On the relationship between climate and Neandertal

Quaternary International 436 (2017) 114e128

Contents lists available at ScienceDirect

Quaternary International journal homepage: www.elsevier.com/locate/quaint

On the relationship between climate and Neandertal fire use during the Last Glacial in south-west France Andrew C. Sorensen Faculty of Archaeology, Leiden University, P.O. Box 9515, 2300 RA Leiden, The Netherlands

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 May 2016 Received in revised form 30 September 2016 Accepted 3 October 2016 Available online 16 January 2017

Both environmental and cultural factors dictate how, when and where hunter-gatherers use fire in the landscape, as well as how well evidence for any one fire will preserve in the archaeological record. Variability in the production and preservation of anthropogenic fire traces can potentially skew our perception of fire use in the past. With this in mind, the research presented in this article weighs in on the debate concerning Neandertal fire use and fire making, specifically, the assertion that Neandertals were unable to produce fire for themselves (Sandgathe et al., 2011a, 2011b). This hypothesis is based on the inferred correlation between climatic deterioration, concurrent lowering of lightning-ignited fire frequencies, and reduced signals for fire use in layers presumably deposited during the Lower Pleniglacial IV (Aquitaine (Marine Isotope Stage 4) at the Middle Palaeolithic sites of Roc de Marsal and Pech de l’Aze Basin, southwestern France), the logic being that if Neandertals could produce fire at will, fire use signals would remain largely consistent throughout the deposits despite there being limited access to natural fires in the landscape during this colder period. This review challenges these assertions at multiple scales by looking at regional lightning and fire regime dynamics, comparing the fire signals observed at Roc de IV to those of other sites nearby and around France, and exploring the various Marsal and Pech de l’Aze environmental and cultural factors likely influencing these signals. Ultimately, the data suggests that estimated reductions in lightning frequencies and fire regime during the Lower Pleniglacial (and colder stadial periods, in general) were not adequate to severely limit Neandertal access to natural fire, while possible artefactual evidence for Neandertal fire making challenges the assumption that they were at all reliant on lightning-ignited fire. Moreover, at the nearby Neandertal site of Combe Grenal, the majority of the layers exhibiting evidence of fire use have cold climatic signals, suggesting the fire use trends IV are potentially local expressions of changes in regional observed at Roc de Marsal and Pech de l’Aze site use patterns, possibly brought on by increased reliance on highly mobile, migratory reindeer prey species and reductions in local woodfuel availability during cold periods. Other factors potentially reducing the archaeological visibility of cold climate fire use are discussed. © 2016 Elsevier Ltd and INQUA. All rights reserved.

Keywords: Fire Neandertals Middle Palaeolithic Last Glacial Lightning Fire regime

1. Introduction Fire has been integral to the human toolkit for a long time. Its uses are manifold and include providing heat for warmth, cooking and tool manufacture, light for night time working and protection from predators, smoke to reduce harassment by flying insects, and flames to burn old bedding, manipulate the landscape or aid in hunting (for an overview, see Clark and Harris, 1985). But just when and how fire entered the human technological repertoire is still subject to intense debate (Goudsblom, 1986; James, 1989; de

E-mail address: [email protected]. http://dx.doi.org/10.1016/j.quaint.2016.10.003 1040-6182/© 2016 Elsevier Ltd and INQUA. All rights reserved.

Lumley, 2006; Wrangham, 2009; Roebroeks and Villa, 2011a; ndez Peris et al., 2012; Gowlett and Sandgathe et al., 2011a; Ferna Wrangham, 2013; Stahlschmidt et al., 2015). Human interaction with fire likely occurred in four stages: 1) conceptualizing and understanding fire (Pruetz and LaDuke, 2010; Parker et al., 2016), 2) passive use of fire, 3) active control, i.e. the collection, preservation, and transportation of fire, and 4) artificial production of fire using stone percussion or wood friction (Frazer, 1930; Goudsblom, 1986). Determining what constitutes each stage and how to recognise it archaeologically is central to the debate and is often the main source of ambiguity when it comes to the positioning of the onset of these stages chronologically (and spatially). Further compounding the issue are the problems

A.C. Sorensen / Quaternary International 436 (2017) 114e128

inherent in differentiating anthropogenic fire from natural fire (e.g., Bellomo, 1993; Gowlett et al., 2005; Mentzer, 2014), as well as our understanding of how various fire residues and other evidences for fire use are produced and preserved (Sergant et al., 2006; Mallol et al., 2007, 2010; Carrancho and Villalaín, 2011; Courty et al., 2012; Henry, 2012; Vallverdú et al., 2012; Mallol et al., 2013; Carrancho et al., 2016; Rhodes et al., 2016). Neandertals were certainly knowledgeable about the properties of fire and not only used it passively for warmth and light, but actively as a tool for cooking food (e.g., Henry et al., 2011; Blasco et al., 2016), or for synthesizing birch bark tar adhesive for hafting (for an overview, see Wragg Sykes, 2015). This latter activity, while not only being complex (i.e., entailing many steps), appeared early in the archaeological record, at least prior to Marine Isotope Stage (MIS) 6 at Campitello Quarry in Italy (Mazza et al., 2006; € nigsaue (Koller Modugno et al., 2006), and again in Germany at Ko et al., 2001; Grünberg, 2002) during the Last Glacial (MIS 5a). Despite demonstrating some Neandertals had developed by at least 200,000 years ago a complex hafting technology for which controlled use of fire was essential, these findings are not proofpositive evidence of fire production, despite perhaps lending credence to the idea (Roebroeks and Villa, 2011b). The case for artificial fire making (see Section 2.4.5.) is further supported, however, by the recovery of Mousterian fire making tools from multiple Middle Palaeolithic (MP) archaeological sites in France dating to the Last Glacial (Rots et al., 2011; Sorensen et al., 2014; Sorensen and Claud, 2016; Sorensen and Rots, 2014; Rots, 2015; Heyes et al., 2016), suggesting at least some Neandertal groups had developed the ability to produce fire. This paper focuses on the transition between active fire use and fire production. This act was arguably the first true instance of domestication by humans (Goudsblom, 1992), wherein all the benefits conferred by fire are instantly made available. Roebroeks and Villa (2011a) contend the increasing regularity with which fire is observed in the MP suggests Neandertals (and their immediate predecessors) were habitual fire users from 400 to 300 ka onwards in Europe and the Near East (see also Gowlett et al., 2005; Shahack-Gross et al., 2014), and likely capable of producing fire for themselves (Shimelmitz et al., 2014). Conversely, Sandgathe et al. (2011b, 2011a) suggests this trend of increasing fire use through the MP is an artefact of taphonomy and does not reflect the more regular use of fire by Neandertals, going so far as to suggest fire was not a fixed component of the Neandertal toolkit. They cite the sporadic nature of fire use at European MP sites as evidence for Neandertals being obligate fire users (i.e., reliant on natural sources of fire, predominately caused by lightning strikes), while anatomically modern humans (AMH), on the other hand, were the first to kindle fire artificially (Sandgathe et al., 2011b). However, firemaking tools are largely absent from the early Upper Palaeolithic (UP) record, as well (Sorensen et al., 2014), and archaeological evidence for fire use by AMH during this period is also sporadic (Roebroeks and Villa, 2011b). Without evidence for fire making, could not any increase in the prevalence of fire traces during the UP also be interpreted as a relic of taphonomy? Moreover, these presumably fire-wielding groups had no noticeable effect on fire regimes in SW France and Spain upon their arrival to the region (Daniau et al., 2010a). Are we to presume then that the earliest AMH inhabitants of Europe were also obligate fire users? If not, what factors might then account for the variation in fire use signals between these two groups, as well as between conspecific groups over time and space? Using the observed trends for fire use at Roc de Marsal and Pech IV as a point of departure (Sandgathe et al., 2011a), the goal de l’Aze of the research discussed here is to test the strength of the concept of obligate fire use by Neandertals during the Last Glacial by

115

comparing and contrasting archaeological signals for fire use during cold and warm climatic periods in SW France with shifts in modelled lightning fire ignition frequencies and palaeofire regimes, and to provide possible explanations for fire signal variability between ‘cold’ and ‘warm’ occupation layers and between sites. 1.1. The environmentally determined Neandertal fire use model Sandgathe et al. (2011b, 2011a) assert that if Neandertals were dependent on natural, lightning-caused conflagrations as their primary source for domestic fire, the prevalence of anthropogenic fire should be linked to prevailing climatic conditions. Thus, changes in the fire use signal within MP archaeological sites should correspond to shifts in climate. The fundamental assertions of this model (both implicit and explicit) are as follows (see also Table 1): 1) Global and regional climate phenomena influence lightning frequency, with cooler climates resulting in reduced lightning frequencies (cf., Rakov and Uman, 2003). 2) Lightning frequency determines the regional and local fire regime, defined as the pattern of frequency, season, type, severity and extent of fires in a landscape (Bond and Keeley, 2005), meaning fewer lightning strikes would result in fewer overall ignitions. 3) Shifts in fire regime directly determined the frequency of Neandertal access to fire and should thus a) manifest similarly in contemporaneous layers at nearby archaeological sites, and b) coincide with climatic shifts indicated by the faunal and floral records. In other words, archaeological layers with strong fire signals should, as a rule, coincide with environmental proxies that indicate warm conditions, while fire-poor layers should coincide with cold climate environmental proxies. 4) If Neandertals were capable of artificially producing fire at will, then the archaeological signals for fire use should be similar throughout the glacial and interglacial cycledif not more pronounced when it is cold. Sandgathe et al. point specifically to the apparent decrease in fire use during colder periods observed at the late MP sites of Roc de IV (Dordogne, France) as evidence for this Marsal and Pech de l’Aze model. 1.2. Site descriptions and fire evidence Roc de Marsal (hereafter, RdM) is a small south-facing cave ze re River situated about 80 m above a tributary valley of the Ve (located 1.7 km to the north-west), approximately 1.5 km southeast of Campagne and 5 km south-west of Les Eyzies (Fig. 1). The site has yielded high numbers of flint tools, cores and flakes (more than 23,000 artefacts 2.5 cm), large numbers of faunal remains, and numerous combustion features (Bordes and Lafille, 1962; Turq, 1979; Sandgathe et al., 2007, 2008; Turq et al., 2008). With a total thickness between 1 and 2 m, the stratigraphic sequence at Roc de Marsal is comprised of seven geological lithostratigraphic units containing 13 archaeological layers (Fig. 2; for more detailed descriptions of the stratigraphy, see Aldeias et al., 2012, and Goldberg et al., 2012). IV (hereafter, Pech IV) is a south-facing collapsed Pech de l’Aze cave situated approximately 50 m above a small, usually dry stream a, a small tributary of the Dordogne valley that runs into the Ene da, River (1.7 km south of Pech IV), 5 km south-east of Sarlat-la-Cane and roughly 20 km east of Roc de Marsal (Fig. 1). Pech IV is one of a series of four caves in the area bearing MP deposits (Bordes, 1972, 1975). Using revised lithostratigraphic designations, eight major stratigraphic layers (17 levels in total, including subdivisions) have

116


Table 1 Outline of the factors potentially influencing archaeological fire signals, including the primary arguments presented by Sandgathe et al. (2011a) and the counterarguments discussed in the text and their associated section number. Natural fire availability Sandgathe et al. (2011a) arguments

p. 235 - "Lightning frequency is directly related to temperature and humidity and drops significantly in cold dry clima c condi ons (Rakov and Uman 2003). Interes ngly, the rela onship between climate and lightning frequency matches what the data show is the case during the Mousterian, namely less fire in cold/dry periods and more during warm/humid ones."

Site funcƟon/use p. 231 - "[T]he lithic and faunal data suggest that the presence or absence of fires is not a reflec on of differences in the site use. The most parsimonious explana on in the face of the ethnographic data and what we see at Pech IV and Roc de Marsal is that both sites served generally as base or residen al camps, and there is nothing to suggest that this func on changed significantly throughout their occupa onal histories."

Seasonality p. 231 - "As yet we cannot exclude the possibility that the other non-fire occupa ons were limited to summer months, but given the long temporal span of these deposits, it would seem unlikely that occupa ons took place only during warm months for many tens of millennia."

Fuel availability/type

Taphonomy/PreservaƟon

p. 235 - "[E]ven during extreme cold periods there were always some trees around, though likely restricted mainly to river valleys."

p. 224 - "[S]trong arguments can be made to show that preserva on was not a significant factor."

p. 235 - "[B]one can also be used as fuel."

p. 224 - "Further evidence that the lack of fires in the upper layers is not a result of preserva on comes from indirect data that reflects the presence of fire even when more direct evidence of actual fire residues are missing."

p. 231 "[C]ondi ons during even the warm months of MIS 4 were s ll far cooler than today’s and, again, modern huntergatherers in even substan ally warmer environments s ll rely on fire to warm themselves, especially at night."

Counterarguments (this paper)

Sampling methods Fig. 6, - "[Artefact] counts are based on p. 227 objects greater than 2.5cm in length and the flint includes only proximal and complete pieces (flakes, tools, and cores)."

p. 227 - "At both sites, the percentage of burned objects, both lithic and fauna, agrees very closely with the frequency of more ephemeral, direct fire residues (charcoal and ash)."

2.4.1. - This assessment does not consider poten al changes in site use frequency or dura on caused by increased mobility.

2.4.1. - Increased spring/summer seasonality at Pech IV in colder layers may reflect shi in site use pa erns based on seasonal local availability of migratory reindeer prey.

2.4.2. - Trees likely not in the direct vicinity of upland sites like Pech IV and RdM during colder periods, so more effort would be needed to collect wood from valley bo oms, likely leading to increased fuel economiza on (i.e., smaller, more task-specific fires of shorter dura on).

2.4.3. - Notable amounts of fire proxies present in layers without combus on features should be an argument for the influence of destruc ve postdeposi onal processes, not against. If taphonomic impacts were minimal, then intact combus on features (just fewer of them) should be present in all layers where fire proxies are present. This is not the case.

2.4.4. - The minimum artefact size of ≥2.5 cm for inclusion in the sample disregards the smaller artefact size classes where most fire affected artefacts fall.

2.1.2. - Marine microcharcoal concentra ons, as a proxy for palaeofire regime, are only weakly to moderately lower during colder stadial events compared to warmer interstadials.

2.4.1. - Roof collapse and/or sedimentary infilling may have caused changes in how a site is used, affec ng fire use and hearth placement.

2.4.1. - Observing diachronic change in seasonality not currently possible at RdM, though layer 4 (Quina) gives a mixed signal.

2.4.2. - Bones likely only used as primary fuel source if heat is more important than caloric intake, or to clear bone debris from a site.

2.4.3. - Lower (hearth) layers at Pech IV protected by rockfall, with resultant 'chemical buffer' helping to preserve ash.

2.4.4. - Fine-grained site-wide sampling necessary to iden fy microcharcoal and highly fragmented bone.

2.4.1. Strong spring/summer seasonality can reduce the frequency of site visits in a year and can affect hearth placement (i.e., outside of the cave/shelter).

2.4.2. - Bone is a weak producer of conduc ve heat, causing less thermal impact on a substrate.

2.4.3. - Evidence for cryoturba on in upper layers and li le to none in lower layers.

2.4.4. - Micromorphology is helpful, but sampling is limited; one cannot easily direct sampling to invisible hearth loca ons, so there is more focus on overt hearth features.

2.4.2. - Increased fragmenta on rate of burned bone generally results in pieces falling below the Sandgathe et al. minimum ≥2.5 cm sample size.

2.4. - Fire proxies are not only about preserva on, but also condi ons for crea on: higher geogenic sedimenta on rates and reduced site use frequencies (perhaps indicated by lower artefact densi es), coupled with reduced frequency, intensity and dura on of fires, decreased the chances of incidental hea ng of extant lithic and bone artefacts.

2.2. - Strong evidence for cold stage fire use is present at a number of Neandertal sites in France da ng to the Last Glacial, including Combe Grenal. Conversely, there are numerous instances of warm layers (at Combe Grenal and elsewhere) with very weak fire signals.

Fire making p. 235 - "European Neandertals did not know how to make fire."

2011b - "We would argue that, if Neandertals had the ability to make fire at will, then evidence for it should occur with much greater frequency in Middle Paleolithic sites and occupa ons and especially, those sites associated with such cold stages."

2.1.1. - Lightning and fire igni on rates are very low in the Dordogne today despite prevailing warm and humid condi ons.

2.1.3. - If these were reduced, it probably wasn’t enough to virtually eliminate archaeological fire use. Disparity not great enough to account for reduc on in fire proxies at RdM and Pech IV.

ExcavaƟon extent p. 229 - "Realis cally it is highly unlikely that fires could have been constructed in these remaining areas without resul ng in residues (even if only burned lithics) bleeding into the adjacent excavated areas."

2.4.4. - This is probable, but it also assumes 1) hearth features would have been preserved in the (cold) upper layers, and 2) that no ceably higher amounts of fire proxies would have been produced.

2.4.5. - Possible fire making tools have been recovered from mul ple Middle Palaeololithic sites da ng to the Last Glacial.

2.4.5. - The ability to make fire at will may have reduced the fire signal by allowing for more economic, taskspecific fires (i.e., there was no need to maintain fires for long periods when fuel is in short supply).

2.3. - Nega ve shi s in fire proxies can be seen either well before the onset of colder clima c condi ons (Pech IV) or poten ally well a er (RdM).

Fig. 1. Locations of archaeological sites discussed in the text.

been identified (Fig. 2; for a more detailed stratigraphic description, see Goldberg et al., 2012); of these, layers 3e8 are Pleistocene deposits containing abundant MP artefacts, faunal remains and fire residues (McPherron and Dibble, 1999; Dibble et al., 2009; Turq et al., 2011). Sandgathe et al. (2011a) point to the comparable stratigraphic succession of these two sites as evidence for their rough contemporaneity. Discrete hearth structures (some stacked) only appear in the lowest occupation levels of the sites (layer 8 at Pech IV, and layers 9 and 7 at RdM), with only diffuse or no fire residues (i.e., charcoal, ash) and reduced frequencies of fire proxies (i.e., heated flint, burned/charred bone) in the layers above (Fig. 2) (Dibble et al.,

2009; Aldeias et al., 2012). The reduction in fire signals coincides with 1) a reduced presence of temperate forest fauna species (e.g., red deer, roe deer) and an increase in reindeer remains (the main faunal proxy indicating cold, dry and more open conditions) to between ~60 and 90% of the fossil bone assemblages in layer 4 at Pech IV and layers 5e2 at RdM (Laquay, 1981; Guadelli, 1987; rin et al., 2012; Morin et al., 2014; Hodgkins et al., 2016; CasGue tel et al., in press), which the authors interpret as indicating the transition from MIS 5 (warmer) to MIS 4 (colder); and 2) a transition from Levallois-based lithic techno-complexes (“Ante-Quina”, Discamps et al., 2011) to the Quina Mousterian in layer 4 at Pech IV and layers 4e2 at RdM (Turq et al., 2008, 2011; Goldberg et al.,

A.C. Sorensen / Quaternary International 436 (2017) 114e128 IV and Combe Grenal, with associated fire proxies, reindeer data and Mousterian typological designations*. The reindeer curves represent the relative taxonomic abundance of Fig. 2. Site profiles for Roc de Marsal, Pech de l’Aze rin et al., 2012; Hodgkins, 2012; Hodgkins et al., 2016; Castel et al., in press; data reindeer remains compared to the Number of Identified SPecimens (% NISP) of other ungulates recovered in each level (Laquay, 1981; Guadelli, 1987; Gue largely compiled by Morin et al., 2014). Layer designations, Mousterian industries and fire proxy data for RdM and Pech IV are after Sandgathe et al. (2011a), while the reindeer curve for Pech IV is based on the older layer designations (right) by Bordes (1975). The fire use signals for Combe Grenal are based on observations by Bordes (1955, 1972) of ‘strong’ (3) and ‘weak’ fire traces (2), while one (1) indicates layers where heated flint or burned bone have been observed (Vogel and Waterbolk, 1967; Bowman and Sieveking, 1983; Faivre, 2011); underlined layer numbers indicate the presence of fire. The Combe Grenal profile and Mousterian industry designations are both after Bordes (1972). The profile photo of Pech IV is after Goldberg et al. (2012). The profile photo of RdM is used with the permission of Alain Turq. *Mousterian typological designations are abbreviated as follows: A ¼ Anisipodian, D ¼ Denticulate, F¼Ferrassie, MTA ¼ Mousterian of Acheulean Tradition, Q ¼ Quina, T ¼ Typical.

117

118


2012). At Pech IV, the fire signal remains weak into the uppermost MP layers (3e2), which contain Discoid lithic techno-complexes assigned to the Mousterian of Acheulean Tradition A/B typology. The authors assert that the observed fire trends cannot be explained by variable preservation of fire traces due to taphonomic agencies, excavation extent or methods, site function or seasonality of site use. 1.3. Testing the model A top-down approach is used to explore the assertions inherent to this environmentally determined Neandertal fire use model, wherein the following questions are addressed: 1. To what degree did climatic shifts affect lightning flash frequencies, and how might these changes in lightning frequency have influenced natural fire ignition frequency? 2. How do these estimated shifts in fire frequency compare in timing and magnitude to diachronic shifts in a) fire regime in south-west France, inferred from fossil marine microcharcoal concentrations from the Bay of Biscay, and b) fire use signals observed at RdM and Pech IV? 3. Do the patterns for ‘cold’ and ‘temperate’ fire signals from these sites indeed capture the same climatic shift from MIS 5 to MIS 4, and how do these signals compare with other presumably contemporaneous archaeological sequences in the region? 4. What other factors (environmental and cultural) potentially create disparities in the production and preservation of fire residues and fire proxies between cold and warm periods, and how might excavation and sampling methods (in general) skew fire signals? 2. Results and discussion 2.1. Lightning, ignition and changes in palaeofire regime in SW France during the Last Glacial 2.1.1. Lightning and fire ignition rates Sandgathe et al. (2011a) contend that lightning frequency, being linked to temperature and humidity (cf., Williams, 1995; Rakov and Uman, 2003), would have been reduced during colder, drier stadial

periods, thus limiting the availability of natural fire sources. If Neandertals were indeed obligate fire users, this reduction in lightningdand the assumed parallel reduction in fire regimedmay have limited Neandertal fire use enough to account for the disparate fire signals encountered between the warm and cold occupation layers at RdM and Pech IV. Discussing in detail all the complexities dictating fire ignition and propagation in any one location (Keeley et al., 2011) is beyond the purview of this article. Among the most important, however, are vegetation/fuel type, which is largely determined by annual rainfall and rainfall seasonality (Archibald et al., 2009; Woillez et al., 2014), fuel state (especially the moisture content of fine fuels), the presence of a high pressure weather system, the amount and positioning of rainfall during a storm event (Nash and Johnson, 1996; Latham and Williams, 2001), the percentage of tree cover/ openness of the landscape (Uhl and Kauffman, 1990; Hennenberg et al., 2006), fuel connectivity (Archibald et al., 2012), and wind speed (see also: Kourtz and Todd, 1992; Wotton and Martell, 2005). Lightning is, of course, integral to this process, but it is important to point out that lightning frequency is only loosely correlated to fire ignition (Court, 1960; Komarek, 1964; Flannigan and Wotton, 1991; Latham and Williams, 2001; Minnich et al., 2009). In other words, areas with higher lightning frequencies do not necessarily have higher fire ignition rates. So how different might fire ignition rates have been during colder glacial conditions compared to the warm interglacial conditions today? According to data compiled by NASA (Cecil et al., 2014), the Aquitaine Basin today only receives on average 2e6 lightning flashes km2 yr1, with the Dordogne receiving around 3 (Fig. 3). While lightning causes most natural fires (Stewart, 1956; Komarek, 1965), not every strike results in an ignition. About 75% of all lightning flashes are restricted to the clouds, meaning only around 25% ever make contact with earth's surface (CIGRE WG C4.407, 2013). Ultimately, only around 1e4% of these cloud-toground strikes ignite a fire (Latham and Williams, 2001, and references therein). This estimate is based largely on (sub)boreal forest environments and can perhaps serve as a mid-range estimate, since fire frequencies are often higher in steppic regions (Sannikov and Goldammer, 1996) and lower in temperate deciduous forests (Bond and Van Wilgen, 1996). Given these rates, one could expect between 0.0075 and 0.03 fires km2 yr1 in the Dordogne

Fig. 3. Global annual distribution of lightning, 1996e2014, from the combined observations of the NASA OTD (1995e2000) and LIS (1998e2014) instruments. Inset: Close-up of France, star indicates location of the Dordogne, which receives on average 3 flashes km2 yr1. Global lightning Image obtained from ftp://ghrc.nsstc.nasa.gov/pub/lis/climatology/ HRFC/browse/HRFC_COM_FR_V2.3.2014.png, maintained by NASA EOSDIS Global Hydrology Resource Center (GHRC) DAAC, Huntsville, AL. 2015. Data for the image were provided by the NASA EOSDIS GHRC DAAC. (For clarity, we refer readers to the color version of this figure on the web.)


(Supplementary Table 1). To compare, historic fire frequency data from the Dordogne (1993e2013) closely align with the modelled range at 0.00916e0.0526 fires km2 yr1, with the inclusion of human-caused fires skewing the data towards the high end of the range (Data provided by the European Forest Fire Information System or EFFIS (http://effis.jrc.ec.europa.eu) of the European Commission Joint Research Centre; San-Miguel-Ayanz et al., 2012). Lightning frequency tends to rise by 12 ± 5% for every 1 C increase in global mean annual temperature (GMAT) (Romps et al., 2014), and likely drops by a similar percentage for every 1 C decrease (Romps, pers. comm. 2015). The modelled 3e6 C reduction in GMAT for the Last Glacial Maximum (MIS 2) (Annan and Hargreaves, 2013, and references therein) provides a conservative lower limit for the coldest stadial periods encountered during the Last Glacial. This would potentially result in a 31.2 ± 11.6% to 51.3 ± 16% drop in lightning frequency, causing between 0.002453 and 0.02412 fires km2 yr1 in the Dordogne (Supplementary Table 1). It is important to note here the overlap between the number of expected fires in colder and warmer periods; it demonstrates the potential (alluded to above) for there to be more fires during a cooler period despite reduced lightning strike frequencies if favourable vegetation and climatic conditions increase ignition rates. Neandertals were not static beings waiting for the fire to come to them, however (Gowlett, 2015). Depending on weather

119

conditions and terrain, smoke plumes are potentially visible up to 50 km away (Horvath, 1995). Using the average single day foraging distance for modern hunter-gatherers of around 15 km (round-trip) (Binford, 2001), one could reasonably expect to encounter between 1.3 and 5.3 natural fires yr1 (of any size) in the Dordogne today within this 176.7 km2 daily foraging area surrounding a site, and between 0.4 and 4.3 natural fires yr1 under full glacial conditions. If whole groups relocated near to active fire sources (again, assuming an average daily walking distance of 15 km), then they could expect between 5.3 and 21.2 fires yr1 within this range (706.86 km2) today, or 1.7e17.0 fires yr1 during glacial periods. Logs left smouldering can potentially burn for days or weeks (Minnich, 1987; Rabelo et al., 2004), while peat fires (usually in boreal zones) can smoulder for months on end (Rein, 2009), meaning if the desire were great enough, any fire within visible range could be reached within a few days and may still be exploitable long after the flaming fire front has been extinguished. Using the 50 km maximum visibility radius, 75e300 fires yr1 could potentially be exploited under interglacial conditions, and 24.5e241.2 fires yr1 under glacial conditions. Indeed, on any given day during the summer months within the boreal zone of Alaska today, about 100 (natural) forest fires will be burning (Farukh and Hayasaka, 2012), which if distributed evenly, would yield at least one active fire within this 50 km distance. Together, the mobile nature of Neandertal groups and their ability to transport and

Fig. 4. Microcharcoal data from deep-sea core MD04-2845 (Bay of Biscay) plotted by age (left), and box plots (right) comparing combined microcharcoal concentrations (CCnb) and total microcharcoal surface areas (CCsurf) for warmer Greenland interstadial (GI) and cooler stadial (GS) periods. The dashed lines (red in the electronic version) denote median microcharcoal values for the entire period displayed. Percentages correspond to the relative distance from median values for the highest and lowest data points for each metric. Microcharcoal data are after Daniau et al. (2009).

120


conserve fire could have helped, in part, to negate the modest differences in fire ignition frequencies between climatic periods. 2.1.2. Palaeofire regime Microcharcoal data obtained from marine core MD04-2845 in the Bay of Biscay has been shown to be a reliable proxy for describing changes in biomass burning that reflect shifts in fire regime in SW France (Daniau et al., 2009, 2010b). This data is expressed as the concentration of microcharcoal (CCnb ¼ number of microcharcoal per gram, nb g1) and as the total surface area of the microcharcoal (CCsurf ¼ the sum of all surfaces of microcharcoal in one sample per gram, mm2 g1). This latter measure helps to correct for inflated microcharcoal concentrations caused by increased fragmentation related to transport (Daniau et al., 2009) or production, where higher intensity fires tend to increase ry-Parisot, 2001). rates of fragmentation (Komarek et al., 1973; The Determining the relative differences in fire regime between these warmer and colder periods may provide a means to test the validity of the Sandgathe et al. model, at least at the larger regional scale. To assess how fire regimes fluctuated during the portion of Last Glacial likely represented at RdM and Pech IV, median microcharcoal values spanning MIS 5c through mid-MIS 3(c. 105e32 ka) were compared after further subdivision into warmer Greenland interstadial (GI) and colder Greenland stadial (GS) periods (North ~ i et al., Greenland Ice Core Project Members, 2004; Sanchez-Gon 2008; Daniau et al., 2009; Svensson et al., 2008). CCnb values vary between 218.4 104 and 1607 104 nb g1, or -75.2% and þ82.2% (respectively) relative to the overall median value of 8820 104 (Fig. 4). CCsurf values vary between 5069 104 and 20,137 104 mm2 g1, or 55.1% and þ78.2% relative to the

median value of 11,299 104. The median CCnb values for GS and GI (respectively) are 8333 104 and 9782 104 nb g1, or -5.5% and þ10.9% relative to the overall median. The median CCsurf values for GS and GI are 11,216 104 and 11,332 104 mm2 g1, or 0.7% and þ0.3% relative to the median. For CCnb, 41.2% of data points during GS are above are above the median, while 63.9% are above the median during GI (Fig. 5). For CCsurf values, 49.1% are above the median for GS, and 51.8% are above the median for GI. As expressed above, microcharcoal values for GI and GS are given both in terms of their relation to median CCnb and CCsurf values for the whole period in question (Fig. 4), as well as the percentage of data points per GI/GS that present values above or below median CCnb and CCsurf values (Fig. 5). This latter measure helps to gauge the relative amount of time fire regimes were above or below average within a given time slice, while largely negating the effects of anomalously high or low data points within a series. For example, three of the five lowest CCnb values and two of the five lowest CCsurf values occur during Heinrich Event 6 (H6), a major cold snap during the Last Glacial. Nevertheless, median microcharcoal values for this period are, perhaps surprisingly, just above the overall median (CCnb ¼ þ0.25%, CCsurf ¼ þ4.42%), while 55.5% of CCnb values and 75.8% of CCsurf values are above the median. These values appear to run counter to the expectations of the Sandgathe et al. model. Overall, the two microcharcoal metrics appear to be slightly at odds, with CCnb values appearing moderately lower during GS compared to GI, while CCsurf values are nearly identical. As alluded to previously, this difference may partially be related to increased charcoal fragmentation caused by higher intensity burning associated with forest fires during warmer GI, thereby inflating

Fig. 5. Microcharcoal data presented as the number of data points above or below median CCnb and CCsurf values (see Fig. 4) within Greenland interstadial (GI) and stadial (GS) periods. Pie charts (left) compare combined GI and GS data for the entire period analysed, while the bar graph (right) provides a breakdown of the relative percentages of data points above and below median values and the number of data points represented per GI and GS period.


microcharcoal concentrations (CCnb) relative to those produced by lower intensity grass and brush fires during GS. 2.1.3. Comparisons with archaeological fire proxy data As it is argued below, treating the amount of fire residues observed within archaeological layers as another environmental proxy is fraught with problems (see Section 2.4.3.). Nevertheless, accepting momentarily the assumption that there is a direct relationship between natural fire prevalence, Neandertal fire use and relative quantities of fire proxies, fire proxy data for RdM suggests there was roughly 23 times more fires in layer 9 (containing the highest proportion of heated lithic artefacts at 30%) compared to layer 4 (the lowest at 1.3%), and at Pech IV, around 23.7 times more fires in layer 8 (21.3%) than layer 4 (0.9%, the average of levels 4A-C according to Turq et al., 2011). By contrast, the model outlined above in Section 2.1.1. suggests the annual number of lightning fires could be at most 13.25 times greater under interglacial conditions than under glacial conditions, assuming the most favourable and least favourable conditions for ignition prevailed during these periods, respectively. While it is true that, compared to the median GS and GI microcharcoal values, greater shifts between individual peaks and troughs can be observed in Fig. 4, the highest CCnb value recorded is 7.36 times greater than the lowest, while the highest CCsurf value is only 3.97 times greater (Figs. 4 and 5). The lack of precision dating and the potential effects of time averaging in archaeological deposits preclude accurate assignment of individual microcharcoal data points to specific archaeological levels or occupations. Ultimately, neither proxy, even at their extremes, suggests rates of fire prevalence in the landscape were ever so low for long enough periods to account for the drastic reductions in fire signal at RdM and Pech IV, indicating other factors are more likely responsible. 2.2. Archaeological signal for cold stage fire use by Neandertals Evidence alluding to the presence of fire on archaeological sites during the MIS 4 cold stage is common (see Roebroeks and Villa, 2011a), but Sandgathe et al. observe that the degree of burning rarely approaches that seen in the levels of RdM and Pech IV attributed to MIS 5, at least as far as preserved combustion features are concerned. While the authors acknowledge that “It is clear that the evidence from these few sites cannot easily be applied to an entire region” (Sandgathe et al., 2011a, p. 234), their conclusions appear to do just that despite strong evidence for fire use through the coldest periods of MIS 4/3 at the nearby site of Combe Grenal. A portion of the 13 m sequence at Combe Grenal (Domme) appears to be contemporaneous with RdM and Pech IV (Discamps et al., 2011; Faivre et al., 2014; Morin et al., 2014), exhibiting similar climatic and archaeological trends (for detailed descriptions of the stratigraphic sequence, see Bordes, 1972 and Mellars, 1996). However, despite being located very near Pech IV (6 km) and Roc de Marsal (22 km) (Fig. 1), the fire signal there seems to run counter to what would be expected according to the Sandgathe et al. model. In all, 127 small and 19 large hearths (foyers, in French) were documented by Bordes in his excavation notebooks (Binford, 2007). Just over one-third of the Combe Grenal layers (22 of 65) are purported to have traces of fire use (Fig. 2), with three-quarters of these in ‘cold’ layers (in bold: 9, 12, 14,17, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 38, 43, 49, 50, 52, 54, 55, 60), as indicated by greater relative presence of reindeer (Guadelli, 1987; Laquay, 1981; Morin et al., 2014) and/or reduced arboreal pollen (AP) (Bordes, 1972). There are nine layers with ‘strong’ traces (54, 52, 50, 43, 29, 27, 23, 20, 14), as described by Bordes (1955, 1972) using terms like “large”, “continuous”, or lacking the qualifiers “scattered”, “small”, “diffuse”, “ashy”, or “thin” used to distinguish the nine layers with ‘weak’ traces (38, 30, 28, 25, 24, 22, 21, 12, 9). Fifty-four heated flint

121

fragments were observed in layer 17 (Faivre, 2011). Five other layers (60, 55, 50, 49 and 20) contained heated flints that were subjected to TL analysis (Bowman and Sieveking, 1983). And layer 12 (E2) contained burned bone that was collected for radiocarbon dating (Vogel and Waterbolk, 1967). Bordes (1975) also notes firereddened patches of limestone bedrock directly underlying layer 55. Despite the current lack of fire proxy data from Combe Grenal for quantitative comparison, the signal of fire use does not at all appear to conform to the Sandgathe et al. model: fire traces occur regularly in the cold layers associated with elevated reindeer presence during MIS 4, with seven of the ten layers assigned to the Quina lithic techno-complex (Faivre et al., 2014) exhibiting evidence for fire use. Other sites in France also exhibit strong fire signals in layers assigned to cold periods. For example, layer 4 at Abris du Maras che and Rho ^ ne Rivers in south-east (near the confluence of the Arde France), exhibits combustion structures, high percentages of burned bone, charcoal and heated flints, with a faunal assemblage comprised of ~85% reindeer (NISP) (Daujeard and Moncel, 2010; Moncel et al., 2015). At La Quina (Charente, SW France), Bed 8 is dominated by reindeer remains (~90%) and contains large amounts of burned bone (~95% < 2.5 cm) (Chase, 1989; as cited in Jelinek, 2013). It should also be noted that the logic of the Sandgathe et al. model could cut both ways: if anthropogenic fire signals weaken when it is colder, then they should become stronger when it is warmer. Yet, this is not always the case, as attested to by the near absence of observed fire traces at any number of Last Interglacial (MIS 5e) and Early Glacial (MIS 5d-a) sites (see Dataset S1 in Roebroeks and Villa, 2011a,b for examples). One is then left to wonder whether the reduced fire signals witnessed in the cold layers at RdM and Pech IV are indeed related to reduced fire in the environment, or instead to reduced use of fire by Neandertals for some other reason, taphonomy, or some combination of these factors. Given the contradictory pattern observed at Combe Grenal, it is possible that the changes in fire signals at RdM and Pech IV reflect local phenomena and are not indicative of a region-wide pattern. 2.3. Contemporaneity: problems and implications for addressing fire and environmental signals The limitations of chronometric dating are rarely more frustrating than around transitional periods. The volatile nature of climatic fluctuations during the Last Glacial and the implications these have on Neandertal lifeways makes this problem of dating particularly vexing, as interpretations of climatic signals are often heavily influenced by newly acquired absolute dates. While the primary concern of this paper is the relationship between environmental and corresponding fire signals, (re)interpretation of climate signals based on chronometric data can potentially cause conflicts with established models. The Sandgathe et al. model hinges on the assertion that the similar environmental, cultural and fire signals observed at RdM and Pech IV together indicate a degree of contemporaneity, in this case, the assumed transition from (firerich) MIS 5 to (fire-poor) MIS 4. On its face, this seems reasonable; rin et al. however, chronometric redating of layers from RdM by Gue (2012, in press) casts doubt on this interpretation by suggesting the fire-rich lower layers 9e7 were deposited early to mid-MIS 4 (70e65 ka). Moreover, these dates coincide with recent thermoluminescence dates (Richter et al., 2013) and single-grain optically stimulated luminescence dates (Jacobs et al., 2016) acquired for layer 4C at Pech IV (68.5 ± 6.6 ka and 71.8 ± 6.7 ka, and 68.3 ± 3.9 ka, respectively), a colder layer virtually devoid of fire traces according to the findings of Sandgathe et al. (2011a). The incongruity

122


between the more temperate environmental signal present in these layers at RdM and the cold snap inferred from multiple climatic ~i proxies during this date range (i.e., GS 18e19; see Sanchez-Gon et al., 2008; Daniau et al., 2009) could suggest regional or local factors (specifically vegetation type and distribution) are more influential on the prevalence of particular kinds of fauna, as well as the prevalence of fire. Indeed, despite prevailing colder temperatures, boreal forests appear to have expanded in the Aquitaine basin ~ i et al., 2008), providing a degree through mid-MIS 4 (Sanchez-Gon of continuity for more forest-adapted species while serving as sheltered wintering grounds for south-expanding reindeer, thus creating the more ‘composite’ faunal spectrum observed in the lower layers at RdM and other sites in the region (Discamps et al., rin et al., 2012). Moreover, retention of the forests (i.e., 2011; Gue abundant fuel resources) in this area would potentially allow for continuity of fire use practices by Neandertals through the first half of MIS 4, prior to the onset of Heinrich Event 6. Ignoring chronometric dates for the moment, the climate and fire signals at RdM and Pech IV also show discrepancies when comparing relative percentages of reindeer remains (Laquay, 1981; rin et al., 2012; Morin et al., 2014; Hodgkins Guadelli, 1987; Gue et al., 2016; Castel et al., in press) to fire proxies (Sandgathe et al., 2011a,b). The earliest archaeological horizons at RdM (layers 9e6), which contain hearth structures (layers 9 and 7) and the highest percentages of heated flint (~30e13%, respectively; see Fig. 2), already have faunal assemblages comprised of 10.6e33.2% reindeer; whereas at Pech IVdhere, hearths are again only present at the base of the sequence (layer 8)dfire proxies drop steadily from 21.3 to 4.6% between layers 8 and 6A, while reindeer presence remains almost nil (0e1.5%, respectively), suggesting the fire signal at the site dropped off already prior to the climatic deterioration signalled by the relative increase in reindeer fauna. Only in layer 5B (Bordes levels J3 and J2) do reindeer percentages (17.1e34.7%, respectively) approach levels seen in the hearth-bearing layers at RdM, but by this point, percentages of heated flint have already dropped to below 1%. Does this support the younger chronology for rin et al. (2012, in press)? While this disparity RdM proposed by Gue could be attributed to differences in when or how the sites were used, perhaps caused by differential positioning of prey species in the landscape (see Section 2.4.1.), the apparent waning of fire use at Pech IV prior to the shift towards a more cold climate faunal spectrumda similar reduction in fire signal occurring well after this transition at RdMdagain calls into question the forces driving these changes in fire signal, and the applicability of Sandgathe et al.'s fire use model to the region as a whole. 2.4. Factors influencing archaeological fire signals Reconstructing the chaîne op eratoire of a fire from its resultant residues is no easy affair (cf., March et al., 2014): primary fire residues rarely preserve (cf., Mallol et al., 2007), and the direct relationship between fire use and archaeological fire proxies is complex and remains poorly understood. Whether or not an individual fire leaves behind evidence of it having burned depends not only on preservation conditions at a site, but also on the nature of the fire itself. External controls, both environmental and cultural, will influence how a fire burns (e.g., large and hot, small and cool, long or short duration), which in turn dictate the strength of the initial fire signal. After a fire is abandoned, attritional processes, both shortterm (e.g., displacement by wind or water) and long-term (e.g., diagenesis), determine to what degree the original fire signal remains intact. Finally, the excavation and analysis methods employed by archaeologists will determine if and to what extent the necessary evidence is collected for identifying and interpreting this depleted fire signal.

With regards to the production, preservation and recovery of fire residues and proxies, Sandgathe et al. (2011a) addresses many of the possible explanations for why fire signals appear weaker in the upper levels of RdM and Pech IV, ultimately discounting site function, seasonality, fuel availability, taphonomic agencies, excavation extent and excavation methods. These options having been exhausted, the authors propose Neandertals lacking the ability to make fire best explains the pattern. Below are some alternative interpretations of the record that should provide insight into the problematic nature of inferring the relative degree of anthropogenic burning from fire residues and fire proxy data between sites and occupation layers. 2.4.1. Site function and seasonality Sandgathe et al. question the notion that a change in primary site function at RdM or Pech IV to one that did not require fire could account for the lack of fire traces in layers with cold climatic signals. They reject this explanation based on roughly comparable carcass transport strategies, butchering activities and compositions of stone tool assemblages between fire-rich and fire-poor layers, which to them suggests these sites remained base/residential camps throughout their use lives. This assessment, however, does not take into consideration potential changes in site use frequency or duration. As colder conditions forced reindeer southward in high numbers through MIS 4/3 (Discamps et al., 2011; Morin et al., 2014), Neandertal subsistence strategies shifted to accommodate this new migratory prey by becoming increasingly more mobile (Niven et al., 2012; Rendu et al., 2012). If the inferred function of RdM and Pech IV as habitation sites remained unchanged, an increase in residential mobilitydpossibly coupled with reduced population densities (cf., Morin, 2008)dwould have led to fewer and shorter site visits, naturally resulting in weaker fire signals due to the presence of fewer fires of shorter duration. The absence of combustion features itself could be an indication that the function of these sites did change with the onset of colder climate (cf., Niven et al., 2012). Or if these sites were never primary residences, but instead perhaps hunting-observation stations or some other short-term campsites within a more logistic system all along, other external factors (e.g., easy access to wood fuel, discussed in the section below) may have been at play. Switching to a reindeer-based subsistence strategy would also have led to changes in when sites were occupied based on the local availability of this resource. Site use at Pech IV (Table 2 in Sandgathe et al., 2011a) shows an apparent increase in seasonality after reindeer become more prevalent that could possibly be related to the position of the site relative to the boundary between forest and steppe vegetation zones (cf., Stewart, 2005). The exploitation of forest species throughout the year in layers 8-6A suggests the site was well within the forested zone. As reindeer often prefer to overwinter just inside wooded areas due to the greater prevalence of lichens for winter forage (Klein, 1982), the hunting of reindeer during the winter and spring at Pech IV during Layer 5A suggests the site may have been located near the forest-steppe ecotone, and then located north(east) of this boundary during the colder period represented by layers 4C-4A, when spring and summer hunting may have been focused on intercepting reindeer migrating northward from more sheltered, southerly wintering grounds. Moreover, the appearance of swarms of biting insects during the spring and summer months (as is usual in the high-latitudes today) may have made more windswept upland locations preferred camping spots (Binford, 1978; Sharp and Sharp, 2015). This pattern could suggest Pech IV was utilised multiple times a year during warmer climatic intervals and perhaps only once a year during colder intervals, again reducing the number of opportunities for fire to have been


introduced into the site. Furthermore, if occupations were indeed largely restricted to the warmer months, it is possible that hearths were placed outside the protected confines of the cave/shelterda common practice among some modern northern hunter-gatherers suline of the northern Canadian Subarctic; Sharp (e.g., the Dene and Sharp, 2015). The few seasonality data reported from RdM are restricted to layer 4, thus not allowing for a diachronic comparison with Pech IV. Foetal reindeer bones suggest a late winter or early spring occupation, while teeth from two reindeer individuals indicate they were hunted in the summer or fall (Castel et al., 2016). This mixed signal could suggesting RdM may have been a multi-seasonal stop, possibly coinciding with the spring and fall reindeer migrations. Roof collapses and sedimentary infilling at Pech IV and RdM also likely altered when and how Neandertals utilised these caves over time. The receding of the porch at Pech IV caused it to function more like a rock shelter and is cited as a possible reason for so few fire remains in the upper layers (Turq et al., 2011). More generally, reduced clearance caused by sedimentary infilling can reduce both mobility and smoke ventilation within a cave (Gentles and Smithson, 1986; Lioubine, 1992), making it less attractive for usedat least with firedlater in its use life. Other enclosed sites where the use of fire appears to have been limited or discontinued in upper deposits possibly due to reduced clearance caused by sedimentary infilling include Combe Grenal (Bordes, 1972), Abri du Brugas (Meignen, 1981) and La Rochette (Soressi, 2002). 2.4.2. Fuel availability Sandgathe et al. argue that the widespread replacement of woodland by grasses during the colder periods does not explain the apparent reduction in fire use during these periods for two reasons: 1) There is always some wood around, though often confined to river valleys. 2) Bone can be used as a fuel when wood is scarce. These points are both valid but carry with them important restrictions. Fuel foraging generally conforms to the principle of least effort, though this can be amended based on the importance of fire to the ry-Parisot, 2001). group in question at any given moment (The During forested interglacial periods, high-quality wood fuel could be easily procured in the immediate vicinity of both RdM and Pech IV. Thus relatively little effort would have been needed to collect an abundance of fuel, allowing the inhabitants to maintain larger, longer-burning fires with minimal extra effort. During colder periods when grasses dominate the landscape, however, the likelihood of there being robust woody fuel sources in close proximity to either of these upland sites is low, as trees would have been largely restricted to valley bottoms. Considering the rough terrain (see Henry et al., in press), and depending on the distance to these fuel sources, the amount of energy required to extract a comparable amount of wood as was used during warm periods likely outweighed the benefits of the fire itself (Henry et al., 2016), likely leading to increased fuel economisation and perhaps shorter residency times as combustible materials in the immediate vicinity of the caves would have been rapidly consumed (Ofek, 2001). This has many negative implications for fire signals. Fires would have been lit more out of necessity rather than convenience or comfort, and would likely have been smaller and/or of shorter duration, i.e., more task-specific (cf., Mallol et al., 2007). Shrubs, grasses and herbaceous plants located in the immediate vicinity of these sites would have likely supplemented meagre wood fuels, potentially reducing fire signals for a number of reasons. Finer fuels tend to burn quickly and are poor producers of conductive heat (Cheney and Sullivan, 2008); they also tend to combust fully (in hearths), leaving behind few if any robust charcoals, and are nearly impossible to maintain in a non-flaming state since they do not produce glowing

123

coals that can be ‘banked’ in ashes and revived later. Much like finer fuels, the combustion of bone results in a flaming fire that extinguishes quickly (i.e., no glowing coals) and favours the production of radiative and convective heat, but is ry-Parisot, completely ineffective at producing conductive heat (The ry-Parisot et al., 2005), meaning fires fuelled primarily by 2002; The bone (and/or fine fuels) were less likely to alter the underlying substrate. Also, if site visits were only short-term affairs, site maintenance would likely have been minimal, and it is unlikely bone would have been extensively burned as a means of disposal. It should be remembered that the combustible portion of bone is also the nutritive portion, meaning the purposeful burning of bones suitable for use as fuel would mean sacrificing internal heat (i.e., calories) for external heat. During periods of nutritive stress, bones were probably only used as fuel if there was a surplus. Given the intensity with which Neandertals often processed bone to extract marrow at RdM and Pech IV (Hodgkins et al., 2016; Castel et al., in press), unless they had a surplus of marrow bones, only trabecular (spongy) bonedgenerally considered a more effective fuel than cortical (compact) bonedwould likely have been sacriry-Parisot et al., ficed to the fire as fuel (Costamagno et al., 1999; The 2005). Trabecular bone only comprises around 20% of a skeleton, so the relative proportion of elements burned may have been low to begin with, while the tendency for burned spongy bone to fragment easily would cause most remaining fragments to fall outside the 2.5 cm size range analysed by Sandgathe et al. (see Section 2.4.4.). However, if Neandertals were stressed to the point where it may have been better to process trabecular bone into bone meal rather than burn it, as postulated by Castel et al. (in press), then this practice would be one more factor potentially weakening the fire signal in the colder layers at these sites. 2.4.3. Taphonomic factors Sandgathe et al. point out the apparent correlation between the presence of combustion features and the greater prevalence of fire proxies in the warmer lower levels at RdM and Pech IV, and conversely, the paucity of hearth traces and low percentages of fire proxies in the colder upper levels. On its face, this observation seems logical, but the actual relationship between combustion features and fire proxies is not nearly so relative due to variability in fire use, depositional settings and preservation. Indeed, comparable levels of fire proxies can occur in archaeological layers devoid of intact combustion features (e.g., the Lower Palaeolithic site Gesher Benot Ya'akov: Goren-Inbar et al., 2004; Alperson-Afil et al., 2007; Alperson-Afil and Goren-Inbar, 2010) as are recovered from layers with multiple well-preserved hearth veyres and structures (e.g., the UP Magdalenian sites of Champre Monruz: Leesch et al., 2010). Hypothetically speaking, variability in fire use (i.e., fire size, duration, intensity, frequency) can result in disparate fire signals; but in some cases, it could lead to similar signals due to equifinality. Regarding this latter point, if conditions are such that hearth features are not preserved, one large fire burning for an extended period could potentially result in a similar proportion of dispersed fire residues or fire proxies as numerous smaller, short-lived hearths. On the other hand, if a fire is built over an extant lithic scatter (Sergant et al., 2006), the resulting number of heated lithic fragments will be much higher than from an identical fire built on a previously unoccupied surface devoid of artefacts (e.g., Johansen and Stapert, 2001). This would also apply to shallowly buried artefactsdperhaps up to 5 cm deep (Stiner et al., 1995)dthough the depth at which temperatures high enough to visibly alter bone and stone can penetrate is dependent, again, on the size, intensity and duration of the fire, as well as the composition and moisture content of the underlying substrate (Bellomo, 1990; Campbell et al., 1995; Bennett, 1999), with even very

124


shallowly buried artefacts remaining largely unaffected by small, short-term, low-intensity fires. Lower geogenic sedimentation rates, coupled perhaps with more intensive fire use and redundant hearth placement in the firerich layers at RdM and Pech IV, left extant artefact scatters directly below the hearths exposed to repeated episodes of heating, thus increasing the frequency of fire proxies (i.e., the palimpsest effect, see Henry, 2012). This effect was possibly amplified by the proximity of these early occupation layers to the cave floors, which may have acted as a reflective surface allowing for more intense heating of the overlying sediments and artefacts after any excess moisture is evaporated away. The smaller available occupation surface area within RdM likely restricted hearth placement to some degree, which in turn increased the chances of later hearths being placed over extant knapping scatters produced near the previous hearth(s). Thus, the smaller size of RdM compared to Pech IV could explain the higher frequencies of fire proxies recovered from the hearth layers at RdM despite possibly greater site use intensity at Pech IV, as suggested by lateral blurring of hearth features (Turq et al., 2011; Aldeias et al., 2012; Goldberg et al., 2012). Increased rates of geogenic sedimentation during colder periods, especially if coupled with reduced frequencies in site use, may have allowed for more vertical separation of individual occupation layers. This additional thermal buffer likely reduced the chances for incidental heating of extant artefacts (e.g., Quebrada Cave: Eixea Vilanova, 2015). The parallel reduction of heated lithics and lithic artefacts densities as conditions deteriorated at RdM could be indicative of this trend (Fig. 7 in Sandgathe et al., 2011a), while the increase in faunal bone densities could reflect minimal site maintenance (i.e., removing or burning bone), better preservation due to colder conditions, or perhaps inflated numbers of bone fragments from more intensive marrow processing (Hodgkins, 2012; Hodgkins et al., 2016). The relationship between fire proxies and artefact densities is less clear at Pech IV, and may reflect a more complex depositional history. Concerning the possibility of differential preservation of fire traces between the lower and upper layers at Pech IV and RdM, the authors assert that “preservation was not a significant factor” (Sandgathe et al., 2011a, p. 224). However, the fact that combustion features are lacking in layers containing notable amounts of fire proxies, by definition, seems to suggest otherwise. The excellent preservation of combustion features (and residues) in layer 8 at Pech IV has been attributed to large roof fall fragments (associated with layer 6B) effectively capping the layer and acting as a chemical buffer, while calcite dissolution and precipitation cycles may have degraded ash deposits in the upper levels (Dibble et al., 2009; Turq et al., 2011). The proximity of the hearth layers at RdM to the underlying bedrock may have acted as a similar buffer. Moreover, rapid localised cementing of the upper portions of ash deposits in layer 8 at Pech IV, as well as in layers 9 and 7 at RdM (here also the formation of mm-thick phosphatic weathering crusts), may have acted as protective caps against the winnowing away of ash and fine charcoal fragments by wind or water runoff and prevented further diagenetic alteration (Turq et al., 2011; Aldeias et al., 2012). Furthermore, the upper layers at RdM and Pech IV effectively shielded the lower horizons from the adverse climatic conditions encountered from MIS 4 onward. Varying degrees of cryoturbation are reported in nearly all layers above layer 8 at Pech IV (Bordes, 1975; Goldberg et al., 2012; Turq et al., 2011), with evidence of solifluction in layer 7 and possibly layer 5B (Turq et al., 2011). At RdM, evidence of ice segregation, platy freeze-thaw structures and ice wedges were observed throughout the sequence into the upper portion of layer 8, but no deeper, while solifluction affected the upper layers outside the cave entrance (Couchoud, 2003; Goldberg, in Sandgathe et al., 2007).

2.4.4. Excavation methods and sampling As suggested by Sandgathe et al., it seems unlikely that obvious hearth features were simply ‘missed’ during excavations at Pech IV and RdM, and the idea that combustion features might remain undiscovered within the unexcavated portions of the sites, while possible, is also unlikely to change the overall trend. However, it appears obvious, based on the evidence provided above, that finegrained excavation, analytical and sampling methods are necessary to aid in the identification and characterization of poorly expressed episodes of Neandertal fire use. Studies have shown that the collection and incorporation of smaller artefact classes (generally 50 mm (0%). Mechanical fragmentation and/or dissolution also affect priry-Parisot, mary fire residues like charcoal, ash or phytoliths (The 2001). Finer fuels (branches, shrubs, grasses) are much less likely to produce robust charcoal fragments that can better withstand these processesdespecially in more basic environments like limestone caves (Cohen-Ofri et al., 2006)dresulting in a scatter of charcoal composed only of microscopic fragments that would potentially go unnoticed without special collection and analytical methods (Cui et al., 2009; Marquer, 2009; Marquer et al., 2010, 2012). Micromorphology is a powerful tool for providing very detailed data on site formation processes and archaeological features (e.g., Albert et al., 2012; Mallol et al., 2013; Mentzer, 2014; Mallol and Mentzer, 2015), and was utilised at RdM and Pech IV (Aldeias et al., 2012; Couchoud, 2003; Goldberg et al., 2012). However, the localised nature of these samples and the propensity for researchers to sample locations where (fire) features are plainly visible is less than helpful when one wishes to know more about possible fire use in layers where burning is not evident. 2.4.5. A case for Neandertal fire making? Based on their interpretation of the fire records at RdM and Pech IV, Sandgathe et al. ultimately point to the Neandertal lack of fire


making technology and reliance on climatically-mediated natural fires as the salient force behind the reduced fire signals observed in the upper layers at these sites (and elsewhere), their logic being that if Neandertals could make fire at will, fire signals should remain consistent between cold and warm intervals. However, in light of the evidence presented above, this may not necessarily be the case. In fact, within the framework of fuel economisation during colder periods, one could hypothesise (perhaps counterintuitively) that possessing the ability to make fire could actually reduce the overall signal of fire use in the archaeological record by allowing fires to be kindled only when they are needed, as opposed to being kept burning constantly so as not to lose one's flameda potentially very labour-intensive task during periods when robust, highquality fuel sources are in short supply. These smaller, short-term, perhaps task-specific fires would likely have been carefully tended to ensure near complete combustion of all available fuel. Since flaming fires are fuel-expensive (Braadbaart et al., 2012), fires of longer duration were probably often maintained as less costly glowing fires that tend to burn at lower ambient temperatures (Rein, 2009), making them less likely to significantly alter the underlying substrate or any shallowly buried artefacts that might be present (Sergant et al., 2006). Moreover, and arguably most importantly, there exists direct and indirect evidence for fire making by Neandertals during the Last Glacial. This includes a probable flint strike-a-light from Bettencourt-Saint-Ouen (75e85 kya) in northern France (Rots et al., 2011; Sorensen and Rots, 2014; Rots, 2015), as well as several Mousterian of Acheulean Tradition (MTA) bifacial tools from ChezPinaud (Jonzac) in south-west France exhibiting microwear traces comparable to those produced on experimental bifaces used in conjunction with pyriteda requisite component of the stone-onstone fire making methoddto make fire (Sorensen and Claud, 2016). The potential for fire making can be inferred from the recovery of allochthonous pyrite nodules and fragments from Mousterian layers at multiple sites (Weiner and Floss, 2004; Sorensen et al., 2014), including a nodule from layer 4 at RdM (Turq, 2016). Moreover, the recurring presence of manganese dioxide (MnO2) at MP sites (see Demars, 1992)despecially abraded I (Soressi et al., fragments like those recovered from Pech de l’Aze 2008)dmay be additional evidence for fire making, as it has recently been shown that this mineral, when powdered and mixed with tinder, lowers the temperature required for ignition by nearly 100 C (Heyes et al., 2016), significantly increasing the efficacy of the stone-on-stone fire production method. 3. Conclusion The pyrotechnical prowess of Neandertals has long been the subject of lively scientific debate. While still sparse in the MP, artefactual evidence of fire making (outlined in the section above) is slowly becoming more prevalent, lending credence to the idea that Neandertals coulddand indeed, likely diddmake fire at least by the Last Glacial. A large body of evidence suggesting regular fire use by Neandertals during this perioddas well as by much earlier Neandertals and their contemporaries (Karkanas et al., 2007; Shimelmitz et al., 2014)dadds further support (Roebroeks and Villa, 2011b, a). However, based on fire residues and proxy data from RdM and Pech IV, Sandgathe et al. (2011a, 2011b) point to weak fire signals (relatively speaking) in archaeological layers deposited during cold periodsdpresumably coinciding with reduced lightning activitydas evidence suggesting Neandertals were reliant on natural sources of fire and did not make it for themselves. The current study has tested the veracity of this environmentally determined fire use model at multiple scales (see Table 1).

125

While signals for fire use do appear suppressed in the layers presenting colder climatic signals at RdM and Pech IV (especially those associated with Quina Mousterian lithic techno-complexes), the apparent chronological offset between the more fire-rich lower layers at these sites, as indicated by chronometric dating and disparities between faunal signals and fire traces, calls into question the relationship between climate and fire use (see Section 2.3.). Furthermore, the faunal assemblages in the fire-rich layers at RdM indicate more temperate conditions despite the most recent dates rin et al., 2012, in for these layers placing them mid-MIS 4 (Gue press), generally considered to be quite cold despite pollen records suggesting an expansion of boreal forests in the region during ~ i et al., 2008). At Pech IV, fire signals drop this period (Sanchez-Gon drastically already prior to the onset of colder conditions, as indicated primarily by the increased prevalence of reindeer remains onsite. Moreover, the trends observed at RdM and Pech IV conflict with that seen at the nearby and coeval site of Combe Grenal, where strong evidence of fire use is noted in the majority of the cold stage Quina layers (see Section 2.2.). The desire to understand the underlying phenomena behind this disparity prompted a reevaluation and reinterpretation of the fire evidence from these sites within a broader environmental context. The similar morphology and geographical positions of RdM and Pech IV (i.e., relatively small caves, located on uplands around 1.7 km from major waterways) may make them comparable to one anotherdperhaps as short-term stops during annual migrationsdbut not necessarily representative of the region as a whole, as demonstrated by Combe Grenal, which may have been more suitable as a longer-term residential site (i.e., a larger cave/rock shelter located lower in the landscape, much closer to a major water sourcedthe Dordogne River located ~300 m to the northwestdand presumably nearer stands of trees), thus making it more likely to exhibit stronger fire signals during colder periods. Generally speaking, a reduction in fire signal during cold climatic episodes does not necessarily indicate a climaticallymediated reduction in fire use; instead, shifts in fauna (i.e., to more mobile reindeer, see Section 2.4.1.) and vegetation (i.e., reduced prevalence of wood fuel, see Section 2.4.2.) triggered cultural responses that resulted in a fundamental change in how and when fire was used, the practice likely becoming more ephemeral but no less important or regular in its use. Moreover, a reduction in fire signal does not indicate Neandertals lacked or lost the ability to make fire. Artificial fire making may have actually allowed for greater fuel economisation in cold periods by shifting from a strategy of long-term fire maintenance afforded by abundant and readily accessible wood fuel to one focused more on short-term, task-specific fires. These low-intensity fires, when coupled with reduced site use frequencies and increased geogenic sedimentation rates, would have resulted in weakened cold stage fire signals that were potentially exacerbated by less favourable preservation conditions (see Section 2.4.3.), ultimately requiring more fine-grained excavation and sampling methods to extract (see Section 2.4.4.). Even if one accepts the premise that Neandertals did not make fire, basic modelling and comparison with marine microcharcoal records suggest it is unlikely that a reduction in lightning frequency and regional fire regime during the Lower Pleniglacial in southwest France would have significantly limited Neandertal access to fire to the point that it virtually disappears from the archaeological record (see Section 2.1). These findings suggest that while climate change may have ultimately been the underlying force behind the variability in fire use and resultant fire signals between colder and warmer periods at RdM and Pech IV, this relationship does not appear to be the direct cause-and-effect scenario of less lightning means less fire posited by Sandgathe et al. Instead, a highly complex interplay between wider environmental and cultural trends and

126


more localised factors together influenced how and when Neandertals used fire, and whether or not evidence of its use would be preserved. Acknowledgements Funding was provided by the Netherlands Organisation for Scientific Research (Grant# PGW-13-42). Many thanks to AnneLaure Daniau for sharing her microcharcoal dataset and for her very helpful instruction. Thank you to David Romps, Graeme Anderson and Martin Uman for taking the time to answer my questions about lightning. I am also grateful for the insightful discussions held with various members of the Leiden Human Origins Group and to Wil Roebroeks, Alexander Verpoorte, Alain Turq and two anonymous reviewers for their constructive comments. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.quaint.2016.10.003. References Abrams, G., Cattelain, P., 2014. Le Trou de l’Abîme a Couvin (Prov. Namur, Belgique): ozoologique pre liminaire de la couche II (fouilles 1984-1987). Bilan arche o-Situla 34, 33e40. Arche Albert, R.M., Berna, F., Goldberg, P., 2012. Insights on Neanderthal fire use at Kebara Cave (Israel) through high resolution study of Prehistoric combustion features: evidence from phytoliths and thin sections. Quat. Int. 247, 278e293. Aldeias, V., Goldberg, P., Sandgathe, D., Berna, F., Dibble, H.L., McPherron, S.P., Turq, A., Rezek, Z., 2012. Evidence for Neandertal use of fire at Roc de Marsal (France). J. Archaeol. Sci. 39, 2414e2423. Alperson-Afil, N., Goren-Inbar, N., 2010. The Acheulian Site of Gesher Benot Ya'aqov, Volume II: Ancient Flames and Controlled Use of Fire. Springer, London. Alperson-Afil, N., Richter, D., Goren-Inbar, N., 2007. Phantom hearths and controlled use of fire at Gesher Benot Ya'Aqov, Israel. PaleoAnthropology 2007, 1e15. Annan, J., Hargreaves, J., 2013. A new global reconstruction of temperature changes at the Last Glacial Maximum. Clim. Past 9, 367e376. Archibald, S., Roy, D.P., Wilgen, V., Brian, W., Scholes, R.J., 2009. What limits fire? An examination of drivers of burnt area in Southern Africa. Glob. Change Biol. 15, 613e630. Archibald, S., Staver, A.C., Levin, S.A., 2012. Evolution of human-driven fire regimes in Africa. Proc. Natl. Acad. Sci. U. S. A. 109, 847e852. Bellomo, R.V., 1990. Methods for Documenting Unequivocal Evidence of Humanly Controlled Fire at Early Pleistocene Archaelogical Sites in East Africa: the Role of Actualistic Studies. Ph.D. Thesis. University of Wisconsin-Milwaukee, WI. Bellomo, R.V., 1993. A Methodological approach for identifying archaeological evidence of fire resulting from human activities. J. Archaeol. Sci. 20, 525e553. Bennett, J.L., 1999. Thermal alteration of buried bone. J. Archaeol. Sci. 26, 1e8. Binford, L.R., 1978. Nunamiut Ethnoarchaeology. Academic Press, New York, NY. Binford, L.R., 2001. Constructing Frames of Reference: an Analytical Method for Archaeological Theory Building Using Hunter-Gatherer and Environmental Data Sets. University of California Press, Berkeley and Los Angeles, CA. Binford, L.R., 2007. The Diet of Early Hominins: Some Things We Need to Know before "Reading" the Menu from the Archaeological Record. In: Roebroeks, W. (Ed.), Guts and Brains. An Integrative Approach to the Hominin Record. Leiden University Press, Leiden, pp. 185e222. Blasco, R., Rosell, J., Sanudo, P., Gopher, A., Barkai, R., 2016. What happens around a fire: faunal processing sequences and spatial distribution at Qesem Cave (300 ka). Isr. Quat. Int. 398, 190e209. Bond, W.J., Keeley, J.E., 2005. Fire as a global ‘herbivore’: the ecology and evolution of flammable ecosystems. Trends Ecol. Evol. 20, 387e394. Bond, W.J., Van Wilgen, B.W., 1996. Fire and Plants. Chapman & Hall, London. Bordes, F., 1955. La stratigraphie de la Grotte de Combe-Grenal, commune de liminaire. Bull. la Socie te pre historique française Domme (Dordogne): note pre 52, 426e429. Bordes, F., 1972. A Tale of Two Caves. Harper and Row, New York, NY. IV: note pre liminaire. Bull. la Socie te Bordes, F., 1975. Le gisement du Pech de l'Aze historique française 72, 293e308. pre couverte d'un squelette d'enfant mouste rien dans le Bordes, F., Lafille, J., 1962. De gisement de Roc de Marsal, commune de Campagne-du-Bugue (Dordogne). Comptes Rendus de l'Academie des Sciences. Paris 254, 714e715. Bowman, S.G.E., Sieveking, G.D.G., 1983. Thermoluminescence dating of burnt flint from Combe Grenal. PACT 9, 253e268. Braadbaart, F., Poole, I., Huisman, H.D.J., van Os, B., 2012. Fuel, Fire and Heat: an experimental approach to highlight the potential of studying ash and char remains from archaeological contexts. J. Archaeol. Sci. 39, 836e847. Campbell, G., Jungbauer Jr., J., Bristow, K.L., Hungerford, R.D., 1995. Soil temperature

and water content beneath a surface fire. Soil Sci. 159, 363e374. Lagunilla, A.H., s, J.M., 2016. Three archaeomagnetic applicaCarrancho, A., Verge tions of archaeological interest to the study of burnt anthropogenic cave sediments. Quat. Int. 414. Villalaín, J.J., 2011. Different mechanisms of magnetisation recorded Carrancho, A., in experimental fires: archaeomagnetic implications. Earth Planet. Sci. Lett. 312, 176e187. Castel, J.-C., Discamps, E., Soulier, M.-C., Sandgathe, D., Dibble, H.L., McPherron, S.J.P., Goldberg, P., Turq, A., 2016. Neandertal subsistence strategies during the Quina Mousterian at Roc de Marsal (France). Quat. Int. (in press). Cecil, D.J., Buechler, D.E., Blakeslee, R.J., 2014. Gridded lightning climatology from TRMM-LIS and OTD: dataset description. Atmos. Res. 135, 404e414. Chase, P.G., 1989. La Quina Bone Fragmentation Project: Report of Research (Unpublished manuscript). Cheney, P., Sullivan, A., 2008. Grassfires: Fuel, Weather and Fire Behaviour. CSIRO Publishing, Collingwood, Australia. CIGRE WG C4.407, 2013. Lightning parameters for Engineering applications. Tech. Broch. 549. Clark, J.D., Harris, J.W.K., 1985. Fire and its roles in early hominid lifeways. Afr. Archaeol. Rev. 3, 3e27. Cohen-Ofri, I., Weiner, L., Boaretto, E., Mintz, G., Weiner, S., 2006. Modern and fossil charcoal: aspects of structure and diagenesis. J. Archaeol. Sci. 33, 428e439. Costamagno, S., Griggo, C., Mourre, V., 1999. Approche experimentale d'un me taphonomique: utilisation de combustible osseux au Pale olithique. proble histoire Eur. 13, 167e194. Pre ologiques de formation du site mouste rien du Roc Couchoud, I., 2003. Processus ge o 51e68. de Marsal (Dordogne, France). Pale Court, A., 1960. Lightning Fire Incidence in Northeastern California, 1945-1956. Technical Report No. 47. U.S. Department of Agriculture, Forest Service, Pacific Southwest Forest and Range Experient Station, Berkeley, CA. Courty, M.-A., Carbonell, E., Vallverdú Poch, J., Banerjee, R., 2012. Microstratigraphic and multi-analytical evidence for advanced Neanderthal pyrotechnology at abric Romaní (Capellades, Spain). Quat. Int. 247, 294e312. Cui, Q., Marquer, L., Arzarello, M., Lebreton, V., 2009. An attempt to separate anthropic and natural fire signals in an archaeological context-The case of the Mousterian site Grotta Reali (Rocchetta a Volturno Molise, Central Italy). Front. Earth Sci. China 3, 171e174. nchez Gon ~ i, M.F., 2010a. Testing the hypothesis of fire Daniau, A.-L., D'Errico, F., Sa use for ecosystem management by Neanderthal and Upper Palaeolithic Modern Human populations. PLoS ONE 5, e9157. ~ i, M.F.S., Duprat, J., 2009. Last glacial fire regime variability in Daniau, A.-L., Gon western France inferred from microcharcoal preserved in core MD04-2845, Bay of Biscay. Quat. Res. 71, 385e396. Daniau, A.L., Harrison, S.P., Bartlein, P.J., 2010b. Fire regimes during the last glacial. Quat. Sci. Rev. 29, 2918e2930. anderthaliens dans le Daujeard, C., 2008. Exploitation du milieu animal par les Ne Lumie re - Lyon 2, Lyon. Sud-Est de la France. Ph.D. Thesis. Universite Daujeard, C., Moncel, M.-H., 2010. On Neanderthal subsistence strategies and land use: a regional focus on the Rhone Valley area in southeastern France. J. Anthropol. Archaeol. 29, 368e391. de Lumley, H., 2006. Il y a 400 000 ans : la domestication du feu, un formidable vol 5, 149e154. moteur d'hominisation. C.R. Pale rien du Pe rigord. L’apport des Demars, P.Y., 1992. Les colorants dans le Mouste fouilles de F. Bordes. Bull. Soc. Prehist. française 47, 185e194. Dibble, H.L., Berna, F., Goldberg, P., McPherron, S.P., Mentzer, S., Niven, L., Richter, D., ry-Parisot, I., Turq, A., 2009. A Preliminary Report on Pech de Sandgathe, D., The IV, Layer 8 (Middle Paleolithic, France). PaleoAnthropology 2009, 182e219. l'Aze Discamps, E., Jaubert, J., Bachellerie, F., 2011. Human choices and environmental constraints: deciphering the variability of large game procurement from Mousterian to Aurignacian times (MIS 5-3) in southwestern France. Quat. Sci. Rev. 30, 2755e2775. n tecnolo gica y uso del espacio en los yaciEixea Vilanova, A., 2015. Caracterizacio n central del Mediterra neo Ibe rico. mientos del Paleolítico medio de la regio Ph.D. Thesis. University of Valencia, Valencia. conomique des syste mes de production Faivre, J.-P., 2011. Organisation techno-e olithique moyen re cent du nord-est aquitain. BAR International dans le Pale Series 2280. Archaeopress, Oxford. Faivre, J.-P., Discamps, E., Gravina, B., Turq, A., Guadelli, J.-L., Lenoir, M., 2014. The contribution of lithic production systems to the interpretation of Mousterian industrial variability in south-western France: the example of Combe-Grenal (Dordogne, France). Quat. Int. 350, 227e240. Farukh, M.A., Hayasaka, H., 2012. Active forest fire occurrences in severe lightning years in Alaska. J. Nat. Disaster Sci. 33, 71e84. ~ udo, P., Fern andez Peris, J., Gonz alez, V.B., Blasco, R., Cuartero, F., Fluck, H., San Verdasco, C., 2012. The earliest evidence of hearths in southern europe: the case of Bolomor cave (Valencia, Spain). Quat. Int. 247, 267e277. Flannigan, M., Wotton, B., 1991. Lightning-ignited forest fires in northwestern Ontario. Can. J. For. Res. 21, 277e287. Frazer, S.J.G., 1930. Myths of the Origin of Fire: an Essay. Macmillan, London. Gentles, D., Smithson, P., 1986. Fires in caves: effects on temperature and airflow. Proc. Univ. Bristol Spelaeol. Soc. 17, 205e217. Goldberg, P., Dibble, H., Berna, F., Sandgathe, D., McPherron, S.J.P., Turq, A., 2012. New evidence on Neandertal use of fire: examples from Roc de Marsal and Pech IV. Quat. Int. 247, 325e340. de l’Aze Goren-Inbar, N., Alperson, N., Kislev, M.E., Simchoni, O., Melamed, Y., Ben-Nun, A.,

A.C. Sorensen / Quaternary International 436 (2017) 114e128 Werker, E., 2004. Evidence of Hominin control of fire at Gesher Benot Ya'aqov. Israel. Sci. 304, 725e727. Goudsblom, J., 1986. The human monopoly on the use of fire: its origins and conditions. Hum. Evol. 1, 517e523. Goudsblom, J., 1992. Fire and Civilization. Viking, London. Gowlett, J., Hallos, J., Hounsell, S., Brant, V., Debenham, N., 2005. Beeches pit: archaeology, assemblage dynamics and early fire history of a Middle Pleistocene site in east anglia, UK. Eurasian Prehistory 3, 3e38. ses Gowlett, J.A.J., 2015. Les origines de l’utilisation du feu par les hommes: hypothe actuelles et indices les plus anciens. In: de Lumley, H. (Ed.), Sur le chemin de . Via humanitatis: les grandes e tapes de l’e volution morphologique et l’hunanite mergence de l’e ^tre humain. Colloque international de culturelle de l’Homme: e du Vatican avril 2013. Acade mie l’Academie Pontificale des Sciences, Cite Pontificale des Sciences/CNRS, Paris, pp. 171e197. Gowlett, J.A.J., Wrangham, R.W., 2013. Earliest fire in Africa: towards the convergence of archaeological evidence and the cooking hypothesis. Azania Archaeol. Res. Afr. 48, 5e30. Grünberg, J.M., 2002. Middle Palaeolithic birch-bark pitch. Antiquity 15e16. l’e tude des zoocoenoses pre historiques en Guadelli, J.-L., 1987. Contribution a Aquitaine (Würm Ancien et Interstade Würmien). Ph.D. Thesis. Universite Bordeaux I, Bordeaux. rin, G., Discamps, E., Lahaye, C., Mercier, N., Guibert, P., Turq, A., Dibble, H.L., Gue McPherron, S.P., Sandgathe, D., Goldberg, P., Jain, M., Thomsen, K., PatouMathis, M., Castel, J.-C., Soulier, M.-C., 2012. Multi-method (TL and OSL), multimaterial (quartz and flint) dating of the Mousterian site of Roc de Marsal (Dordogne, France): correlating Neanderthal occupations with the climatic variability of MIS 5e3. J. Archaeol. Sci. 39, 3071e3084. rin, G., Frouin, M., Tuquoi, J., Thomsen, K.J., Goldberg, P., Aldeias, V., Lahaye, C., Gue Mercier, N., Guibert, P., Jain, M., Sandgathe, D., McPherron, S.J.P., Turq, A., Dibble, H.L., 2016. The complementarity of luminescence dating methods illustrated on the Mousterian sequence of the Roc de Marsal: a series of reindeer-dominated, Quina Mousterian layers dated to MIS 3. Quat. Int. (in press). Hennenberg, K.J., Fischer, F., Kouadio, K., Goetze, D., Orthmann, B., Linsenmair, K.E., Jeltsch, F., Porembski, S., 2006. Phytomass and fire occurrence along forest National park, ivory coast. J. Trop. Ecol. 22, esavanna transects in the Comoe 303e311. Henry, A.G., Büdel, T., Bazin, P.-L., 2016. The Cost of Fire. Book of Abstracts, 6th Annual Meeting of the European Society for the Study of Human Evolution. PESHE 5, Madrid, p. 122. Henry, A.G., Brooks, A.S., Piperno, D.R., 2011. Microfossils in calculus demonstrate consumption of plants and cooked foods in Neanderthal diets (Shanidar III, Iraq; Spy I and II, Belgium). Proc. Natl. Acad. Sci. U. S. A. 108, 486e491. Henry, D., 2012. The palimpsest problem, hearth pattern analysis, and Middle Paleolithic site structure. Quat. Int. 247, 246e266. Henry, D.O., Belmaker, M., Bergin, S.M., 2016. The effect of terrain on Neanderthal ecology in the Levant. Quat. Int. (in press). risson, D., Locht, J.-L., Auguste, P., Tuffreau, A., 2013. Ne andertal et le feu au He olithique moyen ancien. Tour d’horizon des traces de son utilisation dans le Pale Nord de la France. L'anthropologie 117, 541e578. Heyes, P.J., Anastasakis, K., de Jong, W., van Hoesel, A., Roebroeks, W., Soressi, M., 2016. Selection and use of manganese dioxide by Neanderthals. Sci. Rep. 6, 22159. Hodgkins, J., 2012. Tracking Climate-driven Changes in Neandertal Subsistence Behaviors and Prey Mobility Patterns. Ph.D. Thesis. Arizona State University, AZ. Hodgkins, J., Marean, C.W., Turq, A., Sandgathe, D., McPherron, S.J.P., Dibble, H., 2016. Climate-mediated shifts in Neandertal subsistence behaviors at Pech de IV and Roc de Marsal (Dordogne Valley, France). J. Hum. Evol. 96, 1e18. l'Aze Horvath, H., 1995. Estimation of the average visibility in central Europe. Atmos. Environ. 29, 241e246. Jacobs, Z., Jankowski, N.R., Dibble, H.L., Goldberg, P., McPherron, S.J.P., Sandgathe, D., Soressi, M., 2016. The age of three Middle Palaeolithic sites: single-grain opti I, II and IV in cally stimulated luminescence chronologies for Pech de l'Aze France. J. Hum. Evol. 95, 80e103. James, S.R., 1989. Hominid use of fire in the Lower and Middle Pleistocene: a review of the evidence. Curr. Anthropol. 30, 1e26. Jelinek, A., 2013. Neandertal Lithic Industries at La Quina. University of Arizona Press, Tucson, AZ. Johansen, L., Stapert, D., 2001. Vuurmakers uit het Laat-Paleolithicum revisited. Paleo-aktueel 12, 15e19. €ris, O., 2001. Der spa €tmittelpal€ Jo aolithische fundplatz Buhlen (Grabungen 1966-69). Stratigraphie, Steinartefakte und Fauna des Oberen Fundplatzes. Dr. Rudolf Habelt GmbH, Bonn. Karkanas, P., Shahack-Gross, R., Ayalon, A., Bar-Matthews, M., Barkai, R., Frumkin, A., Gopher, A., Stiner, M.C., 2007. Evidence for habitual use of fire at the end of the Lower Paleolithic: site-formation processes at Qesem Cave, Israel. J. Hum. Evol. 53, 197e212. Keeley, J.E., Bond, W.J., Bradstock, R.A., Pausas, J.G., Rundel, P.W., 2011. Fire in Mediterranean Ecosystems: Ecology, Evolution and Management. Cambridge University Press, Cambridge. Klein, D.R., 1982. Fire, lichens, and caribou. J. Range Manag. 35, 390e395. Koller, J., Baumer, U., Mania, D., 2001. High-tech in the Middle palaeolithic: Neandertal-Manufactured pitch identified. Eur. J. Archaeol. 4, 385e397. Komarek, E.V., 1964. The Natural History of Lightning. In: Proceedings of the Third Annual Tall Timbers Fire Ecology Conference. Tall Timbers Research, Inc.,

127

Tallahassee, FL, pp. 139e183. Komarek, E.V., 1965. Fire Ecology: Grasslands and Man. In: Proceedings of the Fourth Annual Tall Timbers Fire Ecology Conference. Tall Timbers Research, Inc., Tallahassee, FL, pp. 169e220. Komarek, E.V., Komarek, B.B., Carlysle, C., 1973. The Ecology of Smoke Particulates and Charcoal Residues from Forest and Grassland Fires: a Preliminary Atlas. Tall Timbers Research Station, Tallahassee, FL, p. 75. Kourtz, P., Todd, B., 1992. Predicting the Daily Occurrence of Lightning-caused Forest Fires. Forestry Canada, Petawawa National Forestry Institute. Chalk River, Ontario, p. 18. Information Report PI-X-112. rigord. Ph.D. Thesis. Laquay, G., 1981. Recherches sur les faunes du Würm I en Pe de Bordeaux I, Bordeaux. Universite Latham, D., Williams, E., 2001. Lightning and forest fires. In: Johnson, E.A., Miyanishi, K. (Eds.), Forest fires: Behavior and Ecological Effects. Academic Press, San Diego, CA, pp. 375e418. Leesch, D., Bullinger, J., Cattin, M.-I., Müller, W., Plumettaz, N., 2010. Hearths and hearth-related activities in Magdalenian open-air sites: the case studies of veyres and Monruz (Switzerland) and their relevance to an underChampre standing of Upper Paleolithic site structure. In: Poltowicz-Bobak, M., Bobak, D. (Eds.), The Magdalenian in Central Europe. New Finds and Concepts. Instytut w, Rzeszo w, pp. 53e69. Archeologii. Universytetu Rzeszowskiego Rzeszo Lioubine, V.P., 1992. Human adaptation in the mountain environments of the caucasus during upper palaeolithic and Mesolithic. Preistoria Alp. 207e219. Mallol, C., Cabanes, D., Baena, J., 2010. Microstratigraphy and diagenesis at the upper Pleistocene site of Esquilleu cave (Cantabria, Spain). Quat. Int. 214, 70e81. ndez, C.M., Cabanes, D., Machado, J., Sistiaga, A., Pe rez, L., Galva n, B., Mallol, C., Herna 2013. Human actions performed on simple combustion structures: an experimental approach to the study of Middle Palaeolithic fire. Quat. Int. Mallol, C., Marlowe, F.W., Wood, B.M., Porter, C.C., 2007. Earth, wind, and fire: ethnoarchaeological signals of Hadza fires. J. Archaeol. Sci. 34, 2035e2052. Mallol, C., Mentzer, S.M., 2015. Contacts under the lens: perspectives on the role of microstratigraphy in archaeological research. Archaeol. Anthropol. Sci. 1e25. March, R.J., Lucquin, A., Joly, D., Ferreri, J.C., Muhieddine, M., 2014. Processes of formation and alteration of archaeological fire structures: complexity viewed in the light of experimental approaches. J. Archaeol. Method Theory 21, 1e45. la reconstitution Marquer, L., 2009. Apport des signaux polliniques et charbons a oenvironnements et des pale a la connaissance des relations homme-milieu au ne dans le bassin de la Loire moyenne (France). Ph.D. Thesis. Muse um Pleistoce National d'Histoire Naturelle, Paris. Marquer, L., Lebreton, V., Otto, T., Valladas, H., Haesaerts, P., Messager, E., an, S., 2012. Charcoal scarcity in Epigravettian settlements with Nuzhnyi, D., Pe mammoth bone dwellings: the taphonomic evidence from Mezhyrich (Ukraine). J. Archaeol. Sci. 39, 109e120. Marquer, L., Otto, T., Nespoulet, R., Chiotti, L., 2010. A new approach to study the fuel used in hearths by hunter-gatherers at the Upper Palaeolithic site of Abri Pataud (Dordogne, France). J. Archaeol. Sci. 2735e2746. Mazza, P.P.A., Martini, F., Sala, B., Magi, M., Colombini, M.P., Giachi, G., Landucci, F., Lemorini, C., Modugno, F., Ribechini, E., 2006. A new Palaeolithic discovery: tarhafted stone tools in a European Mid-Pleistocene bone-bearing bed. J. Archaeol. Sci. 33, 1310e1318. IV McPherron, S.P., Dibble, H.L., 1999. The lithic assemblages of Pech de l'Aze histoire Eur. 15, 9e43. (Dordogne, France). Pre rien du Brugas histoire Meignen, L., 1981. L'abri mouste a Vallabrix (Gard). Gall. Pre 24, 239e253. Mellars, P., 1996. The Neanderthal Legacy : an Archaeological Perspective from Western Europe. Princeton University Press, Princeton, NJ. Mentzer, S.M., 2014. Microarchaeological approaches to the identification and interpretation of combustion features in prehistoric archaeological sites. J. Archaeol. Method Theory 21, 616e668. Minnich, R.A., 1987. Fire behavior in southern California chaparral before fire control: the Mount Wilson burns at the turn of the century. Ann. Assoc. Am. Geogr. 77, 599e618. Minnich, R.A., Ramírez, J.S., Vizcaino, E.F., Chou, Y.-H., 2009. Lightning detection rates and wildland fire in the mountains of northern Baja California, Mexico. Atmosfera 6, 235e253. Modugno, F., Ribechini, E., Colombini, M.P., 2006. Chemical study of triterpenoid resinous materials in archaeological findings by means of direct exposure electron ionisation mass spectrometry and gas chromatography/mass spectrometry. Rapid Commun. Mass Spectrom. 20, 1787e1800. et al., , E., Bailon, S., Barshay-Szmidt, C., Be arez, P., Cre gut, E., Moncel, M.-H., Allue 2015. Evaluating the integrity of palaeoenvironmental and archaeological records in MIS 5 to 3 karst sequences from southeastern France. Quat. Int. 378, 22e39. Morin, E., 2008. Evidence for declines in human population densities during the early Upper Paleolithic in western Europe. Proc. Natl. Acad. Sci. U. S. A. 105, 48e53. Morin, E., Delagnes, A., Armand, D., Castel, J.-C., Hodgkins, J., 2014. Millennial-scale change in archaeofaunas and their implications for Mousterian lithic variability in southwest France. J. Anthropol. Archaeol. 36, 158e180. Nash, C., Johnson, E., 1996. Synoptic climatology of lightning-caused forest fires in subalpine and boreal forests. Can. J. For. Res. 26, 1859e1874. Niven, L., Steele, T.E., Rendu, W., Mallye, J.-B., McPherron, S.P., Soressi, M., Jaubert, J., Hublin, J.-J., 2012. Neandertal mobility and large-game hunting: the exploitation of reindeer during the Quina mousterian at Chez-pinaud Jonzac (CharenteMaritime, France). J. Hum. Evol. 63, 624e635.

128


North Greenland Ice Core Project Members, 2004. High-resolution record of northern hemisphere climate extending into the last interglacial period. Nature 431, 147e151. Ofek, H., 2001. Second Nature: Economic Origins of Human Evolution. Cambridge University Press, Cambridge. Parker, C.H., Keefe, E.R., Herzog, N.M., O'Connell, J.F., Hawkes, K., 2016. The pyrophilic primate hypothesis. Evol. Anthropol. 25, 54e63. Pruetz, J.D., LaDuke, T.C., 2010. Reaction to fire by savanna chimpanzees (Pan troglodytes verus) at Fongoli, Senegal: conceptualization of 'fire behavior' and the case for a chimpanzee model. Am. J. Phys. Anthropol. 141, 646e650. Rabelo, E.R.C., Veras, C.A.G., Carvalho, J., Alvarado, E.C., Sandberg, D.V., Santos, J.C.D., 2004. Log smoldering after an Amazonian deforestation fire. Atmos. Environ. 38, 203e211. Rakov, V.A., Uman, M.A., 2003. Lightning: Physics and Effects. Cambridge University Press, Cambridge. Rein, G., 2009. Smouldering combustion phenomena in science and technology. Int. Rev. Chem. Eng. 1, 3e18. Rendu, W., Costamagno, S., Meignen, L., Soulier, M.-C., 2012. Monospecific faunal spectra in Mousterian contexts: implications for social behavior. Quat. Int. 247, 50e58. pez-Jime nez, A., Lo pez-Martínez, M., Haber-Uriarte, M., Rhodes, S.E., Walker, M.J., Lo ndez-Jalvo, Y., Chazan, M., 2016. Fire in the Early Palaeolithic: evidence Ferna from burnt small mammal bones at Cueva Negra del Estrecho del Río Quípar, Murcia, Spain. J. Archaeol. Sci. Rep. 9, 427e436. Richter, D., Dibble, H., Goldberg, P., McPherron, S.P., Niven, L., Sandgathe, D., Talamo, S., Turq, A., 2013. The late Middle Palaeolithic in Southwest France: new IV. Quat. Int. 294, 160e167. TL dates for the sequence of Pech de l'Aze Roebroeks, W., 1988. From Find Scatters to Early Hominid Behaviour: A Study of de re (The Middle Palaeolithic Riverside Settlements at Maastricht-Belve Netherlands). Leiden University Pressr, Leiden. Roebroeks, W., Villa, P., 2011a. On the earliest evidence for habitual use of fire in Europe. Proc. Natl. Acad. Sci. U. S. A. 108, 5209e5214. Roebroeks, W., Villa, P., 2011b. Reply to Sandgathe et al.: Neandertal use of fire. Proc. Natl. Acad. Sci. U. S. A. 108, E299. Romps, D.M., Seeley, J.T., Vollaro, D., Molinari, J., 2014. Projected increase in lightning strikes in the United States due to global warming. Science 346, 851e854. Rots, V., 2015. Hafting and the interpretation of site function in the European Middle palaeolithic. In: Conard, J., Delagnes, A. (Eds.), Settlement Dynamics of the Middle Paleolithic and Middle Stone Age. Kerns Verlag, Tübingen, pp. 383e410. Rots, V., Van Peer, P., Vermeersch, P.M., 2011. Aspects of tool production, use, and hafting in Palaeolithic assemblages from Northeast Africa. J. Hum. Evol. 60, 637e664. , G., San-Miguel-Ayanz, J., Schulte, E., Schmuck, G., Camia, A., Strobl, P., Liberta Giovando, C., Boca, R., Sedano, F., Kempeneers, P., McInerney, D., Withmore, C., Santos de Oliveira, S., Rodrigues, M., Durrant, T., Corti, P., Oehler, F., Vilar, L., Amatulli, G., 2012. Comprehensive monitoring of wildfires in europe: the European forest fire information system (effis). In: Tiefenbacher, J. (Ed.), Approaches to Managing Disaster - Assessing Hazards, Emergencies and Disaster Impacts. InTech, pp. 87e105. ~ i, M.F., Landais, A., Fletcher, W.J., Naughton, F., Desprat, S., Duprat, J., Sanchez-Gon 2008. Contrasting impacts of DansgaardeOeschger events over a western European latitudinal transect modulated by orbital parameters. Quat. Sci. Rev. 27, 1136e1151. Sandgathe, D., Dibble, H., McPherron, S., Turq, A., Schwortz, S., 2008. Roc de Marsal (Campagne de Bugue, Dordogne). Rapport d’Operation pour l’Annee 2008 gional de l’Arche ologie, Bordeaux). (Report on file with the Service Re Sandgathe, D., Turq, A., Dibble, H., McPherron, S., Goldberg, P., 2007. Roc de Marsal e triannuelle (Campagne-de-Bugue, Dordogne). Rapport de Fouille Programme gional de l’Arche ologie, 2005e2007 (Report on file with the Service Re Bordeaux). Sandgathe, D.M., Dibble, H.L., Goldberg, P., McPherron, S.P., Turq, A., Niven, L., Hodgkins, J., 2011a. On the Role of Fire in Neandertal Adaptations in Western and Roc de Marsal, France. PaleoAnthroEurope: evidence from Pech de l'Aze pology 2011, 216e242. Sandgathe, D.M., Dibble, H.L., Goldberg, P., McPherron, S.P., Turq, A., Niven, L., Hodgkins, J., 2011b. Timing of the appearance of habitual fire use. Proc. Natl. Acad. Sci. U. S. A. 108, E298. Sannikov, S.N., Goldammer, J.G., 1996. Fire ecology of pine forests of northern Eurasia. In: Goldammer, J.G., Furyaev, V.V. (Eds.), Fire in Ecosystems of Boreal Eurasia. Kluwer, Dordrecht, pp. 151e167. , P., Perdaen, Y., 2006. The 'invisible' hearths: a contribution to Sergant, J., Crombe the discernment of Mesolithic non-structured surface hearths. J. Archaeol. Sci. 33, 999e1007. Shahack-Gross, R., Berna, F., Karkanas, P., Lemorini, C., Gopher, A., Barkai, R., 2014. Evidence for the repeated use of a central hearth at Middle Pleistocene (300 ky ago) Qesem Cave, Israel. J. Archaeol. Sci. 44, 12e21. Sharp, H.S., Sharp, K., 2015. Hunting Caribou: Subsistence Hunting along the Northern Edge of the Boreal Forest. University of Nebraska Press, Lincoln. Shimelmitz, R., Kuhn, S.L., Jelinek, A.J., Ronen, A., Clark, A.E., Weinstein-Evron, M., 2014. ‘Fire at will’: the emergence of habitual fire use 350,000 years ago. J. Hum. Evol. 77, 196e203. Sorensen, A., Claud, E., 2016. Neandertal utilisait-il des briquets en silex? In: Turq, A., Faivre, J.-P.H., Maureille, B., Lahaye, Ch., Bayle, P., (coord.) (Eds.), andertal a la loupe. Muse e National de Pre histoire, Les Eyzies, pp. 106e111. Ne

Sorensen, A., Roebroeks, W., van Gijn, A., 2014. Fire production in the deep past? The expedient strike-a-light model. J. Archaeol. Sci. 42, 476e486. Sorensen, A., Rots, V., 2014. Testing the ‘expedient Strike-a-light Model’: an Experimental Assessment Based on the First Identified Middle Palaeolithic Firemaker from Bettencourt (France). Union Internationale des Sciences historiques et Protohistoriques (UISPP) XVII, Burgos, Spain, pp. 1005e1006. Pre rien de tradition Acheule enne du Sud-Ouest de la Soressi, M., 2002. Le Mouste Bordeaux, Bordeaux. France. Ph.D. Thesis. Universite Soressi, M., Rendu, W., Texier, J.-P., Claud, E., Daulny, L., D'errico, F., et al., 2008. Pech I (Dordogne, France): nouveau regard sur un gisement mouste rien de de-l'Aze enne connu depuis le XIX sie cle. In: Jaubert, J., Bordes, J.-G., tradition acheule te s Pale olithiques d'un grand Sud-Ouest: nouveaux Ortega, I. (Eds.), Les socie thodes, nouveaux re sultats. Socie te Pre historique gisements, nouvelles me française, Paris, pp. 95e132. Stahlschmidt, M.C., Miller, C.E., Ligouis, B., Hambach, U., Goldberg, P., Berna, F., Richter, D., Urban, B., Serangeli, J., Conard, N.J., 2015. On the evidence for human € ningen. J. Hum. Evol. 89, 71e91. use and control of fire at Scho Stapert, D., 2007. Youngsters knapping flint near the campfire: an alternative view de re (The Netherlands). Archa €ol. Korresp. 37, of Site K at Maastricht-Belve 19e35. Stewart, J.R., 2005. The ecology and adaptation of Neanderthals during the nonanalogue environment of Oxygen Isotope Stage 3. Quat. Int. 137, 35e46. Stewart, O.C., 1956. Fire as the first great force employed by man. In: Thomas, W.L. (Ed.), Man's Role in Changing the Face of the Earth. University of Chicago Press, Chicago, IL, pp. 115e133. Stiner, M.C., Kuhn, S.L., Weiner, S., Bar-Yosef, O., 1995. Differential burning, recrystallization, and fragmentation of archaeological bone. J. Archaeol. Sci. 22, 223e237. Svensson, A., Andersen, K.K., Bigler, M., Clausen, H.B., Dahl-Jensen, D., Davies, S., Johnsen, S.J., Muscheler, R., Parrenin, F., Rasmussen, S.O., 2008. A 60 000 year Greenland stratigraphic ice core chronology. Clim. Past 4, 47e57. ry-Parisot, I., 2001. Economie des combustibles au Pale olithique. CNRS, Paris. The ry-Parisot, I., 2002. Fuel Management (Bone and Wood) During the Lower The Aurignacian in the Pataud Rock Shelter (Lower Palaeolithic, Les Eyzies de Tayac, Dordogne, France). Contribution of Experimentation. J. Archaeol. Sci. 29, 1415e1421. ry-Parisot, I., Costamagno, S., Brugal, J.-P., Fosse, P., Guilbert, R., 2005. The use of The bone as fuel during the Palaeolithic, experimental study of bone combustible properties. In: Mulville, J., Outram, A.K. (Eds.), The Zooarchaeology of Fats, Oils, Milk and Dairying. Proceedings of the 9th ICAZ Conferences (Durham, August 2002). Oxbow Books, Durham, pp. 50e59. volution du Mouste rien de type Quina au Roc de Marsal et en Turq, A., 1979. L'e rigord: modifications de l'e quilibre technique et typologique. Ecole des Pe tudes en sciences sociales, Toulous. Hautes e Turq, A., 2016. Neandertal et le feu. In: Turq, A., Faivre, J.-P.H., Maureille, B., andertal a la loupe. Muse e National de Lahaye, C.H., Bayle, P., (coord.) (Eds.), Ne histoire, Les Eyzies, pp. 98e101. Pre Turq, A., Dibble, H., Faivre, J.-P., Goldberg, P., McPherron, S.J.P., Sandgathe, D., 2008. rien du Pe rigord Noir: quoi de neuf? In: Jaubert, J., Bordes, J.-G., Le Mouste te s du Pale olithique dans un Grand Sud-Ouest de la Ortega, I. (Eds.), Les socie sultats, nouvelles me thodes. France: nouveaux gisements, nouveaux re moire XLVII de la Socie te Pre historique Française, pp. 83e93. Me Turq, A., Dibble, H.L., Goldberg, P., McPherron, S.P., Sandgathe, D., Jones, H., centes du Maddison, K., Maureille, B., Mentzer, S., Rink, J., 2011. Les fouilles re IV (Dordogne). Gall. Pre histoire 53, 1e58. Pech de l'Aze Uhl, C., Kauffman, J.B., 1990. Deforestation, fire susceptibility, and potential tree responses to fire in the eastern Amazon. Ecology 71, 437e449. , A., Bartrolí, R., Campeny, G., Carrancho, A., Vallverdú, J., Alonso, S., Bargallo sito, I., Fontanals, M., Gabucio, J., Go mez, B., Prats, J.M., San ~ udo, P., Sole , A., Expo Vilalta, J., Carbonell, E., 2012. Combustion structures of archaeological level O and mousterian activity areas with use of fire at the Abric Romaní rockshelter (NE Iberian Peninsula). Quat. Int. 247, 313e324. Villa, P., Bon, F., Castel, J.-C., 2002. Fuel, fire and fireplaces in the palaeolithic of western europe. Rev. Archaeol. 23, 33e42. Vogel, J.C., Waterbolk, H.T., 1967. Groningen radiocarbon dates VII. Radiocarbon 9, 107e155. Weiner, J., Floss, H., 2004. Eine Schwefelkiesknolle aus dem Aurignacien vom Vogelherd, Baden-Württemberg - Zu den Anf€ angen der Feuererzeugung im €ischen Pala €olithikum. Archaologische Inf. 27, 59e78. europa Williams, E.R., 1995. Meteorological aspects of thunderstorms. Handbook of atmospheric electrodynamics. In: Volland, H. (Ed.), Handbook of Atmospheric Electrodynamics. CRC Press, Boca Raton, FL, pp. 27e60. Woillez, M.-N., Levavasseur, G., Daniau, A.-L., Kageyama, M., Urrego, D., S anchez~ i, M.-F., Hanquiez, V., 2014. Impact of precession on the climate, vegetation Gon and fire activity in southern Africa during MIS 4. Clim. Past 10, 1165e1182. Wotton, B., Martell, D.L., 2005. A lightning fire occurrence model for Ontario. Can. J. Forest Res. 35, 1389e1401. Wragg Sykes, R.M., 2015. To see a world in a hafted tool: birch pitch composite technology, cognition and memory in Neanderthals. In: Coward, F., Hosfield, R., Pope, M., Wenban-Smith, F. (Eds.), Settlement, Society and Cognition in Human Evolution: Landscapes in the Mind. Cambridge University Press, Cambridge, pp. 117e137. Wrangham, R., 2009. Catching Fire: How Cooking Made Us Human. Basic Books, New York, NY.

Sandgathe et al. (2011a) arguments

Counterarguments (this paper)

Natural fire availability

Site function/use

Seasonality

Fuel availability/type

Taphonomy/Preservation

Excavation/Sampling methods

Fire making

p. 235 - "Lightning frequency is directly related to temperature and humidity and drops significantly in cold dry climatic conditions (Rakov and Uman 2003). Interestingly, the relationship between climate and lightning frequency matches what the data show is the case during the Mousterian, namely less fire in cold/dry periods and more during warm/humid ones."

p. 231 - "[T]he lithic and faunal data suggest that the presence or absence of fires is not a reflection of differences in the site use. The most parsimonious explanation in the face of the ethnographic data and what we see at Pech IV and Roc de Marsal is that both sites served generally as base or residential camps, and there is nothing to suggest that this function changed significantly throughout their occupational histories."

p. 231 - "As yet we cannot exclude the possibility that the other non-fire occupations were limited to summer months, but given the long temporal span of these deposits, it would seem unlikely that occupations took place only during warm months for many tens of millennia."

p. 235 - "[E]ven during extreme cold periods there were always some trees around, though likely restricted mainly to river valleys."

p. 224 - "[S]trong arguments can be made to show that preservation was not a significant factor."

p. 229 - "Realistically it is highly unlikely that fires could have been constructed in these remaining [unexcavated] areas without resulting in residues (even if only burned lithics) bleeding into the adjacent excavated areas."

p. 235 - "European Neandertals did not know how to make fire."

2.4.1. - This assessment does not consider potential changes in site use frequency or duration caused by increased mobility.

2.4.1. - Increased spring/summer seasonality at Pech IV in colder layers may reflect shift in site use patterns based on seasonal local availability of migratory reindeer prey.

2.1.1. - Lightning and fire ignition rates are very low in the Dordogne today despite prevailing warm and humid conditions. 2.1.2. - Marine microcharcoal concentrations, as a proxy for palaeofire regime, are only weakly to moderately lower during colder stadial events compared to warmer interstadials. 2.1.3. - Reductions in natural fire ignition rates were not so great to eliminate Neandetal fire use and cannot account for the major reductions in fire proxies at RdM and Pech IV. 2.2. - Strong evidence for cold stage fire use is present at a number of Neandertal sites in France dating to the Last Glacial, including Combe Grenal. Conversely, there are numerous instances of warm layers with very weak fire signals. 2.3. - Negative shifts in fire proxies can be seen either well before the onset of colder climatic conditions (Pech IV) or potentially well after (RdM).

2.4.1. - Roof collapse and/or sedimentary infilling may have caused changes in how a site is used, affecting fire use and hearth placement.

p. 235 - "[B]one can also be used as fuel."

p. 231 - "[C]onditions during even the warm months of MIS 4 were still far cooler than today’s and, again, modern hunter-gatherers in even substantially warmer environments still rely on fire to warm themselves, especially at night."

2.4.1. - Observing diachronic change in seasonality not currently possible at RdM, though layer 4 (Quina) gives a mixed signal. 2.4.1. - Strong spring/summer seasonality can reduce the frequency of site visits in a year and can affect hearth placement (i.e., outside of the cave/shelter).

p. 224 - "Further evidence that the lack of fires in the upper layers is not a result of preservation comes from indirect data that reflects the presence of fire even when more direct evidence of actual fire residues are missing."

Fig. 6, p. 227 - "[Artefact] counts are based on objects greater than 2.5cm in length and the flint includes only proximal and complete pieces (flakes, tools, and cores)."

2011b - "We would argue that, if Neandertals had the ability to make fire at will, then evidence for it should occur with much greater frequency in Middle Paleolithic sites and occupations and especially, those sites associated with such cold stages."

p. 227 - "At both sites, the percentage of burned objects, both lithic and fauna, agrees very closely with the frequency of more ephemeral, direct fire residues (charcoal and ash)."

2.4.2. - Trees likely not in the direct vicinity of upland sites like Pech IV and RdM during colder periods, so more effort would be needed to collect wood from valley bottoms, likely leading to increased fuel economization (i.e., smaller, more task-specific fires of shorter duration). 2.4.2. - Bones likely only used as primary fuel source if heat is more important than caloric intake, or to clear bone debris from a site. 2.4.2. - Bone is a weak producer of conductive heat, causing less thermal impact on a substrate. 2.4.2. - Increased fragmentation rate of burned bone generally results in pieces falling below the Sandgathe et al. minimum ≥2.5 cm sample size.

2.4.3. - Notable amounts of fire proxies present in layers without combustion features should be an argument for the influence of destructive post-depositional processes, not against. If taphonomic impacts were minimal, then intact combustion features (just fewer of them) should be present in all layers where fire proxies are present. This is not the case.

2.4.4. - It is indeed improbable that hearth features are only present in the unexcavated portions of the (cold) upper layers, but their statement assumes 1) hearth features would have been preserved, and 2) that noticeably higher amounts of fire proxies would have been produced.

2.4.3. - Lower (hearth) layers at Pech IV protected by rockfall, with resultant 'chemical buffer' helping to preserve ash.

2.4.4. - Fine-grained site-wide sampling necessary to identify microcharcoal and highly fragmented bone.

2.4.3. - Evidence for cryoturbation in upper layers and little to none in lower layers. 2.4. - Fire proxies are not only about preservation, but also conditions for creation: higher geogenic sedimentation rates and reduced site use frequencies (perhaps indicated by lower artefact densities), coupled with reduced frequency, intensity and duration of fires, decreased the chances of incidental heating of extant lithic and bone artefacts.

2.4.4. - The minimum artefact size of ≥2.5 cm for inclusion in the sample disregards the smaller artefact size classes where most fire affected artefacts fall.

2.4.4. - Micromorphology is helpful, but sampling is limited; one cannot easily direct sampling to invisible hearth locations, so there is more focus on overt hearth features.

2.4.5. - Possible fire making tools have been recovered from multiple Middle Palaeololithic sites dating to the Last Glacial. 2.4.5. - The ability to make fire at will may have reduced the fire signal by allowing for more economic, task-specific fires (i.e., there was no need to maintain fires for long periods when fuel is in short supply).

Lightning and fire ignition rates for the Dordogne, France Per 1 km2 Estimates based on modern conditions Flashes/km2/year Flashes in range/year 1:4 Flashes are CG 1% ignition rate 4% ignition rate 2,5% ignition rate (AVERAGE)

Dordogne average 0,6 0,6 0,15 0,0015 0,006 0,00375

0,8 0,8 0,2 0,002 0,008 0,005

1 1 0,25 0,0025 0,01 0,00625

2 2 0,5 0,005 0,02 0,0125

3 3 0,75 0,0075 0,03 0,01875

4 4 1 0,01 0,04 0,025

5 5 1,25 0,0125 0,05 0,03125

6 6 1,5 0,015 0,06 0,0375

7 7 1,75 0,0175 0,07 0,04375

8 8 2 0,02 0,08 0,05

7-17% reduction in flashes for every 1°C drop in GMAT* estimated 3-6°C drop during LGM = min 19,6-35,3% to max 42-67,3% fewer flashes 1% ignition rate with 3°C drop @ 7%/°C 4% ignition rate with 3°C drop @ 7%/°C 1% ignition rate with 3°C drop @ 17%/°C 4% ignition rate with 3°C drop @ 17%/°C 1% ignition rate with 6°C drop @ 7%/°C 4% ignition rate with 6°C drop @ 7%/°C 1% ignition rate with 6°C drop @ 17%/°C 4% ignition rate with 6°C drop @ 17%/°C 2,5% ignition rate with 4,5°C drop @ 12%/°C (AVERAGE)

0,001206

0,001608

0,00201

0,00402

0,00603

0,00804

0,01005

0,01206

0,01407

0,01608

0,004824

0,006432

0,00804

0,01608

0,02412

0,03216

0,0402

0,04824

0,05628

0,06432

0,0009705

0,001294

0,00647 0,008088 0,009705 0,011323

0,01294

0,003882

0,005176

0,00647

0,01294

0,01941

0,02588

0,03235

0,03882

0,04529

0,05176

0,00087

0,00116

0,00145

0,0029

0,00435

0,0058

0,00725

0,0087

0,01015

0,0116

0,00348

0,00464

0,0058

0,0116

0,0174

0,0232

0,029

0,0348

0,0406

0,0464

0,0004905

0,000654

0,00327 0,004088 0,004905 0,005723

0,00654

0,001962

0,002616

0,00327

0,01308

0,02289

0,02616

0,001635

0,00218

0,002725

0,01635 0,019075

0,0218

Fire acquisition distance of 7,5 km = 176,7 km2 Estimates based on modern conditions Flashes/km2/year 0,6 Flashes in range/year 106,02 1:4 Flashes are CG 26,505

0,8 141,36 35,34

1 176,7 44,175

0,0016175 0,003235 0,004853

0,0008175 0,001635 0,002453 0,00654

0,00981

0,00545 0,008175

2 353,4 88,35

3 530,1 132,525

0,01635

0,0109 0,013625

4 706,8 176,7

5 883,5 220,875

0,01962

6 1060,2 265,05

7 1236,9 309,225

8 1413,6 353,4

1% ignition rate 4% ignition rate 2,5% ignition rate (AVERAGE)

0,26505 1,0602 0,662625

0,3534 1,4136 0,8835

0,44175 1,767 1,104375

0,8835 1,32525 3,534 5,301 2,20875 3,313125

1,767 2,20875 7,068 8,835 4,4175 5,521875

2,6505 3,09225 10,602 12,369 6,62625 7,730625

3,534 14,136 8,835

7-17% reduction in flashes for every 1°C drop in GMAT* estimated 3-6°C drop during LGM = min 19,6-35,3% to max 42-67,3% fewer flashes 1% ignition rate with 3°C drop @ 7%/°C 4% ignition rate with 3°C drop @ 7%/°C 1% ignition rate with 3°C drop @ 17%/°C 4% ignition rate with 3°C drop @ 17%/°C 1% ignition rate with 6°C drop @ 7%/°C 4% ignition rate with 6°C drop @ 7%/°C 1% ignition rate with 6°C drop @ 17%/°C 4% ignition rate with 6°C drop @ 17%/°C 2,5% ignition rate with 4,5°C drop @ 12%/°C (AVERAGE)

0,2131002

0,2841336

0,355167 0,710334 1,065501 1,420668 1,775835 2,131002 2,486169 2,841336

0,8524008

1,1365344

1,420668 2,841336 4,262004 5,682672

0,17148735

0,2286498

0,28581225 0,571625 0,857437 1,143249 1,429061 1,714874 2,000686 2,286498

0,6859494

0,9145992

1,143249 2,286498 3,429747 4,572996 5,716245 6,859494 8,002743 9,145992

0,153729

0,204972

0,256215

0,614916

0,819888

1,02486

0,08667135

0,1155618

0,14445225 0,288905 0,433357 0,577809 0,722261 0,866714 1,011166 1,155618

0,3466854

0,4622472

0,577809 1,155618 1,733427 2,311236 2,889045 3,466854 4,044663 4,622472

0,2889045

0,385206

Fire acquisition distance of 15 km = 706,86 km2 Estimates based on modern conditions Flashes km-2 yr-1 0,6

0,8

1

565,488 141,372 1,41372 5,65488 3,5343

706,86 176,715 1,76715 7,0686 4,417875

-1

Flashes in range yr 1:4 Flashes are CG 1% ignition rate 4% ignition rate 2,5% ignition rate (AVERAGE)

424,116 106,029 1,06029 4,24116 2,650725

0,51243 0,768645

1,02486 1,281075

1,53729 1,793505

2,04972

2,04972

4,09944

6,14916

8,19888

3,07458

0,4815075 0,963015 1,444523

7-17% reduction in flashes for every 1°C drop in GMAT* estimated 3-6°C drop during LGM = min 19,6-35,3% to max 42-67,3% fewer flashes

7,10334 8,524008 9,944676 11,36534

2

3

1413,72 2120,58 353,43 530,145 3,5343 5,30145 14,1372 21,2058 8,83575 13,25363

5,1243

7,17402

1,92603 2,407538 2,889045 3,370553

4

5

6

3,85206

7

8

2827,44 3534,3 4241,16 4948,02 706,86 883,575 1060,29 1237,005 7,0686 8,83575 10,6029 12,37005 28,2744 35,343 42,4116 49,4802 17,6715 22,08938 26,50725 30,92513

5654,88 1413,72 14,1372 56,5488 35,343

1% ignition rate with 3°C drop @ 7%/°C 4% ignition rate with 3°C drop @ 7%/°C 1% ignition rate with 3°C drop @ 17%/°C 4% ignition rate with 3°C drop @ 17%/°C 1% ignition rate with 6°C drop @ 7%/°C 4% ignition rate with 6°C drop @ 7%/°C 1% ignition rate with 6°C drop @ 17%/°C 4% ignition rate with 6°C drop @ 17%/°C 2,5% ignition rate with 4,5°C drop @ 12%/°C (AVERAGE)

0,85247316

1,13663088

1,4207886 2,841577 4,262366 5,683154 7,103943 8,524732

3,40989264

4,54652352

5,6831544 11,36631 17,04946 22,73262 28,41577 34,09893 39,78208 45,46524

0,68600763

0,91467684

2,74403052

3,65870736

4,5733842 9,146768 13,72015 18,29354 22,86692 27,44031 32,01369 36,58707

0,6149682

0,8199576

1,024947 2,049894 3,074841 4,099788 5,124735 6,149682 7,174629 8,199576

2,4598728

3,2798304

4,099788 8,199576 12,29936 16,39915 20,49894 24,59873 28,69852

0,34671483

0,46228644

1,38685932

1,84914576

2,3114322 4,622864 6,934297 9,245729 11,55716 13,86859 16,18003 18,49146

1,1557161

1,5409548

1,9261935 3,852387 5,778581 7,704774 9,630968 11,55716 13,48335 15,40955

Fire acquisition distance of 50 km = 10000 km2 Estimates based on modern conditions Flashes/km2/year 0,6 Flashes in range/year 6000 1:4 Flashes are CG 1500 1% ignition rate 15 4% ignition rate 60 2,5% ignition rate (AVERAGE) 37,5

0,8 8000 2000 20 80 50

1,14334605 2,286692 3,430038 4,573384

0,57785805 1,155716 1,733574 2,311432

1 10000 2500 25 100 62,5

9,94552 11,36631

5,71673 6,860076 8,003422 9,146768

32,7983

2,88929 3,467148 4,045006 4,622864

2 20000 5000 50 200 125

3 30000 7500 75 300 187,5

4 40000 10000 100 400 250

5 50000 12500 125 500 312,5

6 60000 15000 150 600 375

7 70000 17500 175 700 437,5

8 80000 20000 200 800 500

7-17% reduction in flashes for every 1°C drop in GMAT* estimated 3-6°C drop during LGM = min 19,6-35,3% to max 42-67,3% fewer flashes 1% ignition rate with 3°C drop @ 7%/°C 4% ignition rate with 3°C drop @ 7%/°C 1% ignition rate with 3°C drop @ 17%/°C

12,06

16,08

20,1

40,2

60,3

80,4

100,5

120,6

140,7

160,8

48,24

64,32

80,4

160,8

241,2

321,6

402

482,4

562,8

643,2

9,705

12,94

16,175

32,35

48,525

64,7

80,875

97,05

113,225

129,4

4% ignition rate with 3°C drop @ 17%/°C 1% ignition rate with 6°C drop @ 7%/°C 4% ignition rate with 6°C drop @ 7%/°C 1% ignition rate with 6°C drop @ 17%/°C 4% ignition rate with 6°C drop @ 17%/°C 2,5% ignition rate with 4,5°C drop @ 12%/°C (AVERAGE)

38,82

51,76

64,7

129,4

194,1

258,8

323,5

388,2

452,9

517,6

8,7

11,6

14,5

29

43,5

58

72,5

87

101,5

116

34,8

46,4

58

116

174

232

290

348

406

464

4,905

6,54

8,175

16,35

24,525

32,7

40,875

49,05

57,225

65,4

19,62

26,16

32,7

65,4

98,1

130,8

163,5

196,2

228,9

261,6

16,35

21,8

27,25

54,5

81,75

109

136,25

163,5

190,75

218

*Change in Global Mean Average Temperature 7% drop

12% drop

17% drop

0 100 -1 93 -2 86,49 -3 80,4357 -4 74,805201 -5 69,56883693 -6 64,69901834

100 100 88 83 77,44 68,89 68,1472 57,1787 59,969536 47,458321 52,77319168 39,39040643 46,44040868 32,69403734

% change 3°C -19,5643 % change 6°C -35,30098166

-31,8528 -42,8213 -53,55959132 -67,3059627