El Mazo cave is located in the town of Andrín, in the municipality of Llanes, in Asturias, Northern Spain (Fig. 1). The rockshelter is approximately 18 m long and 7 m deep. Excavations were undertaken in the inner part of the rockshelter, close to the walls, in 2009, 2010 and 2012. A 2 x 1 m test pit was opened in 2009–2010 in squares V15 and V16 and the resulting north profile was sampled in 2012 in a limited extension (2m in the east-west axis x 0.25m in the north-south axis) of squares X15 and X16 (Gutiérrez-Zugasti and González-Morales 2014; Gutiérrez-Zugasti et al. 2014). The excavation produced several Stratigraphic Units (SUs) composed predominately of shell midden deposits (units 100 to 122). Most of these units were formed by a shell matrix characterized by the predominance of shells over any other remains, and the occurrence of limited amounts of sediment. In contrast, SUs 104, 117 and 122, corresponded to fire structures (hearths) with high densities of charcoal, ashes and burnt shells, which suggest these fire structures were related to shell processing (Fig. 2). Some units were cemented as a consequence of calcium carbonate precipitation in the midden, as was the case of SUs 102 and 106.

The units analysed in this study, were located at the base of the shell midden (Fig. 2). Unit 114 was a discrete shell lense (~2 cm thick) confined to the westernmost part of square X16 and characterised by loose, dark brown sediment, with some shells and considerable numbers of mammal bones, while unit 115 was a shell matrix (~5–15 cm thick) extended along the square X16 and composed by scarce fine grain and loose, brown sediment, and abundant shells (Gutiérrez-Zugasti and González-Morales 2014). Both units have been radiocarbon dated to the early Mesolithic (Table 1). The chronology of these units (including unit 108) suggests the recurrent use of the cave c. 9 kys cal BP and rapid formation of the shell midden units.

Environmental change between the Pleistocene and the Holocene has been recorded by global and regional proxies (North Greenland Ice Core Project Members 2004; Stoll et al. 2013). The onset of the Holocene (~11.6 – 8.7 kys cal BP) in northern Iberia has been characterised as a humid and warm period (Moreno et al. 2011; Rofes et al. 2015) with a considerable expansion of woodland (Muñoz Sobrino et al. 2005). Holocene warming conditions have also been identified in the region through the study of changes in shell assemblages. Gradual substitutions of colder species such as Littorina littorea (Linné 1758) with warmer species, such as Phorcus lineatus (da Costa 1778) and Patella depressa (Pennant 1777), have been recorded between the late Pleistocene and the early Holocene. The substitution process started at the end of the Magdalenian, during the Tardiglacial Interstadial, and ended ~10.8 kys cal BP, when warmer species were well established in the region (Gutiérrez-Zugasti and Cuenca-Solana 2014). In addition, isotopic analyses (δ¹⁸O and δ¹³C) of land snails Cepaea nemoralis (Linné 1758) from Late Pleistocene and Holocene sites in northern Iberia (including samples from unit 105 at El Mazo) suggest the existence of more humid conditions in the early Holocene than in the Late Glacial (Yanes et al. 2012).

The current distance of El Mazo from the coastline is around 1km. During the Mesolithic, this distance would have varied due to the rise in sea level. According to recent studies (Leorri et al. 2012), sea level c. 9 kys cal BP would have been located ~25 m below the present sea level. Taking into account this information, the distance to the coastline at that time was calculated using bathymetric information from the BACO application, developed by the Environmental Hydraulics Institute of Cantabria and the Minister of Environment from the Spanish Government (Gutiérrez-Zugasti 2009). Combination between sea level studies and bathymetry showed that the distance from El Mazo to the coastline c. 9 kys cal BP was ~2.5 km.

The composition of the shell assemblage has provided general information on the environmental conditions in the region c. 9 kys ago. Previous research identified a change of taxa between the late Pleistocene (colder conditions) and the early Holocene (warmer conditions). Cold-loving species such as the periwinkle L. littorea and the limpet P. vulgata were the most abundant in shell assemblages belonging to the late Pleistocene, whilst species adapted to warmer conditions such as P. lineatus, P. depressa and Patella ulyssiponensis (Linné 1758) were predominant in Holocene sites (although P. vulgata was also present in significant amounts) (Gutiérrez-Zugasti 2009; Gutiérrez-Zugasti and Cuenca-Solana 2014; Álvarez-Fernández 2011; 2013). A similar pattern is visible today in the Cantabrian coast, with a predominance of warmer species. Therefore, the presence of P. lineatus and the absence of L. littorea in units 114 and 115 suggest that climatic conditions during the formation of these units were typical of the Holocene in the region. The high percentage of P. vulgata within these units, in comparison with P. depressa, can be explained by the formation of these assemblages at the beginning of the Holocene, when the sea surface temperatures were still increasing. Although the amount of land snails is marginal and their presence at El Mazo is not related to human activity (except perhaps in the case of C. nemoralis), these taxa can also provide information about environmental conditions. The assemblage found at El Mazo, and especially the presence of C. nemoralis and Pomatias elegans (Müller 1774), suggests the existence of temperate conditions at the time of the occupation (André 1975). Therefore, species representation (both marine and terrestrial) at El Mazo suggests the existence of climatic and environmental conditions similar to today. This hypothesis is consistent with evidence from isotopic analyses previously undertaken on C. nemoralis recovered from the site (Yanes et al. 2012), which showed that the environmental conditions were more humid than during the late Pleistocene and were similar to the present day. These data are also consistent with other proxies used to reconstruct the environmental conditions during the Mesolithic in the Cantabrian region (mammals, pollen, seeds, charcoal, ostracods, foraminifera, etc.), which have documented a temperature increase from the start of the Holocene (Cuenca-Bescós et al. 2009; Fano 2007; Martínez-García et al. 2015; Rofes et al. 2015).

A comparison between modern shell assemblages from exposed rocky shores (Gutiérrez-Zugasti 2009) and those from El Mazo showed that the same species predominate (mainly limpets and top shells). Therefore, species representation and collection patterns also suggest that, despite the rising sea level until c. 7 kys cal BP, Holocene coastal areas were very similar to current littoral zones in the central area of the Cantabrian region, which is characterised by the occurrence of exposed shores and a limited existence of estuaries.

The species representation shows that the exploitation during the formation of units 114 and 115 at El Mazo was focused on three taxa (the top shell P. lineatus, the limpets of the Patella genus and the sea urchin P. lividus), whilst the rest of the taxa were collected in limited amounts (Table 2). The predominance of limpets and top shells has been previously recorded in other Mesolithic shell midden sites in northern Iberia, located either in the Asturian area, such as Mazaculos II (Gutiérrez-Zugasti and González-Morales 2010), La Llana (Gutiérrez-Zugasti 2009), Poza l’Egua (Arias et al. 2007), El Toral III (Bello-Alonso et al. 2015) and unit 108 from El Mazo (García-Escárzaga et al. 2015), or in other parts of the Cantabrian region, such as La Garma A (Álvarez-Fernández 2013), El Truchiro (Álvarez-Fernández et al. 2013), La Trecha and La Fragua (Gutiérrez-Zugasti 2009). In all of these sites and within units 114 and 115 at El Mazo, limpets and top shells always account for more than 80% of the MNI of the shell midden. However, despite the clearly established pattern in species representation, there are changes in the abundance of the predominant taxa. In general terms, limpets from the Patella genus are the most exploited species, but the amount of P. lineatus in units 114 and 115 at El Mazo is unusually high when compared to other sites in the region, and only at La Llana does the %MNI reach similarly high proportions (Gutiérrez-Zugasti 2009). These changes could be related to differences in resource availability due to: 1) environmental factors (warmer conditions inducing abundance of warm loving species, such as P. lineatus), 2) human pressure on the limpets (reduced stocks of limpets promoting the collection of alternative species), or 3) human decisions on shell selection at different times (collection of P. lineatus is easier as no tools are needed to detach the shells from the rocks, and they usually inhabit more accessible areas of the shorefront).

The presence of sea urchin as one of the most exploited species is not commonly observed archaeologically in the region. Sea urchins have rarely been documented in any considerable quantity at late Pleistocene and early Holocene sites dated within Cantabrian Spain, with the exception of the late Magdalenian and Azilian levels in Santa Catalina (Gutiérrez-Zugasti 2014), the Mesolithic midden in Arenillas (Gutiérrez-Zugasti 2009) and unit 108 at El Mazo (García-Escárzaga et al. 2015). However, %MNI and MNI/dm³ from units 114 and 115 are considerable larger than in any other assemblage previously reported. The limited numbers of echinoderms compared to molluscs has usually been interpreted from a quantitative perspective, implying that they were opportunistic resources contributing to the survival of the group (Gutiérrez-Zugasti 2011c). However, given the pattern of continuous exploitation exhibited by these resources in northern Iberia (Álvarez-Fernández 2011; Gutiérrez-Zugasti 2011c), some scholars have recently stated that they could be interpreted from a qualitative perspective, and so they should be considered as stable resources with a significant social value (Gutiérrez-Zugasti et al. 2016). Nevertheless, recent studies have demonstrated that the interpretation of the role of echinoderms in subsistence strategies of prehistoric populations is clearly biased.

Investigations of sea urchins have not usually relied on remains recovered from mesh screens smaller than 2mm. Recently, two different investigations have studied the material coming from the 1 `mm mesh in order to test the MNI that would have been lost if only 2 and 4 mm mesh were used (Álvarez-Fernández et al. 2010; García-Escárzaga et al. 2015). However, these investigations reached different conclusions. The results obtained by Álvarez-Fernández et al. (2010) suggested that the MNI would not increase by using remains recovered in the 1mm mesh screen. However, the limited sample size used in that study was probably biasing the results. In contrast, results presented by García-Escárzaga et al. (2015) from unit 108 at El Mazo, where the presence of sea urchins was significant, suggested that the MNI loss would have been considerable (31.5%) if remains recovered in the 1 mm mesh screen were not used. This could explain the lower quantity of sea urchin remains found in other previously analysed shell assemblages. Results presented here confirm that the amount of MNI loss would be higher if only the material recovered in the mesh screen of 2–4 mm were used, and this is especially true for unit 115 where 23.3% of the total sea urchins would have been lost (Table 4). Therefore, using 1 mm mesh screen has been demonstrated to be helpful (and necessary) to increase the accuracy of the MNI calculations and not undervalue the role of sea urchins in human diet. In our case, a correct estimation of the amount of sea urchins was crucial to establish that sea urchins probably were much more important for human populations than previously thought.

Very few land snails were present within units 114 and 115. The occurrence of these organisms at archaeological sites can have a natural or anthropogenic origin (Lubell 2004; Fernández-López de Pablo 2011; Gutiérrez-Zugasti 2011d). When edible species such as C. nemoralis appear in small numbers they are usually considered as natural accumulations and this is probably the case at El Mazo.

In terms of shell collection locations, the human populations occupying El Mazo during the early Mesolithic exploited intertidal areas of rocky exposed shores (Fig. 4 and 5). The most commonly exploited species, the limpets P. vulgata and P. depressa inhabit all levels of the intertidal zone, whereas P. ulyssiponensis occupies the lower zones; top shell P. lineatus inhabits the high-medium areas of the intertidal; and the sea urchin P. lividus is usually found within lower intertidal and subtidal locations, although they can also be found in rock pools located in higher zones of the intertidal range (Fechter and Falkner 1993; Gutiérrez-Zugasti 2009). Data from El Mazo show that high-medium areas of the intertidal were exploited (e.g. collection of P. lineatus), but information obtained from P. vulgata indicates heavy exploitation of lower exposed zones, which is consistent with other Mesolithic sites, contrasts with the exploitation patterns observed during the Upper Palaeolithic in the region, when this species was mainly collected in higher and sheltered shores (Gutiérrez-Zugasti 2009; 2010).

Intensification in the use of coastal resources is thought to have occurred in northern Iberia from the late Palaeolithic to the Mesolithic (Gutiérrez-Zugasti 2011b; Straus and Clark 1986). In the case of El Mazo, taphonomy (high shell fragmentation), chronology (units formed in a short period of time) and shell abundance suggest an intensive use of the cave for exploitation of coastal resources. Data relating to shell biometry and collection areas are also needed to confirm the existence of intensification. Thus, decrease in shell size and exploitation of lower and more exposed areas than in the Upper Palaeolithic have been suggested as potential indicators of intensification during the Mesolithic in the region (Gutiérrez-Zugasti 2011b).

Results based on shell sizes from El Mazo showed that P. lineatus, P. vulgata and P. depressa (Table 3) were smaller than top shells or limpets recovered in Magdalenian and Azilian assemblages from the Cantabrian region (Álvarez-Fernández 2013; Álvarez-Fernández et al. 2011; Gutiérrez-Zugasti 2011b; Ortea 1986). However, shell sizes measured from units 108, 114 and 115 are slightly smaller than those from other Mesolithic assemblages of similar chronology, and are more similar to the sizes recorded from late Mesolithic and Neolithic assemblages (Arias et al. 2007; Álvarez-Fernández et al. 2011; Gutiérrez-Zugasti 2011b). Reduced shell size at El Mazo c. 9 kys cal BP could be related to the intensification in the exploitation of marine resources stated for the Mesolithic by previous research (Gutiérrez-Zugasti 2011b; Straus and Clark 1986), but at the same time the presented data do not fit the model of gradual decreasing sizes through time, with smaller sizes at the late Mesolithic and Neolithic. The explanation could be related to collection strategies and resource management, and periods of intense shell collection could have alternated with times of shell stock recovery through the Mesolithic.

Collection areas could also potentially be linked to the coastal resource intensification proposed for the Mesolithic in the region (Gutiérrez-Zugasti 2009; 2010; 2011b; 2011e). Accessing lower and exposed areas can be a complicated task as these areas are continually wave-beaten, and are therefore more dangerous than higher intertidal zones. A trend exists towards the exploitation of these areas (probably due to depletion of shellfish in higher zones of the intertidal) during the Mesolithic, and data from El Mazo comply with this model. The intensification hypothesis is also consistent with the sizeable amount of sea urchins recovered in these stratigraphic units, as they inhabit the lower intertidal and subtidal zones and the procurement of this species is more complex than other coastal resources. However, with the available information, other explanations (e.g., collection in intertidal rock pools) cannot be discarded.

Differences in size distributions of P. lineatus and P. vulgata between unit 114, and units 108 and 115, suggest the existence of two different collection strategies. Whilst normal distributions were recorded in unit 114, distributions from units 108 and 115 were not normally distributed. Data from the normality test plus data from the univariate statistics (e.g. skewness) clearly demonstrated that larger sized shells were collected during the formation of units 108 and 115, whereas all sizes of shells were collected within unit 114, reflecting that more intense exploitation was performed. However, data on P. depressa from units 108 and 115 (normal distribution and slightly negative skewness) suggest intensive collection within both units. Conversely, distributions of P. lividus were not normal, indicating some kind of size selection. Diversity in size selection between units reflect the existence of changes in collection strategies through time, while differences in size selection strategies between species suggest differential treatment, in terms of intensity of exploitation, depending on the species targeted.

Therefore, although information recovered from species representation, collection areas and biometry from the units analysed at El Mazo reflect some evidence of intensification (reduced shell size, collection in lower areas of exposed shores, no size selection in some units and species), the results suggest the existence of a mixture of strategies, with periods of higher and lower level of intensity in the collection of coastal resources.