We used the collection of modern shells of C. edule and S. plana that were heated in a muffle at controlled temperatures by Aldeias et al. (2019). The shells were inserted into ceramic crucibles and heated at increasing temperatures from 100 to 900°C, for 5 and 20 m heating durations. Mineralogical analyses reported by Aldeias et al. (2019) were done by hand grinding a portion of the heated shells and obtaining infrared transmission and ATR spectra on powdered shell material. The collected results showed that calcitic phases start to appear at slightly different temperatures in each species: 200°C in C. edule and only at 350°C in S. plana. They also observed that the complete transformation of the original biogenic aragonite to calcite occurred at slightly different temperatures for each species: at 500°C for samples of C. edule and at 400°C for S. Plana (Aldeias et al., 2019).

In the present study, we further expand on this previous work by using microFTIR on thin sections produced from the same shells of C. edule and S. plana used by Aldeias et al. (2019) heated for 5 min at the threshold of the transformation interval previously observed, that is, at 250, 350, 400, and 450°C (Figure 2). Since the carbonate phase conversion in C. edule was shown to start at lower temperature than S. plana (Aldeias et al., 2019), two specimens of C. edule heated at 200°C were also used, though one of them was heated for 20 min, a difference that Aldeias et al. (2019) observed to have no effect on the FTIR-ATR results. Two specimens of each shell species were heated to enable double tests at each temperature (repetition one and repetition two in Figures 2, 3), resulting in 18 individual experimental shell samples used in this study (10 samples of C. edule and eight samples of S. plana) (Figures 2, 3). The experimental shells were impregnated with a polyester resin, cut along their growth axis, and produced into 2.7 × 4.6-cm thin sections by Spectrum Petrographics (Vancouver, United States).

To assess the context of heated components in archaeological deposits, we investigate two Mesolithic shell midden sites from Portugal (Figure 1): Cabeço da Amoreira and Poças de São Bento. Both sites are among the largest shell middens in the Muge and the Sado valleys, respectively. These are two major paleoestuaries with brackish conditions formed during the Early Holocene sea-level rise (van der Schriek et al., 2007; Vis et al., 2008; Costa et al., 2019). The beginning of Mesolithic occupations of both the Muge and Sado valleys is characterized by the onset of human burials dated to 8,500–8,000 cal BP, in simple hollows opened in the sandy substrates of both sites (Bicho et al., 2010; Diniz and Arias, 2012; Bicho et al., 2013; Peyroteo-Stjerna, 2020; Arias et al., 2021). The following cumulative anthropogenic accretion of shelly deposits (and further human and animal inhumations) in several sites in the estuaries’ margins spanned until c. 7,400 cal BP, when the first Neolithic farmers were populating the surrounding regions (Bicho et al., 2017; Diniz et al., 2021). In the case of Cabeço da Amoreira and Poças de São Bento, pit features and putative post-holes at times cross-cut the complex interbedded shelly deposits.

Previous micromorphological studies have focused on reconstructing the depositional histories and the range of activities leading to the formation of these shell middens of Cabeço da Amoreira (Aldeias and Bicho, 2016) and at Poças de São Bento (Duarte et al., 2019). These studies stressed human-driven accumulations as the main agents of deposition, with the identification of three main types of deposits (sedimentary microfacies): 1) deposits in primary position, related to single events of direct shell tossing, 2) reworked shelly sediments resulting from mass reallocation of previously discarded debris, and 3) active layers related to more stable occupation surfaces. Table 2 describes the characteristics of the microfacies based on the micromorphological studies of those sites, alongside the indication of the thin sections that are used in the current paper.

The experimentally heated shells can be directly compared to these archaeological sites since C. edule and S. plana are the predominant species at both sites. Micromorphological samples were collected at the two sites through the removal of undisturbed sediment blocks from the excavation profiles and carefully wrapped either with soft paper or with pre-plastered bandages. The blocks were then dried for several days and impregnated with a polyester resin mix. For Cabeço da Amoreira, large 5,7x13,5 cm thin sections were produced by the Servei de Micromofologia i Anàlisi d’Imatges at the University of Lleida (Spain). For Poças de São Bento, 5x7 cm thin sections were produced at the Departamento de Ciencia y Engenieria del Terreno y de los Materiales of the University of Cantabria (Santander, Spain). Thin sections from Cabeço da Amoreira are labelled CAM followed by an unique number identifier, whereas thin sections from Poças de São Bento bear the label PSB, also followed by an unique number identifier (e.g., CAM112 or PSB107). The microscopic observations were carried out under petrographic microscopes with plane- and cross-polarized light at magnifications ranging from 20x to 400x in plane polarized light (PPL) and cross-polarized light (XPL).

MicroFTIR spectra were collected directly on the thin sections (of both experimental shells and archaeological micromorphology samples) using a Thermo Scientific Nicolet iN10 MX microscope. Spectra were collected in attenuated total reflectance (ATR) mode using a germanium crystal tip at a 4 cm⁻¹ resolution in 256 and 64 scans, for experimental shells and archaeological ones, respectively, both within the 675–4,000 cm⁻¹ range, and were compared with available libraries. The germanium tip has 150 x 150 µm diameter, though the measured area can be less than 150 µm depending on the degree of contact between the tip and the thin sections.

In the 18 samples of experimentally heated shells, a standard number of 40 individual measurements were collected in each shell (39 in one of them–see Figure 3), equally distributed in the shells’ thickness (Figure 4), resulting in 1,438 individual measurements (see Supplementary Material). In the archaeological thin sections, several mapping areas of 1 cm² were selected throughout the slides’ surface (Figure 5) to analyze the shells that were visibly specimens of S. plana and C. edule. Depending on the preservation degree and size of the shell fragments, a variable number of measurements were placed in each shell within each map (Figure 5), in the most similar way possible as in the experimental shells, covering the specimens’ thickness along their section. For archaeological shells in six thin sections, a total of 9,038 individual measurements were collected (see Supplementary Material).

The microFTIR mapping on the oven-heated shells yielded the spectra correspondent to the CaCO3 polymorphs of biogenic aragonite and calcite (Figure 4). Biogenic aragonite spectra present the same peak position for shells heated above 200°C as reported by Aldeias et al. (2019) in ATR: the ν ₃ asymmetric stretch at 1,445 cm⁻¹, the ν ₁ symmetric stretch at 1,082 cm⁻¹, the ν ₄ peak at 856 cm⁻¹, and the ν ₂ doublet at 712 and 700 cm⁻¹. Calcite is represented by the ν ₃ peak at 1,397 cm⁻¹, the ν ₄ peak at 871 cm⁻¹, and the ν ₂ peak at 711 cm⁻¹, in accordance with reference data (Weiner, 2010). A third type of spectrum exhibits mixture of both aragonite and calcite polymorphs in the specimens heated at the transitional temperature phases. In the mixed spectra, the peaks of both polymorphs in the same band overlap, usually with one appearing as a shoulder (Figure 4), depending on the relative proportions, as described elsewhere (Loftus et al., 2015; Toffolo et al., 2019).

We observed that the occurrence of all three types of spectra occur in a same shell during the mineral transformation temperature interval (Figure 3). The microFTIR mapping revealed a slightly different pattern between both species in the aragonite conversion to calcite. In S. plana the total conversion is faster since the shells remain mostly aragonitic at 250 and 350°C. At 400°C nearly all measurements provided calcite spectra, with only a few mixed ones, and exclusively calcite spectra at 450°C. In the case of C. edule, the conversion is more protracted, with mixed spectra already at 250°C, co-existence of aragonite and calcite spectra in similar proportions at 350 and 400°C, and only at 450°C all measured points correspond to calcite. These results generally match the trends identified by FTIR-ATR previously obtained by Aldeias et al. (2019).

Both shell species revealed similar petrographic transformations in terms of darkening and longitudinal fissures development with increasing temperature. At 250°C the shells are practically intact, and at 350°C the inner surface exhibits fissures along the different layers, that are progressively intensified at 400 and 450°C. In general, the effects of these alterations are more intense in S. plana than in C. edule. In Figure 2, it can be observed that at 450°C, C. edule exhibits only slightly darkened patches, following the carbonate growth lines, whereas S. plana is homogenously darkened, although the growth lines are still visible.

In CAM112 the areas mapped with micro-FTIR reveal that most shells are calcitic, except at the bottom, where they remain aragonitic or with a mixture of both polymorphs. This shift is progressive, and it is observed also within the mapping areas in the bottom of the thin section, where the totally calcitic shells are above, aragonitic shells are in the bottom, while mixed spectra occur in greater amounts the transitional middle area (Figure 5).

In CAM109, the bottom layer of single tossed shells reveals a totally aragonitic mineralogy in all mapping areas. In the overlying layer of reworked shells, the shells are predominantly aragonitic, except for rare random shell fragments that yielded mixed spectra and calcitic mineralogy, especially towards the top of the deposit. The microFTIR mapping in the active layer at the top of this thin section revealed also mainly aragonitic shells. Overall, in CAM109 the shells are not thermo-altered, being only the layer of reworked sediments that includes few thermo-altered shells (Figure 6).

In contrast to CAM109, in the thin section CAM206, the mapped areas in the active layer revealed completely calcitic shells (Figure 7). In the reworked deposits both under- and overlying the active layer, the shells in the mapped areas exhibit similar proportions of all mineralogical spectra (Figure 7). Thin section CAM210, a sample of reworked shelly sediments, revealed mostly aragonitic shells with rare calcitic fragments (Figure 8).

The sample PSB107 mapping exhibits mostly calcitic shells in the active layer and a mix of all mineralogical spectra in the underlying reworked deposit, though aragonitic ones predominate (Figure 9).

All of the thin sections produced from sample PSB913 provided a mainly aragonitic mineralogy of the shells in all represented layers (Figure 10). Rare fragments of calcitic shells were documented in the direct shell tossing deposit as well as in the reworked shelly sediments deposit (Figure 10).

As it can be observed in Table 3, these results reveal very diverse situations for each type of sedimentary microfacies identified micromorphologically, with several meaningful implications for the circumstances in which the thermo-alteration of shells and their discarding are related in the shell midden formation.

Like other bivalves, S. plana and C. edule are formed by two and three layers, respectively, composed of aragonite crystals that exhibit different crystalline microstructures and grow by precipitating incremental lines controlled by environmental conditions during the mollusk life (tides, temperature, stress) (Schöne, 2008). During shell growth, organic matter is also secreted, filling the space between the aragonite crystals and can represent 0.05–5% of weight (Weiner, 2010).

Milano et al. (2016) demonstrated a progressive microstructure expansion with increasing temperatures in C. edule, associated with the organic matrix loss. Heating is one of the possible causes for disappearance of mollusk shell’s organic matrix pointed out in the literature, which has been recorded at temperatures >300°C and a total replacement of aragonite to calcite at around 500°C (Milano et al., 2016; Milano and Nehrke, 2018). Our results overall agree with previous studies but show that the aragonite to calcite conversion starts at slightly different temperatures in the different species (Figure 3), namely at 250°C for C. edule and 350/400°C for S. plana. However, the temperature of complete mineralogical recrystallization of the shells was recorded at 450°C for both S. plana and C. edule.

Biogenic aragonite is known to reprecipitate as calcite in chemically unstable archaeological contexts (Weiner, 2010; Toffolo, 2021), but not always through recrystallization as it occurs upon heating. Meteoric water induces dissolution of biogenic aragonite and subsequent neoformation of calcite, which means a progressive loss of shell material, eventually until complete disappearance of the shell (Toffolo, 2021). This phenomenon is observed at the Mesolithic shell middens analyzed here by the existence of pendants composed of calcite spar in the underside of components (shells and pebbles) which most probable source are shells in upper layers that were dissolved (Aldeias and Bicho, 2016; Duarte et al., 2019). Previous studies on the transformation of biogenic aragonite into calcite suggest that diagenetic recrystallization observed in the geologic record is likely to be a much slower process than the timespan occurred at the Mesolithic archaeological contexts under study (Curtis and Krinsley, 1965; Brand, 1989; Ritter et al., 2017; Peacock et al., 2020; Toffolo, 2021). The presence of aragonitic shells in the older layers of the studied shell middens further indicates that the alteration from biogenic aragonite to calcite is not time dependent at these sites.

Another cause for aragonite-calcite recrystallization in shells is biological post-depositional activity induced by microorganisms that feed on the shells’ carbonate through micro-boring (James and Jones, 2015). These originate secretion of micritic calcite within micro-tunnels that, with time, replace entirely the skeletal clast microstructure. We can rule out this agent in our archaeological shells, since no micro-boring is observed in thin section, as shells preserve larger crystals in their different layers. Biologically induced calcite at both sites was observed too, in the form of alveolar septal and needle fabric cements hardening some layers, usually at the bottom of the shell mounds, where shells are still well preserved (Aldeias and Bicho, 2016; Duarte et al., 2019).

In our specimens experimentally heated between 250 and 450°C, microFTIR yielded aragonite, calcite and mixed spectra (Figure 3). The previous FTIR-ATR experiments by Aldeias et al. (2019) showed the existence of a mixed stage during the temperature threshold in powdered shell material. Here, the coexistence of areas where aragonite remains unchanged and others where aragonite was completely converted to calcite in the same shell was revealed by microFTIR mapping. This evidence shows that the aragonite to calcite conversion within a same shell can be heterogenous (Figures 3, 4). The reason for this is unknown, but we can tentatively suggest that it might be related to the different aragonite crystal microstructures that constitute each layer of these bivalves. Different crystalline microstructures had been demonstrated to react differently to temperature increase (Milano et al., 2016). Our experimentally heated shells showed that the complete calcite conversion starts from the inner surface and progressively expands to the outer surface in all shells of both species, as exemplified in Figure 4. This suggests a correlation with the different microstructures in the shell layers but needs future targeted research to be confirmed.

Experimental studies on infrared spectra with aragonite and calcite shows that mixtures of both polymorphs is discernable in the ν ₂ and the ν ₃ peaks of CaCO3 where the relative intensities correspond to the predominance of one over the other (Loftus et al., 2015; Toffolo et al., 2019). We observed spectra with these different relative peak intensities in our experimental heated shells (e.g., Figure 4), indicating variable proportions of both aragonite and calcite phases in the measured point in the shell. Therefore, the aragonite-calcite conversion by heating occurs heterogeneously and at different temperatures in each species. The coexistence of aragonite, mixed aragonite-calcite, and calcite spectra in the same shell, was also observed in the archaeological thin sections (Figure 5), a correlation that strongly suggests a thermal cause for their CaCO3 phase conversion.

Thus, we assume that the recrystallization in the archaeological shell middens analyzed here was not cause by diagenetic processes. This is further supported by the preservation of shell’s aragonite, which strongly indicates preservation of the original ontogenically precipitated aragonite, thus absence of major diagenesis (Weiner, 2010). Furthermore, a diagenetic cause for the conversion is incompatible with the existence of aragonitic and calcitic shells in deposits at different depths throughout the sites’ lateral stratigraphies. All these factors strongly indicate that the aragonite-calcite conversion in the archaeological shells analyzed here is caused by thermal alteration related to fire use by humans at the sites.

Previous micromorphological analysis of these shell middens distinguished several different deposition modes that include: direct single-tossing events of shells, reworked load-dumped shelly deposits, and active layers interpreted as occupational surfaces (see Aldeias and Bicho (2016) and Duarte et al. (2019) for details). In this work, we add data on shell’s in situ mineralogical composition to the microfacies analysis.

Our results show that aragonitic shells predominate in microfacies related to reworked shelly deposits. These reworked deposits result, almost exclusive, from accumulation of S. plana specimens, thus indicating that most shells were not subjected to temperatures above 250°C. Therefore, it is more plausible to assume that most reworked deposits are composed by shells generally not subjected to high temperatures (i.e. above 250-350°C) before their deposition. Only in thin section CAM206 aragonite and calcite occur in similar proportions (Figure 7). In this case, about half of the measured shells were exposed to temperatures above 400°C prior to being reworked. The fact that heated shells are still in close association to each other, but not in their primary position, could be indicative of a low degree of mixing, or a reworking within smaller distances. The occurrence in the same microfacies of other components, namely large charcoals and pebbles (Figure 7), further supports this hypothesis, and is in contrast with the lack of charcoals associated with fully aragonitic shells in other reworked deposits analyzed. The fact that shells in CAM206 deposit are very fragmented can be a consequence of heating, which makes shells more brittle, and not necessarily intense reworking.

The microfacies interpreted as having functioned as surface active layers, thus subjected to physical pressure due to intense occupational activity and trampling, showed different cases of shell mineralogy. The shells within the active layer represented in CAM206 are fully calcitic and supported by a charcoal-rich matrix (Figure 7). In PSB107, the shells in a trampled layer are also fully calcitic (except rare aragonitic ones) (Figure 8), but here no charcoals are present in the matrix. These active layers, unlike reworked deposits, do not show mixing of aragonite and calcite polymorphs, indicating that heating in active layers might be syn-depositional, thus related to the occupation activities responsible for their formation. In turn, CAM109 and PSB913 represent occupational debris not affected by high temperatures, thus fire activity was more limited in its formation, as anticipated by the lack of other charred components.

Micro-stratigraphic shell mineralogy suggests that the trampled shell debris in CAM206 resulted from occupation activities involving fire, strong enough to burn shells, although not in situ, but resulting from short-distance dispersion originated by occupation actions such as trampling or sweeping. In sample PSB107, regardless of the microstratigraphic contact between the active layer and the reworked shelly deposit, the thermal alteration of the shells exhibits a gradual vertical pattern through that contact, better observed in CAM112, as discussed below.

Ethnographic accounts suggest that shellfish cooking does not require high temperatures or long exposures to fire, since the main goal is to open the valves and retrieve the eatable mollusk without burning it (Meehan, 1975; Waselkov, 1987). These sources describe that most techniques used for cooking shellfish imply roasting by simply exposing the shellfish to hot embers. The roasting experiments of Aldeias et al. (2019) basically generate aragonitic shells that experience temperatures always below 200°C, which agrees with other experiments (Müller et al., 2017). However, Aldeias et al. (2019) noticed some exceptions in the case of roasting with fire above or directly on embers. With these cooking methods, localized hot spots may entail the exposure of individual shells to temperatures higher than 200°C, producing spatially concentrated conversions to calcite. Still, most shells cooked this way remain aragonitic. Therefore, both ethnographic and experimental data point to low temperature exposure of shellfish, even when using roasting techniques. Thus, archaeological specimens cooked this way should retain their aragonitic structure. Archaeological assemblages of shells exposed to temperatures above 450°C, showing a complete conversion to calcite, are difficult to explain as resulting from cooking and require other explanations. In turn, this raises the question of how can we infer cooking shellfish in archaeological sites?

Contextual data from micromorphology helps in identifying deposits in primary position and is thus optimal to look for traces of activities that formed them. In our studied thin sections, examples of deposits in primary position consist in those formed by direct single tossing of shells and likely relate to the immediate discard of the shell right after the consumption of the mollusk. The mineralogy of tossing deposits analyzed by microFTIR mapping present twofold situations. In samples CAM109 and PSB913, shells within direct tossing deposits are fully aragonitic, with very residual calcitic ones in the latter (Figure 10). The bulk of these deposits are composed of S. plana specimens, with the retention of an aragonitic mineralogy clearly indicating that, when tossed away, the shells were probably not heated to temperatures higher than 350/400°C and definitely not above 450°C.

The deposit in sample CAM112 showed a completely different situation. Here, the mineralogical analysis revealed a gradient, with a progressive conversion of aragonitic to calcitic shells as we move upwards in the deposit (Figure 5). This suggests that the shells in the upper part of CAM112 were affected by homogenous heating at higher temperatures, that progressively decreased downwards. As this sample is composed almost entirely by S. plana, the palaeotemperature experienced by the shells in the deposit can be more precisely pinpointed (Figure 5). The mixed spectra of the lowest shells shows exposures to temperatures mainly below 350°C, with some shells heated to above 400°C. Above this area, all shells are calcitic (that is, heated to more than 450°C). This vertical gradient strongly suggests that the aragonite-calcite conversion occurred in situ. Since the integrity of the deposit has not changed, this evidence is better understood not as the result of cooking of that specific shellfish but interpreted as the effect of a later fire placed on top of the shell tossing deposit. Therefore, the sequence of events would have been an action of discarding shellfish followed by an unrelated fire activity in this locale. No significant charcoals are observed in CAM112, which supports the hypothesis of this deposit having functioned as a combustion substrate. A similar interpretation of burning–and not cooking–can be applied to the gradient conversion observed in PSB107 (Figure 9), where the burned shells were subsequently affected by trampling, causing structural compaction, and crushing.

Our expectations for prehistoric shellfish roasting in the archaeological record involve the presence of mainly aragonitic shells alongside combustion remains, such as abundant charcoals and ashes, which may not preserve the typical microstratigraphy of in situ fires (Table 1). A set of evidence such as this was identified in superimposed deposits represented in sample PSB913 (Figure 10). Spatially, this feature resembled an intact combustion feature, however, after micromorphological analysis, the components here were interpreted as related to direct single tossing event of abundant charcoals and well-preserved shells (Figure 10). The microFTIR mapping revealed fully aragonitic shells in these deposits, with residual exceptions. This led us to consider this context a result of immediate discard of debris directly from roasting activities. The occurrence of few shells that have experienced exceptionally higher temperatures due to localized hotspots occurring in roasting features confirms the expectation formulated in the experiments by Aldeias et al. (2016).

In our approach, micromorphology coupled with microFTIR mapping allowed us to overcome two issues raised by previous works on shellfish cooking. The fact that aragonite-calcite conversion occurs above the burning threshold makes mineralogy by itself not suited to detect pre-depositional heating (Müller et al., 2017). Our investigation shows that, with contextual control and spatial and micro-stratigraphic precision, mineralogical alterations can be useful not only to infer pre-depositional heating, but also for distinguish it from in situ heating.

Staudigel et al. (2019) applied clumped isotopes to specifically distinguish boiling from roasting. They found that a “nonuniform distribution of ∆47 values between shell midden constituents suggests a nonuniform thermal history for shell midden constituents” (Staudigel et al., p. 4), in agreement with data obtained by Müller et al. (2017). Both studies conclude that this may be an effect of the technique used for cooking that consisted either in using only one of the valves or heating the entire bivalve from below, causing an uneven heating of valves, being the lower one significantly more affected than the top. Our results concur that boiling was unlikely and that roasting might have been a preferential cooking technique in the absence of hot stones or organic containers. This is suggested by the existence of some shells heated above 450°C, hence fully calcitic, among unheated shells and charcoals in the remains of interpreted roasting activity (Figure 10). With boiling, one would expect no mineralogical change at all in shells mineralogy (Milano et al., 2016; Milano et al., 2018). Staudigel et al. (2019) raise several effects specific of shell midden formation that hamper interpretation of pre-depositional heating, such as the discarding of uncooked shells, different thermal histories for different shells, or cooking techniques where heating is unevenly distributed. Our approach, by including micromorphological analysis that allow to study the thermal history of shells in their undisturbed archaeological context, considerably overcomes these limitations. We advocate for future integration of the contextual data from sedimentary microfacies to discern cooking from burning in the archaeological record.