The Green, Green Grass of Home: an archaeo-ecological approach to pastoralist settlement in central Kenya

ABSTRACT This paper considers the ecological residues of pastoralist occupation at the site of Maili Sita in Laikipia, central Kenya, drawing links with the archaeological record so as to contribute a fresh approach to the ephemeral settlement sites of mobile herding communities, a methodological aspect of African archaeology that remains problematic. Variations in the geochemical and micromorphological composition of soils along transects across the site are compared with vegetation distributions and satellite imagery to propose an occupation pattern not dissimilar to contemporary Cushitic-speaking groups further north. We argue that Maili Sita exemplifies the broad migratory and cultural exchange networks in place during the mid- to late second millennium AD, with pastoralist occupants who were both physically and culturally mobile.


Introduction
Issues of mobility, ranging from population migration, through generational changes in settlement, exchange networks and human biographies, to annual and seasonal movements of livestock, materials and people, have been prominent themes in archaeological reconstructions and interpretations of the material and biological records of pastoralism in eastern Africa. Spanning over four and a half to five thousand years of human, environment and non-human species interactions, the drivers, consequences and material and biochemical traces of these different kinds of mobility are inevitably varied and manifest themselves at different spatial and temporal scales. That these different forms and scales are in some sense related is often implicit in many studies, although for quite understandable and justifiable reasons there has been a tendency for analyses to focus on single categories of mobility (e.g. the role of population migration in the inception of livestock herding in the region, or the extent of seasonal movements of settlements and livestock) rather than striving for improved understanding of the interplay between these different scales and types. In common with archaeological studies in many other parts of the world, acknowledgement of the importance of mobility to different pastoralist systems, whether in terms of livelihoods, economic strategies and adaptive capacities, or with reference to symbolic constructions, has tended to go hand-in-hand with a conceptualisation of these different mobilities as presenting particular interpretative challenges. The greater difficulty of recognising the settlements and other activity traces of pastoralists (and for that matter also of hunter-gatherer-fishers) relative to those of agricultural populations precisely because of their mobility, for example, has been a major narrative trope in archaeological interpretations of pastoralism in the region and elsewhere on the African continent (see Lane 2016a for a fuller discussion).
While recognising and acknowledging that mobility (regardless of dominant subsistence strategy) can result in highly ephemeral material traces and therefore may limit archaeological visibility and inhibit certain lines of interpretation, we argue for the deployment of a wider range of approaches to the identification of pastoralist settlements rather than a simple reliance on artefactual evidence. We draw inspiration from a number of related recent archaeological and ecological studies (see below) to explore the biocultural significance of ecological indicators (see also Lane 2016b). More specifically, we discuss the notion that archaeologists interested in pastoralist societies, rather than searching for the kinds of material residues associated with sedentary economiessuch as evidence of construction, dense refuse accumulations, etc.and long-recognised as being diminished in the context of mobile groups (e.g. Childe 1936;Gifford 1978;Robertshaw 1978;Banning and Kohler-Rollefson 1992), might instead turn their attention to the ecological changes that such groups effect upon the environments they inhabit. Such impacts are considered here in the context of the semi-arid savanna grasslands of central Kenya, taking as a central case study the site of Maili Sita, 1 probably occupied some time between the mid-seventeenth and early eighteenth centuries AD. 2 Our second concern with African mobility uses the fruits of this archaeo-ecological approach to take a more sociological perspective, whereby we assess how ethnic, cultural and economic fluidity, as well as actual population movement, presaged the emergence of the well-defined identities that characterised the region into the colonial period. We begin by outlining the development and state of current thinking surrounding the position of pastoralism as a component of broader savanna ecologies and then move on to describe how these ideas have influenced our understandings of the occupation of Maili Sita and its relation to the broader social history of the region.

Anthropogenic glades
At a global scale, savanna ecosystems can vary considerably with respect to tree and shrub coverage (Ratnam et al. 2011); in eastern Africa, these variations also occur at a local level such that these environments are mosaics of distinct ecological communities with variably diffuse or abrupt boundaries (Young et al. 1995). One driver of this heterogeneity is pastoralism, a fundamental feature of which is that livestock must be enclosed overnight for protection from predators and theft (Western and Dunne 1979). Among the Maasai, and other pastoral societies in eastern Africa, settlements are constructed first by clearing an area of vegetation, before erecting houses and fences around a central corral (Mbae 1990). Dung accumulates within this enclosure, enhancing nutrient content in local soils. This enrichment fosters distinctive sequences of vegetation succession, with rapid recolonisation of these glade sites by rich grasses that attract grazing animals, both wild and domestic (Young et al. 1995;Augustine 2003a;Veblen 2013); one of these groups of grasses, Cynodon spp., is known as 'manyatta grass' to the Maasai, 3 who recognise its association with former settlements and consider it a key forage species (Muchiru et al. 2009). Indeed, a general preference towards Cynodon spp. has been suggested for the wider grazing ungulate community (Pratt and Gwynne 1977). The presence of grazing animals within glades serves to discourage the regrowth of woody species, and their dung ensures the continued nutrient augmentation of the soil (Augustine 2003a;Muchiru et al. 2008Muchiru et al. , 2009. In response to such processes of nutrient recycling, glades can endure for considerable periods of time. Studies from across eastern and southern Africa (e.g. Blackmore et al. 1990;Causey and Lane 2005;Muchiru et al. 2008) have proposed correlations between glades or other patterns of anomalous vegetationsuch as patches of Acacia-dominated communities within Burkea-dominated environments in South Africa (Blackmore et al. 1990)and concentrations of archaeological material, supposing that these features can endure for centuries or longer. This time depth has encouraged the integration of these features into the ecological functioning of the savanna, such that they have become a vital component; for instance, phosphorus concentrations in grasses in nonglade locations in central Laikipia have been shown to be below the levels required by pregnant and lactating impala (Aepyceros melampus), while these are exceeded within glades (Augustine 2004;Muchiru et al. 2009). The localised concentration of nutrients in abandoned pastoralist settlements can also have beneficial effects at other trophic levels. For instance, invertebrate populations, specifically flies and beetles, often show 'significant positive correlations to [the] amount of dung deposited on abandoned settlements' (Söderström and Reid 2010: 187). Higher populations of insects may also partly account for the higher densities and wider species diversity of birds near these locations (Morris et al. 2008).

The study area
Maili Sita is located atop a low col across one of the northward-trending ridges that characterise the Lolldaiga Hills, a rare instance of raised topography in the southeast corner of the Laikipia Plateau, a 9000-km2 area of high-altitude (some 1500-2200 m a.s.l.) savanna in central Kenya (Figure 1). The hills, which have a maximum elevation of around 2250 m a.s.l., are an outcrop of the Precambrian African basement complex amidst the Miocene-age flood phonolite lavas that dominate the plateau. As well as their slightly anomalous geology, the Lolldaigas are ecologically distinct, falling into Zone III (agricultural highlands), according to the criteria established by Pratt and Gwynne (1977), while the remainder of the plateau is categorised as Zone IV (semiarid with marginal agriculture). That is not to say that farming is practised in the hills today; the area is encompassed by the 200 km 2 area of Lolldaiga Hills Ranch, a privately owned livestock ranch and wildlife conservancy established under the European settlement drives of the 1920s (see Lemoosa 2005;Hughes 2006), which currently holds some 4000 head of cattle.
Previous investigations in Laikipia have pointed to the importance of the region in narratives of both the emergence of early modern humans as well as the spread and development of specialised pastoralism. Middle Stone Age (MSA) and Later Stone Age (LSA) levels have been recorded at Shulumai and Kakwa rock shelters in the Mukogodo Hills on the western edge of the plateau (Kuehn and Dickson 1999;Dickson et al. 2004) and excavations of three rock shelters at Kisima Farm suggested occupation by pastoralist communities during the Pastoral Neolithic (PN, c. 5000-1200 BP) through to the Pastoral Iron Age (PIA, c. 1200-200 BP;Siiriäinen 1984). However, little subsequent work had been done prior to a series of projects initiated by the British Institute in Eastern Africa (BIEA) between 2004 and 2010, which looked at patterns of land use and environmental change across the plateau (Lane 2005(Lane , 2011Taylor et al. 2005;Muiruri 2008), with considerable attention paid towards the Lolldaiga Hills. Survey of the main valley (west of the Maili Sita col) recorded numerous sites on the basis of surface scatters of ceramics and obsidian artefacts (Causey 2005(Causey , 2008(Causey , 2010Causey and Lane 2005;Lane 2005). Based on the presence of Nderit, Maringishu and Narosura waresthe latter case being the first record from the Laikipia Plateau (Causey 2010) 4and formal similarities with sites like Ngamuriak, dated to 2135 ± 140 BP (GX8533, Robertshaw 1990), a number of these sites are attributed to the Pastoral Neolithic (PN), thus representing the first examples of openair sites from that period to be recognised in Laikipia. Other locations yielded Kisima ware, first noted at the eponymous site in levels associated with the PIA and tentatively associated with the emergent Laikipiak Maasai (Siiriäinen 1984). The most intensive investigations, however, were directed towards Maili Sita and on Mugie Ranch to the northwest (see below).

Maili Sita: the story so far
The Maili Sita col, which is approximately 500 m wide, slopes gently to the east and west into the alluvial valleys of seasonally flowing rivers. In contrast to much of the surrounding plains and valleys, which are generally open scrub woodland thickening in the vicinity of river channels, it is covered by open grassland. The setting typifies the kind of settlement location preferred by contemporary Maasai and Samburu communities in Kenya (Spencer 1965(Spencer , 1973Western and Dunne 1979). The locality's diverse archaeological record includes residues of a significant iron smelting industry (Iles and Martinón-Torres 2009;Iles and Lane 2015) and upwards of 55 stone cairns with human interment in at least one (Lane et al. 2007), clustered immediately to the northeast and south of the main site, respectively (though a third, less dense cairn complex follows the ridgeline to the north) ( Figure 2). While dates for the smelting remains match well with those for the main site (post-1670 cal AD, Beta-212297; Table 2), in lieu of independent dating evidence for the single excavated cairn it remains unclear whether these structures are coeval. There is, however, a strong tradition of cairn construction among pastoralist groups in eastern Africa (Davies 2013) and other examples of cairns in association with PIA sites have been recorded in the region (Lane et al. 2007;Lane 2011;Straight et al. 2015).
The main site, which encompasses an area of 250,000 m 2 across the centre of the col, was excavated over several seasons between 2004 and 2010. Perhaps the most notable features of the site at surface level are patches of long grass, initially vaguely recorded as 'buffalo grass' (Lane 2005(Lane , 2011Payton 2005), but recently confirmed as Pennisetum stramineum. These were taken to mark the sites of former livestock enclosures, called bomas in KiSwahili, following the ideas outlined in the previous section of soil nutrient enrichment and distinctive vegetation regrowth at such locations, many of which have been developed at the Mpala Research Station, 20 km west of Lolldaiga Hills Ranch (e.g. Young et al. 1995;Augustine 2003aAugustine , 2003bVeblen 2012Veblen , 2013Donihue et al. 2013).
The 2004 excavations recorded a linear arrangement of stake-holes towards the southern end of the site, alongside a hearth-like feature and in close association with one of the grass patches. These were thought possibly to represent the remains house of a typical Maasai-style house, comprising a dung-plastered, wattle frame (Lane 2005; see Andersen 1977). However, in contrast to the only previous record of a pastoralist domestic unit in eastern Africa, at the PN site of Ngamuriak (Robertshaw 1990), no trace of a floor platform was observed and it is possible that these features mark the fencing of a stock pen. One of the stake-holes yielded a radiocarbon date of 240 ± 40 BP (Beta-189981; Table 2), which, alongside the preponderance of Kisima ware in the ceramic assemblage, is consistent with occupation during the notional period of Laikipiak expansion (Lane 2011), always assuming that Kisima ware was associated with the Laikipiak even if it was not necessarily made by them (Grillo 2012(Grillo , 2014. Additional excavation units initially sought to establish the nature and spatial distribution of archaeological deposits, before subsequently focusing on areas identified during survey as being rich in archaeological material. The faunal assemblage was dominated by cattle and caprines, 60% of which were identifiable to taxon, an unusually high percentage that may indicate specific culinary or butchery practices (K. Mutundu, pers. comm. 2010). 5 Further excavation in 2010 across the southern (close to the 2004 unit A) and central portions of the site offered additional indication of intense activity to the south and east of the site, with dense midden deposits recorded at unit J, while dGPS survey recorded the locations of the grass patches and mapped distribution of ashy deposits visible at surface level. Figure 2 shows the distribution of the excavation units and survey data points. Alongside the 2004 excavations, detailed study was made of the soils across the Maili Sita col (Payton 2005). One transect followed the line of the ridge top, while a second ran west to east from the top of the col at its northern end, to the foot of its eastern pediment slopes, with the aim of elucidating the catenary sequence. The survey revealed a pattern of poorly developed Chromic Luvisols, rarely more than 40 cm deep and lacking a well-developed A-horizon, which appears to have been stripped leaving deposits of quartz residuum. There is evidence that the soils across the ridge and its flanks were subject to considerable erosion in the past; indeed, there are substantial erosion gullies along the ridge flanks to the south of the site, in places over six metres deep, yet of a morphology that suggests relatively recent occurrence. Payton (2005: 51) has speculated that this gullying may have been triggered by the daily movement on livestock in and out of the settlement, creating trackways that facilitated and concentrated run-off; a parallel can be seen across Lolldaiga Hills Ranch today, where graded road surfaces become heavily rutted after periods of rain. In contrast, test pitting along the ridge top has demonstrated enrichment of soils in some areas of the site, in the form of high organic carbon, exchangeable calcium and exchangeable potassium, with evidence of severe sheet erosion elsewhere. Furthermore, particular enrichment of soils underlying the P. stramineum patches (Payton 2005: 54-55) could support their designation as former boma sites and intimate that erosion occurred prior to or concurrent with the principal occupation. Thus, while pastoralist activity may have incurred a cost on the surrounding landscape, in the form of the intense gullying, this is balanced by a pattern of nutrient concentration atop the Maili Sita col.

Recent approaches
Recent research has sought to develop the findings of these earlier investigations at Maili Sita, with the principal aim of better integrating the archaeological and ecological records so as to comment on the enduring ecological effects of pastoralist settlement in a semi-arid savanna environment. The research reported here thus offers an archaeological deep-time perspective on the kinds of research questions being pursued by ecologists working at the Mpala Research Centre (references above), as well as improving our understanding of the development of contemporary pastoralism in an important yet poorly understood location, the Laikipia Plateau. Besides the aforementioned work on glade formation, this research has also been inspired by the geo-ethnoarchaeological investigations undertaken by Ruth Shahack-Gross and colleagues (Shahack-Gross et al. 2003Shahack-Gross 2011), which have considered recently abandoned Kisongo Maasai villages in Kajiado District, southern Kenya, using oral records of spatial arrangement to reveal the geoarchaeological signatures one might expect in particular activity areas. Building on existing work on faecal spherulites (Canti 1997(Canti , 1999 and micro-laminations in the soil structure caused by trampling (Courty et al. 1991;Macphail et al. 1997) as records of degraded dung deposits, Shahack-Gross et al. (2003 attempted to define a taphonomic sequence for livestock enclosure sediments in eastern Africa and to explore whether other activity areas might be similarly identifiable using geoarchaeological techniques such as soil micromorphology, mineralogy, phytoliths and stable isotope analysis. The current work at Maili Sita seeks to bridge these diverse perspectives; where the primary concern of Shahack-Gross and colleagues was with the archaeological record albeit using soil data that could be construed as an ecological resourceand glade ecologists have looked specifically at the impacts of livestock presence with only limited consideration of the historicity and dynamicsi.e. the human elementof that presence, the work at Maili Sita assumes the position of Historical Ecology: that the two are inextricably intertwined (e.g. Crumley 1994;Balée 1998;Lane 2016b).

Field strategy
Several data sources have been pursued in recent work at Maili Sita, the bulk of which relates to the composition of the deposits across the col; this follows the work by Payton (2005) described earlier, though focusing on different and complementary variables, with greater attention given to the association of the soil data with more directly archaeological information. Recent work has also employed a more intensive sampling strategy in order to achieve the resolution needed to access the kinds of potential data advocated by Shahack-Gross et al. (2003. In March 2015, bulk soil samples were obtained using a bucket auger, at depths of 0-5 cm, 5-15 cm and 15-30 cm along two transects, one (NS1) following a similar northsouth path along the ridge top to that used by Payton (2005), with a second (WE1) running east to west across the apparent centre of the site down slope into the Acacia thicket that borders the site on all sides. Samples were taken every 25 m along the 475 m east-west transect, and between 10 and 30 m north-south for 530 m. A third survey transect (EWp1) followed a 24 m path between two of the grass patches close to excavation unit A, with samples taken every 2 m at a depth of 5-15 cm. Regional control data were derived firstly from samples at the respective limits of the transects NS1 and WE1, beyond the limits of the archaeological site (based on ground survey and satellite imagery, see below) and the extent of the Maili Sita 'glade'. For some analyses, these are augmented by data from the wider area obtained during the 2004 soil survey (Payton 2005;unpublished data). Column samples for soil micromorphology were taken from a number of the 2010 excavation units (during the excavation), and stored at the British Institute in Eastern Africa (BIEA) in Nairobi. All bulk samples and three column samplesthose corresponding to units A, B and D, chosen to cover a variety of the contexts recorded under excavationwere exported to the United Kingdom for analysis at the Institute of Archaeology, University College London (UCL).
In addition, a return visit to the site in January 2016, shortly following the rainy season, such that vegetation was near peak levels, allowed detailed survey of species representation across it. This was conducted along the two soil survey transects. Having relocated the original sample locations using a handheld GPS device, the species present within a 0.5 m radius of the sample hole were recorded, along with an approximation of their density. When unknown, detailed photographic records were made to allow later identification using reference texts (Weiss 1989;Dharani 2002;Agnew 2013). Where multiple distinct vegetation communities were present within a sample locale, only that which incorporated the precise location of the soil sample was considered, with a note made of its proximity to varying conditions.

Laboratory methods
Prior to thin section manufacture, sub-samples weighing approximately 100 g of sediment were taken at 5 cm intervals along the column samples with the goal of subjecting them to the same set of analyses as those from the transects. Thin sections were manufactured according to the conventional procedures defined by Courty et al. (1989); impregnation with epoxy resin (3:2 resin to acetone, with MEKP catalyser) was conducted at UCL and blocks sent to Spectrum Petrographics for slide manufacture (50×75 mm slides). Observations were made using a Leica DM EP optical microscope at 2.5x to 50x magnification and recorded using the procedures and terminology defined by Bullock et al. (1986) and Stoops (2003).
Bulk samples were tested for total carbon (Ct) and calcium carbonate (CaCO 3 ) content using the loss-on-ignition method (muffle furnace at 550°C for two hours, and 1000°C for one hour, respectively). Magnetic susceptibility readings were taken on 10 cm 3 of sample using a Bartington MS2 low-frequency sensor and pH is given based on the average of three readings of a 1:2 sample to de-ionised water solution taken using an Hanna HI-9024 instrument, calibrated using pH 4, pH 7 and pH 10 buffers. All the samples were sieved to <2 mm prior to analysis. Preliminary quantitative analysis of faecal spherulites of the unit A samples was undertaken by J. Cogdale, following the protocol outlined by Canti (1997). Table 1 summarises the observations made of the thin sections from 2010 excavation units A, B and D, which generated four, three and three thin sections, respectively. A general fabric of granular excremental pellets across the three sample locations indicates a high level of bioturbation by soil fauna. Termites (Macrotermes spp.) are present in great numbers across the lower northern end of the Lolldaiga Hills Ranch and their capacity for turnover of sediments in eastern Africa, in a manner that could be expected to result in this kind of fabric, is well documented on Lolldaiga (Qandelihle 2010) and elsewhere on Laikipia (Jungerius et al. 1999), as are their more general taphonomic contributions to African archaeological site formation (McBrearty 1990) and savanna ecodynamics (Jungerius et al. 1999). While a detailed discussion of the ecological role of termites is not possible or appropriate here, it is worth noting the similarity of soil composition and nutrient dynamics within and around termite mounds and that noted for livestock enclosures; for instance, both have been noted to exhibit elevated pH, calcium and magnesium levels (e.g. Augustine 2003;Muchiru et al. 2009;Okullo and Moe 2011;Joseph et al. 2013). That termites may be preferentially attracted to the large quantities of deadwood left at abandoned pastoralist settlements is also well attested. Hence, their potential to further enhance soil nutrient levels at and around such sites needs to be acknowledged. We would argue, however, that the features exhibited at Maili Sita show sufficient evidence of anthropogenesis to discount the demonstrable 'ecosystem engineering' capabilities of termites (after Dangerfield et al. 1998) as the primary drivers of landscape transformation.

Micromorphology
Broadly, the samples from unit A are the only ones to give any strong indication of anthropogenic input; in the upper (anthropogenic) levels, unsorted mineral grains (probably basalt-derived) are randomly distributed within a clay-poor, granular groundmass showing no ped development. Also present are occasional fragments of charcoal and Generally 30:70 c:f except in the deepest sediments (60:40, though large pebbles add to the coarse ratio); 80-90% unsorted quartz (10-20% feldspar and mica); subangular, fine sand to coarse sand single grains; rare clay or iron-rich nodules, similar mineral content to main fabric; more rare with depth.
Some charcoal flecking; less than in A; more common in top striated fabric (approximately 2%) than lower down (<1%); spherulites present throughout but more frequent with depth; less frequent than in A; loose infilling of channels with excremental pellets; some micritic calcite on void walls; some micropan clay coating of coarse grains, more frequent with depth.
Slightly elevated pH and magnetic susceptibility, some spherulites and calcium carbonate features suggest this marks the edge of the dense midden deposits in unit A, still showing some input of dung and ash residues. A medium finds density in excavation suggests proximity to domestic or refuse areas. Very rare charcoal flecking (possibly magnetitefine silt grains) only in surface deposits; significant organic staining of fine fraction (dark brown in PPL), more in the top sediments; rare phytolith accumulations in voids; common infilling with pellets, very rare micropan clay coating of coarse grains, less common with depth.
pH is consistent with regional control levels (i.e. 6.5-7.0) and no reliable indication of anthropogenic inputs. Low-medium finds density in excavation suggests this is part of the site, but not an enclosure, midden or intensive activity area.
bone alongside rhomboidal calcium carbonate ash crystals (cf. Canti 2003;Braadbaart et al. 2012; Figure 3a-3b), carbonate accumulations in voids, frequent charcoal flecking commensurate with the kinds of refuse deposits expected at herder settlements (Brochier et al. 1992;Macphail et al. 1997;Shahack-Gross et al. 2003). The samples contained relatively dense accumulations (>30) of faecal spherulitescarbonate minerals formed in the gut of grazing animals (Canti 1997;Figure 3c)and a crystallitic (XPL) b-fabric, indicative of degraded dung deposits. Spherulites were most concentrated in the uppermiddle deposits (10-15 cm), below which there was a sharp rise in the relative proportion of the smallest spherulites, perhaps due to percolation down the soil profile, and a sharp drop in overall quantity (Cogdale 2015). The upper deposits of unit A also showed elevated CaCO 3 levels (up to 4.5%), comparable with the mean (4.9%) for the peaks in the transect data (see below). Both CaCO 3 and spherulite density decrease with depth in the lower half of the profile. These lower, archaeologically sterile levels show greater clay content and a non-crystallic b-fabric with very little charcoal, although they remain comprised of the excremental pellet structure of the upper deposits, indicative of intense bioturbation throughout the profile. Unit B shows some spherulites and charcoal flecking, but at much lower densities than unit A, and lacks clear evidence for ash. This profile also registers a lower pH (7-7.5) than unit A (pH 8-8.5), though this is still above that of off-site levels (pH 6-7). Unit D, contrastingly, registered a pH consistent with these regional controls. Nor were any spherulites observed in unit D, where rare, black, silt-sized flecking (PPL) could not be reliably distinguished as either charcoal or magnetite. It should be noted that across the samples spherulite concentration did not appear to be strongly associated with pH, an indication that presence in certain deposits cannot be explained simply by heightened alkalinity caused by ash inputs, for example (Canti 1997;Demeyer et al. 2001). Although magnetic susceptibility across NS1 and WE1 is elevated at the northern and eastern ends, this continues beyond the perimeter of the site and shows no correlation with the other variables. Of some interest though is the extreme peak at 300 m along WE1, which is most pronounced in the surface deposits. That this anomaly is only registered in the magnetic susceptibility reading, however, supposes that its origin might be an intense localised raising of soil temperature as might be caused by a lightning strike (cf. Maki 2005). Transect EWp1 showed no clear variation in the deposits between the grass patches, a finding at odds with the pattern described by Payton (2005), although the data for all variables is comparable with that observed for the southern peak of NS1. Vegetation growth, however, is significantly different; the patches themselves are comprised predominantly of Pennisetum stramineum, with Cynodon dachtylon, Kyllinga sp. and Panicum sp., and occasional Solanum incanum and Achyranthes aspera ('Devil's horsewhip'). Beyond the patches, areas in the vicinity of the northern, southern, and western peaks (NS1, WE1 and the non-patch samples of EWp1) are fairly bare of vegetation, with the occasional presence of C. dachtylon. 6 The central expanse of the col, however, is dominated by C. dachtylon, though beyond the limits of the site this changes to a mosaic of C. plechtostachyus, P. stramineum and Themeda triandra, with a concomitant increase in woody vegetation, almost exclusively Acacia drepanolobioum ('Whistling thorn'). Dense and lush clumps of C. plechtostachyus and P. stramineum are frequent within the immediate vicinity of these Acacia.

Discussion
Building on the corpus of earlier data relating to the occupation of Maili Sita, the current study shows that there is significant variation in sediment composition across the site.
Concentrations of faecal spherulites, organic carbon, calcite and phosphorus towards the perimeter of the col, as indicated by the transect surveys, are consistent with degradation of the kinds of dung and ash deposits observed at contemporary pastoralist settlements (Western and Dunne 1979;Mbae 1990;Shahack-Gross et al. 2003 and vegetation patterning echoes that noted at boma-related savanna glades in modern ecological studies, with an exclusion of woody species in favour of rich grasses (e.g. Muchiru et al. 2009;Veblen 2012;Veblen 2013). It should be noted that the species represented at Maili Sita and their relative distributions do not exactly match the patterns noted at experimental bomas at nearby Mpala Research Centre. For instance, at Mpala, Cynodon plechtostachyus -'manyatta grass'is observed to dominate species composition after 5-20 years, with Pennisetum stramineum restricted to glade edges, while at Maili Sita it is Cynodon dachtylon that dominatesas it does at Makurian Ranch (see below)while P. stramineum is present in patches across the site. It is difficult to draw firm conclusions as to the root causes and implications of such distinctions, which are likely to be driven by subtle variations in ecosystem dynamics between the two locations. Moving further afield, former herder settlements in South Africa, for example, are marked by circular groves of Acacia tortilis, yet show a comparable soil nutrient composition to that at Maili Sita and Mpala (Blackmore et al. 1990). Such differences within the broader pattern of glade formation do, however, highlight the need for regionally diverse studies if this kind of ecological information is to be useful from an archaeological perspective.
Our study has thus far not recognised any of the more subtle variations that might be used to reconstruct specific elements within a settlement as explored by Shahack-Gross (2003). Yet, if the results of the recent soil and vegetation surveys are considered with respect to high-resolution satellite imagery, a broader pattern becomes apparent. The site plan in Figure 2, which shows the sample locations along transects NS1 and WE1, overlies a Worldview-2 image, acquired in May 2011. Perhaps the most immediately visible feature is the continuous pale ring which encircles the col and, as shown by groundtruthing in support of Soil Adjusted Vegetation Index (Huete 1988;Zhang et al. 2009) conducted on the satellite imagery, is largely composed of bare soils interspersed with patches of P. stramineum. The data peaks in the variables outlined above fall exclusively within this area, presupposing that the entire ring is comprised of those same dung and ash deposits. Crucially, this implies that no significant alteration of the soil ecology occurred within the ring and that, given the continuous and homogenous nature of these ring deposits, activity was concentrated around the entire perimeter of the site in a single, or series of single, sustained episodes of occupation. That the 2010 excavation units D-I located within the ring yielded very little in the way of either artefacts or faunal remains, while the 2004 trenches BA-BE located to the north and at the ring edge on the upper slopes of the col had richer assemblages, lends some support to this view. Even more intriguingly, the scale of such an occupation would exceed that of almost anything previously recorded for pastoralist settlements either archaeologically or ethnographically in eastern Africa, while the layout contrasts with any site-type associated with the Maasai (cf. Mbae 1990) or their immediate neighbours, the Samburu (cf. Spencer 1973).
However, a scan of Google Earth™ imagery across the Laikipia Plateau and its immediate surroundingsthe Leroghi Plateau and Samburu to the north and east, and Lake Baringo and the Rift Valley to the westpresents numerous other locations exhibiting similar characteristics. One of these, Makurian Ranch (37 N 290344.18, E 37639.59), located immediately north of the Lolldaiga Hills Ranch within land belonging to the local community group ranch, was visited briefly during recent fieldwork at Maili Sita. It is clear from the satellite image ( Figure 5) that there are visible similarities with Maili Sitathe site is of a similar size, 400-500 m in diameter, showing pale deposits in a roughly circular ring within a treeless glade dominated by C. dachtylonand scattered across the surface of the site, but particularly concentrated around the perimeter, are obsidian flakes and Kisima ceramics suggestive of a mid-to-late second millennium AD occupation. A second example, Maasai Plains (37 N 234817.95, E 80456.77), which was investigated under the same project as the original excavations at Maili Sita (Lane 2005(Lane , 2011, exhibits concentric rings of organic-rich, ashy deposits between 400 m and 700 m in diameter and has been dated to 480 ± 50 BP (1315-1623 cal. AD, Beta-189982; Table 2).
The disparities shown between Maili Sita and sites like Makurian Ranch and Maasai Plains and ethnographic records of pastoralist settlement in eastern Africa indicate a social landscape distinct from that which developed and was observed on the Laikipia Plateau during the colonial period and throughout the twentieth century. The upheavals of the late eighteenth and nineteenth centuries, when catastrophic regional drought (Bessems et al. 2008), the Iloikop wars (Waller 1979), rinderpest epidemics (Mack 1970) and the Pax Britannica (Hughes 2006) all contributed to the replacement of the Laikipiak with Purko Maasai, belie an earlier cordiality, in which Iloikop communities such as the Momonyot, Il-Chamus and Samburu, as well the Laikipiak and hunter-gatherer groups like the Mukogodo, coexisted and engaged in extensive exchange networks (Waller 1979). It is now widely accepted that these contemporary 'ethnic' distinctions arose from the tailoring of economic practices to the diverse environments these groups inhabit (Galaty 1993;Spear 1993aSpear , 1993b. This interactive sphere seems to have reached at least as far north as Lake Turkana (Lamphear 1993;Sobania 1993) and it is therefore unsurprising that the form of Maili Sita and other sites in Laikipia should present closer similarities with the modern settlements of Cushitic-speaking Rendille camel herders (Spencer 1973) than with those of  the Maasai or Samburu (Mbae 1990;Spencer 2003). Indeed, unverified oral histories have indicated an influx of pastoralist Oromo-Cushitic speakersthe Warra Daaya (sometimes Wardei Daaya)into Laikipia from the eastern Lake Turkana region during the mid-sixteenth century (Tablino 1999;Lemoosa 2005), events and interactions that may account for the Cushitic-based Yaaku language spoken by Mukogodo hunter-gatherers prior to their twentieth-century adoption of Maa (Cronk 1989). Although most scholars believe the Laikipiak to have been Eastern Nilotic speakers (e.g. Waller 1979;Sobania 1993), the evidence presented here revives Jacobs ' (1965: 66) notion of an Eastern Cushitic origin. Spencer (1973: 20) offers the interesting reasoning for the large size of Rendille villages that the strength of consensus opinion made possible by large gatherings aids community cohesion in an essentially heterarchical society, proposing that any economic arguments are 'rationalisations of this essentially social choice'. Following this argument, it may be that the founding of large settlements contributed to the ability of pre-nineteenth-century populations to engage in economically and culturally symbiotic interrelationships within a diverse ethnic mosaic. Equally, McCabe (2004: 53-54) has documented how the Turkana temporarily congregate in large settlements or clusters thereof, known as adakars, in response to inter-ethnic conflict. Were Maili Sita the result of a similar phenomenon, the opportunity to strengthen communal bonds and reinforcement of shared identities amid stressful external pressures may have had a comparable effect, with possible, but as yet unexplored, ramifications for the use of material culture style to create and signal different dimensions of personal and group identity (although see Larick 1986aLarick , 1986bLarick , 1991. Pertinent here also is Anderson's (2016) recent argument that owing to the linked social and ecological catastrophes that affected the Lake Baringo region in the nineteenth century there was significant depopulation of the area and, after these challenges had eased, a re-'assembling' of social practice so radical that the cultural reckoning of time began again. These nineteenth-century disasters certainly would have extended eastwards across the Laikipia and Leroghi Plateaux where cultural memories of their impacts remain strong (Waller 1988;Straight et al. 2016), as well as further north toward Lake Turkana (Sobania 1988). A possible implication, still to be attested archaeologically, is that the abandonment of large sites such as those at Maili Sita may have triggered new forms of mobility and entailed changes in social and material practice of the same kind of order now being suggested for Baringo.

Summary
Maili Sita appears to be emblematic of a distinctive social phenomenon that spanned the Laikipia Plateau during the latter half of the second millennium AD, prior to the disruptive wars and epidemicsalong with the arrival of British colonial ruleof the nineteenth century. Northern Kenya is known to have played host to population movement on a considerable scale during this period, fostering interactions between diverse economic, cultural and linguistic groups out of which emerged the Rendille, Gabbra and Boran, among others (Spencer 1973;Schlee 1985;Kassam 2006). The convergence of these northern Cushites with the Maa-speaking pastoralist communities consolidated in the Rift Valley was surely one of the key drivers behind these ethnic definitions and the results from Maili Sita may go some way to positioning Laikipia at the centre of that interaction.
What is more certain is that in order to recognise and understand population dynamics during this period, or indeed earlier, we must reconsider the reliance of pastoralist archaeologies on fairly rigid settlement typologies as defined by a few key ethnographic works (e.g. Weston and Dunne 1979; Mbae 1990). As a modest step toward such an objective, we hope that this paper helps demonstrate the potential of using soil and vegetation data to investigate ephemeral sites associated with mobile communities without immediate recourse to such narrow analogies. In addition, the ecological studies highlighted here (e.g. Stelfox 1986; Augustine 2003b; Muchiru et al. 2008;Donihue et al. 2013;Veblen 2013) have focused on the effects of modern bomas on the savanna landscapes. The scale of the occupations at Maili Sita and Maasai Plains, and the potential duration of the 'super-sites' as a phenomenonfrom at least the fifteenth until perhaps the nineteenth century ADencourage further reconsideration of the long-term impacts of a pastoralist presence in eastern Africa's arid and semiarid landscapes.  2011,2013) referred to the site as 'Mili Sita', which translates, rather unfortunately, as 'six corpses'. This error was only recently pointed out and we should like to take the opportunity to correct it here. 2. The site lies just north of the Equator, and, as detailed in Table 2, there are some minor discrepancies between the limited radiocarbon dates available when using both the northern hemisphere and southern hemisphere calibration curves. 3. A manyatta is a type of Maasai settlement, associated with age-set graduation ceremonies; the name is frequently erroneously applied to the principal Maasai homestead, called enkang. 4. Prior to this work, the known distribution of Narosura ware was limited to the west of the Rift Valley (Wandibba 1980) and as far south as Narok in the southern highlands of Kenya (Odner 1972), Nasera rockshelter in the Serengeti, around Lake Eyasi (Tanzania) at Mumba and other rockshelters (Mehlman 1989) and possibly on sites to the west of Kilimanjaro (Mturi 1986). More recent survey work has demonstrated its common presence also on open sites further south in Tanzania in the Mang'ola Basin (Prendergast et al. 2014) and along the southern edge of the Mbulu Plateau (Prendergast et al. 2013). 5. The bulk of the analyses of the faunal, botanical and artefactual remains from the different seasons of excavation at the site are ongoing. Among other trends, provisional interpretations of fungal spore data recovered from test-pit profiles at the southern end of the site further supports the orientation of the subsistence economy towards livestock (V. Muiruri, pers. comm., 2008), although more comprehensive analyses are needed to confirm this. 6. Turner (1976)  Recent work at Maili Sita has been conducted as part of OB's ongoing PhD research at UCL, under Kenya Research Permit NACOSTI/P/14/5093/383, funded by the Arts and Humanities Research Council (AHRC). Funding for fieldwork has been received from the AHRC, UCL Graduate School, the Institute of Archaeology UCL and the British Institute in Eastern Africa. Particular thanks must go Julius Mwenda and Kate Jarvis for invaluable assistance in the field, to Andrew Reid, Manuel Arroyo-Kalin and Mark Altaweel for advice, guidance and comments throughout the research process and to the guest editors for the invitation to contribute to this special issue. Finally, we extend our appreciation to the two reviewers for their valuable comments, without which this paper would be greatly inferior.

Notes on contributors
Oliver Boles is (hopefully) in the final stages of his PhD at the Institute of Archaeology, University College London, a thesis that focuses on Maili Sita. His research interests include the historical ecology of pastoralist landscapes in eastern Africa, past human response to environmental change and the zooarchaeological applications of isotope analyses.
Paul Lane is Professor of Global Archaeology at Uppsala University, Sweden, and an Honorary Research Fellow in the School of Geography, Archaeology and Environmental Studies at the University of the Witwatersrand, South Africa. His primary research sites are in Africa and his interests include the transition to food production, the archaeology of pastoralism, the social and symbolic structuring of space and time, the role of the past in the past and the historical ecologies of East African landscapes. He is a former Director of the British Institute in Eastern Africa (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006), and former President of the Society of Africanist Archaeologists (2008-10).