Introduction

Charred remains of hazelnut shells and acorns are frequently recovered from prehistoric sites from the Mesolithic to the Bronze Age, often within pits (Cunningham 2011). There is abundant ethnographic, experimental, historical and archaeological evidence for the purposes and methods of hazelnut and acorn use, for human nourishment and other domestic activities (Kubiak-Martens 1999; Mason 1995; Mason 1996a; McComb and Simpson 1999; Zapata Peña 2000), the use of pits (e.g. Collis 1999) and the deposition of hazelnuts in pits (Carruthers 2000; Cunningham 2011). However, there is disagreement on the dietary significance of hazelnuts in the past (e.g. Holst 2010; McComb and Simpson 1999; Scaife 1992), the functions of pits in which these remains are deposited (DeBoer 1988; Field et al. 1964) and on the significance of hazelnut shells in certain pits, for example distinguishing between a pit-hearth and a pit with floor sweepings (Field et al. 1964) and roasting pits and storing pits (Mithen and Score 2000; Mithen et al. 2001; Cunningham 2011).

The difficulties in interpreting the importance of hazelnuts are partially due to quantification problems; many diverse procedures for quantifying hazelnuts from their charred shell fragments have been used and consensus has not yet been reached (see Berihuete Azorín and Antolín 2013; Lopez-Doriga 2013). Moreover, the abundance of hazelnuts cannot always be compared with that of other resources as they have a greater probability of being carbonised and recovered. Nutshells are often considered domestic waste and thus are likely to be thrown into fires, as fuel or as a way of disposing them (e.g. Scaife 1992; Jones 2000). Some recovery techniques might select and over-represent nutshells due to their high resistance to carbonisation and erosion (Scaife 1992) and their size, which means they are easily seen without flotation or screening with small-sized messes (Zapata Peña 2000). In addition, rather than intensive exploitation, sites with palimpsest-like stratigraphies might show evidence of incidental consumption of hazelnuts (Mithen et al. 2000). Despite these problems, the abundance and ubiquity of hazelnuts are undeniable and need to be better understood.

Although there may be many possible reasons to deposit hazelnuts in pits and considering that hazelnuts might well be staples or snacks, carbonised hazelnut shells in pits (or in other type of concentrations, besides those spread over occupation floors) are by-products of domestic activities, whether or not the deposition in the pit is primary, secondary or tertiary (Fuller et al. 2013). There are a few possibilities that may result in charred nutshells within the context of domestic activities—the hazelnut shells have been discarded into a fire after eating the kernel (Legge 1989), they have accidentally been carbonised while roasting (Zapata Peña 2000) or they have been burnt during storage from a fire.

Undetermined carbonised fragments of parenchymatic tissue, which could possibly be the remains of charred nut kernels, are often retrieved from the same deposits in which nutshells are recovered (e.g. Stevens 2009). Unfortunately, it is difficult to know the extent of this co-occurrence since plant remains that are left undetermined are not included in all site reports. Due to their high oil content, parenchyma does not always survive to carbonisation. When they do survive, they are easily eroded and are not often preserved in a recognisable state. Even when identification is possible, which requires high-power magnification, it is rare to get to species level. Thus, the detection of deposits containing charred whole nuts could easily go unnoticed, as only the shell fragments are recovered and identified.

The research presented here aims to establish criteria for distinguishing between these different domestic activities, the discard of nutshells into the fire after consumption of the kernel and the storing or roasting of hazelnuts for consumption. It is hypothesised that the different actions that turn hazelnuts into charred nutshell fragments leave distinct evidence on their shells. Secondly, it is hypothesised that by reproducing hazelnut processing and recording how nutshells are preserved, it is possible to infer a posteriori the actions that would lead to the deposition of the hazelnuts as charred nutshell fragments, specifically whether the hazelnut shells have been broken before charring (therefore they have been processed with the intention of consuming the kernel and the shells have been discarded into a fire) or have been carbonised unopened (breakage must have been post-depositional), either fresh (possibly falling into a fire while roasting) or dry (thus suggesting long term storage). An experimental collection of charred nutshell fragments prepared from differently processed hazelnuts has been created, and these fragments have been examined with a stereomicroscope in order to determine the critical identification criteria, to test these hypotheses.

An archaeological hazelnut sample from an Early Bronze Age (ca. 1850 cal. bc) deposit from Arangas Cave in Asturias, Northern Spain (Arias Cabal and Ontañón Peredo 1999) was examined in order to compare the modern experimental nutshells and ancient nutshell fragments. The criteria developed from the experimental studies were applied to the archaeological deposits in order to test the hypothesis, suggested during excavation, that this was a storage pit which was burnt.

Methodology

Experimentation

Two experiments were conducted in order to char whole and fragmented wild hazelnuts (Corylus avellana L.), which were collected from Northern Spain. The first was devised as a test experiment (López-Dóriga 2013) and the second as a more formal experiment with controlled variables, including charring temperature, heat exposure length and statistically significant sample numbers.

Test experiment

Half of the hazelnut samples were stone cracked and charred and the other half carbonised first and then cracked. As in previous experiments by other researchers, it was not attempted to control the burning environment, as the aim was to replicate the likely prehistoric conditions, rather than specific laboratory conditions (Church et al. 2007). Two test experiments were conducted. In the first experiment, whole hazelnuts were placed in a pit of 10 cm depth and 20 cm diameter and covered with soil. A wood fire was constructed and burnt above the deposit for 30 min. This method was intended to mimic what would happen if hazelnuts were burnt while stored. In the second experiment, whole hazelnuts and fragments of dry and fresh hazelnuts were placed on a dense layer of fire ash in proximity to fragments of wood ember. This method was intended to mimic what would happen if shell fragments were discarded into a fire or if whole nuts fell into the fire while cooking.

Controlled experiment

Five whole hazelnuts were placed in a muffle furnace at 300 °C for 90 min as an endurance test, with a control stop every 15 min to note the state of preservation of the nuts. This established a time lapse of 60 min as the minimum necessary for the hazelnuts to be carbonised. As there are no macroscopic differences in hazelnut shells charred in oxidising or anoxic atmospheric conditions (Berihuete Azorín and Antolín 2013), the atmosphere of the muffle furnace was always oxidising, with the hazelnuts wrapped in tinfoil. One hundred whole hazelnuts and the fragments of 100 stone-cracked hazelnuts, half of which were fresh and the other half dry, were left roasting in the oven for 60 min at around 300 °C. After charring, the whole hazelnuts were cracked by trampling on them.

Defining criteria

In the test experiment, a small number of hazelnut shell fragments of each of the three groups (fresh broken, dried broken and post-charring broken) were examined (López-Dóriga 2013). In the controlled experiment, the minimum number of statistically significant (n = 100) charred nutshell fragments for each of the three groups was selected with the help of a rifle box to avoid sampling biases. A Leica Z16 stereomicroscope was used to examine the experimental shell fragments at magnifications between ×10 and ×40. Differences in fracture patterns such as those that enable the identification of fresh and dry bones (Outram 2001) were assessed. A presence/absence analysis of fire alteration evidence in the inner edge of the fragments was undertaken, assuming that these would only occur when breakage is prior to fire exposure. All characteristics identified as fire alteration marks have been photographed and recorded using a Canon EOS450D photo camera connected to a Leica S8APO stereomicroscope.

Archaeological comparison

The hazelnut shell fragments (C. avellana L.) from the Early Bronze Age at Arangas Cave were recovered from a cylindrical pit, ca. 140 cm diameter and 60 cm depth, with straight walls covered by clay and stone blocks at the bottom. Other macrobotanical remains including charcoal, seeds and fragments of possible parenchymatic tissue, as well as fragments of pottery and bones, were also present in the pit. The hypothetical interpretation of the taphonomy suggested by excavators is that the pit was a clay-lined hazelnut storage structure that was accidentally burnt, the soil and other archaeological remains (bone and pottery) being either part of the capping of the pit or a post-abandonment filling. Investigation of fire alteration traces was conducted for two subsamples of hazelnut shell fragments extracted from 2 l of floated sediment. The first sample, considered as a preliminary test, was a selection of all fragments above 4 mm, sorted with the help of a 4-mm mesh. Because the size selection of the remains in the sample might produce a biased result, a further sample of 100 fragments was chosen using a rifle box. These are considered in the following section as size-biased and unbiased samples, respectively.

Results

Experimentation

The results of the endurance test in the controlled experiment are presented in Table 1:

Table 1 Observations of experimentally charred hazelnuts
Full size table

The most important find of the controlled experiment concerns the appearance of fire alteration traces, which confirms and expands on previous observations from the test experiment (López-Dóriga 2013), and can now be given a minimum statistical support. Ninety-five percent of the nutshell fragments which were exposed to fire after being broken show at least one of the following fire alteration marks: The outer skin of the pericarp is deformed and curled in itself, outlying the inner woody part of the pericarp (Fig. 1). Transversal fissures appear on the edge surfaces of the pericarp fragments (López-Dóriga 2013), more abundantly in the transversal edge surfaces (in 45 % of cases in fresh hazelnuts and in 55 % in dry ones) but as well in the longitudinal edges (40 % in both) (Fig. 2). The edge surfaces of the pericarp fragments are in general rough and discontinuous, sometimes with vesicles (30 % in longitudinal edges of fresh nutshell fragments and 20 % in dry ones, while only 5 % in both fresh and dry transversal edges) (Fig. 3).

Fig. 1
figure 1

Experimental charred hazelnut shell fragment, in which the outer skin of the pericarp is deformed and curled in itself, outlying the inner woody part of the pericarp (fire alteration mark)

Full size image
Fig. 2
figure 2

Experimental charred hazelnut shell fragment with a fissure on a transversal edge surface (fire alteration mark)

Full size image
Fig. 3
figure 3

Experimental charred hazelnut shell fragment with a bubbled edge surface (fire alteration mark)

Full size image

One hundred percent of the nutshell fragments that were broken after charring present a smooth and continuous edge surface (López-Dóriga 2013; Fig. 4), sometimes presenting negatives, showing that the breakage has occurred after the material has achieved a certain vitrified constitution (Fig. 5). However, 7 % of the fragments have fissures caused by trampling which, isolated, could be mistaken with heat fissures.

Fig. 4
figure 4

Experimental charred hazelnut shell fragment with a smooth and continuous edge surface

Full size image
Fig. 5
figure 5

Experimental charred hazelnut shell fragment with vitrified-type fracture

Full size image

The results from the controlled experimental charring confirm previous taphonomic experiments on fragmentation patterns and the degree of charring of hazelnuts in pits, which can be summarised as follows:

  1. 1.

    When the nutshell is charred, the inner kernel is already overly roasted to be palatable and as such is relatively inedible (Carruthers 2000; Kubiak-Martens 1999; Mithen and Score 2000).

  2. 2.

    Nutshell fragments between ash layers near a high temperature fire carbonise very quickly (López-Dóriga 2013; McComb and Simpson 1999).

  3. 3.

    Roasting hazelnuts in pits is effective and produces only a small percentage of charred inedible hazelnuts (López-Dóriga 2013; Mithen and Score 2000).

  4. 4.

    Uncharred hazelnuts break into halves and smaller fragments, while charred ones break into much smaller fragments and dust (López-Dóriga 2013; Mithen and Score 2000).

  5. 5.

    Impact marks upon charred shells left by stone cracking are difficult to identify (López-Dóriga 2013), as the small and superficial scratches that are left almost disappear while charring (Fig. 6).

    Fig. 6
    figure 6

    Experimental charred hazelnut shell fragment with impact marks

    Full size image

Archaeological assemblage

Size-biased sample

The first sample chosen from the archaeological assemblage consisted of all fragments greater than 4 mm and was examined taking into account only the existence of transversal fissures in the transversal edges (López-Dóriga 2013). The sample consisted of 778 fragments and comprised 30 % of the total weight of the assemblage. Of these fragments, 16 % were not considered useful for the analysis of fire alteration marks due to a combination of recent breakage and fracture patterns. Twelve percent, or 15 % of the determinable fragments, showed evidence on different edges of the two types of events intended to be distinguished, so were considered as discarded nutshell fragments that were re-broken through post-depositional processes. Twenty-five percent, or 30 % of the determinable fragments, showed evidence of breakage prior to charring. When added together with the broken–charred–broken fragments, this gives 37 %, or 44 % of the determinable fragments, that were broken prior to charring. The largest proportion of the fragments, at 47 %, or 56 % of the determinable fragments, showed evidence of charring prior to breakage. The small numeric difference in the evidence for the two types of events makes it difficult to arrive at a definitive conclusion about the assemblage other than the possible mixture of nutshells coming from multiple different charring events (López-Dóriga 2013).

Unbiased sample

As discussed in the “Methodology” section, the archaeological assemblage was rechecked, eliminating the size bias by choosing the sample with a rifle box and including the new characteristics observed in the controlled experiment. The inclusion of more criteria reduced the number of indeterminable nutshell fragments (5 %, because of recent breakage), although many of the potential heat-induced fissures detected (in 35 % of shell fragments, Fig. 7) raised some questions, for example they could have been produced in some cases by trampling. Not a single vesicle was observed in any of the edges examined nor any curled outlying part of the outer skin of the pericarp. In fact, in many cases, this skin was completely missing. The destruction of these characteristics by erosion cannot be ruled out; however, given the circumstances, the results point to the hazelnuts charred whole as the most likely explanation.

Fig. 7
figure 7

Archaeological hazelnut shell fragment with transversal fissure

Full size image

Discussion

Applying the fire alteration identification criteria devised in this experimental work can help to understand the taphonomy of primary and secondary charred hazelnut shell fragment deposits from archaeological sites. Primary and secondary charred assemblages are produced in a single charring event and deposited in situ where they were charred (primary) or displaced somewhere else as a whole (secondary) (Fuller et al. 2013). Tertiary deposits, in which more than one charring event occurred (Fuller et al. 2013), might be also identified when inconsistencies in the presence of pre-charring or post-charring breakage are detected. This is useful as a means of understanding the roles of hazelnuts in human subsistence, allowing the identification of large-scale consumption (massive numbers of nutshell fragments discarded into the fire and charred after breaking open to eat the kernel), and storage or processing with the help of fire (hazelnuts charred whole, due to either the existence of a fire burning stored products or products accidentally falling to a fire while being roasted). These last options seem to be the more likely for the archaeological sample studied in this work. Unfortunately, the criteria identified here did not make a clear distinction between fresh and dry charred whole hazelnuts.

Another observation from this experiment concerns hazelnut processing. It has been wondered if hazelnuts could be opened in large quantities simultaneously, as has been recorded ethnographically for other hard-shelled fruits. These were opened by a quick change of temperature, achieved by heating them with burning straw placed around and pouring cold water over them (Howes 1948). This could be very advantageous for processing high quantities of nuts, in contrast with having to crack them open individually. However, this did not work when roasting hazelnuts at around 300 °C between 5 and 15 min. Roasting at higher temperatures might provide a successful result, but much care is needed to avoid charring if the hazelnuts are going to be in an edible condition afterwards.

Another main taphonomic problem in understanding archaeological hazelnut remains is quantification, which is necessary to know precisely how many hazelnuts are being recovered, but very difficult to achieve when what is left are only small nutshell fragments (most of times). As in the case of other plant macroremains (van der Veen 2007), with nutshell fragment quantification, it has often been necessary to adopt a specific criterion for each type of site with its preservation particularities. However, there are certain unavoidable problems: a potential selection of the hazelnuts present at the site (Scaife 1992) or the high morphological variation between hazelnuts from different hazel populations and even within single trees (Kosina 1991). There are quantification methods relying on a reference quantity of fragments per hazelnut obtained in different ways (Buxó-Capdevila 1997; Berihuete Azorín and Antolín 2013; Hosch and Jacomet 2001; Martin 2010; Scaife 1992; Tolar et al. 2011); others on different weight references (Bouby and Surmely 2004; Carruthers 2000; Lopinot 1984; McComb 1996; apud McComb and Simpson 1999; Perry 1999; apud Mason 1996b; Perry 2005) or volume references (Holst 2010; Testart 1982). Accordingly, some authors consider it unreliable to apply any quantification method of those devised so far (Zapata Peña 2000). Ideally, an agreement could be reached between researchers, if not to use the same quantification method, at least to publish raw data in order to allow others to re-quantify accordingly with their proposed methodologies, allowing comparisons between different sites (Berihuete Azorín and Antolín 2013).

This problem has not been tackled within this work, but we have confirmed previous observations that nutshells fragmented after charring break into much smaller fragments than when broken before charring (Mithen and Score 2000). This means that the depositional history of hazelnut shell fragments has consequences for the quantification of the remains present in archaeological deposits and thus in the evaluation of the quantitative importance of hazelnuts as food resources. This is particularly important when the charred assemblages have not been appropriately sampled, such as those extracted by hand picking by eye from the sediment or by sieving with large-sized meshes. In these cases, only large pieces of hazelnut shell are retrieved, thus contributing to underestimate the numbers of nuts and introducing biases in potential taphonomic studies. It is very important to apply appropriate recovery sampling techniques, even when only some hazelnut shell fragments are retrieved. These prove to be very useful pieces of information for understanding taphonomic processes as well as the importance of plant food resources in the past.

Conclusions

This experimental work has established criteria for understanding the taphonomic processes involved in the formation of charred hazelnut shell fragments, by distinguishing between charred nutshells which have been fragmented post-depositionally (as in the case of stored deposits accidentally burnt or hearth accidents while roasting food resources) and charred nutshells fragmented pre-depositionally (as a result of discarding food by-products into fires). This can help to understand the formation processes of deposits, i.e., whether they are primary or secondary: whether they are produced from a single event of charring but not necessarily being in situ. This is also useful for identifying tertiary deposits in which the charred items are a result of multiple charring events under different conditions. This in turn can help our understanding of the role that hazelnuts played in past human subsistence.