Vegetation History and Archaeobotany

, Volume 20, Issue 3, pp 223–239

Plant use at an early Islamic merchant town in the West African Sahel: the archaeobotany of Essouk-Tadmakka (Mali)

Authors

  • Sam Nixon
    • Institute of ArchaeologyUniversity College London
  • Mary Anne Murray
    • Institute of ArchaeologyUniversity College London
    • Institute of ArchaeologyUniversity College London
Original Article

DOI: 10.1007/s00334-010-0279-6

Cite this article as:
Nixon, S., Murray, M.A. & Fuller, D.Q. Veget Hist Archaeobot (2011) 20: 223. doi:10.1007/s00334-010-0279-6
  • 321 Views

Abstract

We present archaeobotanical data from the early Islamic era (ca. a.d. 750–1400) obtained from excavations at Essouk-Tadmakka, an important trans-Saharan trading town site in the West African Sahel and an early centre of the Tuareg. The paper provides insight into a little researched area of arid zone medieval West Africa and presents practically the only substantive archaeobotanical evidence of the medieval Tuareg. The evidence firstly enables us to shed greater light on the Arabic historical references to traditions of wild cereal gathering at Essouk-Tadmakka. It also establishes the presence at the site of a range of important taxa, including pearl millet, date, balanites, cotton and linseed, as well as a host of other fruits, legumes (Fabaceae) and wild plants. Perhaps the most striking finding is the earliest and largest archaeobotanical data set for wheat in West Africa. In addition to providing the first archaeobotanically based discussion of Essouk-Tadmakka’s gathering traditions, agriculture, and grain importation, we also seek to highlight certain evidence for change over time in the archaeobotany recovered. The data seems to suggest that towards the end of the site’s occupation (ca. a.d. 1300) there was a shift to increased presence of fruit and legumes and more limited presence of cereals, and we attempt to relate this to wider shifts in Sahelian culture at this time.

Keywords

PalaeoethnobotanyAgricultureWild cereal gatheringTriticumTuaregSaharan Berber

Introduction

In the early Islamic era (ca. a.d. 650–1500) trans-Saharan trade experienced a surge beyond levels seen in Carthaginian and Roman times as Islamic powers sought to further develop access to the vast gold, slave and ivory resources of West Africa. Associated with this surge, a network of trading towns developed across the trans-Saharan landscape. Early Arabic sources provide much insight into these towns and their trade, particularly in the 11th to 15th century ‘high’ period of trade during the eras of the Almoravids, and the Empires of Mali and Songhay (Levtzion and Hopkins 2000). The early Arabic documents do however also leave many gaps in our understanding and therefore archaeology has been essential for furthering knowledge of these towns and their trade (e.g. Devisse 1983; Berthier 1997; Insoll 1996, 2000). Seeking to improve our understanding of the trans-Saharan trading town system, in 2005 the first professional excavations were conducted at the site of Essouk-Tadmakka in northern Mali. Essouk-Tadmakka was one of the most significant early locations of trans-Saharan trade and an important early centre of the Tuareg, the Saharan nomads associated with the control of Saharan trade caravans (Figs. 1, 2, 3; Nixon 2009). This paper focuses on the project’s archaeobotany. Given the limited archaeobotanical investigation of early Islamic trans-Saharan trading towns—systematic studies only having been conducted at Jarma (Pelling 2005, 2008), Dia (Murray 2004) and Gao (Fuller 2000)—these new data represent an important addition of knowledge. The data is also important as it provides insight into the plant use of the early Tuareg, a cultural group for which we have very little archaeobotanical knowledge.
https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig1_HTML.gif
Fig. 1

Map illustrating Essouk-Tadmakka’s location within West Africa and Mali, and its position within the early Islamic trans-Saharan system; other sites referred to in the text are also marked (developed from data in Levtzion and Hopkins 2000)

https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig2_HTML.jpg
Fig. 2

View across the central area of the Essouk-Tadmakka ruins (looking east–west; the wadi and the island within it are seen in the middle distance)

https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig3_HTML.gif
Fig. 3

Map of the Essouk ruins with locations of archaeobotanical sampling indicated (excavation units A, B and C)

From Arabic historical texts written between the 10th and 14th centuries a.d. we learn Tadmakka (or ‘Tadmekka’) was one of the greatest West African trans-Saharan trading cities, and an important Sahelian political centre (Levtzion and Hopkins 2000). Throughout this period the sources clearly identify it as a site within the Berber world and more specifically as a centre of the early Tuareg. In the 10th century, Tadmakka is described as a southern Saharan locality with an Islamic cultural element (Levtzion and Hopkins 2000, pp. 50–51) and from the 11th century as a flourishing trade centre on routes between contemporary Libya/Tunisia/Algeria and the West African trading powers of the Niger Bend region (Levtzion and Hopkins 2000, pp. 84–87). It is likely that from this early period onwards its population was at least in part composed of traders from North Africa engaged in the trans-Saharan commerce. Tadmakka seemingly experienced certain changes in the late 11th century with the rise of the Almoravids, a militantly Islamic Saharan Berber group who seized control of trans-Saharan trade. However, it seemingly continued as a trade centre in the 12th and 13th centuries, although descriptions of the town are unclear at this time (Levtzion and Hopkins 2000, pp. 98–99, 193). The final significant reference to Tadmakka comes in the 14th century, it being described as an important locality for powerful Tuareg pastoral groups but with no reference being made to an urban centre or trade (Levtzion and Hopkins 2000, p. 274).

References in the Arabic histories to plant usage at Tadmakka do exist but are limited. Firstly, in a.d. 967–968, Ibn Hawqal describes people living beyond the southern territory of the Maghreb Kingdoms, including those of the Fazzan (southwest Libya) and Tadmakka, as ‘unacquainted with cereals (ta’am) and hav[ing] never seen wheat or barley or any other kind of grain’ (Levtzion and Hopkins 2000, pp. 50–51). A century later al-Bakri (writing a.d. 1068) stated Tadmakka’s inhabitants lived ‘on meat and milk as well as on a grain which the earth produces without being tilled’; also supposedly ‘Sorghum and other grains were imported for them from the land of the Sudan [the savannah zone]’ (Levtzion and Hopkins 2000, pp. 84–87). Lastly, in the 14th century (a.d. 1337–1338) al-’Umari describes Tadmakka’s inhabitants as ‘liv[ing], as desert dwellers do, on meat and milk; grain [being] scarce with them’ (Levtzion and Hopkins 2000, p. 274); while they were described as ‘short of food’ they supposedly inhabited ‘mountains which produce many fruits’ (Levtzion and Hopkins 2000, p. 274).

The ruins of Tadmakka, today called Essouk (Figs. 2, 3), are situated in the mountains of the semi-autonomous Tuareg region of northeast Mali, at the most arid extremes of the Sahel (Neumann and Schulz 1987, p. 151; Ozenda 1991, p. 640). Essouk consists of approximately 50 ha of stone structural remains (mainly houses and commercial premises) as well as numerous early Islamic cemeteries bearing Arabic epigraphy (de Moraes Farias 2003). The ruins are situated within a valley surrounded by rocky cliffs. A wadi, which experiences flooding in the summer months (July–August), runs through the middle of the site. The wider environment around Essouk is predominantly sandy desert with rocky outcrops and with only very limited vegetation cover, mainly patches of wild grasses and occasional acacia trees. Limited garden cultivation of fruits and salad plants takes place within the region but not extensive crop cultivation, with cereals being imported from the north and south. At Essouk itself there is only a single, hand-watered tomato garden (approx. 100 m²)—next to a house built at the site—but also occasional date palms (author’s observation).

Rainfall in the region is low with precipitation at Essouk being around the 100 mm isohyet mark (Rodier 1963, p. 183; Boudet 1976, p. 137; Grove 1978, pp. 15, 146; Gischler 1976, p. 86). The Essouk locality itself is one of the more environmentally favourable spots in the region, due to a relatively high water table (author’s observation). In the Essouk valley and immediately around the site, stands of wild grasses are more frequent and acacia trees more abundant. The paleoenvironmental data for this area of the Sahel is poor, but we do have some data from other regions of the Sahel, including northern Nigeria (Street-Perrott et al. 2000) and further south in Mali (Mayor et al. 2005). Consensus seems to indicate an only slightly less arid environment in the early medieval period (Grove 1993, pp. 39–42; Brooks 1998), although regional drought induced crises have also been postulated (e.g. ca. a.d. 750–950—Street-Perrott et al. 2000). The historical data for Tadmakka (Levtzion and Hopkins 2000; see above) certainly suggests this was broadly a similarly arid, sub-desert zone in early medieval times.

Three areas of the site were excavated, EKA, EKB and EKC (see Fig. 3). The excavations recorded and radiometrically dated a 5 metre occupation sequence associated with permanent architecture, dated ca. a.d. 750–1400, and a further 1.5 m of undated cultural deposits below this (Nixon 2008, 2009, 2010). Essouk’s occupation was seemingly uninterrupted. Four major occupation periods are apparent from the excavations:
  • Period 1 (pre ca. a.d. 750; ca. 670–510 cm below surface): semi-permanent structures and relatively intensive occupational deposits; no trans-Saharan or other long-distance trade evidence.

  • Period 2 (ca. a.d. 750–950; 510–290 cm below surface): commences with the first permanent architecture (well-built stone and mud structure) associated with trans-Saharan trade evidence including glass; later evidence includes gold coin production evidence and large quantities of vessel/bead glass.

  • Period 3a (ca. a.d. 950–1100; 290–230 cm below surface): continued rebuilding of permanent architectural structures and site expansion; trans-Saharan trade evidence including large quantities of vessel/bead glass and glazed ceramics.

  • Period 3b (ca. a.d. 1100–1300; 230–160 cm below surface): construction of ‘luxury’ buildings and trade goods evidence including large quantities of sub-Saharan imported pottery; noticeable changes include limited North African material culture imports, evidence for some site retraction, and material culture as well as faunal changes.

  • Period 4 (ca. a.d. 1300–1400; 160–0 cm below surface): continued permanent architectural occupation, seen by reference to surface remains to be over a wide area, but limited artefactual evidence for trade; radical cultural changes in the nature of the excavated evidence by comparison with the previous period, including changes in architecture, material culture and faunal remains.

The excavations pushed back the evidence for trade and permanent settlement at Essouk-Tadmakka to the 8th century a.d. and demonstrated for the first time continual permanent occupation from ca. a.d. 750–1400 (Nixon 2009, 2010). They recorded a mixed Sahelian/North African immigrant culture at the site up until ca. a.d. 1300 and the subsequent, previously unidentified infiltration of this by a Saharan Berber culture. They also improved understanding of Essouk-Tadmakka’s commodity trade, most importantly recording gold coin production. More broadly, the excavations allow new light to be shed on ideas of the early chronology and geography of trans-Saharan trade routes, the socio-cultural shifts which occurred in the West African Sahel and the nature of trans-Saharan commodity movement (Nixon 2009, 2010). The archaeobotany we now detail has been a key element of the improved understanding of Essouk-Tadmakka.

Materials and methods

Fifty archaeobotanical sediment samples of 5 l were collected during excavations and floated in the field using washover bucket flotation into a finely meshed cloth (of approximately 300 μm) obtained in Mali. The limited size of the samples taken was due to the difficult field conditions, principally the lack of large quantities of water available for flotation. Various context types were sampled, including hearths, pit fills, occupation floors and ash deposits (Table 1).
Table 1

Archaeological contexts of flotation samples processed at Essouk

Type of deposit

Period 1

Period 2

Period 3a

Period 3b

Period 4

Wall collapse (stone/earth)

    

3, 4a, 5, 16, 17, 30, 45, 43

Floor

    

22

Ashy gravel

119

    

Ash deposit

  

6

62

13, 18

Pit

  

85, 83

 

19

Clay floor

    

39

Coiled reed mat and surround

   

76

 

Loose rich mid brown sandy silt deposit

   

77, 78

 

Compact mid brown clay deposit

   

79

 

Hearth

118

89

4b

 

29

Loose sand

 

103

   

Compact clayey silt

 

113

   

A concentration of silicified grass husks from what was probably a threshing context was also hand collected during excavation (see below). Visible wood charcoal samples were hand-collected from nearly all contexts excavated primarily for purposes of radiocarbon dating, but were also investigated for seeds and other smaller plant remains they might contain (study of the wood charcoal itself has not yet been conducted). The majority of the archaeobotanical samples collected were from the long sequence of unit EKA but two samples were also taken from unit EKB. Of the samples within EKA, half come from Period 4, a result not only of the fact that more soil was excavated from this period but likely also the greater presence of ashy deposits.

The macro-remains were identified using comparative collections (both modern and ancient) housed at the Institute of Archaeology, University College London. These contain a good range of African samples and in cases where limited African comparative samples exist (e.g. Brachiaria and Echinochloa) there is good representation of these genera from Asia. A selection of specimens were mounted and gold-coated for conventional SEM imaging, while additional samples were subjected to ESEM imaging.

In addition to macro-remains, plant impressions from ceramics were also analyzed (Nixon 2009). A small number of ceramics with impressions of vegetable temper were examined by binocular microscope and those deemed to have potentially identifiable impressions (five sherds) were studied further. Casts were made using polyvinylsiloxane dental casting material, then mounted and sputter-coated with gold for SEM examination, following Fuller et al. (2007).

Results

While some samples contained high quantities of plant remains, the Essouk sample as a whole is characterized by relatively low density of specimens (6 items/l on average) with 1,539 plant items recovered in total and many sediment samples containing only wood charcoal (22/50). However, the assemblage features a wide diversity of remains, including cereals, legumes (Fabaceae), fruits, oil/fibre plants and wild/weed taxa. Only 28 contexts produced plant remains other than wood charcoal. Most plants were preserved by charring, although silicified grass husks, and desiccated fibrous fruit, a leaf and inflorescence fragment, as well as several desiccated wild/weed items were also present. In addition to the flotation samples a separate concentration of silicified grass husks was a noteworthy exception to the general rule of low density of plant remains in the archaeology. A ca. 100 ml sample from a deposit observed to be rich in silicified remains was taken (from context 6, unit EKB) and ca. 75% (estimated) of the sample volume was seen to be composed of highly fragmentary silicified grass husk, together with sand and fine wood charcoal. Some of the larger pieces represent fragments of the lemma and palea of Echinochloa-type and Brachiaria-type as found in other more complete forms in the other samples. Because of the highly fragmentary state of preservation no attempt was made to quantify this sample, although it is clearly dominated by smooth husk fragments likely to be from Echinochloa with only a minor admixture of Brachiaria (the latter probably less than 5%). In addition, the remains of a coiled reed mat, textile fragments and a lens of what was believed to be semolina (coarse flour) (all from EKA Horizon 10), attest to the seeming high level of preservation at Essouk (Nixon 2009). While the sample size is relatively small and further sampling would undoubtedly provide new insights into the plant use regime at Essouk-Tadmakka, these results provide a good first sampling of the site.

Most plant remains were charred, and therefore preserved, when waste products from the processing of crops were used as fuel, including weed-rich remains from cereal processing. The remains recovered are generally well-preserved, being heavily carbonized. A small number of carbonized dung pellets and dung fragments of caprines were recovered. When broken, these pellets rarely contained seed remains and the seeds they did contain were not from the main taxa identified from the archaeobotanical samples. We have not tried to quantify these dung remains other than noting their presence or absence.

Obviously the plant remains recovered result from specific cultural and economic processes. Recognizing this, we must also recognize certain plants could have been used at Essouk-Tadmakka in a way which means they do not show up archaeobotanically. Likewise, the relative quantities of taxa recovered do not necessarily entirely reflect the importance of these taxa at Essouk-Tadmakka as the sampling program does not purport to provide a definitive account of plant remains at the site.

Full details of species counts of all specimens recovered are shown in Table S1 (Supplementary material). Taxa identification follows the nomenclature of The Useful Plants of West Tropical Africa (Burkill 1985–2000). For species not included in Burkill, the nomenclature is that of Ozenda (1991). Unless otherwise stated, a specimen indicated in the table refers to the seed/grain of the plant. Also, unless a specimen is noted as desiccated (desicc.) or silicified (silic.) it is carbonised. Table 2 provides basic counts of important taxa identified at Essouk, their relative frequency and ubiquity, and presence/absence by period. Fig. 4 provides a more detailed illustration of relative frequency and ubiquity, comparing remains from Period 4 with the combined remains from all other periods. Table 3 compares the presence/absence of a selection of taxa at Essouk with evidence from other relevant sites. Figures 5 and 6 provide SEM, ESEM and photographic images of selected plant remains excavated.
Table 2

Taxa counts, relative frequency (percentage of all counted seeds) and ubiquity (percentage of samples in which species is present), and presence/absence by period of important plant taxa of Essouk. See Table S1 for full taxa counts

 

Taxa counts

Frequency (%)

Ubiquity (%)

Period 4

Period 3b

Period 3a

Period 2

Period 1

(a.d. 1300–1400)

(a.d. 1100–1300)

(a.d. 950–1100)

(a.d. 750–950)

(pre ca. a.d. 750)

Cereals

141

9.2

60.7

X

X

X

X

X

 Brachiaria/Setaria

17

1.1

25

X

X

X

X

 

 Digitaria

2

0.1

3.6

 

X

   

 Echinochloa

64

4.1

35.7

X

X

X

X

 

 Panicum

17

1.1

21.4

X

X

X

X

 

 Pennisetum

5

0.3

10.7

  

X

X

X

 Triticum

16

1.0

28.6

X

X

X

X

 

Legumes

387

25.2

42.9

X

X

X

 

X

 Acacia

8

0.5

17.9

X

 

X

 

X

 Small legumes

376

24.4

25

X

X

  

X

Fruits

58

3.8

39.3

X

X

   

 Balanites

3

0.2

10.7

X

    

 Citrullus

1

0.1

3.6

X

    

 Hyphaene

3

0.2

3.6

X

    

 Phoenix

3

0.2

7.1

X

    

 Ziziphus

47

3.0

14.3

X

    

Oil/fibre

4

0.3

10.7

X

X

   

 Gossypium

3

0.2

7.1

 

X

   

 Linum

1

0.1

3.6

X

    

Wild/weed

107

7.0

46.4

X

X

X

X

X

 Cyperus

31

2.0

14.3

X

   

X

 Chenopodium

19

1.2

14.3

X

X

X

  

 Tribulus sp. seed

21

1.4

14.3

X

X

  

X

 Tribulus sp. pod

13

0.8

14.3

X

X

   

Other

842

54.7

82.1

X

X

X

X

X

https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig4_HTML.gif
Fig. 4

Illustration comparing relative frequency and ubiquity of plant taxa from Essouk Period 4 versus combined remains from all other periods. Frequency indicates the percentage of all counted seeds. Ubiquity indicates the percentage of samples in which the taxon is present

Table 3

Presence/absence of selected taxa at Essouk and other relevant sites (Fuller 2000; McIntosh 1995; Murray et al. 2007; Murray 2004; Pelling 2005, 2008)

Taxa

               

Site:

Cubalel (Senegal)

Jarma (Libya)

Dia-Shoma

Dia-Shoma

Jenne-Jeno

Dia-Shoma

Jarma (Libya)

Essouk

Jenne-Jeno

Jenne-Jeno

Dia-Shoma

Dia-Mara

Gao

Jenne-Jeno

Dia-Shoma

Period:

Iron Age

Iron Age

Iron Age

Late Iron Age

Late Iron Age

Early Islamic/LIA

Pre Islamic

Early Islamic

Early Islamic

Mid Islamic

Islamic

Islamic

Islamic

Late Islamic

Post medieval

Oryza sp.

X

 

X

X

X

X

  

X

X

X

X

X

X

X

Pennisetum glaucum

X

X

X

X

X

X

X

X

X

X

X

X

X

X

 

Pennisetum sp. wild

  

X

X

      

X

X

  

X

Sorghum bicolor

 

X

  

X

X

X

**

X

X

X

X

 

X

 

Triticum aestivum/durum

 

*

    

X

X

  

X

X

   

Brachiaria sp.

X

X

  

X

 

X

X

 

X

X

X

 

X

X

Digitaria sp.

X

     

X

X 

X

 

X

X

 

X

 

Echinochloa sp.

X

      

X

     

X

 

Panicum laetum (incl. cf.)

       

X

 

X

 

X

 

X

 

Panicum turgidum (incl. cf.)

      

X

   

X

X

   

Other Paniceae/Panicum sp.

X

 

X

  

X

X

X

  

X

X

   

Paspalum sp.

X

 

X

X

X

X

    

X

X

   

Setaria sp. (incl. cf.)

X

     

X

    

X

   

Citrullus lanatus (incl. cf.)

      

X

X

    

X

X

 

Gossypium sp.

 

X

   

X

X

X

  

X

X

X

 

X

Linum usitatissimum

      

X

X

       

Balanites aegyptiaca

       

X

    

X

  

Phoenix dactylifera

 

X

    

X

X

    

X

  

Hyphaene thebaica

  

X

X

 

X

 

X

  

X

X

X

 

X

Ziziphus sp.

X

 

X

X

 

X

X

X

  

X

X

X

  

* Free-threshing wheat not found but Emmer wheat (Triticum dicoccum) found

** Sorghum not found in macro remains but found in the temper of pottery at the site

https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig5_HTML.jpg
Fig. 5

SEM, ESEM and photographic images of selected Essouk cereal remains: aBrachiaria/Setaria sp. grain (EKA 77). bBrachiaria cf. lata lemma, interior (EKA 5). c, dDigitaria sp. grains (EKA 79). egEchinochloa sp. grains (EKA 79, 79, 77). hEchinochloa sp. lemma (EKB 6). iEchinochloa sp. palea (EKA 85). j Close-up image of silicified grass husk-rich sample—mainly Echinochloa sp. (EKB 6). kPanicum sp. grain viewed obliquely from ventral and lateral views (EKA 62). l, mPanicum sp. lemma and lemma apex detail (EKA 85). nPennisetumglaucum grain (EKB 4). oqTriticum durum/aestivum grains, ventral views (EKA 62, 5, 45). r, sTriticumdurum/aestivum grains, dorsal views (EKA 77, 62). tTriticum durum basal rachis (EKA 77). uTriticum durum rachis (EKA 77). vTriticumaestivum rachis (EKA 83). w, x Cast of ceramic impression of Sorghumbicolor cf. race bicolor lemma with attached rachis and lemma detail showing rows of unicellular trichoms (EKA 75, 77)

https://static-content.springer.com/image/art%3A10.1007%2Fs00334-010-0279-6/MediaObjects/334_2010_279_Fig6_HTML.jpg
Fig. 6

SEM, ESEM and photographic images of selected Essouk legumes, fruits, oil/fibre, wild and other plant remains: a cf. Acacia sp. frag. (EKA 29). b cf. Fabacae type A seed (EKA 3). c cf. Crotalaria sp. seed (EKA 13). dBalanites aegyptiaca endocarp (EKA 30). eCitrullus lanatus seed tip (base) (EKA 18). f, gHyphaene thebaica seed fragment of hard endosperm (‘g’ shows the interior surface of ‘f’) (EKA 16). hPhoenix dactylifera seed (EKA 43). i, jZiziphus sp. seed and endocarp (EKA 3). kGossypium sp. seed with charred lint (EKA 77). l, mGossypium sp. seed fragment in cross-section and detail of seed coat cross-section (EKA 76). n, oLinum usitatissimum seed tip (base) and detail of interior surface of testa—visible are testa fibre cells, (EKA 18). prCyperus sp. nutlets (EKA 29). sChenopodium sp. seed (EKA 79). t Portulacaceae cf. Talinum sp. (EKA 4). u Portulacaceae/Caryophyllaceae (EKA 4). v, wTribulus sp. seed and pod (EKA 79). xHackelochloa granularis spikelet (EKA 89)

Details of individual taxa follow.

Cereals

The only identified cereal in the earliest deposits (Period 1) is pearl millet (Pennisetumglaucum), the most important dry savannah to Sahel cultivar (Fig. 5n). Pearl millet is seen again in Period 2 and 3a but is not recorded in the large numbers found in Periods 3b and 4. This species is inferred to have been domesticated in the third millennium b.c. in the far western Sahel, perhaps in Mauritania and/or northeast Mali (Fuller 2003, 2007; Kahlheber and Neumann 2007). The grains of Pennisetum recovered at Essouk fall within the size range of cultivated pearl millet (P. glaucum) (Brunken 1977; Zach and Klee 2003; Fuller 2007); no millet chaff was recorded at Essouk. Pennisetum glaucum has been found on the Malian Neolithic sites of Karkarichinkat from 2500 to 2000 b.c. (Manning et al. 2011) and Windé Koroji from ca. 2000 b.c. (MacDonald 1996), and is known from Mauritanian ceramic impressions from ca. 1700 b.c. onwards (Amblard and Pernes 1989; Fuller et al. 2007). In addition to being a staple cereal, the species is used as animal fodder, for beer and for medicine (Burkill 1994, pp. 313–317).

The Essouk assemblage contains two seeds of Digitaria sp. (Fig. 5c, d), found in Period 3b, most closely resembling fonio (D. exilis), one of the two cultivated Digitaria species used as cereals in West Africa. Fonio has several possible centres of origin across the West African savannah belt (Harlan 1992; Blench 2006, pp. 209–210). Digitaria likely comes from further south in the savannah belt, mostly in regions with ~1,500 mm rainfall (Portères 1976). This is in contrast to Pennisetum glaucum which is seen to have its origin in the more northerly Sahel and grows comfortably with only ca. 200 mm of rainfall. Elsewhere, south of the Sahel, cf. D. exilis and P. glaucum are reported to co-occur at Cubalel by a.d. 500 (Murray et al. 2007) and at Jenne-Jeno some time after a.d. 800 (McIntosh 1995). In this region, the genus Digitaria contains generally good fodder plants (Dalziel 1937; Burkill 1994, pp. 222–231) and other wild species are collected for food (Harlan 1989). It is also used for beer and medicine (Burkill 1994, pp. 222–231). Hutchinson notes that there are twenty six species of Digitaria in West Africa (Hutchinson and Dalziel 1927–1936) and there needs to be more work on distinguishing Digitaria species in the region. There is one species (D. commutata) that is known from the Saharan massif regions (Ozenda 1991, p. 158).

Although not recorded in Period 1, Echinochloa sp. is present in all other periods and is the single most common grass recorded at Essouk (Fig. 5e–j)—following a high in Period 3a/b it is less common in Period 4. Of the five species present in West Africa (Hutchinson and Dalziel 1927–1936), the most economically important is E. stagnina, a swamp grass often prepared like rice and collected in particular in the inland Niger Delta (Harlan 1989; Cissé 1991). All of these Echinochloa species are associated with pools, wet places and water edge habitats (Hutchinson and Dalziel 1927–1936). Other species, e.g. E. pyramidalis, are also gathered in the Sahel—along water courses—under the general term ‘kreb’ (Burkill 1994, p. 86; cf. Harlan 1989, p. 86). E. colona is also appreciated as a minor cereal, and is used to produce beer in central Africa (Burkill 1994, p. 233). Ozenda (1991, p. 158) reports E. colona to be found occasionally in the southern Sahara, and we expect this to have been focused on locally wet zones (e.g. escarpment seepages, semi-sheltered hill slopes). Echinochloa has been found in the late Islamic levels at Jenne-Jeno (McIntosh 1995). Wild Echinochloa featured in the plant collection tradition of the Early-Mid Holocene Sahara, evidenced from sites in the Tadrart Acacus (Wasylikowa 1992; Mercuri 2001) and the Egyptian Western Desert (Barakat and Fahmy 1999; Wasylikowa and Dahlberg 1999). The sample of husk fragments of Echinochloa (with some minor admixture of Brachiaria) from EKB 6 (see Fig. 5j) attests to the dehusking of this taxon, presumably for human consumption.

Plant remains of the Brachiaria/Setaria group were identified from Periods 2–4 (Fig. 5a, b). Analysis of the husk remains allows more specific identification, suggesting gathering of a Brachiaria species. The wideness and flatness of the spikelet suggest Brachiaria rather than Setaria, seemingly confirmed by examination of the lemma and palea rugose patterns. The lemma rugae pattern and form of the acuminate lemma apex both are a better match for B. lata than for B. deflexa or B. ramosa in the limited reference material examined. While B. deflexa, black fonio, is known in a domesticated form (Harlan 1992; Blench 2006), wild populations of all three Brachiaria spp. are reported to be collected as wild food in West Africa (Harlan 1989; Burkill 1994, pp. 191–194). Of these species, B. lata has the more northerly Sahelian distribution where it is widely gathered. Most often today it is found in disturbed and agricultural soils, so it is not restricted to water edge habitats to the degree that Echinochloa is, but to locally richer soils which might have included wadi margins, seepages and areas around the settlement. Brachiaria remains referred to as B. ramosa (potentially difficult to distinguish from B. deflexa) were in abundance at Jenne-Jeno (McIntosh 1995), but it is unclear if these were cultivated or gathered black fonio. As one moves southwards into the savannahs and forest margins of West Africa with higher rainfall, there are numerous (about 15) additional Brachiaria spp., often rainfed and less restricted to richer and locally wetter soils (cf. Hutchinson and Dalziel 1927–1936).

Panicum remains are found in consistently low quantities at Essouk in Periods 2–4 (Fig. 5k–m). With over forty species in West Africa, specific identification is problematic, but several are reported to be used as food, ranging from the more southerly savannah P. laetum and P. subalbidum, to the northerly Saharan P. turgidum (Cissé 1991; Harlan 1989; Burkill 1994, pp. 296–306). Panicum repens, which has somewhat narrow grains, is reported across the southern Sahara (Ozenda 1991, p. 157). Essouk specimens, which are fairly broad, are comparable to, but not definitively, P. laetum. P. laetum in particular is an important wild resource in the region, especially during times of scarcity and in areas of marginal subsistence (Burkill 1994). P. laetum is also present in the Islamic levels at Jenne-Jeno (McIntosh 1995) and Dia (Murray 2004). P. laetum occurs on locally wet soils in the Sahel zone and is reported to be gathered by Tuareg groups (Burkill 1994). We expect it to have been found occasionally in the wettest micro-environments around Essouk-Tadmakka.

Wheat, both grains and rachis remains, is found at Essouk from Period 2–4 (Fig. 5o–v). For a cultivar not normally encountered in West Africa the quantities of remains recorded are significant. Also a lens of what was believed to be semolina was found in Period 3b, although unfortunately no images or sample was taken. When well-preserved, grains appear to be identifiable as free-threshing (Triticum durum/aestivum s. l.). A free-threshing wheat grain was AMS dated to cal a.d. 1020–1150 (see Nixon 2009 for all AMS dates quoted herein) but wheat is also found in lower contexts dated back to ca. a.d. 750. Identifiable rachises were few, but both T. durum and T. aestivum rachises were clearly identified. These wheats are not traditionally West African cultivars, but were established in prehistory in the Mediterranean (Zohary and Hopf 2000). The earliest wheats in the Sahara and the Nile Valley were predominantly the glume wheat emmer (T. dicoccum) which became established in Saharan oasis areas by ca. 1000 b.c. (van der Veen 1992; Pelling 2005). The first occurrence of free-threshing wheats in these regions is in Hellenistic or Byzantine/Late Garamantian times (Pelling 2005, 2008). A shift to durum wheat in North Africa is associated with the Islamic period based upon data from Setif in Algeria (Palmer 1991). There is minimal ethnographic evidence for traditions of cultivating wheat south of the Sahara (Burkill 1994, p. 372), but there do exist early Islamic historical references to wheat cultivation in the Sahel (Lewicki 1974, pp. 22–26; Murray 2007). Free-threshing wheat has previously been seen from Islamic period Dia (Murray 2007). Couscousiere vessel fragments, recorded from Periods 2–4 and first recorded in the same context as the earliest recording of wheat, might have been used for wheat preparation, although it is possible they were used for preparing small-grained millets. If grown locally, presumably over winter, wheat must have been watered, although importation over the desert from North Africa is also possible.

Sorghum was recovered but only as temper in a pottery type found in large quantities, mainly in Period 3b (ca. a.d. 1100–1300). Details of the morphology of the sorghum in the pots are comparable to modern Sorghum bicolor race bicolor, in terms of overall shape and size (Fig. 5w), but also in terms of uniseriate/unicellular trichomes on the interior of the glume (Fig. 5x). While the later trait is a good match for sorghum, as distinct from other millet grasses and crops considered here, further comparative research is needed to establish whether this is diagnostic of sorghum alone. All examples came from a low-fired ware (Cord-Wrapped-Stick Grid-Impression ware) stylistically similar to pottery from the Niger Bend region (Nixon 2009). Given that these pots have been shown to be imports to Essouk-Tadmakka from the Niger Bend, this is not evidence of sorghum processing or consumption at Essouk-Tadmakka. The macro remains showed no sorghum despite references to this in the Early Arabic sources.

Legumes

Small legumes were found in great quantity in Period 4 and were also present in Periods 1 and 3b. These present a good match for Crotalaria seeds, although we are not aware of diagnostic criteria for assigning these with certainty to that genus, and we have therefore recorded them in Table S1 as cf. Crotalaria. It is possible these were weeds from cereal fields, though numerous taxa in this family provide ideal graze for animals (Burkill 1995). Several native small legumes are important amongst Tuareg pastoralists for their use as pasturage, for instance Trigonella anguina and Crotalaria saharae (Benchelah et al. 2000, pp. 168, 175). While it is possible that certain of the legume remains came from dung fuel that has broken down, this seems an inadequate explanation as recovered carbonised dung pellets rarely contained seed remains. It is therefore possible that groups at Essouk-Tadmakka were gathering and eating wild, small-seeded legumes, but also possible they gathered the plants for fodder but removed the potentially toxic mature seeds. Crotalaria saharae favours wadi beds in the Sahara (Benchelah et al. 2000), and is endemic across the western Sahara (Ozenda 1991).

Cf. Acacia (Fig. 6a) is present in Periods 1, 3a and 4 but not in great quantities. An Acacia specimen from Period 3a was AMS dated to cal a.d. 1020–1150. Acacia trees are typical of the Sahel and open dry savannah and there are numerous Acacia species in Mali, and across the Sahara (Ozenda 1991). Acacia trees are used in various ways, such as for animal fodder, building material, fuel, tanning leather, rope making (tree fibres), and medicine (Burkill 1995, pp. 177–203).

Fruits

Fruit is a very common component of the Period 4 assemblage but only 1 specimen of fruit is seen prior to Period 4. The remains consist primarily of fragmented seeds and endocarps of Ziziphus (Fig. 6i, j). A Ziziphus specimen from Period 4 was AMS dated to cal a.d. 1295–1390. Five species are present in West Africa: Z. abyssincia, Z. lotus, Z. mauritiana, Z. mucronata and Z. spina-christi, all of which have edible fruits (Dalziel 1937; von Maydell 1986; Burkill 1997). Ziziphusmauritiana and Z. mucronata are both reported as more common in the Soudanian savannah zone, while the other three species are all reported to occur in the Sahel. Ziziphus fruits are edible fresh, dried, or ground into flour that makes long lasting bread, cakes and beverages (Cissé 1991; Dalziel 1937; von Maydell 1986). The genus is also used as a living hedge for cattle enclosures and is important for shade (Burkill 1997). Ziziphus is widely reported from other sites in the region (Table 3).

Other fruits present include small numbers of watermelon (Citrullus lanatus), date (Phoenix dactylifera) and balanos (Balanites aegyptiaca), all found only in Period 4 (Fig. 6d, e, h). Interestingly, watermelon was earlier present in the central Sahara at Jarma (Pelling 2005, 2008). Watermelon is a native of tropical and subtropical Africa, and may have been formerly distributed in the Early/Mid Holocene Sahara (Wasylikowa 1992; van der Veen 1992; Wasylikowa and van der Veen 2004). Nevertheless it appears to have been reintroduced as a cultivar in the Iron Age (cf. Pelling 2005, 2008). As in many parts of Africa today, watermelon may have been cultivated primarily for its oil producing seed and cooked rind, rather than its juicy flesh, since this was likely to have been bitter and unpalatable prior to the development of improved varieties (Watson 1983; FAO 1988; Murray 2000). Watson (1983) however suggests improved juicy-fruited varieties may have spread widely in the early Islamic period. As noted by Harlan (1992, p. 64), watermelon and pearl millet are cultivars associated with the driest margins of agriculture in the Sahel and the fringes of the Kalahari.

As for the date palm (Phoenix dactylifera), virtually all parts of the tree are utilised (e.g. Murray 2000). The nutritious date fruits have considerable food value and are widely consumed in the region today. Phoenix dactylifera had already reached the Sahel in earlier times of Arabic trans-Saharan trade, finds at Marandet in Niger, for example, have been dated to the 7th to 9th century a.d. (Magnavita et al. 2007). All parts of balanos (Balanites aegyptiaca), the ‘desert date’, are useful, for crafts, firewood, charcoal and penning cattle, for its edible leaves, but especially for its oil bearing seed. The tree also produces useful resin and strong fibre and has many medicinal qualities (Burkill 1985, pp. 242–246). Balanos is indigenous in the Sahel and there are records of wood charcoal as well as of fruit remains from sites in Burkina Faso and Nigeria from Late Stone Age times on at the latest (e.g. Höhn 2002; Kahlheber 2004; Kahlheber and Neumann 2007; Klee et al. 2000; Magnavita et al. 2002). The only other identifiable fruit found was Hyphaene thebaica, three fragments of which were found in Period 4 (Fig. 6f, g) (see Hillman et al. 1989 for comparison). The Hyphaene mesocarp is sweet and aromatic, and may be pounded into a meal, while the hard white endocarp is used as vegetable ivory (Burkill 1997, pp. 371–373).

Oil and fibre plants

Important oil and fibre plants are present at Essouk although in limited quantity, consisting of three cotton specimens from Period 3b (Fig. 6k–m), one of which was AMS dated to cal a.d. 1020–1150, and a linseed specimen from Period 4 (Fig. 6n, o).

The cotton seeds and seed fragments need not indicate local cultivation or cotton ginning (seed removal) on site. Instead it can be suggested that raw cotton was imported to the site for spinning, but that the imperfections of ginning meant that some seed waste material remained in the cotton. The later processing of cotton is testified to by the presence of spindle whorls (in Periods 3b and 4). Cotton is susceptible to drought, but thrives in temperatures between 20 and 40°C (Burkill 1997, p. 20), which certainly would have occurred at Essouk-Tadmakka. Cotton cultivation at Essouk-Tadmakka is possible, but water requirements make large scale cultivation unlikely. It is worth noting that cotton is known from several Saharan oases by Late Roman times, such as the Libyan Fezzan, supplied by foggera irrigation (Pelling 2005), and the Egyptian oases (Clapham and Rowley-Conwy 2009), so some cultivation at Essouk-Tadmakka could be seen to extend to such practices. However, if spindle whorls are taken to indicate routine spinning, it is perhaps more reasonable to attribute the small and fragmentary cotton remains to impurities in imported raw cotton that had been grown and ginned elsewhere.

Despite attempts with SEM to identify distinguishing features of seed coat cross-section among Gossypium spp. (as suggested by Chowdhury and Buth 1971), we were unable to satisfactorily determine species (cf. Fig. 6m). While G. herbaceum is regarded as truly wild in sub-Saharan Africa (further South) (Hutchinson 1959; Wendel 1995), G. arboreum is generally regarded as a domesticate from Pakistan (Zohary and Hopf 2000, p. 134; Fuller and Madella 2001, p. 337; Fuller 2008). Cotton became a crop in North/Eastern Africa from the Roman period, based on finds in Nubia, Egypt and Libya (Pelling 2005, 2008; Clapham and Rowley-Conway 2007, 2009; Wild et al. 2007), but sub-Saharan West African finds are all from Islamic contexts (Fuller 2000; Murray 2007). Corresponding with historical evidence (Kriger 2005), cotton is commonly found on Malian sites of early Islamic date onwards (Table 3).

Linseed/Flax (Linum usitatissimum) is present at Essouk, although only a single specimen was recovered (it is recognizable from the curved seed base and the morphology of fibre cells on the interior of the seed coat—see anatomical comparisons in Vaughan 1970, p. 142; Boesewinkel 1984, p. 446). It is an interesting find as, thus far, it appears to be absent from other West African sites, although it is known from the central Sahara at Jarma (Table 3). It is an important species due to its usefulness as fibre (flax) and its edible oilseed (linseed). Like wheat, it is normally a winter grown crop. As we found no reports of traditional cultivation in sub-Saharan West Africa (cf. Burkill 1994) we must at this point infer it to have been imported from the north. As this is a solitary and fragmentary specimen, and the first of its kind from the region, further evidence should be sought before concluding that this species was of importance in Islamic West Africa.

Additional wild/weed taxa

Identifiable wild/weed taxa consistently occur throughout the sequence in low numbers. The presence of sedges (Cyperus sp.—Fig. 6p–r) suggests wet localities near the site, although many Cyperus spp. can grow without standing water. It seems probable these seeds might have been gathered while gathering millet grasses. Chenopodium seeds (Fig. 6s), found throughout, could be weeds collected with crops or other millet grasses, but may have been processed as food in their own right. C. murale in particular is reported to be consumed in parts of the Sahara and distributed in the Sahel, being found especially in disturbed ground (Benchelah et al. 2000, p. 155; Burkill 1985, p. 367). The Portulacaceae type is seen in Period 4 in relatively high quantities. Tribulus sp. also appears to be present in reasonably frequent occurrence in Periods 3b and 4 (Fig. 6v, w). While having broad similarities to a number of illustrated comparative samples from Africa and elsewhere (Ozenda 1991, p. 321; Le Maout and Decaisne 1873, pp. 303–304; Townsend and Guest 1980, pp. 289–293) it was not possible to more specifically identify the specimens present. In the Saharan regions however, T. alatus, T. ochroleucus, T. terrestris and T. bimucronatus are recorded as being important taxa, in desert wadis and the Sahel zone (Benchelah et al. 2000, pp. 225–226; Ozenda 1991, pp. 320–322). They are noted both as an animal browse (Benchelah et al. 2000; Ozenda 1991; Burkill 2000) and occasionally as flour for human consumption (Benchelah et al. 2000, pp. 225–226). Hackelochloa granularis (the only Hackelochloa species known from West Africa), identified based on visual comparison with a number of accurate drawings (Gardner 1952, pp. 306–308; Soerjani et al. 1987, pp. 430–431), is known as a weed of cultivation and makes good fodder for horses, although probably not for other animals (Burkill 1994, pp. 257–258; Hutchinson and Dalziel 1927–1336, pp. 505–506). It is short-lived and therefore never abundant in the landscape. It is reported mainly amongst crops in the Soudanian savannah zone and would be expected to be scarce in the Sahel, but might occur in gardens there. Other wild/weed taxa present have so far only been identified to probable family.

Discussion

Defining broad patterns in the Essouk evidence

The archaeobotany, while quite limited, has identified some of the taxa that were present and seemingly important at Essouk-Tadmakka in a period for which historical records provide practically no information. It also identifies potential evidence for change in plant use over time.

During the earliest period of Essouk-Tadmakka’s occupation (Period 1/pre-ca. a.d. 750), at around the time the site was likely to have started functioning as a trans-Saharan trade centre, the cultivar pearl millet was identified, as well as small legumes and certain wild taxa. With the first clear evidence for permanent settlement and significant trans-Saharan trade (beginning of Period 2/ca. a.d. 750) we see continued evidence for pearl millet, but also wheat. The remainder of Period 2 through Period 3b (ca. a.d. 750–1300) sees continued evidence for wheat, wild taxa and legumes, and also a strong presence of the wild millet-grasses, Echinochloa, Brachiaria and Panicum. In Period 3b we also note cotton as a new presence and the absence of pearl millet.

The archaeobotany from ca. a.d. 1300 to 1400 (Period 4) provides far more evidence. While there is some continuity with the pre ca. a.d. 1300 data, certain changes do seem to be apparent, although we must be cautious in interpreting these patterns in light of the fact that this is not an extensive sampling of the site. Firstly we see a markedly reduced cereal presence with many samples containing no cereals. We also seem to see a different emphasis on the wild grasses collected with Brachiaria and Panicum being more common than the previously ubiquitous Echinochloa, and wheat notably less present. Small legumes are far more common, in particular Crotalaria-type. Fruit is ubiquitous, with Ziziphus seen in the largest quantities and the first recordings of date, balanos, watermelon, and Hyphaene. Cotton is not present and linseed is seen for the first time. Amongst the wild/weed taxa Tribulus is seen in significantly reduced quantities and Cyperus, not seen since Period 1, is a noteworthy presence. The latter, which correlates with the shift towards Panicum and Brachiaria over Echinochloa, could suggest a shift in the nature of moist microenvironments, perhaps towards lower water tables, if one assumes that Echinochloa prefers more water than these other species. This remains speculative until reliable species level identification and autoecological data are available. Interestingly the Period 4 archaeobotanical record corresponds with the 14th century historical evidence referring to grain shortage and fruit richness (Levtzion and Hopkins 2000, p. 274).

Given that we see a radical break ca. a.d. 1300 in the architecture, fauna and material culture (Nixon 2009), it would not be surprising to see signs of a change in the archaeobotanical evidence as well. The archaeological changes at Essouk have been interpreted as suggesting that 14th century Essouk-Tadmakka saw a major change, with incoming Saharan Berber groups infiltrating the Sahelian merchant culture there (Nixon 2009). This shift probably relates to a broader pattern of movement of Saharan Berber groups into the Sahel as the Empire of Mali loses its control over Sahelian trading towns (Nixon 2009). Interestingly a “Saharan plant package” is evident later at Gao and Jenne-Jenno associated with the Songhay dynasty, of Saharan Berber descent (Fuller 2000; McIntosh 1995). We take this “Saharan plant package” to include the fruits Phoenix dactylifera, Balanites aegypitaca and Citrullus lanatus. Hyphaene, although available in West Africa may also be considered to have become more widely used. The extent to which the availability of Gossypium sp. may also be connected also deserves further consideration.

Cultivation and gathering at Essouk-Tadmakka

Essouk’s contemporary environment is extremely arid and offers poor soils for cultivation. Even allowing for the existence of a slightly wetter environment in early medieval times, it seems likely there was limited opportunity for cultivation without irrigation, probably by wells. The archaeology at Essouk has so far not demonstrated any evidence for irrigation. Historical sources do not mention local cultivation and moreover refer specifically to ‘gathering of grain which grows without being tilled’ (Levtzion and Hopkins 2000, pp. 84–87). All this said, a complete absence of cultivation at such a large settlement seems unlikely.

The evidence for pearl millet, a highly drought tolerant crop, is in line with the long tradition of its cultivation in the Sahel. Also, while it is likely some or even most of the wheat recorded was shipped across the Sahara—grain imports being a necessity for such marginal sites as Essouk-Tadmakka—the presence of rachises (indicating threshing) raises the possibility that wheat was cultivated locally with irrigation. Given the comparative historical evidence from early Islamic Tegdaoust in Mauritania (Lewicki 1974, pp. 22–26) and the obvious local demand for wheat at Essouk-Tadmakka, it does seem highly likely that some form of experimentation with irrigated wheat was attempted. Cotton was also possibly a limited local cultivar. However for both wheat and cotton we must be aware of the possibility that crop-processing waste (rachises, ginned seed fragments) could have been unintentionally included amongst imported processed crops—we are at present unclear to what extent this would have occurred.

The Period 4 fruit evidence also possibly indicates cultivation, including date palm and watermelon, and the shade from date trees would have provided a context for small fields and gardens of cereals and vegetables. A similar format is seen in the contemporary irrigated cultivation of tomato plants at Essouk, and occasional date palms (author’s observation). While there were probably no flowing fields of cultivated crops in Essouk-Tadmakka’s surroundings, we infer a patch scale cultivation in and around the town.

The evidence provides greater insight into the nature of wild gathering at Essouk-Tadmakka, vaguely referred to by the historical sources. The evidence does suggest the Essouk-Tadmakka population used taxa which were not cultivated and begins to provide us with an idea of the types of cereals being gathered. These include Echinochloa, Brachiaria and Panicum which, while just possibly semi-cultivated or tended, were most likely to have been simply wild-gathered. These grasses would have been patchily available in wetter microenvironments, where either ground water seepage occurred or where light summer rains accumulated. Slightly wetter climatic conditions that were probably prevailing at the time (Grove 1993, pp. 39–42; Brooks 1998) would have been likely to have made these more common than is the case at present. Echinochloa, which is the most common wild grass through the first three periods, seems the most probable contender for being al-Bakri’s ‘grain which grows without being tilled’ (Levtzion and Hopkins 2000, pp. 84–87).

While it is possible to suggest these grasses result from other processes than human consumption, on balance this seems unlikely, especially in view of the sample (from EKB 6) rich in silicified grass husks which is almost certainly the remains of dehusking of these grasses for human consumption. An argument suggesting the small Panicoid grasses could be weeds of pearl millet cultivation seems unlikely as pearl millet is taller than most other grasses and is typically harvested by cutting spikes, therefore meaning shorter weeds growing amongst it do not tend to be accidentally harvested when it is collected (Reddy 2003). The small grasses are not typical weeds of wheat cultivation, as Panicoids are normally summer flowering and wheat is normally winter-grown. Also, they do not show any strong pattern of co-occurrence with wheat. As these wild genera are well-known ethnographically recorded sources of food on several continents, their occurrence with the apparent absence of other taxa of savanna grasses points towards deliberate selection of useful food grain species. It cannot of course be ruled out that these were for fodder as seems likely to have been the case with the small legumes and several of the other wild grasses.

Sorghum and other grains from the land of the Sudan

The historical records refer to the import of sorghum and other grains from further south in West Africa (Levtzion and Hopkins 2000, pp. 84–87). This does seem a very likely scenario for this relatively marginal settlement to provision itself, as zones of cereal production in the south were relatively nearby and trade caravans travelled from there. However it was striking that in the flotation samples, neither sorghum nor any other obvious crops from the south were found. Sorghum was only evident from seed impressions in pottery imported from the south. While it is possible that the archaeology is indicating that grain import from the south was less extensive than suggested by the historical records, the archaeobotanical absence of a range of grains from the south does not mean that they were not at Essouk-Tadmakka. We conclude that it is most likely any southern grain imports were arriving at Essouk-Tadmakka ‘clean’, requiring no processing, and were less likely therefore to be preserved. Also grain imports would be valued, reducing the likelihood of their loss and consequent archaeobotanical recovery. We must obviously also accept the possibility that future investigations might show that imported southern grains are archaeologically recoverable in a different area of the site.

Wheat

The assemblage of wheat from Essouk is remarkable in being the earliest and largest archaeologically recovered from West Africa (cf. Murray 2007). The finding is surprising as not only is there little historical evidence for wheat in West Africa and limited archaeological precedent, but certain historical sources explicitly state that wheat was not at Essouk-Tadmakka (Levtzion and Hopkins 2000, pp. 50–51). As explained above, the extent of local cultivation of wheat, versus its importation, is unclear. It is quite possible that the rachises recovered were simply included in sacks of imported wheat. However, as discussed above, it does seem likely that some form of experimentation with irrigated wheat was attempted at Essouk-Tadmakka.

Regardless of whether it was imported or grown locally, the wheat evidence is significant new data reflecting transfer of cereal food practice from North Africa. For discussion of the formation of Islamic identity in the Sahel this evidence is important, as the early Arabic writers clearly identify wheat as an important element of civilized Islamic dietary regimes (Levtzion and Hopkins 2000, pp. 50–51). While the historical evidence for wheat consumption at Tegdaoust (Lewicki 1974, pp. 22–26) suggests wheat was also consumed at other sites in the arid zone Sahel, it is unclear whether more significant archaeobotanical evidence for wheat at any sites further south in West Africa will be found (Table 3; Murray 2007). One further point to make is that from our limited sample we found evidence for both Triticum durum and Triticum aestivum species of free-threshing wheat. The field record of possible semolina is also of interest.

Conclusions

The archaeobotanical finds from Essouk-Tadmakka both precede historical evidence for plant usage at this important Sahelian trans-Saharan trading site and move beyond the inaccuracies and limitations of early Arabic documentation of plant usage there. The data provide new insights into Essouk-Tadmakka’s gathering traditions, providing clear identifications of significant gathered taxa (including Echinochloa, Brachiaria and Panicum) only ever imprecisely described in historical documents. It also provides insights into possible cultivation at Essouk-Tadmakka, including pearl millet, wheat and various fruits. Additionally, we observe that imported crops from the south said to be present were not evidenced archaeologically. A key element of the assemblage is the earliest and largest archaeological assemblage of wheat from West Africa, a species important within Muslim dietary practice but until now perceived to have been practically absent from the West African Muslim diet in the medieval period. A range of other taxa important within the history of the region were also identified, including date, balanos, cotton, watermelon, and the earliest recorded linseed in West Africa. As well as providing broad insight into some of the plants used at the early Tuareg merchant centre of Essouk-Tadmakka, we also identified potential changes over time in the plant use regime that were not recorded historically, most crucially a key shift in plant usage at the site ca. a.d. 1300 identified with the arrival of new Saharan populations.

Acknowledgments

This study is part of the Essouk archaeological research project led by Sam Nixon and funded by the Arts and Humanities Research Council (PhD studentship program), University College London Graduate School, the Institute of Archaeology (UCL) and the University of London Central Research Fund. Thanks are due to the Institut des Sciences Humaines and the ‘Direction Nationale de la Patrimoine Culturelle’ in Mali for respectively authorizing and assisting in fieldwork at Essouk. Thanks are also due to the reviewers for their constructive comments, including aiding with identification of certain taxa. Some of the writing of this paper was completed while DF was a visiting researcher with Professor Sato at the Research Institute for Humanity and Nature, Kyoto.

Supplementary material

334_2010_279_MOESM1_ESM.xls (40 kb)
Table S1Species counts of plant remains excavated at Essouk (*indicates taxa observed in the bulk silicified sample from EKB 6 which was not fully studied and quantified) (XLS 40 kb)

Copyright information

© Springer-Verlag 2011