Content uploaded by Ian Watts
Author content
All content in this area was uploaded by Ian Watts on Jul 30, 2015
Content may be subject to copyright.
4 Red ochre, body painting, and language:
interpreting the Blombos ochre
Ian Watts
4.1 Introduction
Whereas language leaves no material trace, collective ritual—with its
formal characteristics of ampliWed, stereotypical, redundant display—
might be expected to leave a loud archeological signature. Does
the archeological record of ochre use provide such a signature, and can
it indirectly contribute to our understanding of the evolution of language?
I begin by highlighting the formal diVerences between language and
ritual as modes of communication. Why, despite having opposed charac-
teristics, is ritual widely regarded (Durkheim 1961; Rappaport 1999;
Knight 1999) as establishing the social conditions for language? I then
turn to the principal theories and inductive hypotheses that can be
brought to bear on the interpretation of early (pre-45 ky) ochre use. In
addition to being the Wrst major theorist to posit a link between language
and ritual, Durkheim drew attention to the role of body-painting in
grounding the collective representations central to ritual action. Subse-
quent theoretical perspectives can be distributed along a spectrum. At one
extreme is the innatist view that biology provides suYcient constraint to
account for universal features of color labeling (Berlin and Kay 1969).
Although this ‘‘Basic Colour Term’’ (BCT) theory is biological, it is not
evolutionary and generates no predictions as to when pigments should be
expected to emerge. It has, however, been used to predict the order in
which diVerent pigments should appear (Hovers et al. 2003). At the other
Funding from the British Academy and the South African National Research Foundation is
gratefully acknowledged, as is Professor Chris Henshilwood’s permission to work on the
Blombos material andAlison Brooks’ permission to cite unpublished resultsfrom Olorgesaille.
extreme is the ‘‘Female Cosmetic Coalitions’’ (FCC) model (Knight et al.
1995; Power and Aiello 1997; Power this volume). This sets out from
premises in human behavioral ecology, prioritizing the role of reproduct-
ive strategies in driving early pigment use and generating archeologically
testable predictions. Between these two poles is the qualiWed innatism of
Deacon (1997: 119), who treats the evolution of BCTs as subject to
constraints from both neurophysiology and ‘‘pragmatic constraints of
human uses.’’ Deacon’s model speciWes a ritual and a temporal context,
but is indistinguishable from BCT theory with respect to the sequence in
which terms should arise. Finally, challenging the presumption that ochre
was a pigment, several utilitarian hypotheses have been proposed (Klein
1995; Wadley et al. 2004; Wadley 2005a). I evaluate these perspectives and
their implications in the light of a survey of early potential pigments and
my research on the Blombos Cave ochre assemblage.
4.2 Context
Our species evolved in Africa sometime between !150 ky and !200 ky
(Ingman et al. 2000; McDougal et al. 2005). Shell beads and geometric
engravings on red ochre (d’Errico et al. 2005; Bouzouggar et al. 2007;
Henshilwood et al. in press) indicate that symbolic traditions were present
in Africa by the time a small subgroup of Homo sapiens migrated beyond
the continent, between !80 ky and !60 ky (Oppenheimer 2003; Mellars
2006). All of this occurs within a technological stage known in Sub-
Saharan Africa as the Middle Stone Age (MSA), a prepared core technol-
ogy that evolved out of the Acheulian around 300 ky, and lasted until
!25 ky in southern Africa (Clark 1997), !45 ky in eastern Africa (Am-
brose 1998). The beads and engravings double the conventionally accepted
antiquity of symbolic traditions, previously regarded as restricted to the
Upper Paleolithic (Eurasia) and Later Stone Age (Africa). These Wndings
have been used as a proxy for language (e.g. Henshilwood and Marean
2003: 636; Henshilwood et al. 2004: 404; Mithen 2005: 250; but see Botha,
this volume). In most Middle Stone Age contexts, the only recurrent
artifact class other than stone tools is red ochre.
Archeologists commonly use ‘‘ochre’’ as a generic term for any
rock, earth, or mineral producing a reddish or yellowish streak when
abraded, attributable respectively to hematite (an iron oxide) or one of
Red ochre, body painting, and language 63
the iron hydroxides (typically goethite). Ethnographically and archeo-
logically, red ochre is the most widely reported earth pigment. Use of
ochre and other potential earth pigments such as black manganese is
not restricted to Homo sapiens. The Weld may, therefore, provide com-
parative insights on the signaling strategies of closely related species
(cf. d’Errico 2003).
4.3 Language and collective ritual
In the animal world, signals vary according to whether theyencounter strong
or weak resistance. High resistance from receivers prompts costly multi-
media display; by contrast, low resistance permits low-cost ‘‘conspiratorial
whispering’’ (Krebs and Dawkins 1984). Translating this general principle
into the domain of human social communication, ritual (‘‘costly signaling’’)
and language (‘‘conspiratorial whispering’’) have the expected diametrically
opposed formal characteristics (Knight 1998, 1999: table 12.1), summarized
in Table 4.1. With resistance minimal, language evolves to be conventionally
coded, low cost, generally of low amplitude, digitally processed, interper-
sonal, focused on underlying intentions, and allowing for potentially inWnite
creativity. Being cheap and intrinsically unreliable, words are unable tosignal
social commitment (Rappaport 1999). Language leaves no direct archeo-
logical trace (Botha this volume), and is biologically unprecedented
(Chomsky 2002). Designed to overcome high levels of resistance and to
cement social contracts, ritual signals are multimedia indexical displays,
costly to produce, of high amplitude and redundancy, evaluated on an
Table 4.1 Signals: speech versus ritual (adapted from Knight 1999: Wg. 12.1).
Speech Ritual
Cheap signals Costly signals
Conventionally coded Iconic & indexical
Low amplitude High amplitude
Digitally processed Analog scale evaluation
Productivity/Creativity Repetition/Redundancy
Interpersonal Group-on-group
Focus on underlying intentions Focus on body boundaries and surfaces
64 Watts
analog scale and focused on body boundaries and surfaces (Sperber 1975;
Rappaport 1979: 173–246,1999). These features suggest that collective ritual
might leave a loud archeological signature. Unlike language, human ritual
has clear evolutionary precedents in the ritualized displays of other animals
(e.g. Laughlin and McManus 1979; Maynard Smith and Harper 2003).
Although ritual and language represent opposite extremes, a long theor-
etical tradition holds that they are mutually interdependent. This position
holds that collective ritual created the supportive framework for contractual
understandings and associated symbolic communication between group
members to become established (Durkheim 1961; Turner 1967: 93–111;
Rappaport 1979, 1999; Gellner 1988; Maynard Smith and Szathma
´ry 1995:
272–273; Searle 1996; Deacon 1997: 402–407). The central function of ritual
is to create the intensity of in-group trust necessary for symbolic commu-
nication to be possible (Knight 1998). Deacon (1997: 402) adds that ritual
facilitates the transition from concrete sign–object associations (indices and
icons) to abstract sign–sign associations. Alcorta and Sosis (2005) explain
that it is the costliness of ritual that enables it to demonstrate commitment
and deter freeriders.
4.4 A prediction ahead of its time
The basic premise concerning the role of ritual in sustaining symbolic
culture can be traced back to Durkheim. Less well known, Durkheim
also proposed that ‘‘the Wrst form of art’’ consisted of geometric designs
painted on sacred objects and on the bodies of ritual performers (1961:
149 fn.150, see also pp. 148, 264–265, 417), these designs bearing
witness to the participants’ ‘‘moral unity’’ (1961: 432). Typically, accord-
ing to Durkheim (1961: 159–161), such designs would be executed in
red ochre, a substance of ‘‘equal importance, religiously’’ as blood.
While drawing heavily on Aboriginal Australian ethnography, Durkheim
advanced the following arguments on the basis of general theoretical
considerations:
.The ‘‘emblems’’ of collective representations had to be abstract be-
cause the representations concerned ‘‘social facts’’—things that have
no real-world likenesses but exist only by virtue of collective agree-
ment (1961: 236).
Red ochre, body painting, and language 65
.Such emblems had to appear Wrst on the body because collective
ritual is a bodily display of participation ‘‘in the same moral life’’
(1961: 264–265).
.Red ochre inevitably symbolized blood, this substance being reiWed to
a ‘‘sacred principle’’ (1961: 159–61).
1
4.5 Innatist theory
The theory of ‘‘basic color terms’’ (e.g. Berlin and Kay 1969) has come to
exemplify the more general proposition that perceptually grounded semantic
categorization is a direct projection of innate cognitive universals, structured
by hard-wired neural mechanisms. This view, endorsed by some linguists
(Landau and JackendoV1993), is associated with cognitive psychology
(Fodor 1975; Pinker 1994) and evolutionary psychology (Tooby and Cos-
mides 1992). While a strong form of innatism is found in some ‘‘basic color
term’’ (BCT) literature (e.g. Berlin and Kay 1969: 109; Kay and McDaniel
1978: 611; Kay and Berlin 1997: 201), elsewhere the more limited claim is that
‘‘the semantics of basic color words . . . is partially constrained by parameters
of the visual system’’ (Kay et al. 1991: 24; see also Kay in Ross 2004).
A tacit assumption of BCT theory has been that the color domain is a
natural and universal semantic Weld of such salience that all languages will
have dedicated terms exhaustively partitioning the domain (cf. Kay 1999:
76). Among the original criteria for BCT status (Berlin and Kay 1969: 6)
were that terms should be monolexemic (excluding referent-based similes
for hue, e.g. ‘‘blood’’) and that application should not be restricted to a
narrow class of objects (as in the case of ‘‘blond’’). The principal cross-
cultural Wndings informing the theory were:
.All languages have between two and eleven BCTs.
.There are high levels of agreement within and between cultures as to
the focal points of the extremes of the achromatic scale (black and
white) and the four unique hues of red, yellow, green, blue (where
languages have BCTs in the appropriate hue area).
.The foci of BCT terms can be predicted from their number.
1
The argument as to why blood is reiWed to a sacred principle is minimally developed
in Elementary Forms (1961: 161 and footnote 50), the reader being referred to an earlier
paper (Durkheim 1898) in which menstruation plays the central role.
66 Watts
The last Wnding led to the hypothesis of an implicational scale of seven
cultural evolutionary stages (Berlin and Kay 1969: 4). Stage I comprised
Black versus White, followed by Red (Stage II), followed by either Green or
Yellow (Stages IIIa and IIIb). In outlining this successive encoding of new
foci, labels were used loosely, sometimes referring to category foci (e.g.
pure black), sometimes to the foci plus its extension (e.g. pure black and
all dark hues). In a revised formulation (Kay and McDaniel 1978), the
initial BCTs comprise two composite categories with several potential foci,
successively fractioned out in subsequent stages. Stage I comprised ‘‘light/
warm’’ versus ‘‘dark/cool’’ terms (respectively focused on white or red or
yellow versus black or green or blue). The new composite of Stage II was
‘‘warm’’ (focused on red or yellow). The vague biological explanation of
Berlin and Kay (1969: 109) was reformulated such that six ‘‘fundamental
neural response categories’’ (FNRs) of black, white, red, yellow, blue, and
green are ‘‘encoded as the universal semantic categories’’ (Kay and McDa-
niel 1978: 625, 627). FNRs were the postulated output of the achromatic
vision provided by rod cells (cf. Kandel et al. 2000: 507–513), and of
trichromatic vision—where signals from three types of cone cell are pitted
against one another (opponent processing) to derive a diVerence signal,
enabling the four unique hues to be discriminated from the wavelength
continuum (De Valois and De Valois 1993; Abramov 1997; but see Jame-
son and D’Andrade 1997). Trichromatic vision evolved with Old World
monkeys (Jacobs 2002).
How far back in time are these posited stages projected? Stage I terms
could potentially date to when vocabularies were of a size comparable to ‘‘the
repertoire of discreet [sic.] verbal signs used by apes and monkeys’’ (Berlin
and Kay 1969: 16). However, contrary to the impression given in some
commentaries (e.g. Hovers et al. 2003: 493), a biological mechanism
accounting for the order in which BCTs are labeled (as distinct from the
category foci) forms no part of theoriginal theory (Berlin and Kay 1969: 17).
Despite red being invariably labeled before other hues, red and yellow are
treated as equally plausible potential foci of ‘‘light/warm’’ and ‘‘warm’’ terms
(Kay and McDaniel 1978: Wg. 13; Kay and Berlin 1997: 201). Sahlins (1976:
3–8) outlined possible mechanisms and biases underpinning the ‘‘natural-
perceptual logic’’ to the emergence of BCTs, while arguing persuasively that
the terms are ‘‘codes of social, economic, and ritual value’’ (1976: 8), not
labels for natural categories.
Red ochre, body painting, and language 67
The ambiguous classiWcatory status of red in any binary lexicon is
evident in Berlin and Kay’s (1969) discussion of Stage I languages. Both
their examples concerned New-Guinea Highland cultures of the Danian
language group. Among the Jale
´, ‘‘the appearance of blood is sih‘BLACK,’
exactly as ‘blood (red)’ should be at Stage I because of its low brightness’’
(1969: 24, citing a seminar presentation by K. F. Koch). However, their
second source (Bromley 1967) reported that related languages divided
colors into ‘‘brilliant’’ and ‘‘dull,’’ the ‘‘brilliant’’ category including most
reds, yellow, and white (Berlin and Kay 1969: 24). The 1978 revision of
Stage I categories largely arose from Heider’s (1972) research with the
Dugum Dani. Presented with a saturated array of Munsell color chips, the
best exemplar of the ‘‘light/warm’’ term selected by her informants
(n ¼40) was not white but dark red (selected by 69%), followed by light
pink (most informants selecting pink already having a term for red that
denoted a red clay pigment).
One of the few other published studies of an arguably Stage I language
concerns the Gidjangali of Australia (Jones and Meehan 1978). The
Gidjangali ‘‘light’’ term—gungaltja—‘‘refers to light, brilliant and white
colors, and also to highly saturated red’’ (1978: 27). The authors empha-
sized the element of ‘‘brilliance’’ or ‘‘animation’’ in the gungaltja concept.
Asked to identify examples of this term among saturated Munsell chips,
their principal informant responded that there were no proper gungaltja
colors there, pointing instead to some silver foil. Subordinate to this
universal binary classiWcation, the color of a restricted range of objects
or states could be described using the terms for four pigments (pipe-clay,
yellow ochre, red ochre, and charcoal), constituting the four ritually
recognized colors. Jones and Meehan considered that djuno (red ochre)
was the color that excited most interest (1978: 31). The best Gidjingali
exemplars of this term were two types of ochre, with Munsell hues of
Purple and Red-Purple, both decidedly dark (brightness levels 4 and 3).
The darker type was ‘‘a high grade haematite with a lustrous purple
streak’’ which, when burnished on objects, gave ‘‘a metallic sheen’’
(Jones and Meehan 1978: 32). These ritually deWned and recognized colors
are at least as salient to color lexicalization as the two Gidjingali BCTs.
Both Jones and Meehan (1978: 27) and Heider (1972: 464) noted that
red’s inclusion in the ‘‘light’’ or ‘‘light/warm’’ term was paradoxical given its
low brightness (cf. Solso 1994: Wg. 1.7). Heider speculated that the original
meaning of the term glossed as ‘‘light/warm’’ was centered on ‘‘warm’’ dark
68 Watts
colors (i.e. red). Her Wndings led Berlin and Berlin (1975: 84) to conclude
that the foci for the two primordial categories are ‘‘Xuid and unstable,’’ and
to speculate that red might be the principal focus of ‘‘light/warm’’ terms,
with black dominating ‘‘dark/cool’’ terms. Neither speculation has been
pursued in subsequent BCT-inspired research, but one would not predict
speciWcally dark red to be exemplary in a red-versus-black opposition.
Red plays a more prominent role in well-documented Stage I languages
than is conveyed by the BCT glosses. Factors contributing to its exclusion as a
BCT may include active nominal reference and restricted usage; but, given
the critical role of metaphor in the evolution of language (Deutscher 2005),
nominal reference is likely to provide clues as to the domain in which color
lexicons arose. The most common analyzable root of any BCTor referent-
based simile for hue is ‘‘blood’’ (Greenberg 1963: 134; Berlin and Kay 1969:
38; Nash 2001; Everett 2005: 627; Deutscher 2005: 237), probably followed
by ‘‘red earth pigment’’ (see above; Koch in Berlin and Kay 1969: 23).
Analyzable ‘‘black’’ and ‘‘white’’ terms (e.g. Berlin and Kay 1969: 38–39,
citing Rivers 1901; Levinson 2000: 10; Everett 2005: 627) also often refer to
things with partible color (e.g. cockatoo feathers, charcoal, cuttleWsh ink).
Important exceptions to this tendency are ‘‘black’’ and ‘‘white’’ terms such as
‘‘night,’’ ‘‘darkness,’’ or ‘‘to see,’’ taking us beyond the labeling of surface
color. Such exceptions notwithstanding, ritual display would appear to be
deeply implicated in simple forms of color lexicalization.
Paul Kay himself now grants that there is no physiological evidence for or
against neural processing determining BCTs (Ross 2004). While there are
constraints from visual perception, it seems that perception can itself be
biased by linguistic categorization (Kay and Kempton 1984; DavidoVet al.
1999; Levinson 2000, 2003). Some cultures with simple classiWcatory systems
have no universal partitioning of the color domain, no composite color
terms, and referent-based similes for hue may circumscribe BCTs (Levinson
2000). In societies with simple coloring technology, other aspects of surface
appearance such as brightness/dullness, freshness/dryness, brilliance, or
pattern may be at least as salient as hue (e.g. Conklin 1955; MacLaury
1992; Lyons 1995; Casson 1997; Lucy 1997; Levinson 2000). Addressing
some of these challenges, BCT theory has been further revised (Kay and
MaY1999), but as Levinson (2000, 2003) points out, the revision is incom-
patible with the innatist claim that universal perceptual constraints directly
determine semantic universals (Kay and McDaniel 1978: 610; Durham 1991:
281; Shepard 1992: 522; Pinker 1994: 63; Hovers et al. 2003: 493).
Red ochre, body painting, and language 69
4.5.1 Archeological application of BCT theory
Using Berlin and Kay’s (1969) original formulation, Hovers and colleagues
(2003: 493) attempt to apply BCT theory to archeological data.
2
One would
predict on this basis that the earliest pigments should be black and white,
followed by red, and then yellow. They claim: ‘‘red and black pigments are
relatively ubiquitous in Paleolithic . . . sites, from the Plio/Pleistocene to Upper
Paleolithic’’ (2003: 491). Archeologists areurgedtore-examineexistingcol-
lections for black and white pigments, as their presence would be ‘‘in line with
linguistic studies of color terms’’ and the ‘‘infrastructure of trichromatic
vision’’ (2003: 518).
3
If, instead, Kay and McDaniel’s (1978) formulation had
been used, there is no theoretically grounded predictable order in which red,
black, yellow, and white (the most common earth pigments) should appear,
since all are potential foci of Stage I composite terms.
4.6 QualiWed innatism
Deacon (1997) accepts the BCT hypothesis concerning the stages of color
lexicalization; he, too, presents the unrevised stages (1997: 117).
4
However,
challenging innatism, he argues (1997: 119) that the process by which
shared—perceptually based—semantic categories emerge is determined
both by hard-wiring and ‘‘the pragmatic constraints of human uses.’’ He
goes on to make a more general argument, positing the demands of ritual
action as the earliest pragmatic constraint in ‘‘symbol discovery’’ (1997: 402).
SpeciWcally, Deacon argues for the primacy of rituals cementing sexual
contracts, arguing that these extend back !2my (1997: 384–401). He con-
cludes: ‘‘Out of the ritual processes for constructing social symbolic rela-
tionships, symptoms of the process itself (exchanged objects,body markings,
etc.) can be invested with symbolic reference’’ (1997: 406).
Discussing a probable association of red ochre with early Homo
sapiens burials at Qafzeh (Palestine), !92 ky, Hovers and colleagues
(2003: 508-509) invoke Deacon’s argument about the role of ritual in
constructing symbols (his critique of innatism goes unremarked). No
2
Although Kay and McDaniel (1997) and MacLaury (1992) are cited, Hovers and
colleagues do not refer to revised BCT stages.
3
Trichromatic vision does not concern achromatic perception.
4
Deacon’s account incorrectly states that the simplest classiWcations comprise three
terms and that green necessarily follows the labeling of red.
70 Watts
ritual context is proposed for any other early pigment occurrences. They
conclude (2003: 509) that the record of early use of red and black pigments
(purportedly extending back !2 my) accords with Deacon’s claim for
early beginnings to the gradual co-evolution of the brain and symbolism.
However, they continue, ‘‘normative social constructs’’ can be inferred
only when co-associations of the kind argued for at Qafzeh occur. The use
of Deacon to theoretically justify a focus on ritual represents a welcome
development in archeological discussion of early pigment use and sym-
bolism in general. However, uncritical adherence to BCT theory precludes
the possibility of ritual displays themselves inXuencing pigment choice.
4.7 The ‘‘Female Cosmetic Coalitions’’ model
The FCC model (Knight et al. 1995; Power and Aiello 1997; Power 1999;
Power this volume) has much in common with Deacon’s model of the
origins of symbolic culture. Both approaches stress conXicting male-
versus-female reproductive interests in the context of encephalization,
maternal energy budgets, and access to meat and mating opportunities.
Both identify ritual as the basic mechanism for resolving these conXicts.
Finally, both agree that as brain size increased with increasing group size,
females had to bear the costs of producing and maintaining increasingly
slow-maturing, energetically demanding babies. While Deacon envisages
wedding ceremonies stretching back to the Plio-Pleistocene, the FCC
model envisages initiation rituals of much more recent date. According
to this model, pressure to reward investor males at the expense of philan-
derers favored concealed ovulation, extended receptivity, and enhanced
capacities for ovulatory and menstrual synchrony. With signals of ovula-
tion phased out, menstruation was left salient as a signal of imminent
fertility. Males are expected to compete to bond with females perceived to
be cycling, doing so at the expense of current partners who are pregnant or
nursing. Females threatened by corresponding loss of male investment
should respond by scrambling the signal. Building on standard explan-
ations for ovulation concealment (Alexander and Noonan 1979; Hrdy
1981; Sille
´n-Tullberg and Møller 1993), a similar logic is applied to mens-
truation. How might females scramble the information divulged by this
biological signal? ArtiWcial pigments suggest one possibility (Plates 6–8). In
this scenario, menstrual onset prompts pregnant/lactating females to paint
Red ochre, body painting, and language 71
up as ‘‘imminently fertile’’ on the model of their cycling relatives. This leads
to the following archeological predictions concerning pigment use:
.The initial focus should be on red rather than black, white, or yellow.
.Pigment use should not predate the marked increase in encephaliza-
tion that begins in the middle of the Middle Pleistocene (between
!400 ky and !550 ky, RuVet al. 1997; Rosenberg et al. 2006). It should
predate the achievement of modern encephalization quotients, be-
tween !200 ky and !100ky (De Miguel and Henneberg 2001).
.Within this time-window (c. 500 to 150 ky), there should be a shift
from irregular to regular use of red cosmetics (accompanied by rapid
spread of such usage) as an initially context-dependent ‘‘sham men-
struation’’ strategy was raised to the level of a regular monthly
ceremony, performed whether or not a menstruant was present.
.Coalitions living in areas lacking blood-red earth pigments would be
expected to incur heavy costs to procure them from elsewhere.
4.8 Utilitarian hypotheses
Challenging the presumption of use as pigment, some archeologists have
suggested alternative general explanations for early ochre use—foremost
being the hypothesis that ochre was used as a tanning agent and/or as a
functional ingredient in cements for hafting stone tools (Klein 1995;
Wadley et al. 2004; Wadley 2005a).
The tanning hypothesis arises from a misunderstanding of basic chemis-
try, where the properties of certain soluble iron salts (e.g. iron sulphate) have
been assumed to be shared by relatively insoluble iron oxides (e.g. Keeley
1980: 172; Knight et al. 1995: 88; Wadley et al. 2004: 662; all citing Mandl’s
[1961] experiments with metal salt solutions). Iron salts have been used as
tanning agents (Tonigold et al. 1990), but no ethnographic or leather
industry sources conWrm similar use of iron oxides.
5
This hypothesis can
therefore be dismissed.
5
The two claimed ethnographic precedents for use of ochre as a tanning agent (cf. Wadley
et al. 2004: 662; Wadley 2005a: 589; Audouin and Plisson 1982: 57) do not bear scrutiny.
Steinmann’s (1906: 78) inference is contradicted by more detailed observations on Tehuelche
hide working (Cooper 1946: 148, with refs.). Sollas (1924: 275) made no functional claim; his
uncredited primary source (Mathews 1907: 35) simply stated that the mixture of ochre and
grease made garments water resistant. The claimed experimental support (Wadley et al. 2004:
662; Wadley 2005a: 589, citing Audouin and Plisson 1982) can be more parsimoniously
accounted for by the desiccating action of red ochre (Phillibert 1994: 450).
72 Watts
The hafting-cement hypothesis is consistent with archeological reports
from relatively late (post-80 ky) MSA assemblages, where ochre residues on
stone tools were predominantly restricted to parts whichwould have been in
a haft (Lombard 2007 with references). Replication experiments (Allain and
Rigaud 1986; Wadley 2005a) conWrmed that the inclusion of either yellow
ochre or hematite in resin-based cements made them more manageable
during use, helped drying and hardening, and made them less brittle.
However, no property of ochre has been identiWed that might make
it preferable to the wide range of ethnographically documented Wller/
loading agents, most of which would be easier to procure and
process. Australian accounts mention the use of plant Wber, dung, calcined
powdered shell, powdered charcoal, dirt, sand, and ochreous dust (Dick-
son 1981: 67–69, 164; Helms 1892–6: 274, 280). The primary require-
ment appears to have been for substances that were desiccant but
otherwise inert.
6
Presenting this as a plausible general account even for early large MSA
‘‘pigment’’ assemblages such as Twin Rivers (where 60 kg of pigment is
estimated to have been recovered in the original excavations—Barham
2002b), Wadley (2005a: 599) has suggested that such assemblages might
resemble the material used in her hafting experiments. This comprised
3 kg of ironstone nodules, only the weathered cortices of which were
pigmentaceous. Consequently, seven hours’ ‘‘vigorous’’ grinding (Wadley
2005b: 5) exhausted the nodules, but produced just 70 ml of powder
(enough to haft 28 tools). However, available evidence is that MSA
assemblages overwhelmingly comprise homogenously pigmentaceous
material (Watts 1998; Barham 2002b; and see below).
7
Additionally, if
2.5 ml of powder (representing 15 minutes’ work) was required to haft one
tool, one would not predict pieces of ochre with solitary, small grinding
facets. That both yellow ochre and hematite were experimentally success-
ful implies that the hypothesis is null with respect to the hue and chroma
of raw materials. At present, ochre in hafting cements is more parsimo-
6
Burnt shell may additionally have served as a polymerising agent (Dickson 1981: 70).
7
Twin Rivers is currently the best described MSA pigment assemblage; there is nothing
in Barham’s (2000, 2002) accounts indicating non pigmentaceous associated material. In
Watts’ examination of over 4,000 pigments from 11 southern African MSA assemblages,
only three pieces are reported as predominantly non pigmentaceous material (Watts 1998:
plates 5.1, 6.81 and tbl. 6.46).
Red ochre, body painting, and language 73
niously interpreted in terms of symbolic considerations determining
functional choices.
4.9 Early pigment occurrences: diVerences between African
and Eurasian hominins
There are two claims (Leakey 1958: 1100; Beaumont and Vogel 2006: 222)
and one suggestion (Clark and Kurashina 1979) for ochre use in the Lower
Pleistocene (790 ky to 1.8 mya) and early Middle Pleistocene (c. 500 ky to
790 ky), but these are not compelling.
8
Middle Pleistocene (130 ky to
790 ky) occurrences are listed in Table 4.2. Current evidence suggests
initial use in the middle of the Middle Pleistocene, between !300 ky and
!500 ky (Howell 1966; de Lumley 1969; Barham 2002b; Tryon and
McBrearty 2002; Brooks 2006a; Beaumont and Vogel 2006). However,
only one of these early occurrences (Barham 2002b) has been adequately
published, and doubts have been raised whether the material at two of the
European sites was pigment (Butzer 1980 re. Ambrona; Wreschner 1983,
1985 re. Terra Amata). A stronger case for initial European use can be
made at !250 ky (The
´venin 1976: 984; Marshack 1981).
Initial occurrences of red ochre may be broadly coeval, but European and
African records for the later Middle Pleistocene and earlier Late Pleistocene
diVer dramatically. For Middle Pleistocene Europe, there are at most Wve
occurrences, three of which are questionable. All are thought to predate
220 ky, and are followed by a Wnd gap of at least 100,000 years (Wreschner
1985: 389). Even after this gap, I know of only two cases from the earlier Late
Pleistocene, between 128 ky and 75 ky (Demars 1992 re. Combe Grenal
layers 57/8; Marshack 1976 re. Tata). The great majority of the 40 or so
European Mousterian sites with pigment date to the Last Glacial (beginning
74 ky), most post-date 60 ky, and manganese predominates over red ochre
(Demars 1992; d’Errico and Soressi 2006: 86). Forty is a small proportion of
8
Citing Mary Leakey (Leakey 1971), Dickson (1990: 42 43) states that the two pieces of
red ochre (subsequently identiWed as rubiWed tuV) from Olduvai Bed II at site BK ‘‘show
signs of having been struck .. . by hammerstone blows’’. The basalt manuports at the Lower
Pleistocene site of Gadeb, Ethiopia (Clark and Kurashina 1979) showed no signs of use and
pigmentaceous material was weathered cortex. Beaumont and Vogel (2006) claim that
hematite use at Wonderwerk extends to the initial Middle Pleistocene; however, the
hematite is thought to derive from the cave host rock and no use wear is reported, so
the claim remains to be substantiated.
74 Watts
Table 4.2. Middle Pleistocene potential pigment occurrences.
Site Country Unit
Approximate
age
Technological
association Pigment
References
(dating references in
parentheses)
Europe
Terra Amata France 380+80 ky (ESR)
or
Acheulian Questionable de Lumley 1966, 1969;
Wreschner 1983, 1985
214 & 244 ky (TL) (Falgue
`res et al. 1991;
Wintle & Aitken 1977)
Ambrona Spain >350 ky Early Acheulian Questionable Howell 1966; Butzer
1980
(Pe
´rez Gonza
´les et al.
2001)
Maastricht
Belvedeer
Holland Site C, Unit
4
c. 250 ky Middle Paleolithic Probable Roebroeks 1988
Achenheim France Middle
Loess (lvl
19)
c. 250 ky Middle Paleolithic Good The
´venin 1976; Wernert
1952
(Buraczysky & Butrym
1984)
Bec¸ ov 1A Czech Rep. c. 222 ky? Middle Paleolithic Good Fridrich 1976, 1982;
Marshack 1981
India
Hunsgi southern
India
c. 200 300 ky Acheulian Good Bednarik 1990
(Continued)
Table 4.2. (Continued).
Site Country Unit
Approximate
age
Technological
association Pigment
References
(dating references in
parentheses)
Africa
Sai Island Sudan BLG/TLG
gravel
c. 200 ky? ‘Lower’ Sangoan Good van Peer et al. 2004
RS sand 152+10 ky,
182+20
‘Middle’ Sangoan Good
OlorgesaillieyKenya B OK 1 >340 ky, <493 ky Post Acheulian Probable Brooks 2006a
Kapthurin
(GnJh 15)
Kenya K3 Sedi
ments
>285 ky Fauresmith/MSA Good McBrearty 2001 (Tryon
& McBrearty 2002)
Mumba
^
Tanzania Stratum VIB 132 ky Sanzako (MSA) Probable Mehlman 1979:91
(Mehlman 1991)
Twin Rivers Zambia A Block 266 ky to >400 ky
(?)
Early Lupemban Good Barham 2002
F Block 140 ky to 200 ky Early Lupemban Good
Mumbwa Zambia Unit X >172+22 ky MSA Good Barham 2000
Kabwe Zambia c. 300 400 ky ? Charama? Probable Clark 1950 (Barham
et al. 2002)
Kalambo Falls* Zambia Lupemban (Sis
zya)
Probable Clark 1974
Pomongwe Zimbabwe Area 1, lyrs
22 27
Charama (MSA) Good Cook 1963, 1966; Watts
1998
Bambata Zimbabwe Lower Cave
Earth
Charama (MSA) Good Jones 1940:17 (cf. Cook
1966 re. Charama)
Wonderwerk S.A.
(Northern
Cape)
Major Unit
7
Major Units
34
c. 790 ky
276+29, >350 ky
Acheulian
Fauresmith
Questionable
Probable
Beaumont & Vogel 2006
Kanteen Koppie S.A.
(Northern
Cape)
Stratum 2a Fauresmith Probable Beaumont 2004
Nooitgedacht 2 S.A.
(Northern
Cape)
Fauresmith Good Beaumont 1992a
Pniel 6 S.A.
(Northern
Cape)
Stratum 3 MSA/Fauresmith? Good Beaumont 1992b, Watts
1998
Kathu Pan 1 S.A.
(Northern
Cape)
Stratum 4a Fauresmith Probable Beaumont 1992c
Stratum 4b Acheulian Questionable Beaumont & Vogel
2006}
Kathu Townlands 1 S.A.
(Northern
Cape)
Acheulian Questionable Beaumont & Vogel
2006}
Duinefontein 2 S.A.
(Western
Cape)
>270 ky, <290 ky Late Acheulian Good Cruz Uribe et al. 2003
(Continued)
Table 4.2. (Continued).
Site Country Unit
Approximate
age
Technological
association Pigment
References
(dating references in
parentheses)
Border Cave S.A. (Kwa
Zulu
Natal)
6BS
5WA
5BS
>227 ky
227+11, 174 +9
166+6, 147 +6
MSA
MSA
MSA
Good
Good
Good
Watts 1998 (Gru
¨n&
Beaumont 2001)
Pinnacle Point 13B S.A.
(Western
Cape)
LC MSA
Lower
164 ky MSA Good Marean et al. 2007
Blombos Cave S.A.
(Western
Cape)
Layers
CL CP
>143 ky MSA Good Watts this paper (Jacobs
et al. 2006)
yPending geochemical analysis, the artifactual status of the ochre at Olorgesaillie remains indeterminate (Brooks pers. comm.)
^
The traits of the Sanzako industry suggest that the U series date on bone (from approximately the same level as the oldest ochre), may be a minimum age.
* The Siszya Lupemban is only assigned a Middle Pleistocene age on the basis of comparison with Twin Rivers
}The authors cite Beaumont & Morris 1992 for this claim, but the relevant paper (Beaumont 1990c) does not mention pigment in stratum 4b at Kathu Pan or in the
Townlands site
excavated Mousterian sites. It is not until the arrival of modern humans
(!40 ky) that pigment use in Europe becomes ubiquitous, when it over-
whelmingly takes the form of red ochre. The last (Cha
ˆtelperronean) Nean-
derthals, living alongside the newcomers, also start using much larger
quantities of red ochre (Harrold 1989: 696; Couraud 1991).
In Africa, it is estimated that the number of excavated MSA sites is only
a tenth of the European Mousterian ones (McBrearty and Brooks 2000:
531). Despite the less intensive history of research, excavation units from
at least 18 sites dated or believed to date to the Middle Pleistocene (three
to six times the number of European sites) have provided probable earth
pigments. Most of the earliest occurrences span the transition from the
Acheulian to the MSA. Contrary to some authors (Wadley 2005b: 2;
d’Errico et al. 2003: 4), Barham’s Twin Rivers excavation did not provide
evidence for use of a wide range of colors. Use-wear was restricted to nine
pieces of specularite (laminar crystalline hematite) and a piece of pedo-
genic, earthy ‘‘hematite’’ (Barham 2002b: table 1). The specularite is
thought to have come from further aWeld than the hematite; it produced
‘‘a darker, purple shade of red (Munsell 10R 4–3/3–3) that sparkles’’
(Barham 2002b: 185). Of the 302 potential pigments (1,617 g), 93.1%
were red (92.4% by weight), these being overwhelmingly specularite.
None of the yellow limonite was utilized, although a limonite ‘‘crayon’’
is reported from the original F Block excavations (Clark and Brown 2001:
Wg. 20, no. 23). Barham treats the tiny amount of manganese as intro-
duced to the site, but this could readily have come from autochthonous
concretions (Barham 2002b: Wg. 3). The only well-supported case for
Middle Pleistocene black pigment is a small fragment of graphite associ-
ated with ‘‘Middle’’ Sangoan material at Sai Island. This site is also unique
among both Middle and earlier Late Pleistocene assemblages in that
yellow predominates over red. All other Middle Pleistocene reports exclu-
sively concern red ochre in one form or another.
Pigment use is not ubiquitous in the early MSA, nor is it necessarily a
regular behavior in the early assemblages where it is documented. For
example, at Kalambo Falls in Zambia it is absent in the early Lupemban
but present in the later Lupemban (Clark 1974: table 10). In the long cave
sequences of Mumba, Pomongwe, and Bambata, it is rare in the basal
assemblages, becoming more frequent in overlying layers. In South Africa,
it is absent in the large, early (undated) MSA assemblages at Peers
Cave and Cave of Hearths, but is a recurrent feature of overlying Late
Red ochre, body painting, and language 79
Pleistocene MSA layers (Volman 1981: 325; Mason 1957: 135; pers. obs.
regarding Peers Cave Late Pleistocene).
Currently the most informative South African site for this period is
Border Cave (Watts 1998). Figure 4.1 shows the relative frequency of
pigment for the Wrst Wve stratigraphic aggregates. The basal unit (6BS) is
>227 ky (Gru
¨n and Beaumont 2001); a sample of almost 10,000 lithics
provided just one piece of ochre, on the threshold of archeological visi-
bility. The overlying unit, with a similarly sized lithic sample, provided
just three pigments. The youngest of the Middle Pleistocene units (5BS)
provided inverted dates of 166 ky and 147 ky, placing it in the middle of
the penultimate glacial. This witnesses a Wvefold increase in relative
frequency, with overlying Late Pleistocene samples providing comparable
percentages. At this site, use of red ochre only became regular between
!170 ky and !150 ky.
Pinnacle Point (approximately 85 km east of Blombos) conWrms regular
use of red ochre from !164 ky (Marean et al. 2007). There is suggestive
evidence, therefore, that red ochre use in southern Africa only became
habitual and ubiquitous to cave/rockshelter occupations with the spread
of Homo sapiens. It remains so thereafter (Watts 1999). In the Late
Pleistocene MSA of southern Africa, non-red pigments are rare, with
black, white, and yellow largely restricted to a few Still Bay and Howiesons
Poort contexts (!75 ky to !60 ky) (Watts 2002: 10–11).
To conclude, archeology provides no support for revised or unrevised
versions of BCT Stage I, for Deacon’s proposed Plio-Pleistocene weddings,
or for use of red and black pigment (‘‘ubiquitous’’ or otherwise) extending
back to the Plio-Pleistocene. The African Middle Pleistocene record is
probably of greater antiquity than non-African counterparts, is certainly
much more extensive, and—unlike the European record—is continuous.
The habitual use of red ochre can be considered a species-deWning trait.
Occasional use may have occurred among all post Homo erectus/ergaster
lineages, but it is no longer acceptable to suggest that the pigment record
of Neanderthals and their immediate ancestors is comparable to that of
early Homo sapiens and their immediate ancestors (e.g. Klein 1995: 189).
Having discounted the two principal utilitarian hypotheses as alternative
general explanations, it is the habitual nature of the behavior from the end
of the Middle Pleistocene in southern Africa (probably earlier in the
African tropics) that permits the inference of habitual collective ritual,
80 Watts
0
0.04
0.08
0.12
0.16
0.2
6BS (>227ky) 5WA (174-
227ky)
5BS (166-
147ky)
4WA (120ky) 4BS (82ky)
Percentage pigment relative frequency
Fig. 4.1 Changes in the relative frequency of pigment in the earlier MSA (Pietersburg) units at Border Cave
(KwaZulu-Natal). Excavation units from Beaumont’s 1987 excavation. Relative frequency pigment counts
as a percentage of the combined pigment and lithic assemblages. Source: Watts 1998: fig.7.23, dates from
Grun & Beaumont 2001.
with applications of red pigments to the body playing an integral part in
ritual displays.
Given the posited relationship between collective ritual and language
(Knight 1998), the higher-level inference is that, at least by the terminal
Middle Pleistocene, speech communities were distributed across Africa,
with roots probably going back at least 250 ky within the tropics (see also
Barham 2002b). The temporal and color focus predictions of the FCC
model are particularly consistent with the summarized African data,
although a detailed account of the claimed early hematite at Wonderwerk
(footnote 8) is awaited, and the predominance of yellow ochre at Sai
Island is surprising. This raises some intriguing questions in relation to
Europe. Why should a lineage ancestral to Neanderthals have brieXy and
sporadically engaged in a behavior consistent with the FCC hypothesis,
only to abandon it? Why should a more varied form of pigment use
reappear with late classic Neanderthals, only to converge with modern
human practice during the brief period of co-existence (see Power this
volume)?
4.10 Red ochre use at Blombos Cave
The coastal site of Blombos Cave has provided some of the earliest
compelling evidence for symbolic traditions: shell beads (some bearing
ochre residues), dated to !75 ky (d’Errico et al. 2005), and geometric
engravings on ochre spanning the period from !100 ky to !75 ky
(Henshilwood et al. 2002, in press). I know of no hunter-gatherer society
without some form of body marking—predominantly body painting, but
including also tattooing and scariWcation. As predicted on theoretical
grounds by Durkheim, the designs are invariably non-Wgurative, compris-
ing geometrically arranged lines or shapes (e.g. Spencer and Gillen 1899;
Teit 1927–8; Drury 1935: 102; Marshall 1976: 276; Lewis 2002, Plate 9.4,
9.5; Fiore 2002). It is almost inconceivable that the MSA occupants of
Blombos were engraving such designs onto pieces of ochre while not
doing similar things with ground ochre powder on their bodies (grinding
being the predominant form of use-wear).
The MSA sequence spans the period from >143 ky to !70 ky (Jacobs
et al. 2006). Ochre is present throughout. At least in the younger occupa-
82 Watts
tions, its use seems to have permeated many aspects of life. As well as
appearing on some beads, it may have been used as a polishing agent to
lend ‘‘added value’’ to some of the bone tools from the !77 ky layers
(Henshilwood et al. 2001b). My own cursory examination of selected
lithics found variably distributed ochre residues on a variety of tools—
predominantly from the younger layers.
Over 1,500 pigment pieces $1 cm in length were analyzed, weighing 5.6
kilos.
9
Shale, siltstone, and coarse siltstone predominate (Figure 4.2). Fewer
than a dozen pieces (c. 150g in total) were associated with signiWcant non-
pigmentaceous material. While pigment use was habitual, the quantities vary
enormously through time (Layer CI, for example, accounts for half the
assemblage mass). This is attributed to changes in sea level and sand cover,
exposing and then maskinga local exposure of Bokkeveld shale and siltstone.
The inference is based on two observations. First, in layers where pigment is
most abundant (CJ-CH)—culminating around 100ky—much of it shows
borings by pholadid molluscs and carbonate tests of marine organisms (e.g.
Plate 4.1), testifying to procurement from the wave-eroded coastal pene-
plane. Second, although there is currently no exposed Bokkeveld within
c. 15 km of the site, there is an extensive, largely masked contact between
the Bokkeveld and (non-ochreous) quartzitic sandstone (Table Mountain
Group), running parallel to the coast (Rogers 1988: 411); the closest coastal
intercept to Blombos is estimated 3–5 km WNW (masked by beach sand).
Where ochre is less abundant (underlying CL-CP, and overlying CF-BZ),
tracesof marine organisms are rare (absent above CF), and hematite and Wne
sandstone are better represented (Figure 4.2). Color proWles
10
track the raw
material changes (Figure 4.3), with the combined representation of ‘‘satur-
ated reddish-brown,’’ ‘‘very red,’’ and ‘‘very dark’’ values tracking hematite
9
The data presented here supersedes the preliminary site report (Henshilwood et al.
2001a). They will be presented more exhaustively in a forthcoming report.
10
The Natural Color System Index (2nd edn., 1999) was used to code streaks. This uses a
percentage metric for blackness, chroma, and hue. Values were grouped along two axes.
Nuance (combined blackness and chroma) was divided into pastel, intermediate, and satur
ated groups. ‘‘Saturated’’ cases have the highest chroma for given levels of blackness. Pastels
have the lowest chroma (in the range of 10 25%) for the same levels of blackness. Intermediate
nuances fall between these poles. Blackness and chroma values above the 5th percentile in the
10% intervals are rounded up. Hue groupings were ‘‘yellow brown’’ (<50% redness), ‘‘reddish
brown’’ (50 74% redness), and ‘‘very red’’ ($75% redness). Values with $56% blackness
were grouped into a ‘‘very dark’’ category, regardless of chroma or redness.
Red ochre, body painting, and language 83
and Wne sandstone, and ‘‘intermediate’’ reddish-brown and yellowish-brown
tracking Wne-grained sedimentary materials.
11
When ochre was scarce, Blombos occupants could have abandoned its
use; or traveled 15–20 km east to obtain similar material from the lower
Goukou Valley; or travelled 35–40 km north (inland) to obtain higher-
quality materials (an area commercially quarried for red and yellow
ochre). The inland exposures are also Bokkeveld, but, being beyond the
0%
20%
40%
60%
80%
100%
BZ-CB n=148, 319g
CC (73 ky) n=57, 266g
CD/CE n=49, 236g
CF (77 ky) n=56, 152g
CG n=18, 30g
CL-CP (>143 ky) n=46, 51g
CK n=32, 113g
CJ n=312, 833g
CH (100 ky) n=131, 809g
CI (100 ky) n = 685, 2772g
Other Haem' Md Snd Fn Snd
Crs Silt' Siltstone Shale
Fig. 4.2 Pigment raw material profiles by excavation aggregate at Blombos Cave
(1998/1999 excavations). Percentages based on frequency, column headings also
providing total mass (grams).
11
Twenty two of the 23 ‘‘very dark’’ values had $70% redness and just four had $70%
blackness; most can, therefore, be considered extensions of intermediate and saturated
reddish brown and very red groupings. Among intermediate yellowish browns (n¼156),
78.2% fall within 10% of the yellow/red cut point, and can be treated as an extension of
reddish browns.
84 Watts
0%
20%
40%
60%
80%
100%
BZ-CB n=148
CC (73 ky) n=57
CD/CE n=49
CF (77 ky) n=56
CG n=18
CH (100 ky) n=130
CI (100 ky) n=685
CJ n=312
CK n=32
CL-CP (>143ky) n=46
Very dark
Sat v. Rd
Sat Rd-Brn
Int dk v. Rd
Int Rd-Brwn
Int Yllw-Brn
Pstl Rd-Brn
Pstl Yllw-Brn
Grey
Fig. 4.3 Grouped colour (streak) profiles by excavation aggregate at Blombos Cave (1998/1999 excavations).
Percentages based on frequency of grouped Natural Color System (NCS) values.
reach of Cainozoic marine transgressions, they are more deeply weath-
ered, with more pronounced secondary alteration (e.g. hematization).
Additionally, sandier expressions of Bokkeveld are more common than
to the south (Theron 1972: 18, 58).
12
Judging by raw material and color
proWles, the last option was frequently chosen. This complements results
from Qafzeh, where local ochre was ignored in favor of more distant, more
hematite-enriched material (Hovers et al. 2003).
The incidence of use-wear is strikingly correlated with redness
(Figure 4.4). Only about 10% of yellowish-brown pieces were utilized; as
soon as red predominates, utilized percentages incrementally increase, peak-
ing at !50% of pieces with $80% redness. Not only were the redder pieces
more likely to be used; saturated reds were more likely to be used than
intermediate counterparts (Figure 4.5). The pattern persists even in layers
such as CI, where local procurement prevailed (peak rates of utilization shift
to ‘‘intermediate very red’’ and ‘‘saturated reddish-brown,’’ owing to small
sample size bias among ‘‘saturated very red’’ [n ¼10]). The great majority
(82.7%) of ‘‘saturated very red’’ values were moderately dark ($35%, <56%
blackness). The preferential use of what might be dubbed ‘‘blood-reds’’ is
borne out by estimates of the intensity of grinding (Figure 4.6). The reddest
and the most saturated pieces were more likely to be intensively ground than
less red, less saturated counterparts. Intensively ground pieces were multi-
faceted, typically with facets converging to a point (e.g. Plate 4.2, 4.3, 4.4a–b),
such pieces generally being described as ‘‘crayons’’ (but see Wadley 2005b).
Harder pieces could only have acquired this morphology through prolonged
processing, probably involving multiple episodes of use. This in turn would
imply safekeeping between episodes of use, further supporting the inference
of high esteem. The color selection is consistent with that reported from
Twin Rivers, from the terminal Middle Pleistocene at nearby Pinnacle Point
(Marean et al. 2007), and with Watts’ (2002: 10) more subjective observa-
tions on a multi-site MSA sample. It also accords with the two previously
discussed ethnographic accounts of the best (indigenous) exemplars of what
have been glossed as ‘‘light’’ or ‘‘light/warm’’ color terms.
While ‘‘saturated very red’’ pigments (typically fairly dark) were the
most valued, the margins of the utilized distribution deserve comment.
12
Because coastal and inland ochreous exposures are the same substrate, and because
Bokkeveld shale formations are mineralogically uniform (Danchin 1970), it is unlikely that
geochemical and mineralogical analyses could test this interpretation.
86 Watts
0%
20%
40%
60%
80%
100%
15-25% (n=19)
30-35% (n=41)
40-45% (n=146)
50% (n=262)
55% (n=23)
60% (n=480)
65% (n=82)
70% (n=349)
75% (n=66)
³80% (n=65)
Percentage redness
Unmodified
Possibly
Probably
Definitely
Fig. 4.4 Percentages of utilization confidence assessments by percentage redness (redness as a ratio to
yellowness).
Utilized ‘‘pastel yellowish-browns’’ (n ¼5) have #25% blackness; light-
ness rather than hue seems to be the most perceptually salient attribute of
this group (which includes two geometrically engraved pieces from CI).
Of 18 utilized ‘‘intermediate yellowish-browns,’’ three had >60% yellow-
ness, the rest were close to reddish-browns. One of the four utilized ‘‘very
dark’’ values had 70% blackness and was described as brownish-black.
These three peripheral subgroups are signiWcant in showing that the focus
on reds was not exclusive; light, very dark, and yellowish materials were
occasionally used. Assuming a color lexicon, such pieces may have been
distinguished from the vast bulk of the assemblage; but, as at Twin Rivers,
their rarity underlines just how preoccupied MSA people were with red.
As with the overall survey, these few pieces do not support the binary
oppositions predicted by either the original or the revised versions of BCT
theory. Short of invoking untestable propositions about use of white ash
and charcoal, the only recurrent opposition that might be archeologically
0%
20%
40%
60%
80%
100%
Grey n=7
Pstl Yllw-Brn n=47
Pstl Rd-Brn n=40
Int Yllw-Brn n=156
Int Rd-Brwn n=979
Int v. Rd n=43
Sat Rd-Brn n=163
Sat v. Rd n=75
Very dark n=23
Unmodified Possibly Probably Definitely
Fig. 4.5 Percentages of utilization confidence assessments by grouped NCS values.
88 Watts
0%
20%
40%
60%
80%
100%
Intnsv
Grnd
n=37
Mod'
Grnd
n=60
Grnd
Frag'
n=42
Lgt Grnd
n=52
Unutilized
n=1146
Very dark
Sat v. Rd
Sat Rd-Brn
Int dk v. Rd
Int Rd-Brwn
Int Yllw-Brn
Pstl Rd-Brn
Pstl Yllw-Brn
Grey
Fig. 4.6 Percentages of grouped NCS values by intensity of grinding compared to unutilized pieces
(excludes non-ground forms of utilization and non-definite assessments).
inferred would be ‘‘signal on’’ (ritual display with red pigment) versus
‘‘signal oV’’ (no pigment use).
Two Wnal features worth noting are the small size of many utilized
pieces, and the high proportion of lightly utilized ones. Of 307 deWnitely
used pieces, 80 were judged to be $90% complete. A quarter of these
were between 1.5 cm and 2.5 cm long (mean 19 mm, s.d. 2.9 mm, n ¼22),
just large enough to be held between foreWnger and thumb. Some are
intensively utilized, others less so (e.g. Plates 4.5, 4.6). That individual
episodes of use often only produced tiny amounts of powder is evident
among lightly utilized pieces (e.g. Plates 4.7, 4.8). Eleven lightly ground
pieces (n ¼52) were judged $90% complete with single facets; facet
widths were recorded for seven of these, providing an average of 2.9 mm
(s.d. 0.8mm). Ten of the eleven cases had $70% redness. The high propor-
tion of saturated and very red values among lightly ground pieces
(Figure 4.6) suggests that, rather than representing mere trials in the search
for the reddest, most saturated pigments, these were used similarly to more
intensively ground counterparts. The tiny amounts of powder produced
would surely have been insuYcient for just about anything other than design
purposes.
In summary, MSA people at Blombos preferred saturated red earth
pigments. These were more likely to be ground and ground intensively,
probably involving multiple episodes of use and curation. At the same
time, individual episodes of use often only produced tiny amounts of
powder (with similar selective criteria), a practice probably inconsistent
with anything other than making designs on the face, body, or some other
organic surface. Together with the geometric engravings (from c. 100 ky),
this provides good circumstantial evidence for the use of typically ‘‘blood-
red’’ ochre in the painting of abstract designs on the bodies of ritual
performers, from at least 143 ky. That a nearby assemblage shows identical
selective criteria from "164 ky (Marean et al. 2007), suggests that this
cultural tradition was already established by the time of our speciation,
between "150 ky and "200 ky.
4.11 Discussion
With the exception of the tanning hypothesis, all of the theoretical per-
spectives and inductive hypotheses considered here have some explanatory
90 Watts
merit. The hafting hypothesis partially explains the archeological observa-
tions from which it arose. However, until functional properties additional
to those of known alternative and cheaper Wller/loading agents are dem-
onstrated, it is more parsimonious to infer that this was a case of symbolic
considerations inXuencing a functional choice. As a general explanatory
hypothesis, hafting cannot account for large assemblages, the hue and
chroma-based selective criteria, or pieces with solitary, small grinding facets.
The cross-cultural Wndings associated with BCT theory are fairly ro-
bust, and few doubt that biology constrains both color categorization and
naming. What is contested is whether biology provides suYcient con-
straints for coordinating perceptually grounded categories codiWed in
language (e.g. Deacon 1997; Jameson 2005; Steels and Belpaeme 2005).
The paradox of dark, saturated red being selected as exemplary of what
is glossed as a ‘‘light/warm’’ term in Stage I color lexicons remains
inadequately addressed. Although not designed to address archeological
data, the Middle and earlier Late Pleistocene record of pigment use
presents several challenges to BCT theory. Why the overwhelming use of
just one color rather than the predicted binary opposition? Why a focus
on one term of the predicted pair rather than the other? Why red rather
than white or yellow? And why the focus on relatively dark reds?
Deacon’s qualiWed innatism opens the door to cultural factors—spe-
ciWcally ritual—impinging on color lexicalization, but it does no more
than this. The projection of wedding rituals back into the Plio-Pleisto-
cene—and with it the implicit antiquity of BCTs—makes it implausible
that the pragmatics of color terminology in extant cultures have any
bearing on the evolution of color terms.
Hovers and colleagues are oblivious to any contradiction in presenting a
thoroughly innatist model, and then (in discussion) invoking Deacon’s
arguments about the role of ritual in learning symbols. As with Deacon’s
hypothesized rituals, the pigments probably deployed in Qafzeh mortuary
rituals could as readily have been black or white as red.
Like BCT theory, Durkheim’s theory of collective representations is
non-Darwinian. However, his predictions regarding the form of early
ritual performance—involving the painting of geometric designs on the
bodies of ritual performers with red ochre—seem remarkably prescient in
view of the Blombos engravings. A necessarily circumstantial case has been
made for these predictions being met in the late Middle Pleistocene and
Red ochre, body painting, and language 91
early Late Pleistocene African record of ochre use. What is missing is an
evolutionary account that might account for this blood symbolism.
The Wve archeological predictions made by the FCC model are met by
the evidence outlined above. Several mid-Middle Pleistocene lineages may
have engaged in something like ‘‘sham menstruation,’’ but habitual col-
lective ritual can only be inferred among our immediate African ancestors,
perhaps initially restricted to local populations within the tropics, but
becoming generalized across Africa towards the end of the Middle Pleis-
tocene. Not only is there an almost exclusive focus on reds, but blood reds
seem to have been especially esteemed. When not locally available, people
would go some distance to procure them.
To summarize: Because collective rituals are costly, they demonstrate
commitment. A consequence of commitment is the generation of trust.
Once you have a ritual community within which there is suYcient trust,
you no longer need costly signals for internal use—you can aVord to
develop cheaper, coded forms of communication. Costly ritual continues
to be required for signaling to an ‘‘out-group’’ (e.g. potential mates), and
for the incorporation of new members (e.g. girls reaching reproductive
age) into the ritual coalition. Human speech communities were born out
of the regular performance of such costly displays.
92 Watts
Plate 3 Artifacts from the Still Bay levels at Blombos Cave: a) silcrete bifacial
point b) formal bone tool c) engraved ochre SAM-AA 8938 d) Nassarius kraus-
sianus shell beads e) ochre deposit on a shell bead (Images by C. Henshilwood
and F. d’Errico)
Plate 4.1 CI Siltstone, 56.3g, 73.9mm length, NCS streak 3545 Y60R. Two
adjacent pholadid borings, one with both valves of the pholadid in-situ. Uniden-
tified tests of marine organisms attached to adjacent surface.
Plate 4.2 CF Hematite ‘‘crayon’’. 3.7g, 34.1 mm length, NCS streak 3357 Y80R,
9 facets.
Plate 4.3 CFC Coarse siltstone ‘‘crayon’’. 0.6g, 17.3 mm length, NCS streak 3257
Y80R, 3 facets.
Plate 4.4 (a and b). CI Siltstone ‘crayon’. 3.9g, 41.3 mm length, NCS streak 4040
Y75R, 7 facets.
Plate 4.5 CB Coarse siltstone, intensively ground. 1g, 16.9 mm length, com-
plete, NCS streak 3550 Y70R, 9 facets.
Plate 4.6 CF Hematite, 1.5g, 21.6 mm length. NCS streak 5030 Y80R. Ground
on part of one main surface and two edges, illustrated facet 3.2 mm wide.
Plate 4.7 CI Coarse siltstone, lightly scraped. 13.6g, 38.3 mm length. NCS streak
3258 Y60R.
Plate 4.8 CGB Shale, lightly (edge) ground 1.3g, 33.9 mm length, NCS streak
3555 Y70R. Facet width 3 mm.
