CHRONOLOGY OF THE BURIAL ACTIVITY OF THE LAST HUNTER-GATHERERS IN THE SOUTHWESTERN IBERIAN PENINSULA, PORTUGAL

ABSTRACT For most of human history, funerary burial has been unusual. Archaeology shows a shift in funerary practices in postglacial hunter-gatherers, in parts of Europe during the Late Mesolithic. This is documented by the burial grounds in the Tagus and Sado valleys in the southwestern Iberian Peninsula, Portugal, where ca. 376 burials were excavated. This study presents a chronology for the burial activity in these sites and contextualizes the start and end activity phases within regional environmental changes and cultural developments. The dataset consists of 76 14C dates on human bone (19 new, 57 published) including new dates from contexts in Portugal outside these valleys. Bayesian chronological models were defined in OxCal, and protein carbon contributions of marine foods were estimated by the Bayesian mixing model FRUITS. The results indicate a broader timeframe for the Late Mesolithic in Portugal, than previously suggested, starting during a period of significant environmental changes, ca. 8500–8300 cal BP, and ending ca. 7000 cal BP. The burial activity decreased during the establishment of Neolithic farmers in southwestern Iberia from ca. 7450 cal BP, however, these burial grounds continued to be used by Mesolithic hunter-gatherers, showing that diverse social structures and worldviews coexisted for several generations.


INTRODUCTION
Death rituals involving the burial of the deceased were unusual until the Late Mesolithic (Pettitt 2011). The Mesolithic shell middens of the Tagus and Sado valleys in Portugal are remarkable because of the large number of human bodies buried in these sites. These are some of the largest and earliest burial grounds known, arranged and maintained by the last populations of hunter-gatherers in southwestern Europe (Jackes and Lubell 2012;Umbelino and Cunha 2012;Peyroteo Stjerna 2016a). The development of open-air burial grounds contrasts with earlier practices in the region; burial was practiced in the Iberian Peninsula, however, it was not a common treatment (Peyroteo Stjerna 2016b).
This study examines the chronology of the burial activity of the last hunter-gatherers in southwestern Iberia, Portugal, and contextualizes the start and end activity phases within regional environmental changes and cultural developments. The broad chronology has been established but there are several outstanding issues that this paper aims to address: 1. The nature and context of the 14 C dates has not been systematically assessed in terms of the human activity they are supposed to date. Samples used in this study come from buried human bone only, and the reliability of the isotopic measurements is assessed. 2. Stable isotopes of carbon ( 13 C) and nitrogen ( 15 C) indicate the consumption of foods from terrestrial and marine reservoirs (Lubell et al. 1994;Fontanals-Coll et al. 2014;Guiry et al. 2015;Peyroteo Stjerna 2016a), which may influence the calibration of 14 C measurements and their chronological interpretation (Stuiver et al. 1986). This study applies a new approach to estimate the protein carbon contribution to diet with the Bayesian mixing model FRUITS (Fernandes et al. 2014), and uses these measurements to calibrate the 14 C dates with the Bayesian chronological models in OxCal (Bronk Ramsey 2009).
3. The burial activity of the last hunter-gatherers in Portugal clusters in the Tagus and Sado valley, and the evidence for human remains outside these regions is scarce and largely undated (Peyroteo Stjerna 2016b). This paper investigates human remains in Late Mesolithic contexts outside these river valleys and provides new 14 C dates. 4. Hunter-gatherer and farmer relations at the onset of the Neolithic in Portugal, ca. 7450-7250 cal BP (Carvalho 2010;Martins et al. 2015) are poorly understood. A recent study suggested that one of the Tagus burial grounds was used simultaneously by Mesolithic and Neolithic people (Bicho et al. 2017). This paper tests this hypothesis from a multisite perspective by examining all human remains dated at the end boundaries of the burial activity.
The chronological analysis is based on a dataset of 76 14 C dates on human bone (19 new, 57 published), excavated in Late Mesolithic contexts in Portugal. While the focus on the human burials allows the establishment of chronologies based on strong indicators of mortuary activity, other uses of the sites were not considered in this study.

ARCHAEOLOGICAL BACKGROUND
Most human remains dating to the Late Mesolithic in Portugal are in open-air shell midden sites in the valleys of two major southern rivers, the Tagus and the Sado (Peyroteo Stjerna 2016b). The Tagus middens are located on the lower valley, ca. 80 km northeast from the river mouth in Lisbon, on the banks of three tributaries to the Tagus River: Fonte da Moça, Muge, and Magos. The sites of the Sado valley are located ca. 120 km south of Muge and ca. 80 km east from the river mouth ( Figure 1). From the 13 middens identified in the Tagus valley, nine have human remains with a minimum number of individuals (MNI) of 263 (Paço 1938;Santos et al. 1990; Cunha and Cardoso 2001;Roksandic 2006;Jackes and Meiklejohn 2008;Meiklejohn et al. 2009;Bicho et al. 2013). Similarly, in the Sado valley, there are 11 known middens, six of which with human burials to a total of ca. 113 individuals (Cunha and Umbelino 1997;).
Late Mesolithic human remains found outside these river valleys are rare. One tooth dating to the Late Mesolithic was excavated in an open-air site in southern Portugal . Skeletal remains were also found in a shell midden located in the Mira valley on the coast of Alentejo (Arnaud 1988;Lubell et al. 2004), and new 14 C data confirm the Late Mesolithic chronology of the remains. Lastly, one skeleton excavated in the rock shelter Abrigo Grande das Bocas was interpreted as possibly Mesolithic based on the analysis of the original field notes (Bicho 1995). 14 C measurements obtained for this study  date the death of this individual to 133-330 cal CE (Table S1.1).
At the Tagus and Sado valleys, the burial of the deceased was a common way of handling the cadaver. Mortuary practices in both valleys were based on common principles indicating a shared mindset on the treatment of the dead (Peyroteo Stjerna 2016a). There were slight variations, but these did not affect the overall pattern of the treatment of the dead. The cadavers were typically placed in individual graves, soon after death, while retaining their anatomical integrity. The bodies were normally laid on the back, with the upper limbs arranged close to the body, and with the lower limbs in flexion at various degrees, with the feet often placed towards the buttocks ( Figure 2). Typically, the cadavers lay underground and undisturbed in their graves. In some instances the bones of a previous burial were partially removed to accommodate the burial of a new cadaver, indicating the continued preference of certain areas for the burial of the dead within a larger area of the site (Roche 1972;Peyroteo Stjerna 2016a). The relation between these burial clusters and the remaining areas of the sites remains an open question.
In rare cases, burial analyses confirm the multiple nature of primary deposits containing the remains of two individuals Peyroteo Stjerna 2016a). Recently, the context CAM-01-01 at Cabeço da Amoreira was referred to as a multiple burial, possibly in successive episodes, corresponding to changes in Mesolithic funerary patterns, in line with Neolithic practices (Bicho et al. 2013;Bicho et al. 2017). This interpretation is not supported by archaeological data. While CAM-01-01 contained the skeletal remains of one main individual and bone fragments of at least three others, this observation alone is not evidence of a multiple burial. The bones were distributed over an area of 2 × 2 m and excavated between 0.2-0.6 m from the surface of the shell midden. The material was heavily disturbed by recent activities, but it was possible to reconstruct the burial position of the main individual (Roksandic 2006). The scattering and fragmentation of the bones, as well as the lack of key observations (Duday 2009) such as the identification of a common burial context, or the common arrangement of the remains does not allow the determination of CAM-01-01 as a multiple burial of Neolithic tradition.
These middens are also rich in faunal and botanical remains providing evidence for the consumption in situ of a varied range of food resources, terrestrial, estuarine and marine (Lentacker 1986;Lentacker 1994;Detry 2007;Marques-Gabriel 2015;Rowley-Conwy 2015;López-Dóriga et al. 2016). There is no evidence for domesticated plants or animals, except for the domestic dog (Detry and Cardoso 2010;López-Dóriga et al. 2016). Stone tool production is also well documented for all stages of the chaîne opératoire (Araújo 1997;Marchand 2001;Nukushina 2012;Cascalheira et al. 2015). The few fragments of pottery found in some areas of the sites show evidence for occasional Neolithic activity (Ferreira 1974;Bicho et al. 2011;. In southwestern Iberia, the Late Mesolithic settlement was clustered in the innermost areas of large estuaries, in contrast with the scattered pattern known for the Early Mesolithic (Araújo 2015). Explanations for this reconfiguration (Zilhão 2003;Bicho et al. 2010) are based on paleoenvironmental evidence. Here, the rapid sea level rise during 12,000-7000 cal BP resulted in the drowning of the Tagus and Sado valleys creating inland sheltered seas and tidal environments several kilometers upstream (van der Schriek et al. 2007;Vis et al. 2008;Costa et al. 2019). Data from Muge show the rapid establishment of inner estuarine environments ca. 8050 cal BP (van der Schriek et al. 2007), although the tidal mudflat and saltmarsh conditions were settling in earlier (Vis et al. 2008). Research in the Sado valley indicates a similar pattern (Costa et al. 2019), and preliminary data suggest that estuarine conditions existed as early as 8500 BP (A. Costa, pers. comm.). These new ecosystems were particularly attractive not only because of their natural resources, but also because of their sheltered inland position with easy access to the interior and to the Atlantic coast through connected waterways. As the sea level rise declined, the river systems prevailed, and the estuaries began to contract. In Muge this occurred between ca. 7450-5750 cal BP and an open landscape was established around ca. 6950 cal BP, although the same rich and diverse ecological systems remained until at least ca. 5750 cal BP (van der Schriek et al. 2007). In the Sado valley, the estuarine environment and the transitional marine-fluvial habitats were still productive until ca. 3300 cal BP, at least at Arapouco (Costa et al. 2019).

MATERIALS
The analysis was based on 14 C dates on buried human bone exclusively. Dates on other material are available, such as on charcoal and shells, but with few exceptions (Bicho et al. 2013;López-Dóriga et al. 2016) the context is unclear and the relationship with the burials is unknown.
A total of 76 sets of measurements of 14 C and stable isotopes of 13 C and 15 N were selected for analysis, corresponding to 70 individuals. This material was excavated since the 19th century from the shell middens in the Tagus (Table 1).
Moita do Sebastião is located on a low terrace, 20 m.a.s.l., on the left bank of Muge (Roche 1972). It has excellent access and visibility towards the river and towards Cabeço da Arruda on the opposite margin. The neighboring site of Cabeço da Amoreira is ca. 700 m east. It was excavated in the 19th century (Ribeiro 1884; Paula e Oliveira 1888), and in 1952-1954 after the destruction of the top layers of the shell midden which were estimated to be at ca. 2.5 m from the bottom river-terrace sands (Roche and Ferreira 1967). Despite the depth of the midden, it has been suggested that possibly all skeletons were excavated from the lower layers (Jackes and Alvim 2006).
Cabeço da Amoreira is located 15 m.a.s.l. on the left bank of Muge, on the opposite side to Cabeço da Arruda, ca. 700 m east of Moita do Sebastião (Roche 1965 1961-1964(Roche 1965), 2001-2003(Roksandic 2006, and 2008-present (Bicho et al. 2013). The shell midden has a depth of ca. 3.5 m in the central area and human remains were excavated from the top layer down to the bottom sandy layer on the base of the midden .
Cabeço da Arruda is located on the right bank of Muge, 8 m.a.s.l., on the edge of the floodplain. The site has excellent visibility towards the river and towards Moita and Amoreira on the opposite margin. It was excavated in the 19th century (Ribeiro 1884; Paula e Oliveira 1888), 1933 and 1937 (Cardoso and Rolão 2000), 1964-1965(Roche 1967, and in 2000 (Roksandic 2006). Arruda is the largest shell midden in Portugal reaching ca. 5 m in depth in the central area (Roche 1967), with human remains excavated from various depths of the midden (Peyroteo Stjerna 2016a).
Cova da Onça was the first shell midden discovered in Portugal, in April 1863 (Ribeiro 1884). The site is located 2 to 8 m.a.s.l. on the right bank of Magos near another midden named Cabeço dos Ossos. Both sites are destroyed. There is no documentation known for this site and the bones are stored in disarticulated state and mixed in containers without identification or stratigraphic information Meiklejohn et al. 2009;Peyroteo Stjerna 2016a).
Arapouco is the westernmost shell midden in the Sado valley. The site is located on the left margin of the river at 47 m.a.s.l., and has excellent visibility over the valley (Diniz and Arias 2012). It was excavated in 1961-1962, but there is no information about the location of the burials in the midden. According to Dario de Sousa who did the site drawings, which are missing from the site archives, all burials were in the bottom sandy layer beneath the layers with shells (Peyroteo Stjerna 2016a).
Poças de S. Bento stands on a plateau ca. 3 km from the Sado River at 85 m.a.s.l. with no visibility towards the valley, unlike most sites (Diniz and Arias 2012). The site was excavated in 1960in , 1964in , 1986in -1988in -2017in (Arnaud 1989Larsson 1996;Arias et al. 2015). All burials were identified in the basal sandy layers ca. 1.6-1.8 m deep, at the bottom of the shell midden (Peyroteo Stjerna 2016a).
Cabeço das Amoreiras is located on the left margin of the Sado River at 52 m.a.s.l. (Diniz and Arias 2012). The site was excavated in 1958in , 1985in -1986in (Arnaud 1989. The human remains were excavated from the bottom sandy layer in a small area of ca. 25 m 2 , at depths ranging from 1.2 to 1.5 m from the top of the shell midden (Peyroteo Stjerna 2016a).
Vale de Romeiras is located on the eastern side of the Sado valley, on the right margin of the river at 55 m.a.s.l. It has excellent visibility over the valley, Cabeço das Amoreiras on the opposite margin, and Cabeço do Pez on the neighboring hilltop (Diniz and Arias 2012). The site was excavated in 1959-1960, and most burials were found lying on the bottom sand or soft bedrock at depths ranging from 0.1 to 0.84 m due to the significant erosion of the top layers of the midden (Peyroteo Stjerna 2016a).
Cabeço do Pez is one of the largest sites in the Sado valley. It is located on the eastern part of the valley, on the right margin of the river at 52 m.a.s.l., on a hilltop next to Vale de Romeiras, with good visibility over the river (Diniz and Arias 2012). It was the first shell midden discovered in the valley (Barradas 1936), but was excavated only 20 years later in 1956and 1958-1959(Peyroteo Stjerna 2016a, as well as in 1983 and 2010 (Arnaud 1989;Arias et al. 2015). Documentation is limited but most skeletons seem to have been excavated from the sand layer and on top of the bedrock beneath the layers with shells. The two individuals identified in 1959 were excavated from a middle layer with shells 0.18-0.8 m deep (Peyroteo Stjerna 2016a).
Várzea da Mó is located on the eastern side of the Sado valley on the right margin of the river, 8-25 m.a.s.l., but the precise location of the site is unknown (Diniz and Arias 2012). It was excavated in 1959 and only one skeleton was identified. The human remains were excavated on the bottom of the shell midden at a depth of ca. 0.65 m (Peyroteo Stjerna 2016a).
Fiais is located 10 km from the modern coast, on a high terrace of the Mira River at 100 m.a.s.l., and was excavated in 1986-1988 (Arnaud 1988;Lubell et al. 2004). Fragments of human bones of possibly several individuals were excavated in 1986-1987, found scattered and mixed with faunal remains. In 1988, the remains of one individual were found clustered but disarticulated in one area of the site (Arnaud 1988

Estimate of the Marine Carbon Contribution towards Bone Collagen
The concentration of 14 C in the sea, rivers, and lakes is generally lower than that in the contemporary atmosphere due to a slower mixing rate between the chemical elements in the atmosphere and the water surface. Because of this relative lower concentration in 14 C, the organisms grown in water reservoirs will variably present relatively older 14 C ages, than those from coeval terrestrial environments. Thus, the foods consumed from different primary reservoirs, terrestrial or from the water systems, will determine the 14 C incorporated in each archaeological sample (Stuiver et al. 1986). This is particularly important when calibrating 14 C measurements from humans that may derive their carbon from foods of terrestrial and/or marine origin.
Protein carbon from marine and terrestrial sources in the specimen's bone collagen are estimated using FRUITS 3.0 (Food Reconstruction Using Transferred Signals, Fernandes et al. 2014). FRUITS introduced the Bayesian mixing model approach to diet reconstruction studies, and it is used to provide more accurate estimates of food intake using simple or complex models, depending on the nature of the data and on the archaeological question. To correct chronologies obtained from 14 C dating of human bone collagen that potentially have a reservoir effect, a simple concentration-independent and non-routed model can be defined by using the measured δ 13 C value of each sample (Fernandes 2015). This is a powerful method not only because it provides an estimate of the protein carbon from marine and terrestrial foods for each consumer, but also the associated uncertainties, which can be included in the chronological calibration. The output of the model is presented on Table 2 (% marine). See Supplementary Material (S2) for specific details on model parameters and assumptions.

Calibration and Modeling
The dataset excludes individuals with 14 C measurements published without independently measured isotope-ratio mass spectrometry (IRMS) isotopic values (Table S1.2) because only IRMS-based values can be reliably used for isotopic-based environmental analysis such as reservoir corrections (Millard 2014;Taylor and Bar-Yosef 2014). The parameter collagen yield (%C, %N) was not always provided by the laboratories, and for this reason, the atomic carbon and nitrogen ratios (C:N), and δ 13 C and δ 15 N were used for quality checks. The acceptance ranges of the atomic ratio of carbon and nitrogen (C:N) are within the 2.9-3.6 range (DeNiro 1985; van Klinken 1999), although C:N ratios between 3.4-3.6 may indicate some contamination (Ambrose 1993). See Supplementary Material (S3) for specific details on 14 C laboratory protocols.
All chronological models are defined by a Bayesian approach to the analysis and interpretation of 14 C measurements in archaeological contexts (Bayliss et al. 2007;Bayliss 2009). In the case of sites with one 14 C date an independent event calibration is used.
The prior beliefs (the archaeology) are limited to the surviving documentation. Stratigraphic relations of the dated material are limited, and the Bayesian models are primarily defined by simply incorporating the information that the dates derived from a coherent archaeological activity (i.e., burial activity). Despite the limitations, the application of the method allows the estimate of posterior densities of the 14 C measurements, which are more precise than single date estimates (Bayliss 2009).
The models are outlined in OxCal 4.3 (Bronk Ramsey 2009) which provides the posterior beliefs for each model and event. For this study, the most suitable kind of model is the site-based model. Site-based models are defined as phase models framed by two undated events, i.e. boundaries, for the start and end of activity. The introduction of boundaries in a model is a way to make the events relate statistically to each other by accounting for the scatter of dates due to measurement uncertainty. This approach is particularly useful when we do not have explicit archaeological data indicating a clear order of a group of elements (Bronk Ramsey 1995), such as in the case of historical collections with scarce documentation. This is a reliable approach because each sample corresponds to a burial event which is part of the burial activity at the site.
Two statistical criteria produced by OxCal are used for reliability and consistency testing: the index of agreement (A) and convergence (C) (Bronk Ramsey 1995). A is high (≥ 60%) when the posterior distribution is situated in a high probability region of the prior distribution. If A is low (< 60%) the result is regarded problematic, or it may indicate the presence of a statistical outlier. The overall index of agreement (Aoverall) provides a measure of the consistency between the prior information and the 14 C results. Aoverall is typically ca. 100% with a threshold value of 60%, and models which produce values lower than this should be re-examined (Bronk Ramsey 1995;Bayliss et al. 2007). The index of convergence indicates how quickly the sampler algorithm can produce a representative and stable solution to the model. In practice, a model is unstable when C is poor (< 95%) and the results should not be used (Bronk Ramsey 1995).
Calendar ages are reported as "cal BP" (Before Present, where Present is 1950 CE). All calibrated ranges are given at 95% probability. The calibrated ranges quoted in italics derive from mathematical modeling and are posterior density estimates. In each plot, the large square brackets and the OxCal keywords define the overall model. For each of the dates two distributions have been plotted, one in outline which is the result produced by the scientific evidence alone, and a solid one which is based on the chronological model used. Calibrated ranges of 95% and 68% probability are plotted by the lower and higher square brackets, respectively.

Moita do Sebastião
Two chronological models were built on 19 14 C dates, one of which is new, on skeletons from all excavation areas (Table 2). One published measurement (Beta-127449) was excluded from the models because its δ 13 C value was obtained by AMS (C. Umbelino, pers. comm.) (Table S1.2).
Chronological model 1 assumes one general phase of burial activity and does not impose a relative order to the dated events. The model shows good overall agreement and the agreement index for each dated event is high, except for the sample Ua-46264 (9, 1952Ua-46264 (9, -1954 which is in poor agreement but in good convergence ( Figure 3). This sample is the earliest dated burial in the site and its greater age suggests an earlier event relatively older than the group of events in the considered phase. Its low agreement and the visual observation of the plot suggests the possibility of more than one archaeological phase of burial deposition. As a working hypothesis, and based on the visual observation of the plot, a second model is proposed, by explicitly defining three main periods of burial activity as three phases, with boundaries defined for each phase (Figure 4). Model 2 shows good overall agreement, and the agreement index for each dated event is high, indicating that the 14 C dates are in accordance with the model. Sample Ua-46264 which is in poor agreement in model 1 (A = 38%, 8292-7996 cal BP) is in good agreement in model 2 (A = 62%, 8344-8017 cal BP). Nevertheless, model 2 should be considered with caution.
Burial Activity of the Last Hunter-Gatherers 273 The archaeological data is insufficient, and we know only that the burials were excavated from the basal sand layer, at the bottom of the shell midden. This is not to suggest that the burial activity could not have happened in distinct phases, but to caution against the manipulation of the data, which, without a strong archaeological basis, can be significantly misleading.  Overall, the available 14 C measurements can be considered representative of the burial assemblage. The chronological sequence suggests that the burial activity clustered in three chief moments, possibly discontinuously, with relatively short hiatuses between them without significant burial activity. At Moita do Sebastião the burial activity was frequent and continuous between ca. 8000-7600 cal BP (mod. 1) or 7900-7700 (mod. 2), supporting earlier suggestions of a main group of burials clustering at ca. 7800 cal BP (Jackes and Lubell 2012). Burials from all excavated areas fall into this time span, indicating that when the burial practice was more frequent, the extent of the burial area was as large as the total burial area known. The earliest mortuary activity is estimated to have been between ca. 8300-7900 cal BP, when the local estuarine environment was expanding, as indicated by burials 9, 1, and 7. The burial activity seems to end ca. 7650-7450 cal BP, while the favorable estuarine ecosystem was still predominant, as illustrated by the dates of individuals 10, 33, and CT.

Cabeço da Amoreira
Two chronological models were built on 16 14 C dates, eight of which are new, on skeletons from all excavation areas (Table 2). One published measurement for individual CAM-00-01 (TO-10218) was excluded from the models because it was replaced by a reanalysis (TO-11819-R)   (Table S1.2). Two 14 C measurements available for CAM-00-01 (TO-11819-R and UOC-913) were calibrated using the OxCal function R_Combine and the average between the marine carbon estimates.
Chronological model 1 assumes one general phase of burial activity and does not impose a relative order to the dated events. The model shows satisfactory overall agreement and the agreement index for each dated event is high, except for the samples CAM-00-01 (TO-11819-R, UOC-913) and CAM-01-01 (Wk-26796), which are in poor agreement but in good convergence ( Figure 5). These are the oldest and youngest dates available and may indicate an earlier event relatively older than the group of events in the considered phase, as well as a more recent event.
Chronological model 2 tests the previously proposed 3-burial phase chronology for Cabeço da Amoreira (Bicho et al. 2013), which was based on the stratigraphic sequence as documented in recent excavations, and on AMS measurements on human bone (n = 7), shells (n = 28) and charcoal (n = 1): • Phase 1, bottom layer: burials 2011.1 (Wk-32143) and CAM-00-01 (TO-11819-R); • Phase 2, upper layers: burials 2011.2 (Wk-32142) and 7, 1933 (Beta-127450); • Phase 3, upper layers: burial context CAM-01-01 (Wk-26796, TO-10225, and TO-10218, note that this latter measurement has been rejected, see above). Both models 1 and 2 show that burial activity was more frequent and continuous than previously suggested (Peyroteo Stjerna 2016a). However, data from the new excavations seems to indicate that the burial activity clustered in three chief moments (mod. 2), possibly discontinuously, with relatively short hiatuses without significant burial activity. While model 2 may be well resolved in the light of the new documentation for the site, it is unclear how it fits with previous archaeological data. The models are comparable at the start and end of the burial activity at the site, but without further examination of the earlier excavations the hiatuses estimated by model 2 should be considered with caution ( Figure 6).

Cabeço da Arruda
The chronological model was built on 18 14 C dates, three of which are new, on 16 skeletons excavated from all areas ( Table 2). Two published measurements on individuals CA-00-01 (Wk-26795) and CA-00-02 (Wk-26794) were excluded from the model because were measured on dental enamel (Table S1.2). Despite improving methods, dating and calibrating measurements on enamel is problematic (Hedges et al. 1995;Hopkins et al. 2016). Fortunately, both skeletons were also collagen dated (TO-10217 and TO-10216). Two 14 C measurements available for skeleton 6 (Beta-127451 and AA-101343) were calibrated using the OxCal function R_Combine and the IRMS-based δ 13 C value (AA-101343). Likewise, the two 14 C measurements for individual III (Beta-447681 and TO-360) were calibrated using R_Combine and the average between the marine carbon estimates.
The model assumes one general phase of burial activity and does not impose a relative order to the dated events. It shows good overall agreement, and agreement index for each dated event is high, except for skeleton N (TO-356) which is in poor agreement but in good convergence (Figure 7). This individual is the most recent burial known in the site and its younger age may suggest a later event relatively more recent than the group of events in the considered phase.
A second chronological model could be proposed by defining the main phases of burial activity in relation to the documented depth of each burial. Nevertheless, the posterior density estimates in model 1 show that depth and antiquity of the burial are not always correlated. This is the case for CA-00-02 (TO-10216) found in the base of the midden. This burial in the bottom sandy layer, presents a relatively more recent 14 C date than burial 6 (Beta-127451, AA-101343) which was found 0.8 m from the base of the midden, as well as more recent than individual 3 (UA-46273) at 1.2 m from the bottom. Other dated burials, however, seem to follow a vertical chronology, with individual CA-00-01 (TO-10217) at 2 m from the bottom presenting one of the most recent dates known at the site.
This is a large site with many human burials in a complex stratigraphy. Nevertheless, 14 C data indicates that the different burial areas excavated in the 19th century, 1930s, and 1960s were synchronic, and can be analyzed as one series considered representative of the assemblage.
The earliest burial activity at Arruda is estimated to have been between ca. 8150-7850 cal BP, while the estuarine environment was established in the region, as indicated by the 14 C dates for individual 6 (Beta-127451, AA-101343). Despite doubts expressed about the early date of this burial (Jackes and Meiklejohn 2004;Jackes and Lubell 2012) the new 14 C date on skeleton B (Ua-56457) strengthens the case for the antiquity of these burials. This earlier activity is followed by the main period of burial, where most 14 C dates cluster, indicating that burial

Cova da Onça
The chronological model was built on four new 14 C dates ( Figure 8, Table 2). One published measurement (Beta-127448) was excluded from the model because its δ 13 C value was obtained by AMS (C. Umbelino, pers. comm.) (Table S1.2).
Burial activity at Cova da Onça seems to be continuous, possibly with one short hiatus between the earliest and main burial activity, which should be tested with further data. Nevertheless, the current model shows good overall agreement and estimates the main activity to have been between ca. 7600-7350 cal BP, with an earlier moment at 7920-7617 cal BP, when the inner estuarine conditions were highly favorable. The later burial activity at the site, estimated at 7488-7258 cal BP, is coeval with the earliest Neolithic phases in Portugal. More dates are necessary to refine this chronology; however, the new data securely confirms the Late Mesolithic chronology of these burials.

Arapouco
Only skeleton 2A provided 14 C measurements within the accepted quality ranges (Sac-1560 and Beta-447689), despite multiple attempts (Cunha and Umbelino 2001;Peyroteo Stjerna 2016a). The 14 C dates were calibrated using the OxCal function R_Combine and the IRMS-based δ 13 C value (Beta-447689). These measurements do not allow the construction of a chronological model for the burial activity at Arapouco but date the death of individual 2A to have been between 8158-7971 cal BP (Table 2). This is a large site with high density of human burials, and despite the poor collagen preservation further 14 C measurements should be tried in the future.

Poças de S. Bento
Two individuals buried at Poças de S. Bento have been 14 C measured. One measurement is from the skull excavated in 1986 (Ua-425) which reliability is uncertain because the isotopic values are not available (Table S1.  Table 2). The construction of a chronological model for the burial activity at Poças de S. Bento requires further 14 C dates on human bone.

Cabeço das Amoreiras
The chronological model was built on three 14 C dates (Figure 9). The measurement Beta-125110 was calibrated with the recently published IRMS-based δ 13 C value (Table 2). Despite the low number of dates this model can be considered representative of the assemblage, because the six burials from this site were found in one stratigraphic layer and lie in proximity (Peyroteo Stjerna 2016a). These burials were expected to be coeval, however, the model suggests that the burial episodes were spread over time. Despite the scattering, all dates introduced in the model show good agreement and the model did not consider any outliers. The earliest burial (5, Beta-125110) is estimated to have been between 8153-7880 cal BP, while the most recent use of the site for burial practice is dated several centuries later, 7439-7275 cal BP (4, Ua-47973). At Cabeço das Amoreiras, burials were not frequent, but the practice continued over a long period of time.

Vale de Romeiras
Only three skeletons out of 15 provided 14 C measurements within the accepted quality ranges (Cunha et al. 2002;Peyroteo Stjerna 2016a). Furthermore, the chronological model is limited to two 14 C dates ( Figure 10, Table 2), because one of the skeletons (8, Ua-46968) has a surprising modern age and is out of the scope of this study (Table S1.1). This is a small number of reliable measurements compared with the MNI (26) buried in the site. Although new measurements should be tried in the future, this preliminary model provides valuable indications about the burial activity at Romeiras. The earliest period is estimated to have been between 8543-8334 cal BP, as indicated by the burial of individual 19 (Ua-46972) found on the basal sand outside the shell midden area (Peyroteo Stjerna 2016a). The second date (9, Ua-47983) is several centuries later, with a posterior density estimate of 7575-7429 cal BP. This burial is in the shell midden area and was excavated in the bedrock. The two measurements indicate a long duration of burial activity but it is unknown if the practice was frequent during certain periods or scattered over a long period These results indicate that the burials at Vale de Romeiras do not represent one coeval group, contra earlier suggestions of a synchronic burial ground, based on the spatial arrangement of the burials, organized in a semicircle (Arnaud 1989). Nevertheless, these dates do not invalidate this suggestion, which should be reformulated to focus on possible synchronic burial subgroups within the burial ground.

Cabeço do Pez
The chronological model was built on six 14 C dates, one of which is new, from skeletons excavated in 1956 from the two main burial areas ( Figure 11, Table 2). The three 14 C measurements available for individual 4 (Sac-1558, Beta-125109, and Beta-447690) were calibrated using the OxCal function R_Combine and the only IRMS-based δ 13 C value (Beta-447690). This is a relatively low number of samples when compared with the MNI (32-36) buried in the site. Nevertheless, the 14 C measurements obtained from both areas are coeval within a calibrated range of ca. 7680-7500 cal BP. At Cabeço do Pez, during the Mesolithic, the burial activity concentrates in a restricted period, indicating a remarkable high frequency of burial practice at the site, over a relatively short period of time.
The site was also active in the Middle Neolithic for burial, as indicated by the date of the isolated burial of a child (2, Ua-46930), 6446-6220 cal BP. The chronology of this burial is a strong argument supporting the already suggested interpretation (Arnaud 2000) of Cabeço do Pez as a shell midden developing during the Late Mesolithic with later activity at the site, as indicated by the fragments of Neolithic pottery excavated from the upper layers on top of the shell midden .

Várzea da Mó
The burial activity at Várzea da Mó is dated by the death of the only individual buried at the site, which is estimated to have been between 7315-7026 cal BP (Table 2). Burial Activity of the Last Hunter-Gatherers 283

Fiais
The new 14 C measurement dates the death of this individual to have been between 7693-7481 cal BP (Table 2) and confirms the Mesolithic chronology of the human activity.

Vale Boi
The loose human tooth excavated at Vale Boi is dated between 8397-8028 cal BP (Table 2). This date demonstrates the Late Mesolithic activity at the site; but its funerary nature is unknown.

Chronological Boundaries
Archaeological and 14 C evidence suggest that the burial activity of the last hunter-gatherers of southwestern  (Figure 12).   Bronk Ramsey 2009). In the case of sites with one 14 C date, an independent event calibration is used (95% confidence). Protein carbon contribution of marine foods towards bone collagen and associated uncertainties (% marine, SD) were estimated from the measured δ 13 C value and outlined in FRUITS 3.0 (Fernandes et al. 2014). Abbreviations: not published (n/p), not determined (n/d), standard deviation (SD), not available (n/a) but reported to be within the accepted quality range of 2.9-3.6 (Lubell et al. 1994 Burial Activity of the Last Hunter-Gatherers 289 Burial Activity of the Last Hunter-Gatherers 291 The oldest human burials lie on the sterile sandy layers at the bottom of the shell middens suggesting that the site development was associated with the use of the place for burial practices. Nevertheless, the 14 C data also highlights that depth and antiquity of the burial are not always correlated. This is illustrated by older burials in upper layers at Cabeço da Arruda and Amoreira in the Tagus valley, or by the burials with different time ranges but found at the same depth at Cabeço das Amoreiras and Vale de Romeiras in the Sado valley, suggesting a pattern of spatial continuities and discontinuities within the sites.
These early burials were followed by the most active period of burial practice in the Tagus  The last phase of burial activity in the shell middens is contemporary with the earliest Neolithic sites in Portugal and with human remains found in Early Neolithic contexts, mostly in caves, ca. 7400-7100 cal BP, e.g. Caldeirão, Correio-Mor, Cisterna, and Lameiras (López-Dóriga and Simões 2015; Martins et al. 2015;Carvalho 2018) (Figure 12). It has been argued that the latest burial activity at Cabeço da Amoreira integrated Mesolithic and Neolithic populations (Bicho et al. 2017). This interpretation was based on several observations about the burial context CAM-01-01. Unlike other individuals of the same chronology, the main individual in this context (bone 139, TO-10225) had a terrestrial based diet (Table 2) and was likely non-local as suggested by strontium values (Price 2015 Rather than the simultaneous use of these burial grounds by Mesolithic and Neolithic people, the consistency of later dates in several Mesolithic burial contexts strongly indicates that these sites were repeatedly used for mortuary activities by Mesolithic hunter-gatherers at the onset and establishment of the Neolithic lifestyle in southwestern Iberia, ca. 7400-7000 cal BP. Later burial activity is evident only at Cabeço do Pez in the Sado valley, by the burial of a child between 6446-6220 cal BP (Ua-46930) outside the two areas with Mesolithic burials. This event is coeval with the early stages of the Middle Neolithic in Portugal ca. 6450-5450 cal BP  and with Neolithic activity at Pez (Arnaud 2000). This episode requires further exploration as it suggests that some sites may have been used occasionally for burial activity after their abandonment by Mesolithic hunter-gatherer populations.

Environmental and Cultural Explanations
The well-defined chronological and spatial boundaries of the mortuary practice suggest that this relationship with death was particular to these communities. The beginning of the phenomenon coincides with the reconfiguration of the settlement pattern, as well as with other cultural changes in the region that define the Late Mesolithic in the southwestern Iberian Peninsula.
The start of burial activity at the Tagus and Sado valleys corresponds to the first moments of occupation of these sites, ca. 8500-8300 cal BP, and is coeval with sea level rise related environmental changes. In fact, the burial activity was more frequent between ca. 7800-7600 cal BP, when the inner estuarine environments were well established in the valleys. Although the chronological match between environmental changes and the new settlement pattern is not a coincidence, the social mechanisms allowing this territorial and social reconfiguration remain unclear (Araújo 2015), and need further research. Likewise, the processes leading to the decline and end of burial activity around 7450-7000 cal BP, at both the Tagus and Sado valleys, are poorly understood, and in this case, the environmental factors seem to play a less important role. The decrease of burial activity in the two valleys is coeval with the first phase of the Early Neolithic in Portugal ca. 7450-7250 cal BP, and coincides with environmental changes documented in the Tagus valleys, defined by the contraction of estuarine habitats ca. 7450-5750 cal BP and the establishment of an open landscape around 6950 cal BP. Nevertheless, the socio-cultural explanation for the end of the phenomenon, defined by the expansion of the Neolithic lifestyle, seems to be more compelling, because despite the decline of the environmental conditions, the inner estuaries were still productive and the same mix of favorable habitats remained until ca. 5750 cal BP in the Tagus valley, and ca. 3300 cal BP in the Sado valley.

CONCLUSION
This chronology of the burial activity of the last hunter-gatherers in southwestern Iberia starts with a paradigm shift contrasting with previous and contemporaneous responses to death in the Burial Activity of the Last Hunter-Gatherers 293 region. This change in funerary practices emerged in the context of settlement reorganization and cultural transformations that define the Late Mesolithic in the region.
Overall, these sites were used for burial activity more or less synchronously. In both the Tagus and Sado valleys, burial activity was more frequent between ca. 7800 and 7450 cal BP. In both valleys, there were early episodes of human burial activity, around 8500-8300 cal BP, defining the first phases of the Late Mesolithic in Portugal. After 7450 cal BP, the frequency of burial practice decreases dramatically, in both valleys, with a few last episodes not later that ca. 6950 cal BP.
Mechanisms of change that led to the disruption of this paradigm are poorly understood.
Archaeological and 14 C evidence indicates that local hunter-gatherer burial activity decreases during the expansion of the Neolithic in southwestern Iberia from ca. 7450-7250 cal BP onwards, but hunter-gatherer burial practices continued, though less frequently, showing that diverse social structures and worldviews coexisted for several generations.
The multi-site approach offers a comprehensive perspective, contextualized within regional environmental and cultural developments. This study provides the chronological boundaries for the burial activity of the last hunter-gatherers in southwestern Iberia based on current archaeological and 14 C evidence. Burial 14 C data suggest a longer chronology for the Late Mesolithic in Portugal than previously suggested, starting around 8500-8300 cal BP and ending ca. 7000 cal BP. This timeframe is based on mortuary activity of Late Mesolithic tradition. The focus on the human burials allows the establishment of chronologies based on strong indicators of human activity, however, other uses of the sites were not considered in the analysis and require further investigation.