dated events can be organized in various ways as I will show in the following.

Calculating memory durations: radiocarbon data from Monjukli Depe and Bayesian modelling

The radiocarbon dates from Monjukli Depe were analysed in 2019 using Bayesian statistics aimed at modelling the Neolithic and Aeneolithic site occupation (Heit 2019). At that time the models obtained in a multi-step procedure—one for the Neolithic and one for the Aeneolithic occupation—were regarded as the most appropriate to represent the chronology of the settlement. New stratigraphic insights into the settlement history, particularly of the time between the abandonment of the Stratum IV houses and erection and use of houses of the subsequent Stratum III (Heit 2021), an updated radiocarbon age calibration curve (IntCal20: Reimer et al. 2020) and an additional radiocarbon date from the burial context in focus of this study allow me to re-adjust the existing Aeneolithic model (Heit 2019, 85–97) and propose two further alternative scenarios for the dated events in Strata IV and III. The comparison of several scenarios resulting from different archaeological interpretations of the sequence of events has been frequently applied in Bayesian chronological modelling to select the most appropriate and plausible one from several possible models (Bayliss, Bronk Ramsey, et al. 2007).

Dataset: charcoal outliers and their impact on the modelling results

Thirty-eight ¹⁴C dates from charcoal samples from the central part of the Aeneolithic settlement build the framework for the investigation of the timing and duration of the commemorative event in focus of this study (Table S1). Most samples derive from the branch fragments of Tamarix shrubs and in fewer cases from Populus/Salix species and can be regarded as short-lived charcoal samples. Some fragments could not be identified, and three samples derive from tree-stem parts and might potentially have been old wood samples. One date, Poz-53195, from a charred Juniperus stem fragment was identified together with five other dates as an outlier using visual inspection of the posterior density estimates and operating with individual agreement values (Bronk Ramsey 2009a, 1024–5). It, as well as radiocarbon measurements for Poz-53195, Poz-82804, Poz-82801, UGAMS-11395, Poz-53196, and Poz-67928, yielded significantly older dates than the other measurements in phases in which these samples were included for stratigraphic reasons. Whereas the latter two dates (Poz-53196, Poz-67928) derive from botanically indeterminate samples and may potentially belong to Juniperus stem fragments, the offsets of other samples—Tamarix branch fragments—are seemingly caused by reasons other than large radiocarbon age of the tree samples. The dated wood pieces might have had a longer use history before being burnt, or the charcoal context might have been re-deposited during the settlement occupation. Both are especially likely for tell settlements marked by strong spatial continuity. The life of the settlement plays out on top of the deposits of older settlement phases, which can be re-deposited in the course of activities such as making of installations or buildings and thereby land in the upper levels. Bayesian models from other densely and long occupied tell settlements such as Çatalhöyük East in Anatolia (Marciniak et al. 2015) or Vinča-Belo Brdo in Serbia (Tasić et al. 2016) contain a considerable number of residual samples in their sequences.

Dating the burial MDB11

In 2021 the series of radiocarbon datings on charcoal from Monjukli Depe was complemented by a single determination from the human remains of burial MDB11, which directly indicates the time of death of the old male individual and therefore should be quite close in time to the burial event. Osteological analyses (Steadman 2019) showed no evidence of excarnation or exhumation that would point to a greater temporal distance between the individual's death and his interment.

Even though AMS dating requires only minuscule amounts of bone of about 60–200 mg (Cersoy et al. 2017), sampling and age determination is a destructive process. In the case of human remains, ethical considerations need explicitly to be kept in mind. Sensitization for this topic has increased in recent decades in biological anthropology and archaeology, especially in connection with destructive sampling of human bones (cf. Scarre 2006; Squires et al. 2019). The respectful treatment of human remains by scientists should not violate legal aspects, should use minimally invasive methods and carefully balance scientific benefits against the invasiveness of the methods. With the permission of the Turkmenistan antiquities authority, samples of human bones from burial contexts of Monjukli Depe were transferred to Berlin for further analysis. To minimize the loss of material, a fibula fragment of the MDB11 skeleton, which had been previously used for isotopic analyses (Eger 2022 and the ethics discussion therein: 153–5), was selected for radiocarbon dating. The sample remains were sent back by the laboratory after sampling. The isotopic signals from the individual MDB11 (isotope sample MD62) indicate a mixed terrestrial C3 plant-based diet (δ¹³C –18.6‰; δ¹⁵N 12.9‰: Eger 2022, 257, appendix XXI), as is generally the case for other human individuals sampled from Monjukli Depe. It is unlikely that the individual regularly consumed aquatic resources, which would potentially entail reservoir effects and impact the radiocarbon age (for dietary offsets in radiocarbon dating, see e.g. Bayliss et al. 2004; Higham et al. 2010). This is also supported by the long distance of the settlement from major water sources and the scarcity of the fish bones in the assemblage from Monjukli Depe despite the regular use of dry and wet screening as a recovery method (Eger 2022, table 3.2).

Three models in comparison

The ¹⁴C data are irregularly distributed across the Aeneolithic strata of the settlement. Most of the samples (16) come from the early settlement phase, Stratum IV, into which the use, destruction and after-use of House 14 fall. The large number of dates, as well as good stratigraphic resolution of the events related to House 14, allow for a finer division of this settlement period by data modelling. The Stratum IV sequence was divided into three phases: ‘Usephase IV’ (use period of House 14 and other houses of Stratum IV); ‘Destruction House 14’ (the conflagration from which several dated wood samples originated); and ‘Ruins IV’ (post-use above the ruins of House 14 and other houses of the early settlement phase). The position of each data point remains unchanged in the phases of all three models. The differences lie in the conception of the sequences of these three phases and their relation to the following period, Stratum III (Fig. 4).

Figure 4. Overview of the prior information of the three models. Phases are indicated by square brackets, boundary events between the phases are indicated in italics. The durations in question between the events are indicated by a dashed line: (a) interval between the destruction of House 14 and burial MDB11; (b) interval between the complete covering of the ruins of House 14 with settlement debris and burial MDB11.

In Model 1 (Fig. 5) applied in the initial modelling (Heit 2019), the three phases of Stratum IV as well as the following Stratum III were arranged in a sequence of contiguous phases (Bronk Ramsey 2009b, 348) separated by a transitional event and directly following each other without a hiatus. Model 2 (Fig. 6), in contrast, presupposes a temporal gap between the ruin phase of Stratum IV and the beginning of Stratum III, marked by two boundary events between the phases. In Model 3 (Fig. 7), the phases of use, abandonment and after-use of the ruins of the houses in Stratum IV were grouped as sub-phases in a large phase within which discrete events of the early settlement phase took place and ended prior to the start of the next subphase(s).

Figure 5. Posterior probability densities of Model 1. The date of the burial MDB11 is marked in red.

Figure 6. Posterior probability densities of Model 2. The date of the burial MDB11 is marked in red.

Figure 7. Posterior probability densities of Model 3. the date of the burial MDB11 is marked in red.

All three models correspond well with the ¹⁴C data, with agreement indices (A_model) well above the significance threshold at 90–91 per cent. Thus, there is no statistical reason to prefer one model over the other. Do archaeological observations argue for a higher plausibility of one of these models? At least two observations speak in favour of models 2 and 3, those with a clearer distinction and greater temporal distance between Stratum IV and Stratum III. Firstly, all dated contexts contributing to Stratum III result from contexts of later use and destruction of houses in this stratum. An earlier segment of Stratum III, for example, building and early use of House 10, is missing in the radiocarbon evidence. Thus, there should be a temporal gap between the contexts of the last ruin phase in Stratum IV and dated contexts from Stratum III. Secondly, although the general settlement history of Monjukli Depe appears as a spiral stratigraphy (Pollock et al. 2013, 54), with new houses being built next to already abandoned ones during the occupation, there is one major stratigraphic break between Strata IV and III when all houses in the centre of the settlement stood in ruins before new houses were erected there (Heit 2021; Pollock & Heit 2023).

Durations in models

Two time intervals have been estimated that are important for memory issues. The first is the time between the abandonment and destruction of House 14 and the burial, which provides information on whether the people involved in the burial ritual could have witnessed the use of the house or were even its inhabitants. For this duration, the difference command was used to determine the temporal difference between the skeletal sample of the buried person, Poz-139097, and the boundary event Start afteruse phase IV (end of the destruction of House 14: see duration a in Figure 4). In the Model 1 duration estimate (36–152 years, 95 per cent probability or 70–121 years, 68 per cent probability, Fig. 8a), the man, who was 50 or more years old at the time of his death, could have seen the house in use. In Model 2 (55–154 years, 95 per cent probability or 79–129 years, 68 per cent probability: Fig. 8c) and Model 3 estimates (70–160 years, 95 per cent probability or 90–135 years, 68 per cent probability: Fig. 8e), such a scenario is impossible. In general, it seems less likely that any of the persons involved in the ritual of burial had witnessed the house.

Figure 8. Estimated durations: (a) Model 1, time between the destruction of House 14 and the burial; (b) Model 1, time between the complete covering of the ruins of House 14 with settlement debris and the burial; (c) Model 2, time between the destruction of House 14 and the burial; (d) Model 2, time between the complete covering of the ruins of House 14 with settlement debris and the burial; (e) Model 3, time between the destruction of House 14 and the burial; (f) Model 3, time between the complete covering of the ruins of House 14 with settlement debris and the burial.

Did these people at least see the ruins of the building? To address this question, we need to consider the time between the end of after-use of the ruins of House 14 and the burial MDB11. Depending on the model, this period is defined by different boundary events (see duration b in Figure 4). In Model 1, the difference was calculated between the boundary event Transition IV/III between two contiguous macrostratigraphic units Stratum IV and III and the skeleton sample Poz-139097. In Model 2, the corresponding boundary event was End IV marking the end of Stratum IV. In Model 3, the boundary event End afteruse phase described the end of the after-use subphase of House 14 within the larger Stratum IV phase.

The results show significantly larger differences between the three models compared to the previous estimate of the time between House 14 in use and the interment. The relatively wide ranges make the interpretation of these durations challenging. They can, however, be interpreted at the generational scale. In Model 1 (–2 to 90 years, 95 per cent probability or –2 to 51 years, 68 per cent probability: Fig. 8b), which is least plausible from an archaeological point of view, the time between house destruction and burial offers all possible scenarios, namely that the interment might have taken place shortly after the complete covering of the ruins with settlement debris (or according to the negative lower values of the span even before the destruction of House 14, which is impossible from a stratigraphic point of view) as well as after generations of residents had passed. Model 2 estimates (14–124 years, 95 per cent probability or 43–100 years, 68 per cent probability: Fig. 8d) imply that considerable time has passed since the complete disappearance of the ruins of House 14 beneath the settlement deposits. However, both the buried person and the ritual participants may have seen the ruins of the house. Likewise, the time between two investigated events could have included one or more generations of residents based on a mean generation interval of 25–32 years (Fenner 2005). According to Model 3 estimates (51–152 years, 95 per cent probability or 81–131 years, 68 per cent probability: Fig. 8f), it is rather unlikely that the buried person ever saw the ruins of the house. The time between two events in this case spanned several generations of inhabitants.

House memories

The estimated time interval between the destruction of House 14 and the burial on its ruins, which exceeds the realm of individual life, shows that the house and its physical layout were part of long-standing collective memory. The importance of the house as an organizing social unit in past communities and its role in creating and maintaining social memory have been highlighted in archaeological works dealing with social dimensions and temporal depths of dwellings. Studies that build on the ‘house societies’ concept of Claude Lévi-Strauss and focus on its intersection with material culture (cf. contributions in Joyce & Gillespie 2000), but also comparable concepts such as the ‘history house’ (Hodder and Pels 2010) emphasize the agency of the house and its ability to shape identities and generate histories transcending the lives of its inhabitants. These studies often consider the spatial continuity or durability of houses as hints at their significant role in the organization of social relations and commemorative potential. Houses serve as arena for memory practices and storage of past events, with ancestors kept inside them as spiritual or even physical remains. The intramural burials occurring in long-standing structures have been interpreted as part of a ‘strategic material language’ (Kuijt 2001, 93) that emphasizes affinity and cements intra-generational social relations in space.

Both spatial continuity of houses and burials in domestic space seem to have been part of the material language of the inhabitants of Monjukli Depe; however, they were not a universal strategy in management of social space. The houses at Monjukli Depe appear to have been central places in the social life of the inhabitants. This is indicated by the elaborateness of house interiors, especially particular parts of the constructions such as the buttresses. The relatively large (20–25 sq. m) inner spaces contained various installations and were regularly cleaned and renovated. While the form of the structures remained strikingly uniform over the centuries, the houses show considerable variability in their durability and use and in their trajectories in terms of house histories (Heit 2021; Pollock & Heit 2023). Besides houses or sequences of houses which were used longer or built directly on top of each other and thereby indicated a strong commitment to place, there were also other houses with shorter durations that, once abandoned or destroyed, were never rebuilt.

Burials occur in many but not all houses at Monjukli Depe. Despite high variability in mortuary practices (Rol 2019), a general trend regarding house trajectories can be noted. The houses with long biographies contain burials of younger individuals—infants, children, or adolescents—interred in the ‘intermission’ phases between house uses: during temporary abandonment, before or during house construction as well as during rebuilding and possible temporary abandonment, on ruins of an older and under the floors of a younger structure. The houses with short biographies did not include any burials or, in two cases, senior individuals were placed on the ruins of completely destroyed houses: in addition to MDB11 in House 14, an older individual was buried in House 7 at a similarly prominent location of the former house (Pollock & Heit 2023). This suggests that the house in Monjukli Depe remained a spatial anchor in the living landscape, historicized through oral transmission of social memory not only after its abandonment, but even after its disappearance from everyday life. The commitment to past houses is not only evident in the vertical sequences of houses and burials in (former) houses. Stones strewn on the floor during the abandonment of some houses with short biographies, as well as caching objects in the ruins of houses, especially near the buttresses, indicate domestic structures that were no longer used for living purposes but still remained a part of social space and memory (Heit 2021).

The long period between the destruction of House 14 and the commemorative event of the interment indicates that the house remained in the collective memory of the Monjukli community as an important place of the past for several generations. No less interesting is the question of how the spatial knowledge of a house that was no longer visible was maintained or passed on. If the time span between the complete disappearance of the ruins of House 14 under the settlement surface and the burial was less than 50–60 years (which corresponds to the age of the oldest individuals at Monjukli Depe as known from the excavated burials), then it is possible that among those who performed the interment there was an older person who witnessed the house ruins and retained his or her embodied, sensory experience with the house and its layout and location. If the period between the two events was longer than the maximal life span of the Monjukli inhabitants—which is possible according to the statistical results of Model 2 and very likely according to Model 3, both of which are stratigraphically preferable models—then also the knowledge of the house location must have been passed on to younger people orally and possibly also visually (describing and showing the exact location).

In any case, this act of remembrance required abstraction and topological thinking, since the place where the house had been located was transformed over time by building new structures, and the relief and heights of its walking surface changed due to accumulation of trash and building debris. The long-lived neighbouring House 3, which was built in Stratum IV along with House 14 and still existed at the time of the burial, may have served as a spatial anchor in this regard. This case can be seen as a prime example of the conscious engagement of the inhabitants of the tell settlement with their own past underfoot, conceptualized by McAnany and Hodder (2009) as ‘social stratigraphy’.

Conclusion

The temporality of collective memories and experienced and transmitted histories of a settlement can be approached by Bayesian chronological modelling, which provides duration estimates of social practices including memory work. The analysis of alternative scenarios shows that the modelling duration estimates depend not only on radiocarbon dates, but significantly on archaeological interpretation of excavated stratigraphic sequences and samples. These estimates cannot achieve an accuracy of a few years. However, the precision of duration estimates can be assured at the generation or human lifespan level, and this is an important benchmark in the transmission of memories. In considering duration and temporality of memory practices embedded in material culture, Bernbeck et al. (2017, 25) speak of a ‘relational duration’, referring to the relationship between the ‘lifetimes’ of things in relation to human life (cf. Miller 1994). This relational duration can be applied to the duration of memories, although the question would be whether these are shared within one or across one or more generations. As the examples of modelling stratigraphic sequences with different, sometimes very short, durations of houses and levels in tell settlements show (cf. Nishiaki et al. 2018), a generation cannot be determined as a reference value from stratigraphic information alone. The houses with different durations in Monjukli Depe are a further indication that house lifespans should not be equated with generations of inhabitants. As the example of Monjukli Depe shows, Bayesian chronology can help to clarify questions of whether or not specific, temporally displaced events were experienced within a human lifetime. For this benchmark of human life the experienced time of past people makes a major difference: while in the former case what is directly experienced is preserved in the memory of one person or a group of people, in the latter case the narrative of an experience becomes a collective intellectual possession transmitted from older to younger people and in a sense an indicator for the importance of particular events or past realities in the life of a community. In Monjukli Depe, such past realities of great importance were the old houses, whose materiality and spatiality were preserved in the collective memory of the village community after the demise of the houses themselves and their witnesses.