Neandertal dietary behavior has been explored previously (El Zaatari et al., 2011, El Zaatari et al., 2016, Estalrrich et al., 2017, Fiorenza et al., 2011, Fiorenza, 2015, Fiorenza et al., 2015, Karriger et al., 2016, Krueger et al., 2017, Pérez-Pérez et al., 2003 and Williams et al., 2018) and non-dietary behaviors related to the use of “teeth-as-tools” have been documented through the analysis of anterior dental wear (Estalrrich and Rosas, 2015, Krueger et al., 2017, Lozano et al., 2017 and de Lumley, 1973). Analyses of occlusal microwear textures (El Zaatari et al., 2011 and El Zaatari et al., 2016) and macrowear (Fiorenza et al., 2011) indicate that the diet of Neandertals from cold steppe habitats falls within the range of Holocene hunter-gatherers who consumed a large amount of meat. However, the dental calculus of Spy I suggests starchy plant foods were subjected to heat treatment before being consumed, including the remains of underground storage organs and grass seeds (Henry et al., 2011). The inference that plant foods were regularly consumed by the Neandertals of Spy cave was supported by nitrogen isotope signals (Naito et al., 2016 and Naito et al., 2018). Meanwhile aDNA evidence from dental calculus intimates that Spy II consumed more meat than plants compared to individuals from El Sidrón in which only plants were detected (Weyrich et al., 2017), and helps corroborate evidence of meat consumption previously inferred from carbon isotopes (Richards and Trinkaus, 2009), particularly in northern Europe (Bocherens et al., 2001 and Wißing et al., 2016).

Carbon and nitrogen isotopes have been recovered from Spy I and II as well as five individuals from Goyet cave and these have been compared to the cave faunal remains from Spy, Goyet and Scladina Mousterian levels (MIS 3). Mammoth was an important component of the diet in every individual (Wißing et al., 2016), and the Neandertals uniformly fell within the carnivore guild with respect to carbon isotope values. Noteworthy however was that Spy I and Spy II were less enriched than the individuals from Goyet. Spy I had a relatively low contribution of mammoth meat (30%) whereas Spy II and the Goyet individuals had higher values (Wißing et al., 2016). Other resources consumed by Spy I included rhinoceros, reindeer and bovids (Bocherens et al., 2001 and Wißing et al., 2016).

While carbon isotopes and stone tools provide a signal of meat consumption, nitrogen isotopes can infer the degree to which the diet comprised plant proteins (Naito et al., 2016 and Naito et al., 2018), which at Spy cave was apparently quite elevated. Direct evidence of underground storage organ tissue and seeds preserved in the dental calculus of Spy I contradicts a diet devoid of plants (Henry et al., 2011). These different dietary reconstructions suggest a reexamination of Spy I is indeed warranted.

Neandertals from the terminus of MIS 3 may have differed in dietary proclivities from other chronological periods and ecogeographic zones, and the cold steppe habitat coupled with extreme frigid temperatures and aridity may have limited the availability of plant resources. To further expand the reconstruction of the diet of this late MIS 3 Neandertal adult, Spy I is compared to climate groupings that correspond to two reconstructed ecological zones. These include emergent coastal Mediterranean plains, represented by five individuals from Hortus cave, and a continental grouping typified during MIS 3 as comprising mixed conifer and deciduous forests (Fiorenza et al., 2015). This continental grouping is inclusive of La Quina 5 (Southwest France), Malarnaud (French Pyrennees), Kůlna 1 and Svédův stůl (Moravia) as well as 19 individuals from Krapina and five individuals from Vindija (Croatia). The Mediterranean region was comparatively warmer and dryer than continental inland habitats, which experienced severe winter conditions resembling those in the northern European shrub tundra (Fiorenza et al., 2015). These are used as broad paleoenvironmental proxies to examine whether microwear texture has a relationship with habitat, which would imply that Spy I would be more similar to continental than Mediterranean zones. Ecogeography has been invoked previously to account for variation in Neandertal diets (Bar-Yosef, 2004, El Zaatari et al., 2011, Estalrrich et al., 2017, Fiorenza et al., 2011, Hardy, 2010 and Krueger et al., 2017). Neandertals from northern Europe are known for a diet heavily reliant on meat resources (El Zaatari et al., 2011, Fiorenza et al., 2011, Fiorenza et al., 2015, Richards and Trinkaus, 2009, Weyrich et al., 2017 and Wißing et al., 2016). Since people with a smaller proportion of plants in their diets have lower textural complexity (Asfc) values (e.g., Holocene pastoralists; Schmidt et al., 2016), we expect Spy I to have reduced Asfc compared to Neandertals from comparatively warmer regions with a greater abundance of plant resources (e.g., El Zaatari et al., 2016). Since anisotropy (epLsar) separates Hortus young adults with elevated values from other ages (juvenile and 50 + years) (Williams et al., 2018), it is expected that the Spy I adult will exhibit high anisotropy (epLsar), either from the mastication of specific dietary items or lack of heterogeneity in jaw movements.

In addition, 12 Holocene groups from Eurasia and North America (Karriger et al., 2016) inclusive of 173 individuals are included to infer the diet of Spy I using DMTA. We anticipate that the textural values of Spy I will be more similar to the published means for complexity (Asfc) and anisotropy (epLsar) for Holocene high-meat consuming hunter-gatherers and dissimilar to agriculturalists and pastoralists (Karriger et al., 2016).

The right M² of Spy I is examined. Spy I is aged to be about 35 years with full eruption and substantial attrition of the molars, which are worn to a single functional plane (Twiesselmann, 1971 and Williams, 2013). The comparative sample includes Hortus IV, Hortus V, Hortus VI, Hortus VIII and Hortus XI, all of which are from Hortus cave located about 30 km from the Mediterranean coast of France and are dated to MIS 3 (Lebègue et al., 2010, de Lumley, 1972, de Lumley and Licht, 1972, de Lumley, 1973 and Pillard, 1972) (Table 1). We also include La Quina 5 from Level 3 of Station Amont of La Quina cave in southwestern France, dated to 47–43 kya and falling within MIS 3 (Debénath and Jelinek, 1998, Discamps and Royer, 2017 and Petite-Marie et al., 1971). An additional site considered is Malarnaud from the French Pyrenees, provisionally dated to MIS 5 (Petite-Marie et al., 1971). The Moravian Neandertal sites of Kůlna and Švédův stůl (Ochoz 1) of the Czech Republic, both dated to MIS 3 (Krueger et al., 2017) (Table 1) are included, as well as a sample of isolated molars from Krapina (n = 19) dated to MIS 5e (Rink et al., 1995) and Vindija (n = 5) from MIS 3 (Wild et al., 2001), both of which are from Croatia. The Hortus (n = 5), Krapina (n = 19) and Vindija (n = 5) assemblages, together with isolated sites (n = 4), represent much of the known temporal and ecogeographic distribution of European Neandertals (Table 1).

A dental mold of the Spy I right M² was made with Coltène President light body dental silicone and loaned to FLW from the “Institut royal des sciences naturelles de Belgique” by PS. Additional dental molds were created using polyvinylsiloxane (Coltène-Whaledent) at the “Centre européen de recherches préhistoriques de Tautavel” and the “Musée de l’Homme.”. The resulting casts were created from epoxy resin and hardener (Buehler). Epoxy (Epo-Tek 301) dental casts of Kůlna and Švédův stůl from Erik Trinkaus were also included.

We studied Phase II wear facets on or adjacent to the protocone (e.g., Krueger et al., 2008). Emphasis was placed on facet 9 (Kay, 1981). We did not study Phase I facets, which bear more shear-related features. Instead, observations were restricted to the planar surfaces of Phase II facets, which indicate more of the crushing and grinding action of molars during occlusion (Kay and Hiiemae, 1974). We collected microwear data at 100× magnification using a white-light confocal profiler (Sensofar Plμ, Solarius Development Inc.) housed at the University of Indianapolis (Schmidt et al., 2019). Only surfaces devoid of irregularities from postmortem processes, such as deposition, excavation or casting, were included. Four scans contributed to a total sampling area of 276 × 204 μm which was reduced to 242 × 182 μm after they were digitally stitched together; the data collection process was standardized for all individuals. Surface morphology was visualized to validate whether the scanning process produced a surface showing dental microwear free of postmortem taphonomy. Specifically, we created 2D photosimulations and then 3D representations to search for irregularities (Fig. 2).

Textural properties of the enamel surface were described using two texture variables which were extracted from the point cloud using scale-sensitive fractal analysis (Scott et al., 2006). These included complexity (Asfc), or area-scale fractal complexity, which compares the coarseness of a surface at different scales of observation, from 7200 μm² to 0.02 μm². Complex surfaces result from the mastication of mechanically resistant food particles which tend to puncture the enamel matrix (Scott et al., 2006, Scott et al., 2012, Ungar, 2015 and Ungar et al., 2012). Tough foods are expected to require more homogeneous masticatory regimes than hard-brittle foods, and will tend to leave striated occlusal enamel surface damage which can be approximated using a measure of anisotropy (epLsar) or the “exact proportion of Length-scale anisotropy of relief” (Scott et al., 2006 and Scott et al., 2012).

A bivariate plot contrasts complexity (Asfc) and anisotropy (epLsar) values for Spy I, four isolated Neandertal sites and five individuals from Hortus cave coupled with a 100% convex hull demarcating the outermost values for this site. This comparison demonstrates the proximity of Spy I to the anisotropy (epLsar) values for young adults preserved at Hortus. The Spy I complexity (Asfc) value is compared to that of Neandertals from continental and Mediterranean zones, including the paleoecologically distinct phases of Hortus cave.

Spy I is additionally compared to the means and standard deviations for complexity (Asfc) and anisotropy (epLsar) of three Neandertal sites (Hortus, Krapina and Vindija) and 12 Holocene groups (n = 173) including Natufian hunter/gatherers from Chiu et al. (2012), North-Central Indiana Middle Woodland, Eastern Indiana Middle Woodland and Indiana Middle to Late Archaic hunter/gatherers from Frazer (2011) as well as Kentucky Archaic hunter/gatherers from Karriger et al. (2016) (Table 2). We also include agriculturalists such as Neolithic farmers from Israel (Chiu et al., 2012), Early Bronze Age England, Late Bronze Age England, Iron Age England, Iron Age Nepal (Mebrak) and Bronze and Iron Age Greece (Karriger et al., 2016). In addition, we consider the textural values for pastoralists of Mongol Xiongnu (Xiongnu Period Mongolia) and Bronze Age/Iron Age Mongolia from Schmidt et al. (2016) (Table 2). Although a variety of foraging proclivities are known from historical and archaeological hunter-gatherers, this is only a minor fraction of the variation that once existed (Kenneth, 2005). For these reasons, we acknowledge the high possibility of equifinality in all of the Holocene microwear signatures in comparison to the Middle Paleolithic samples. Microwear signals are directly linked to the mechanical properties of the foods rather than the food resources per se, and highly dependent on non-food abrasives some of which may be habitat specific. In other words, a similar microwear signature does not necessarily imply a similar diet, but it implies one with similar mechanical properties.

Since dental microwear texture data for Spy I were reported by El Zaatari et al. (2016) we compare them to the values obtained in this study. La Quina 5 is added to the comparison as both studies include this individual, albeit from molar antimeres. To provide context, Neandertal samples from open, mixed and wooded habitats are included (El Zaatari et al., 2016).

Spy I has a more elevated complexity (Asfc) value (2.22) than sampled from Neandertals from a Mediterranean habitat and is closest to Kůlna 1 from the continental zone (Fig. 3). The elevated anisotropy (epLsar) of Spy I (0.0032) is similar to those of young adults from Hortus cave where a division by age separates young adults from juvenile and older individuals (Williams et al., 2018) (Fig. 3).

When Spy I is compared to Neandertals (n = 33) as well as 12 Holocene groups (n = 173) for complexity (Asfc), this adult is more than one standard deviation above the mean value for the Hortus, Krapina and Vindija assemblages, and all of the Holocene humans with the exception of some of the early farmers from the Neolithic period of Israel (Fig. 4). The complexity (Asfc) value for Spy I is most similar to the means for prehistoric temperate hunter-gatherers of the Americas. For anisotropy (epLsar), Spy I falls within the Hortus, Krapina and Vindija assemblages, and it is higher than the means for prehistoric temperate hunter-gatherers of the Americas (Fig. 5). With respect to Holocene groups, Spy I is unlike the means for farmers and pastoralists with more fibrous diets (Fig. 5).

When the textural values for Spy I are compared to those from El Zaatari et al. (2016), nearly identical positions are apparent for anisotropy (Fig. 6). For complexity, however, our Spy I value is higher. The El Zaatari et al. (2016) value hints at plant consumption and the data from our study support that interpretation. The overlapping values of the wooded and mixed Neandertal habitats are inclusive of the values for complexity obtained for Spy I. For comparison, the values for La Quina 5 are more similar for complexity than for anisotropy between the studies yet both group this specimen with open-habitat Neandertals (Fig. 6).

With respect to the Hortus assemblage, Spy I may have consumed or processed more fibrous plants than individuals from the cold/arid paleoclimate of Sub-Phase Vb, all of whom may have had a diet richer in meat. Spy I is closer to Neandertals of the continental group than to the Mediterranean zone.

It is possible to suggest that Spy I replicates the division of diet with respect to age existing within the Hortus assemblage, in which anisotropy (epLsar) is low in juvenile Hortus III and older Hortus XI and highly elevated only in the four young adults from different paleoecological phases of Hortus cave (Williams et al., 2018). In a study of dental microwear texture analysis of the Neandertals from El Sidròn in northern Spain, most of the adults show elevated anisotropy (epLsar) whereas El Sidròn Juvenile 1 exhibits one of the lowest values of the assemblage (Estalrrich et al., 2017). Perhaps the lack of heterogeneous masticatory movements of the jaws in Neandertal adults is evidence of considerable effort to chew tough, fibrous foods, a pattern noted to occur as well at Krapina (Karriger et al., 2016).

Although the open steppe was a harsh climate, isolated stands of oak and pine existed which contained a variety of plant resources perhaps exploited by Spy I. There are many edible plants on steppe landscapes. For example, geophytes are common as are certain open woodland trees such as terebinth and oak (Hillman, 1996, Power and Williams, 2018 and Primavera and Fiorentino, 2013). Oak acorns, hazelnut and low-lying chenopods may have been available at least seasonally on the open steppe during the late Pleistocene, allowing for a considerable amount of plant food to be incorporated into Middle Paleolithic diets (Power et al., 2018). Although Mediterranean habitats are more productive and exhibit greater tree cover, much of the biomass can be found in inedible tree trunks. In comparison, open-steppe habitats present lower productivity; however, these ecologies may exhibit a greater diversity and abundance of edible plants compared to warmer climate regimes (Power et al., 2018). Although warmer climates may have reduced levels of preservation compared to colder ones, Neandertals in both the northern and southern regions of western Eurasia appear to have utilized plant food resources in comparable frequencies (Power et al., 2018).

For example, at Spy cave, the roots of water lilies or other underground storage organs appear to have been exploited (Henry et al., 2011). Water lilies in the diet of Spy I could represent a significant part of the carbohydrates intake. This is not visible in stable isotopes given the fact that water lily roots are very poor in proteins (and DNA) and to a selective fractioning of animal proteins by human metabolism. Nevertheless, the work of Naito et al., 2016 and Naito et al., 2018 demonstrates that the stable isotope signatures of Spy I and II are also compatible with a protein intake of 10% of plant proteins, which means 80% of the diet. Furthermore, starchy grains in Neandertal dental calculus have been discovered as early as MIS 7/8 and are increasingly prevalent after 50,000 years BP, including those from Spy cave, suggesting a regular dietary intake of plant foods (Hardy and Moncel, 2011 and Hardy et al., 2012). Neandertal consumption of these starchy plant foods does not contradict the results of isotope analyses, because nitrogen isotopes record only the consumption of meat and protein-rich plant foods (Power et al., 2018).

A study of microwear by Garcia Martin (2000) on the vestibular or buccal molar surface (method of Mollesson) using scanning electron microscopy also found considerable evidence of plant food consumption, grouping Spy II with the Neolithic inhabitants of Belgium which in addition to agricultural products included substantial hunted and gathered resources in the diet (Semal et al., 1999). Spy I differed from the Belgian Neolithic cave burials and had a greater resemblance to medieval monastic cemetery populations such as Coxyde (Garcia Martin, 2000). Our Spy microwear signature is similar to that of El Zaatari et al. (2016), particularly in terms of anisotropy. The values from each study are almost identical. Our complexity value is higher than what El Zaatari et al. (2016) report, but this difference is difficult to explain. The studies used different profilers and interprofiler differences have been reported (Arman et al., 2016), but the system at the University of Indianapolis has been calibrated to the profiler used by El Zaatari at the University of Arkansas (Schmidt et al., 2016). In any event, the two studies are likely telling similar stories in that both values indicate some level of dietary hardness and neither clusters with, for example, soft food eating groups like the Holocene pastoralists. This confirmation of microwear signatures is important because it shows that different researchers can make similar determinations, but they can also detect subtle nuances that may open up opportunities to explore new interpretations of diet.

The elevated complexity (Asfc) of Spy I indicates the consumption of hard particles such as nuts, seeds, shells and exogenous grit given the association between this textural property and hard plant particle consumption (Calandra et al., 2012, DeSantis et al., 2013, Schmidt et al., 2016, Schmidt et al., 2019, Scott et al., 2005, Scott et al., 2006 and Scott et al., 2012). Since meat is a soft food, it is unlikely to damage the surface enamel such that pastoral Holocene populations exhibit much lower values (Schmidt et al., 2016). The elevated complexity (Asfc) of Spy I could derive from the consumption of the hard parts of seasonally available plants, such as seeds, or possibly from underground storage organs (Hardy, 2010), which may have contributed inadvertent grit to the occlusal surface during mastication. Since virtually no food processing technology has been recovered from the site, it is likely that the natural hardness of the foods was not mechanically reduced prior to consumption. A combination of foods, some of which were rather hard and brittle, and other resources that the individual could acquire likely led to the relatively high degree complexity (Asfc) of the enamel surface.

Although the aDNA analysis of dental calculus by Weyrich et al. (2017) suggest Spy II had a meat-based diet, the earlier analysis of plant microfossils contained in the Spy I and II calculus indicates the consumption of starch-rich plants and grass seeds (Henry et al., 2011). Such a finding is consistent with a dental microwear texture analysis of Spy I indicative of the mastication of plant materials on a level similar to that of the Chumash (El Zaatari, 2007, cf. El Zaatari et al., 2011), a mid-Holocene archaeological population from Santa Cruz Island, California who were hunter/gatherer/fishers, exploiting clams, abalone, mussels, fish, and shark as well as migratory birds, otter, sea mammals, duck, quail, bear, deer, acorns, herbs, seeds, nuts, bark, berries, leaves and bulbs (Colten, 1996, Kenneth, 2005, Timbrook, 1986 and Timbrook, 1993). The Spy I Neandertal could have consumed hard seeds, underground storage organs and other plant parts or grit resulting in an elevated complexity (Asfc) value. Yet hard foods can mask meat consumption. Given that the Spy I microwear texture signature indicates both hard and fibrous foods, it is likely that most plant foods were poorly processed if at all, and could have been consumed in their natural state.

Thus, the microwear data from the current study contributes to an important multi-faceted approach that adds to existing studies of the Spy I teeth; specifically it has both verified the anisotropy and clarified the complexity values provided in previous works.