Introduction

Plants used and transformed by people can produce a diverse record that can be considered as a proxy of their choices and activities, and in certain cases of ecological conditions too. In a broad sense, the term ‘proxy’ is used to define a representative or intermediary. In palaeoclimatology, a proxy is defined as “a local record that is interpreted using physical or biophysical principles to represent some combination of climate-related variations back in time” (Folland et al. 2001, p. 130). Noise and possible biases make it necessary to calibrate and cross-validate proxies in order to obtain more accurate and reliable palaeoclimatic reconstructions. A multi-proxy approach is also commonly adopted in palaeoecological studies, particularly in palaeolimnology (Birks and Birks 2006 and references therein). In palaeoecology, a proxy is understood as a record of changes that can be measured or analysed to reconstruct past ecosystems and biotic responses to natural or human-caused changes (Birks and Birks 2006). Palaeoecological proxies include fossil organisms such as diatoms, phytoliths and pollen grains, as well as sediment characteristics which are measured through physico-chemical analyses.

Despite the fact that archaeobotany shares several methodological approaches with palaeoecology, the concept of proxy is not much theorised and, in general, the major evidence (proxy) is considered to be the charred remains record. This is due to (a) visibility, because charred remains can be seen by naked eye, (b) relatively easy methods of recovery either handpicked or by flotation, and (c) direct analysis without previous chemical processing.

An example of the single-proxy approach in archaeobotany, based on charred remains, is the study of plant-related subsistence strategies and agricultural practices. However, as a single-proxy approach, plant microremains offer a valuable alternative. Phytolith and starch grain analyses have made major contributions to the understanding of plant domestication processes worldwide (e.g. Piperno et al. 2009). Moreover, direct evidence of past human diet has been gained from plant microremains recovered from human dental calculus (Henry and Piperno 2008) or artefacts involved in food processing such as grinding stones (Piperno et al. 2004). Similarly to plant macroremains, phytoliths and starch grains have also been used to establish the practice of irrigated agriculture (Madella et al. 2009; Rosen and Weiner 1994), dry farming (Lu et al. 2009) and vegeculture (Barton and Denham 2011), as well as to study crop-processing activities (Harvey and Fuller 2005; Yang et al. 2013).

Although macro and microremains can be recovered from the same contexts, only a few studies have actively pursued an integration of data from both lines of evidence (Delhon et al. 2008; Dickau et al. 2012). The integrated analysis of charred macroremains, phytoliths and starch grains can widen the information spectrum at several levels:

Taxonomy

The combined use of macro- and microbotanical remains increases the number of taxa identified, independently of the preservation pathways. Microremains allow for the recognition of taxa from leaves (Out and Madella in press; Yang et al. 2013), roots and tubers (e.g. Chandler-Ezell et al. 2006) and fruits such as banana (Denham et al. 2003 for starch and Mindzie et al. 2001 for phytoliths). Charred seeds and related floral parts, on the other hand, are often strongly taxonomically diagnostic. For example, charred small millets can usually be identified to species level, whereas starch grains are, at best, diagnostic to genus level (Krishna Kumari and Thayumanaban 1998; Liu et al. 2011; Yang et al. 2012). The potential of phytoliths to differentiate between small millets has only started to be evaluated outside the two main genera, Panicum and Setaria (Madella et al. 2013 and references therein).

Anatomy

A multi-proxy approach allows for the identification of different plant parts, useful for both dietary and non-dietary investigation of plant use (Fig. 1). Plant parts seldom preserved in the macrobotanical record such as chaff of small grasses, leaves or culms can be identified from plant microremains (Lu et al. 2009; Madella et al. 2013; Out and Madella in press; Yang et al. 2013; Zhang et al. 2011).

Fig. 1
figure 1

Idealised drawings and examples of proxies (macro and microbotanical remains produced by different plant parts) from a Setaria italica (L.) P. Beauv. (foxtail millet), and b Triticum turgidum ssp. dicoccon (Schrank) Thell. (emmer wheat)

Full size image

Taphonomy

Plants and plant parts are preferentially preserved depending on their intrinsic characteristics (soft vs. hard tissues), processing technique (roasting, boiling, etc.) and post-depositional settings (dry vs. wet environments, bioturbation, etc.). A multi-proxy approach offers the possibility to analyse a wider spectrum of plant residues, therefore allowing more precise evaluations of the original assemblages. Phytoliths are usually preserved regardless of the depositional conditions, since they are not dependent on fire for preservation as most macroremains are. Starch grains can be easily degraded by enzymes, bacteria and other organisms of the soil (Haslam 2004). However, when trapped in dental calculus or pores in artefacts they can be preserved for thousands of years in diverse environmental settings (Torrence 2006, Table 1). Starch taphonomy and preservation has been experimentally assessed in a set of tests by Lu (2003; in Haslam 2004).

Table 1 Fire-related contexts analysed in this study. Phytoliths are expressed in concentration per g of AIF (Acid Insoluble Fraction). Starch grains are expressed in concentration per g of original sediment
Full size table

To summarise, a multi-proxy approach offers the possibility of wider anatomical and taxonomical identification as well as overcoming some taphonomic effects, resulting in a more representative set of data from the original input. The aim of this paper is to illustrate how the analysis of multiple archaeobotanical proxies from the same archaeological contexts improves and enriches our understanding of plant use. The case study is from the Chalcolithic Harappan settlement of Shikarpur (ca. 2500–1500 bc), located in Kachchh, Gujarat, northwest India, at the fringe of the Indus valley. The paper discusses the evidence from integrated macro and microbotanical analyses from fire-related contexts and grinding stones.

Archaeobotanical background

Harappan subsistence strategies varied between the core area in the main Indus valley and the peripheries. In the Indus valley, subsistence relied on winter crops (rabi) such as Triticum spp. (wheat), Hordeum vulgare L. (barley), Cicer arietinum L. (chickpea), Lens culinaris Medik. (lentil) and Pisum sativum L. (pea). In the southern peripheral areas such as Gujarat, crops were mainly cultivated in summer (kharif), including large and small millets, Oryza sp. (rice) and tropical pulses such as Vigna radiata (L.) R. Wilczek (mung bean), Vigna mungo (L.) Hepper (black gram) and Cajanus cajan (L.) Huth (pigeon pea) (Fuller and Madella 2001).

Previous studies of plant-related subsistence strategies in Harappan Gujarat focused mostly on charred macroremains. Archaeobotanical research at Kanmer, a settlement in Kachchh occupied from the Early to the Post Urban Harappan Period (Pokharia et al. 2011), shows a switch from a predominance of winter crops, mainly barley, in the earlier phases towards a more diversified strategy at the end of the Harappan occupation, when the assemblage is dominated by summer crops. Plant remains from Rojdi, located in the Saurashtra peninsula in Gujarat, show the predominance of summer crops (millets and pulses) during both the Urban and Post Urban phases of occupation. Post Urban Harappan, 2nd millennium bc Oriyo Timbo has a plant assemblage dominated by summer crops, especially millets such as Eleusine coracana Gaertn. (finger millet), Setaria and Panicum spp. and to a minor extent, pulses (V. mungo) (Reddy 1997). In contemporary Babar Kot the crop assemblage additionally included winter pulses, such as L. culinaris or Vicia sp. (Reddy 1997).

Macrobotanical analyses from previous field seasons at Shikarpur by Chanchala and Saraswat show the presence of Triticum aestivum L. (bread wheat) and Eleusine coracana and Setaria sp. (small millets) (IAR 1995, 2002). However, Fuller (2003, 2006) questioned the presence of African crops such as E. coracana at Shikarpur and other Harappan settlements during the 3rd millennium bc. The author also pointed out the possible confusion between Setaria sp. and Brachiaria ramosa (L.) Stapf (browntop millet). Therefore, the identification of E. coracana and Setaria sp. at this settlement cannot be taken as conclusive.

Materials and methods

Shikarpur (N 23°14′15″, E 70°40′39″) is located in Kachchh, a semi-arid region with an average annual rainfall of ca. 400 mm, most of which falls during the monsoon period between June and September. The materials analysed for this study were collected during the 2012 field season, when the eastern part of the fortified area was excavated to expose the structures of occupational Phase II, Late Urban Harappan, ca. 2200–1900 bc (Bahn and Ajithprasad 2008).

A total of 923 l of sediment were floated of which 240 l from nine fire-related contexts such as hearths, ashy patches or areas with burning activity were analysed for this study. For each context, the sampling strategy consisted on collecting a minimum of 10 l of sediment for flotation and a 5 × 3 cm zip-lock bag of sediment for microremains extraction. Bucket flotation was carried out with a 0.25 mm mesh. It was not possible to collect a flotation sample from one oven due to a mishap during the excavation. A total of 11 phytolith and starch grain samples were analysed (Table 1).

Furthermore, 20 microremain samples were analysed from 18 grinding tools (Table 2). Quern 7 presented two used surfaces (a and b) which were sampled and analysed separately. Quern 9 was broken into two fragments (a and b), which were also separately sampled and analysed. In addition, 11 control samples from the same contexts of the grinding tools were checked to assess them for contamination, which was preliminarily appraised by checking differences in microremains concentration, where higher concentrations were assumed for stone tools; further experiments on this issue are ongoing. Residue recovery consisted of a two-step process in which the outer layer of sediment was first dry brushed from the used surface(s) of the grinding stone (dry sample), and then the inner layer of sediment was brushed with deionised water (wet sample) (Hart 2011). Microbotanical remains were extracted from the wet sample with a combination of the methods described in Madella et al. (1998) and Horrocks (2005), adapted for small samples recovered from grinding stones (generally < 1 g). Loose sediments from fire-related contexts and control samples were processed using the same extraction protocol to allow for comparison. Phytoliths, both single cells and silica skeletons, and starch grains were observed at 200 and 630 magnifications with a Leica DM 2500 microscope equipped with a Leica DF 470 camera. Macroremains were identified using a Leica EZ4 D stereoscope. Taxonomical identification of all plant remains relied on the plant reference collection of the BioGeoPal Laboratory (CaSEs, Barcelona) and on relevant literature (Fuller and Harvey 2006; Madella et al. 2013; Neef et al. 2012). Phytoliths were described using the International Code for Phytoliths Nomenclature (ICPN—Madella et al. 2005), whereas starch grains were described according to the International Code for Starch Nomenclature (ICSN 2011).

Table 2 Grinding stones analysed in this study. Descriptive terms after Wright (1992). Phytoliths are expressed in concentration per g of AIF (Acid Insoluble Fraction). Starch grains are expressed in concentration per g of original sediment
Full size table

Results

Fire-related contexts

Charred macroremains from Shikarpur were scarce and generally poorly preserved (Table 3; Fig. 2). Charred grass (Poaceae) remains included 31 grains of small millets, of which 22 belonged to the general SEB type (Setaria, Echinochloa and Brachiaria) (Fig. 2a) and nine were identified no further than small millets due to severe damage. Chaff from Oryza (Fig. 2b) and Hordeum were also found. Further recovered grasses included half a charred grain of Coix lacryma-jobi L. (Job’s tears) (Fig. 2c) and 10 mineralised full inflorescences of B. ramosa (Fig. 2d-e). Pulses (Fabaceae) were also present, and 19 seeds morphologically comparable to Vigna radiata and V. mungo were found (Fig. 2f). Morphometric analyses were not conclusive, but the generally small seed size (average 1.15 × 0.87 mm) suggests a wild species of Vigna (Fuller and Harvey 2006). Moreover, leaf fragments from an unidentified member of the Fabaceae were recovered from four contexts. Other finds included several weeds (Trianthema sp. and Chenopodium sp.) (Fig. 2g–h), six sedge grains and several parenchyma fragments from tubers.

Table 3 Results of the macroremains analysis from fire-related contexts. + = present
Full size table
Fig. 2
figure 2

Macrobotanical remains recovered from Shikarpur. a charred caryopsis of a SEB type small millet, b, charred spikelet base of Oryza sp., c half charred caryopsis of Coix lacryma-jobi L. (Job’s tears), d dorsal and e ventral view of a mineralised inflorescence of Brachiaria ramosa (L.) Stapf. (browntop millet), f charred seed of Vigna sp., g charred grains of Trianthema sp., h charred grains of Chenopodium sp. Scale bar 1 mm in a–f and 2 mm in g–h

Full size image

The analysis of single-cell phytoliths (Table 4, Fig. 3a) showed a predominance of grass morphotypes (91.99 % Poaceae) over non-grass (1.51 %). Undetermined phytoliths, a group that includes non-diagnostic as well as unidentified phytoliths, accounted for 6.50 % of the single-cell phytoliths. The non-grass phytoliths were mainly morphotypes from dicotyledons and monocotyledons, such as palms (Arecaceae) and sedges (Cyperaceae) (Fig. 4a–c). Among grass phytoliths, long (elongated) cells and bulliforms offer anatomical information, whereas short cells are taxonomically diagnostic at subfamily level. Anatomical and taxonomical information is considered separately, so percentages presented below are calculated independently for the anatomically and the taxonomically diagnostic phytoliths. Leaf/culm phytoliths (bulliforms and psilate/sinuate elongated cells; 75.38 %) (Fig. 4d) outweigh inflorescence phytoliths (11.08 %) and anatomically non-diagnostic elongated cells (13.54 %). Short cell panicoid morphologies predominate (46.09 %) over chloridoid (15.42 %) and pooid (12.47 %), with non-attributable morphologies accounting for 26.02 % of the total. A total of 111 grass multi-cell phytoliths (silica skeletons) were encountered in the fire-related contexts, with 65 from leaves/culms, 44 from inflorescences and 2 anatomically non-diagnostic. Based on the morphology of elongated and short cells, 10 inflorescence silica skeletons were identified as panicoids. In particular, this silica skeleton morphology typically occurs in the external parts of the inflorescence (glumes, lower lemma and lower palea) of small millets (Fig. 5).

Table 4 Results of phytolith and starch grain analyses from fire-related contexts
Full size table
Fig. 3
figure 3

Percentages of single-cell phytoliths and starch grains recovered from a fire-related contexts, b grinding stones. i single-cell phytoliths, ii anatomically diagnostic grass phytoliths (elongated cells and bulliforms), iii taxonomically diagnostic grass phytoliths (short cells), iv starch grains

Full size image
Fig. 4
figure 4

Single-cell phytoliths recovered from Shikarpur. a irregular phytolith from a dicotyledonous plant, b, globular echinate phytolith from a palm (Arecaceae), c scrobiculated cone phytolith from a sedge (Cyperaceae), d bulliform phytolith from a grass (Poaceae) leaf. Scale bars are 20 µm

Full size image
Fig. 5
figure 5

Modern (a–b) and archaeological (c–d) multi-cell phytoliths (silica skeletons) recovered from Shikarpur: a lower lemma from Brachiaria ramosa (L.) Stapf. (browntop millet), b lower glume from Echinochloa colona (L.) Link (shama millet), c–d, panicoid silica skeletons from fire-related contexts. Scale bars are 50 µm

Full size image

Starch grains were also scarce (n = 46) and only four morphotypes were identified (Table 4; Fig. 3a). The most common typology (52.17 %), further divided into three sub-types according to size, has a Panicoideae faceted polyhedral morphology. Type 1 grains are small (5-10 µm) to very small (<5 µm), characteristic of small millets (Fig. 6a) and rice, whereas Type 3 grains (>20 µm) occur mostly in big millets such as Sorghum bicolor (L.) Moench. (sorghum), Pennisetum glaucum (L.) R.Br. (pearl millet) and C. lacryma-jobi (Fig. 6b) (Madella et al. 2013 and references therein). Type 2 grains are of medium size (10–20 µm) and they can represent any of the previous taxa within the Panicoideae. The second-most frequent type (28.26 %) has discoidal grains, with a smooth surface and lamellae, characteristics of the tribe Triticeae (Pooideae, Poaceae) (Fig. 6c–d). Other finds include seven ovoid grains with a smooth surface, lamellae and a linear hilum diagnostic of the Faboideae (Fabaceae) (Fig. 6e–f); and two small grains that could belong to the Triticeae but which are also present in other taxa, and are therefore classified as cf. Triticeae.

Fig. 6
figure 6

Modern starch grains from the reference collection. a Brachiaria ramosa (L.) Stapf. (browntop millet), b Coix lacryma-jobi L. (Job’s tears), c Hordeum vulgare L. (barley), d Triticum aestivum L. ssp. sphaerococcum (Perc.) MK. (dwarf wheat), e Vigna radiata (L.) R.Wilczek (mung bean), f Vigna mungo (L.) Hepper (black gram). Scale bars are 20 µm

Full size image

Grinding stones

Plant microremains were abundant in the grinding implements from Shikarpur. Phytolith concentrations were very high compared to other published analyses of grinding stones (e.g. Portillo et al. 2009). The pestle showed particularly high values, whereas Hand 3 is an exception with an extremely low presence of phytoliths (Table 2). The phytolith assemblage from Hand 3, unlike other grinding tools, was dominated by bulliforms and trichomes (included within the undetermined taxa). These morphotypes are, overall, thicker than elongated and short cells, and thus more resistant to taphonomic processes that may have affected the phytolith assemblage from Hand 3 (Madella and Lancelotti 2012). Moreover, this is the sole sample where phytolith concentration is lower than its control. For these reasons, this assemblage might not represent the original phytolith input and has therefore been excluded from the percentages presented below (Fig. 3b).

The analysis of single-cell phytoliths (Table 5) shows similar results to samples from fire-related contexts. Grass morphotypes predominate (91.81 %) and, among grasses, panicoids (39.23 %) are more represented than chloridoids (16.85 %) and pooids (19.60 %). The anatomical analysis also shows the predominance of leaf/culm phytoliths (66.56 %) over inflorescence morphotypes (26.43 %). The presence of sedge and palm phytoliths is worthy of note (although the latter were only encountered in Hand 3). Silica skeletons were scarce and only 25 of these were encountered, among which one was identified as being from pooid grasses.

Table 5 Results of phytolith and starch grain analyses from grinding stones
Full size table

A total of 650 starch grains were recovered from grinding stones (Table 5; Fig. 3b), most of which belong to the Panicoideae (51.69 %) (Fig. 7a–b) and the Faboideae (30.92 %) (Fig. 7c). Triticeae (2.46 %) (Fig. 7d) and cf. Triticeae (4.15 %) grains were also recovered, although the former were much less common. Three morphotypes that were not encountered in fire-related contexts were recovered from grinding stones: (a) 30 spherical grains with a linear hilum and lines radiating from the centre, attributed to cf. Panicoideae (Fig. 7e); (b) 18 ovoid grains with a smooth surface, a regular extinction cross and an eccentric small vacuole hilum (Fig. 7f–g), most probably from a tuberous plant; and (c) one bell-shaped grain with an eccentric, linear hilum (Fig. 7h–i) that occurs in roots and palms. Finally, 21 starch grains could not be identified due to severe damage.

Fig. 7
figure 7

Starch grains recovered from Shikarpur. a Panicoideae Type 1 (<10 µm) grains, b Panicoideae Type 2 (10–20 µm) and Type 3 (>20 µm) grains, c cf. Panicoideae grain, d Triticeae grain, e Faboideae grain, f–g tuberous grain, h – i root/palm grain. Scale bars are 20 µm. Images g and i are under cross-polarised light

Full size image

Discussion

Integrating macro- and microbotanical remains

The scarcity of charred macroremains at Shikarpur, linked to the dry-wet cycles caused by the monsoon climate of this region and the high salinity of the soils, highlights the difficulty of assessing the plant use strategy solely based on this evidence. The few data gathered from macroremains show that small millets appear together with common weeds of millet, such as Trianthema sp. and Chenopodium sp., suggesting that small millets were being cultivated and not simply gathered. Most charred small millets are identified as SEB type, a group within the tribe Paniceae (Panicoideae, Poaceae). Several species within this group are native to South Asia, whereas Setaria italica (L.) P. Beauv. (foxtail millet) was domesticated in China (Nasu et al. 2007). The timing of the arrival of S. italica in northwest South Asia is controversial. Some researchers claim that it was already present in Early Harappan times, before 2600 bc (Pokharia et al. 2014 and references therein), while Fuller (2006) suggested that the finds reported as S. italica are instead Brachiaria ramosa, which is morphologically very similar to Setaria spp. Fuller advocates that S. italica and Panicum miliaceum L. (common millet) were not present in South Asia until Post Urban Harappan times, after 1900 bc. Charred remains from Shikarpur are not conclusive, but the presence of 10 mineralised B. ramosa grains suggests the undeniable presence of this species, although its importance is difficult to estimate.

The evidence from other proxies, phytoliths and starch grains, also suggests that small millets were staples for the inhabitants of Shikarpur. No silica skeletons from the upper lemma and palea of small millets were encountered, which are diagnostic to species level (Lu et al. 2009; Zhang et al. 2011). However, the presence of silica skeletons from the external parts (glumes, lower lemma and lower palea) suggests that small millet processing was taking place at the settlement. The significant presence of Type 1 Panicoideae starch grains (<10 µm) and the scarcity of inflorescence silica skeletons in grinding implements implies that these tools were used to mill clean small millet grains for flour and not for dehusking. This highlights the use of millets in Shikarpur as flour, possibly for bread making, but not as whole grain food. These starch grains are also produced in Oryza (Yang and Perry 2013), but the only rice evidence from the macroremains is a single charred spikelet base found in Pit 2 and there is no evidence of rice phytoliths. Macroremains and missing phytoliths therefore reinforce the hypothesis that Type 1 Panicoideae starch grains are from small millets.

The presence of Type 3 Panicoideae starch grains (>20 µm) seems to suggest that big millets were also milled at Shikarpur. The presence of African millets (Sorghum bicolor, Pennisetum glaucum and Eleusine coracana) in South Asia during the 3rd millennium bc seems to be dubious (Fuller 2003) and it is possible that Type 3 morphologies are from Coix lacryma-jobi, the grains of which were found in Pit 1, Shikarpur, and other Chalcolithic sites in northern Gujarat (authors’ unpublished data). This plant is still a minor food and fodder crop in some parts of India (Arora 1977).

Triticeae starch grains were also present at Shikarpur. The damage caused by grinding and the small number of grains recovered, 29 grains in total, prevents a more specific taxonomical identification based on surface features as suggested by Yang and Perry (2013). Most of the recovered grains (75.86 %) were larger than >20 µm, suggesting that they were from Triticum, Hordeum, Secale, Agropyron or Aegilops (Yang and Perry 2013). Moreover, plants from the Triticeae are not native to Gujarat (Fuller 2006) and, according to macrobotanical evidence from excavations of other archaeological sites, the only crops from this tribe which were consumed in this region during Harappan times were Triticum and Hordeum (Fuller and Madella 2001). Therefore, the Triticeae starch grains identified in both fire-related contexts and grinding stones can be attributed to Triticum/Hordeum. The macrobotanical evidence for the processing and consumption of these cereals is limited to some charred chaff but, once more, the multi-proxy approach highlights their presence and use in the settlement, even as minor components of the diet.

The combined botanical evidence demonstrates that pulses also played an important role in the diet of the inhabitants at Shikarpur, who seem to have consumed wild Vigna taxa. The presence of Fabaceae leaf fragments in fire-related contexts could suggest that the wild Vigna grains had entered the archaeobotanical record as fuel, either directly or via animal dung (Lancelotti and Madella 2012). However, the significant presence of Faboideae starch grains on grinding stones makes clear that at least one wild Vigna taxon was part of the people’s diet. Similarly, sedge nutlet phytoliths from grinding stones, despite being scarce, suggest that sedges may have been processed in small quantities for human consumption. Finally, tuberous plants and roots also appear in the assemblage from Shikarpur and they were probably consumed both whole (charred parenchyma) and ground (starch grains from grinding tools and globular echinate phytoliths).

Subsistence strategies at Shikarpur

The multi-proxy approach shows that the subsistence strategy of the inhabitants of Shikarpur was based on local summer crops, mostly small millets (B. ramosa and some other taxa) but also a wild Vigna legume. Taking into account the environmental settings of this region, with low water availability, high inter-annual variability including droughts, short cropping period and high salinity, the combination of small millets and wild pulses would probably have constituted the most profitable land use strategy. The possibility of the cultivation of wild Vigna taxa, which could also have improved the soil by nitrogen enrichment, cannot be discarded completely. Other resources such as rice and sedges and also some kinds of roots and tubers also seem to have been consumed, possibly as condiments or spices.

Triticum and Hordeum, which were staple crops in the core Harappan areas, were scarcely present at Shikarpur. Macrobotanical evidence is limited to one Hordeum rachis. Pooideae phytoliths were only marginally present both in all analysed contexts and tools. The best evidence for Triticum/Hordeum consumption comes from Triticeae starch grains which were recovered from grinding stones. This minor presence of big grain C3 cereals at Shikarpur might represent a local, small-scale cultivation, which seems unlikely in the absence of chaff phytoliths or, most probably, the trading between Harappan settlements set in different ecological regions of which Shikarpur was part (Bahn and Ajithprasad 2008; Gadekar et al. 2014). Triticum and Hordeum were not the main staples in northern Gujarat, and their use may be related to cultural preferences of the inhabitants of Shikapur as part of the area of Harappan influence.

Conclusions

We believe that a combined approach, in which several botanical proxies and a broad-spectrum sampling strategy are used together, is the best possible way to explore diet and plant use strategies in past societies. This paper has shown how effective this method can be and how the information obtained can be enhanced. The combined information from the different deposits and grinding tools at Shikarpur highlights not only the presence at the site of various taxa, both cultivated and wild, but also the pathways of their use. The macrobotanical evidence helped, regardless of its paucity, in identifying some of the staple grains, such as the small millets, and some secondary grains such as the Vigna sp. wild pulses, which could have been interpreted as part of the weed or fuel (dung) assemblage. However, the starch from the grinding stones undeniably shows that these seeds were ground to flour and therefore that they were part of the diet. The microbotanical remains broadened the information on the plant spectrum used for food such as sedges and tubers, as the remains were connected to processing with grinding stones. Finally, it is remarkable to see how the different proxies can reinforce and complement each other; as is the case of the wild Vigna identified in the macroremains for which the microremains strongly highlighted their pre-domestic character.