Database 1: Southern African archaeological organic database, 800 cal BC- cal AD 1500

Database 1 provides an inventory of the six types of organic remains recovered from the southern African Later Stone Age context from cal 891 BC to cal AD1512 (for ease of reference and because of the lack of precision in radiocarbon dates, we refer to this as 800 cal BC to cal AD 1500 in the headings) (S1 Map). The geographic scope includes regions between latitudes 17 to 34 degrees south and between longitudes 11 to 33 degrees east. This includes coverage of Namibia, Botswana, Zimbabwe, South Africa, Lesotho and the Kingdom of Eswatini (formerly Swaziland). We excluded Zambia because of the limited reporting/research and/or the absence of organic remains from sites in Zambia for the period in which we are interested. We could find no evidence of ostrich eggshell containers although there are sites with ostrich eggshell beads [58–62].

For purposes of discussion within the hunter-herder debate, the database also contains radiocarbon-dated and geo-referenced data from all known occurrences of Later Stone Age pottery, livestock and ostrich eggshell beads for the same period.

Database 2: Southern African historical database of organic container use, 16^th century onwards

Database 2 presents the spatio-temporal distribution of organic container use amongst extant and historical Khoesan populations, as well as Bantu-language speaking agro-pastoralists, as derived from ethnographic and historic accounts ranging in date from the 16^th century AD to the present.

Compilation of the southern African archaeological organic database, 800 cal BC- cal AD 1500 (Database 1)

This database was collated from multiple sources, namely, site excavation reports, academic publications, unpublished dissertations and radiocarbon laboratory lists (S1 Table). Key texts were [63] for Later Stone Age sites, and [64] for sites related to the spread of livestock keeping. The database contains the following fields:

1. A unique site reference number.
2. The official site name as designated by the excavator(s).
3. The type of site (open or shelter, coastal or inland).
4. Country and/or province location.
5. Geographic co-ordinates.
6. The subsistence base of the site’s occupants.
7. Stratigraphic context of radiocarbon date and/or material remains.
8. Uncalibrated radiocarbon date and laboratory number. Direct dates are highlighted in bold alongside the material dated.
9. Calibrated date ranges. Dates were calibrated using Oxcal version 4.3. All materials were calibrated using the Southern Hemisphere Calibration Curve ShCal 13 [65]. Dates derived from uncorrected and uncalibrated marine shell are calibrated using Marine13 [66]. The mean marine carbon reservoir correction (∆R) is taken [67], with a weighted mean of 146±85 for the western coast of southern Africa (this applies only to the coastal regions of Namaqualand in the Northern Cape and the Western Cape Province). For the southern coast of South Africa, including the coastal regions between Mossel Bay and Port Elizabeth, the mean marine carbon reservoir correction (∆R) is taken, with weighted mean ∆R of 187±18 [68]. All calibrations are to one sigma and reported as calibrated years cal BC/cal AD. A reservoir correction is not used for dates derived from ostrich eggshell, since they are fossilised and thus old (Higham pers comm. 2019).
10. Radiocarbon dates were given a rating:
    1. Good. A direct date.
    2. Fair. Sequence of dates with good stratigraphic control and the excavator is confident in the association between dates and material remain(s).
    3. Poor. Date derives from poor stratigraphic contexts or there is only a single date obtained for a site.

The sites are grouped temporally into four 600-year time slices:

1. Group 1: 891–291 cal BC.
2. Group 2: 290 cal BC- cal AD 310.
3. Group 3: cal AD 311–911.
4. Group 4: cal AD 912–1512.

These four 600-year time slices capture patterns that have been identified previously [25, 28, 38, 42, 64, 69]–an early first millennium AD appearance of livestock and ceramics (Groups 2 and 3) and a later first millennium to second millennium AD appearance of lugged ware, livestock and large ostrich eggshell bead diameters associated with Khoe pastoralists (Group 4). Each map has an accompanying table (arranged as separate sheets in S1 Appendix) showing the finer chronology of the material culture for each time slice.

The reporting style of published excavated material assemblages is uneven. Gaps occur particularly in the reporting of the quantity, size and presence or absence of the manufacture of ostrich eggshell beads. Where individual counts of material items could not be established for each stratigraphic level, total counts at a site are reported. The database was filtered to remove sites with no clear association between stratigraphic context and dates, or which fell outside of the geographic scope of the database collection–e.g. northern Mozambique where sites lie above 17 degree latitude. A total of 49 sites and 98 radiocarbon dates were thus excluded from the database (S1 Appendix). Post-filtering, the database comprises 187 sites, associated with 489 radiocarbon dates. Of these 77 sites contain organic remains (S1 Appendix).

For ease of reference and because there are so few organic remains, the finds in Database 1 are condensed into five categories:

1. Category 1: Later Stone Age pottery (type of temper).
2. Category 2: Ostrich eggshell containers (mouth/rim fragments and/or decorated pieces).
3. Category 3: All other organic remains, namely leather, twine, gourd (Lagenaria sp.), wood or tortoise carapace.
4. Category 4: Livestock.
5. Category 5: Ostrich eggshell beads–quantity, and presence or absence of manufacture (this was based on the presence of unfinished beads).

On the maps, these categories are reflected in a symbol (Fig 4). The size of the symbol reflects the number of ostrich eggshell beads recovered at a site. Ostrich eggshell was captured as a container only if it showed either, (a) evidence of modification in the form of a mouth or rim piece, or (b) an engraved or painted pattern. The capturing of tortoise carapace was based on whether or not it was described as ‘modified’ by the excavator. The quantities of other items of material culture at each site are recorded in Database 1. Those sites with a poor temporal rating (3) have a symbol with a 60% transparency setting on the maps for easy identification.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. The symbol used to display the five categories of archaeological material found at a site.

https://doi.org/10.1371/journal.pone.0235226.g004

For simple comparison with containers made from organic materials, we took the decision to make a single category of Later Stone Age pottery (category 1) and to look at one of its variables (temper) and to simply consider its presence or absence at a site. All the pottery in this category is thin-walled (5–8 mm) and found exclusively at sites of herders, pastoralists and hunter-gatherers [69]. There are regional subsets to this group that differ in terms of their shape and decoration (For example, in the southwestern Cape [21, 28, 70, 71], the South African interior [42, 72, 73]; south-eastern southern Africa [43, 45, 74]; northern Cape, South Africa [75], Eastern Cape, South Africa [76], Namibia [18], Free-State, South Africa [77], Zimbabwe and Botswana [3, 78–80]. These subsets are not considered here.

To document the types of engraving observed on ostrich eggshell fragments and to confirm the presence of ostrich eggshell rim pieces and other organic material from the excavated record, a sample of six sites were selected for further documentation by visiting the museum collections in South Africa (S2 Table). The six sites were chosen because they have evidence of organic containers and pottery (mineral and grass tempered) (S2 Appendix, S3 Appendix, S4 Appendix).

Compilation of the southern African historical database of organic container use, 16^th century onwards (Database 2)

This non-exhaustive geo-referenced database comprises 89 entries from historical and ethnographic sources for the use of organic containers amongst non-agro-pastoralist and agro-pastoralist groups in southern Africa (S1 Appendix, S3 Table). References to the use of non-agro-pastoralist mineral- and grass-tempered pottery were also recorded. The database has a date range from 1593 AD to the present (S1 Appendix).

Database 2 contains the following fields:

1. A unique site reference number.
2. Name of record taker.
3. Documentation type (e.g. traveller account, museum object or photographic collection etc.).
4. Date of record.
5. Cultural group.
6. Mode of subsistence.
7. Region and country of record.
8. The area from which geographic co-ordinates were taken for mapping. Geographic co-ordinates are estimates: none of the accounts provided latitude and longitude information.
9. Geographic co-ordinates.
10. Container type recorded.
11. Container function.
12. Other information (e.g. plant species, where provided).
13. Verbatim record.
14. Reference/source.

Ethnographic sources

Six ethnographic works were consulted. Five of these are records made among Khoe- and non-Khoe-speaking Bushmen groups between 1950 and 1993 [56, 81–84]. They cover a geographic region that extends from the Nyae Nyae Conservancy in north-eastern Namibia to Ghanzi, in Botswana. The sixth source is Hoernlé’s ethnographic study of the Nama of Walvis Bay, Namibia (1912 and 1923) [85].

Historical sources

Two online museum archives were accessed: 1. The Oswin Köhler Archive from the Goethe University, Frankfurt (https://www.uni-frankfurt.de/62970506/Collection_Oswin_Köhler). A simple search was carried out by object type or cultural group name between 1911 and 1996. 2. The Pitt Rivers Museum collection (https://www.prm.ox.ac.uk/collections-online). Search parameters included the photographic collections, Africa continent, southern Africa and ‘Khoisan’ cultural group.

The following research publications were consulted: (1). Bollong et al. [86] provides a review of early colonial traveller and settler accounts for the use of pottery amongst historically observed Khoe pastoralist groups and Bushmen occupying the drier interior and south-western coastal regions of South Africa. (2). Rudner [87] offers additional historical references for the use of pottery amongst non-ago-pastoralist groups. (3). Sydow [88] references the use of organic containers by historically documented Khoe pastoralists, (4). Guenther [84] describes container use amongst the Namib Bushmen, Namibia. (5). Shaw and van Warmelo [89] record the use of baskets, gourds, calabashes and wooden containers amongst Bantu language-speaking agro-pastoralists in the Eastern Cape, South Africa. (6). Hooper [90] records containers used by the Zulu and Swazi of KwaZulu-Natal, South Africa and Eswatini (formerly Swaziland) respectively.

The geographic coordinates in Database 2 are approximate, based on maps where provided, and the geographic region as described in each source. A map was compiled by overlaying point data (each symbol refers to a historic or ethnographic account) onto the freely available UNEP Africa average annual precipitation (mm) map (UNEP, 2002 http://geodata.grid.unep.ch).

Map 1, 891–291 cal BC

Fifty-three (53) archaeological sites fall within this period (Fig 5). The organic materials present are ostrich eggshell (beads at 48 sites, containers at 8 sites), fibre/twine (6 sites of which 4 are graves) and leather (1 grave site). Five sites contain no evidence of ostrich eggshell beads, four of which are gravesites (this concurs with the observation amongst hunter-gatherers of the Kalahari that all possessions with the exception of beadwork are taken to a person’s grave on death [82:41]) and one is a scatter site. There is no evidence for pottery or livestock during this time slice.

Map 2, 290 cal BC- cal AD 310

Sixty-three (63) sites fall within this time slice (Fig 6). The organic materials recorded are ostrich eggshell (55 sites with beads, 17 sites with containers), fibre/twine (10 sites, of which 2 are graves), leather (3 sites, of which 1 is a grave), tortoise containers (6 sites), gourd (3 sites), wood (2 sites), and modified marine shell (1 site). Eight sites have no evidence of ostrich eggshell beads. 36 sites contain mineral-tempered pottery; of these sites, 12 also contain ostrich eggshell containers. Five sites contain fibre-tempered pottery–but with no evidence of ostrich eggshell containers, and with very little evidence of ostrich eggshell beads (2 of the sites have less than ten beads each). No fibre-tempered pottery is found with livestock. Five sites have ostrich eggshell containers without pottery. One of these is a grave, three are coastal middens and one is a small coastal rock shelter. All ten sites containing livestock have pottery and beads (all but one of them are bead manufacture sites). Six of the sites with livestock have ostrich eggshell containers (of which four are decorated; one may have been used to store ground ochre) (site 167, Ai Tomas [91: 214]). There is no evidence of bead manufacture at sites in the southern eastern parts of South Africa. Gourds are found only on the eastern side of South Africa.

Map 3, cal AD 311–911

Seventy-six (76) archaeological sites fall within this period (Fig 7). The organic material recorded is ostrich eggshell (beads at 61 sites, containers at 18 sites), fibre/twine (12 sites), leather (8 sites), tortoise carapace (4 sites), gourd (2 sites), organic plugs (2 sites), kelp (1 site) and marine shell (1 site). Thirteen (13) sites have no evidence of ostrich eggshell beads. 61 sites have mineral-tempered pottery, of these 15 sites also contain ostrich eggshell containers. Four sites have fibre-tempered pottery, 2 of these have ostrich eggshell containers. Fibre-tempered pottery is not found with livestock. Seven sites have agro-pastoralist (farmer) pottery in a Later Stone Age context. There are 28 sites with livestock, of these 25 also have pottery and beads (all but four of these are bead manufacture sites). Eight of the sites with livestock have ostrich eggshell containers. There is no evidence of bead manufacture in KwaZulu-Natal. Gourds are found in the eastern half of southern Africa. Fibre- and mineral-tempered ware are found together at sites for the first time (4 sites only, all without livestock).

Map 4, cal AD 912–1512

Ninety-two (92) sites are recorded in this period (Fig 8). The organic material recorded is ostrich eggshell bead (64 sites), ostrich eggshell containers (25 sites), fibre/twine (13 sites), leather (12 sites), tortoise carapace (10 sites), gourd (3 sites) and wood (2 sites). Twenty-six (26) sites have no evidence of ostrich eggshell beads. Seventy-two (72) sites have mineral-tempered pottery, 22 of which also have ostrich eggshell containers. Fifteen (15) sites have fibre-tempered pottery, 4 of which also have ostrich eggshell containers. For the first time fibre-tempered pottery is found with livestock (4 sites), at all of these sites the fibre-tempered pottery is found with mineral-tempered pottery. Two sites contain fibre-tempered pottery only. Eleven (11) sites have agro-pastoralist (farmer) pottery in a Later Stone Age Context. Thirty-five (35) sites have livestock, of these 33 have pottery as well as beads (17 of these have evidence of bead manufacture). Seven sites with livestock have ostrich eggshell containers. There is no evidence of bead manufacture in KwaZulu-Natal. Gourds are only found in the eastern portion of South Africa.

Ethnographic and historic observations of container use from AD 1600 onwards

Mistakes are likely to have been made in assigning identity to people encountered on the southern African landscape during the historic period [92]. The map showing the ethnographic/historic use of organic and pottery vessels largely coincides with what is understood from the archaeology: (1) Fibre-tempered pottery was identified only among hunter-gatherers; (2) Both hunter-gatherers and pastoralists used mineral-tempered vessels; (3) Gourds were used by both Bantu-language speakers and some pastoralists. The distribution of gourd use is concentrated in the summer rainfall area of southern Africa; and (4) Ostrich eggshell container use is almost exclusively by hunter-gatherers (Fig 9).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 9. Map to show the ethnographic and historic observations of container use from AD 1600 onwards.

https://doi.org/10.1371/journal.pone.0235226.g009

Pottery: Was pottery exchanged? Was it made by hunter-gatherers?

We suggest that pottery was introduced into southern Africa by a small group of migrating pastoralists around 2300 years ago [131, 132, cf. 42]. Their sites stand out as having evidence of livestock, bead manufacture and mineral-tempered pottery (Map 2 Fig 6, Table 1). They may have moved into the western part of South Africa, from Namibia and Northern Botswana, through the interior Nama biome and the succulent Karoo biome and the coastal areas (cf. [15, 38]). The evidence from sites containing both ostrich eggshell containers and pottery suggests that hunter-gatherers quickly adopted pottery making (cf. [42]). It is possible that pottery was subsequently dispersed among hunter-gatherers in a combination of local innovation and exchange along existing exchange networks. This accounts for its widespread occurrence at many sites during the period 290 cal BC to cal AD 310 and for its variability in temper, decoration and shape (cf. [21]).

Acquisition of stock and the transition to pastoralism

In the period 290 cal BC to cal AD 310, some hunter-gatherers, through their interactions with incoming pastoralists, acquired stock and became livestock-keepers (Table 2). Using ostrich eggshell containers as a proxy for hunter-gatherer presence, we suggest that most sites with livestock (60%) represent hunter-gatherers with livestock. It is upon this data that we base our reasoning that the incoming pastoralists were small in number (40% of sites represent immigrant pastoralists) (see [38] who reaches a similar conclusion based on stone tool analysis). From this point we present two scenarios, neither of which can yet be verified. Scenario 1 is that these hunter-gatherer turned livestock-keepers then gradually became the dominant pastoralist group on the landscape and are the ancestors of the historically observed Khoe pastoralist groups. Scenario 2 is that this early demic and cultural diffusion process was followed by a later migration of pastoralists and that the numbers of hunter-gatherer people with livestock remained consistently low on the landscape. Scenario 2 fits with the time line offered by geneticists for the first appearance of pastoralists with an east African origin approximately 1200–1300 years ago [133, 134].

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Sites that potentially represent pastoralist presence in southern Africa in the given time periods.

https://doi.org/10.1371/journal.pone.0235226.t002

The distinctiveness of southeastern South Africa

The southeastern part of South Africa stands out as distinct in the mapping exercise (Fig 11). In the period 290 cal BC to cal AD 310, Later Stone Age sites in the area show no evidence of ostrich eggshell bead manufacture, although almost all of the sites have evidence of ostrich eggshell beads (Fig 12). There is also no evidence of the use of ostrich eggshell containers. This area is not a natural habitat for ostriches (see [53, 55, 129] nor were these items commonly exchanged between groups (Fig 12). These sites also contain no evidence of livestock (with the exception of undated, painted rock art images of fat-tailed sheep found in low numbers) [45]. Yet they nearly all have pottery. Thin-walled mineral-tempered vessels are found in the Maloti-Drakensberg and the Thukela river valley, whilst the interior sites in the succulent Karoo biome have fibre-tempered thick-walled vessels. The mapped evidence suggests that mineral-tempered pottery was diffusing along already established social exchange networks that carried finished ostrich eggshell beads into the area. The likely source for beads and mineral-tempered pottery for sites in KwaZulu-Natal and Lesotho are area (iii) (for example, the site of Jubilee shelter, 365) approximately 400–600 km away, or area (ii) (e.g the site of Limerock 2, 240) which is a distance of approximately 400–640 km away (Fig 11). It is likely that fibre-tempered, bowl-shaped pottery in this area represents a local innovation amongst hunter-gatherers. This is the first appearance of the use of fibre temper in southern Africa. It is unlikely the pottery came from elsewhere. Mineral-tempered, wide-mouthed bowls are found at two sites in the surrounding area at approximately the same time (at sites 269 (Boomplaas) and 294 (Die Kelders) (area i, Fig 11)). The fibre-tempered bowls slightly pre-date those with mineral-temper. The area with fibre-tempered pottery might also have been more socially isolated: not all sites have finished beads and they occur in low counts of 1 to 10 beads.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 11. Map showing the southeastern part of South Africa and possible source areas for pottery and beads.

https://doi.org/10.1371/journal.pone.0235226.g011

Download:

• PNG
  larger image
• TIFF
  original image

Fig 12. Map showing the distribution of ostrich eggshell finished beads, manufacture sites and containers against the natural habitat of the ostrich.

https://doi.org/10.1371/journal.pone.0235226.g012

Continuing the hunter to herder debate

In recent research, Sadr [38] presented a new dataset presenting the ratios of backed stone tools to scrapers at 123 southern African archaeological sites. By comparing their ratio prior to the arrival of livestock (4000 to 2000 years ago) to that after the arrival of livestock (2000 to 1000 years ago), he concludes that a scraper-rich toolkit can be seen as a “cultural emblem” for southern Later Stone Age hunter-gatherer populations, whilst backed rich toolkits are linked to hunter-herders [38:9]. Webley [135] made a similar suggestion, that herder sites would lack formally retouched scrapers. Using this pattern, Sadr [38] recognises two small-scale migrations, one of which brought livestock and pottery along the Atlantic seaboard to the southern coast of South Africa. The other originated in the Zambezi watershed, with livestock and pottery introduced by hunter-herders but later adopted by hunter-gatherers moving in a westward direction across northern Botswana.

We have compared our data, where possible, to that of Sadr [38: Table 2]. Our results are complimentary: sites that we would classify as hunter-gatherer based on the presence of ostrich eggshell containers have scraper rich toolkits. Sadr works with a much longer time span than we do. He distinguishes two periods, before and after the arrival of livestock to the south. Further work using these two approaches in combination, for tighter and overlapping time spans, may prove productive.

Only two sites, Die Kelders (site 294) and Spoegrivier (site 165), contradict Sadr’s [38] findings. He argues that the furthest south the northern backed toolkit makers (i.e. people bringing livestock and pottery) reached was the site of Die Kelders in the southern Cape, with its backed tool rich assemblage (250 cal BC–cal AD 350, 8.9% scrapers and 58.9% backed tools) [38: Table 2]. Our interpretation is different: between 290 cal BC to cal AD 310, the site was occupied by hunter-gatherers who used ostrich eggshell containers, as well as marine and tortoise shell containers. They also used, and possibly made, mineral-tempered thin-walled pottery and manufactured ostrich eggshell beads. Later the site reflects a livestock-keeper signature, of mineral-tempered pottery and livestock. Ostrich eggshell containers are no longer used. This pattern can be interpreted in two ways: either the hunter-gatherers adopted livestock, the interpretation we favour, or an incoming group replaced them at around 300 cal AD. However, the stratigraphy at Die Kelders is disturbed [136]. It has been shown, by directly dating sheep bone, that smaller items, such as this bone, had moved from above through the Die Kelders stratigraphy into older layers [137]. It is possible that tiny backed tools, measuring 10 to 19 mm in height and a few mm in width have infiltrated the lower levels at Die Kelders. Backed tools occur in many layers at Die Kelders (layers 1 (the youngest level), 5, 7, 8, 9 and 12 (the oldest level) at the site [136: 173]) whereas scrapers (a hunter-gathers item [38]) are only found in the oldest layer, Layer 12, and are thus likely to be in situ.

At the site of Spoegriver Cave, Sadr [38: 11] notes that in Phase 1 the site had a scraper-rich stone tool assemblage (2030–1630 cal BC, 33% scrapers, 33% backed tools), and that this was replaced with a backed rich one that arrived with sheep and pottery (10–970 cal AD, 0% scrapers, 70% backed tools). Our data presents Spoegrivier as occupied by hunter-gatherers becoming livestock keepers from 290 cal BC to cal AD 310. The assemblage contains decorated ostrich eggshell containers, mineral-tempered pottery, evidence of bead manufacture and fibre/twine, and of sheep. Level 6b, the oldest level with pottery (38: 27) has one scraper (cf. [38]) and dates to this time bracket. This level has just one backed tool. In this case the large time slice used [38] may mask subtle changes through time. From approximately 300 AD, the only change to this signature is that the ostrich eggshell containers are undecorated and tortoise carapace bowls are found. We would interpret this as a case of hunter-gatherers becoming herders.

These results suggest the early presence of immigrant pastoralists at around 2300–2100 years BP. This is at odds with the earlier date of 1300 years BP for a pastoralist migration with origins in eastern Africa as suggested by geneticists–who state that this date may be an under-estimate[133](cf. [134, 138, 139]). The only second migration event we detect in the data presented here, is that of Bantu-language-speaking farmers, whose pottery appears at sites in the eastern half of southern Africa (Map 3, Fig 7 and Map 4, Fig 8). The possibility of a more recent migration (circa 1200–1300 years BP) needs further exploration, as does the possibility that the genetic clock is incorrect or an underestimate, as stated by Breton et al. [133].

Vedder’s [88] observation of Nama pastoralists with ostrich eggshell water containers warns us not to get too prescriptive about the significance of their occurrence, and is a reminder that in these interpretations we cannot rely on any one piece of evidence.